Dams in Section

A book in thirteen chapters

Dams in Section

From a handful of sand to a dam you design

050 m
An earth and rockfill dam, cut across and drawn to scale: the kind of section this book builds up a piece at a time, from the ground to the finished dam.

Soil holds only as well as its grains press on one another, and the water between them can take that press away.

A dam is a heap of earth and rock, or a block of concrete, that holds back water or mine waste. This book cuts dams open the way engineers draw them, in section, and asks of each one the two questions in the sentence above. It starts with a handful of sand on a beach and ends with a dam you design and test yourself. Most of the drawings are live. Drag them, play them, and try to break them.

First edition · 2026

Contents

Preface

Stand on the crest of an earth dam and look down the far side. A slope of grass, a gravel road along the top, a ditch at the bottom where a little clear water runs. Behind you is a lake, perhaps tens of metres deep and several kilometres long. Between the two is a heap of soil and broken rock, much of it dug from the hills nearby and rolled flat in layers. That a heap of loose grains can hold a lake back for a century, and the few ways it stops doing so, is what this book is about.

It is written for a curious adult with school arithmetic: someone who can work out a percentage, read a graph and is willing to meet a tangent. The toolkit after Plate 0 checks every piece of maths the book uses, and that is all of it. If you are an engineering student you will find here the ideas of a first course in soil mechanics and dam engineering, worked with real numbers, along with plain statements of where the models are simplified.

Dams hold back more than drinking water and the water that turns turbines. They protect towns from floods, and at mines they hold what is left of the ore once the metal is out, ground to sand and silt and mixed with water. That waste has to stay put long after the mine closes. When a dam of either kind fails, everything it held comes out at once.

Most of the examples come from British Columbia. More than 1,900 dams there are regulated under the province's Water Sustainability Act.BC Dam Safety (2026). Tailings dams at mines are not in that count; they are regulated under the Mines Act. Four dams from the province run through these pages. Mount Polley, whose night in August 2014 opens the book, is the one the whole book works toward. The W.A.C. Bennett Dam on the Peace River shows what each zone inside an embankment is for. Coquitlam shows that the way a dam was built can matter more than what it holds. Testalinden, a small dam near Oliver, shows that inspection and regulation are part of a dam as much as its soil is.

By the last chapter you will be able to read a dam's cross-section the way an engineer does, work out how hard the water pushes on it and where, follow water through and under it, and say why one soil is strong and another weak. You will be able to check whether a slope will stand, name the ways embankments fail, and design a small dam of your own. You will also be able to say what your own check cannot see.

One sentence holds the book together. It is on the title page: soil holds only as well as its grains press on one another, and the water between them can take that press away. The first chapters build it from a beach, a bucket and a block on a ramp; Chapter 4 proves it; every chapter after that applies it. It does not cover everything that brings a dam down. A flood that pours over the crest can wash a heap away however hard its grains press, and Chapter 6 deals with that separately.

The figures run in your browser, among them a seepage solver, a slope-stability solver and a box of simulated grains. Each number the text quotes from a figure is checked by a test that recomputes it from the code behind the figure, so the words and the drawings can't drift apart. Numbers from the literature carry a note in the margin that names their source. The seven narrated films were made separately; each is captioned, and each has a transcript.

How to read this book

This is a book to read and a set of instruments to play. A few habits get the most out of both, and each is shown here once, working.

Commit before you look

Some figures wait behind a question. Pick an answer before you open them. A wrong guess you committed to teaches more than a right answer you only nodded at, because it shows you exactly which picture in your head needs mending. Try this one:

A glass is filled to the brim with dry sand. Can you still pour water into it without spilling any?

Figures you can play with

Most figures are instruments. The strip across the top gives each one its number, and on the right says whether it is drawn to scale or simplified. Plate 0, on the next pages, is a still drawing; Fig. 2.1 is live, and its sliders move a sensor through a lake. Every caption opens with the figure's claim in bold, says how to read the drawing, gives the one number worth keeping and ends with a short list of things to try. On a phone, every figure wider than the text has an Expand button that opens it full screen; turn the phone sideways to read it.

Each chapter opens with a small drawing of a dam that grows a piece at a time. The Prologue starts with bare ground, and by Chapter 13 it is the whole section; the words under it say what that chapter adds.

Every number the text quotes from a figure is the figure's own. A test recomputes each one from the code that draws the figure, so when Chapter 12 says a dam's slope has a factor of safety of 1.12 (its strength divided by the push on it; the Prologue explains it), the workbench in the last chapter shows 1.12 for the same dam.

Worked examples, and Set the figure

Each chapter works two to five problems with real numbers, one step to a line, with the reason for each step beside it (under it, on a phone). Most end with a Set the figure button, which loads the example's numbers into the live figure and takes you there; a pill at the foot of the screen brings you back. Here is the smallest one in the book:

Example 0

How hard does the water press 10 m down?

Given
A sensor 10 m below the surface of still water. Water weighs 9.81 kN per cubic metre.
Find
The water pressure on the sensor.
The idea
Every square metre at that depth carries the weight of the column of water standing on it.
  1. column above one square metre: 10 m × 1 m² = 10 m³the water standing on the sensor's patch
  2. its weight: 10 × 9.81 = 98.1 kNeach cubic metre weighs 9.81 kN
  3. p = 98.1 kN ÷ 1 m² = 98.1 kPapressure is force per area; 1 kPa is 1 kN/m²

The water presses 98.1 kPa on the sensor.

Check. Set the figure: it rounds to whole kilopascals and shows 98. Double the depth and the column doubles, so at 20 m it shows 196 kPa; at the surface, with no column, zero.

What it means. Ten metres of water press about as hard as the whole height of the air above us. The size of the lake appears nowhere in the sum. Chapter 2 asks whether that can really be right.

Notes in the margin

Notes sit in the right-hand margin beside the line they belong to, and on a phone they fold away behind a small number or ⊕ that you tap to open.Like this one. Nothing the argument needs is ever in a note; you can skip every one and lose nothing but the pleasure. They come in four kinds. Source notes name where a number from the literature comes from, with the page or table when there is one; the Prologue has several. Units notes translate a unit into something you can feel.A kilopascal is a kilonewton spread over a square metre: about the weight of a 100 kg mass, 0.981 kN, resting on a square metre. History notes say who found something out, and when. For the curious notes hold a derivation or an aside for anyone who wants to go a step further.

Check yourself, and the questions at the end

Inside each chapter, a few one-line questions fold open to show their answer:

Check yourself A dam's crest is 24 m above the ground, and the lake behind it is 20.4 m deep. How far below the crest is the water?

3.6 m. That margin has a name, the freeboard, and Plate 0 shows it on this very dam.

Each chapter ends with a box of what to remember and then four to six questions. A tag says what kind each one is, and the dots say how hard:

Every question has a hint under it, and the answers, with their working, are collected at the back of the book. After the questions come two or three readings, each marked gentle, serious or primary source, and a card that points to the next chapter.

Films

Seven narrated films, 58 minutes in all, retell the story with moving drawings. Each sits in the chapter it retells most closely, and several range over other chapters as well; the note before each film says where. Before it plays, a film asks its question. Under it, chapter marks jump to the part you want, each linked back to the figure it retells, and a transcript is there for reading instead of watching. Film 2.1 is the first you will meet.

Words, scale and typical values

A dotted underline marks a word defined in the Glossary at the back: hover over it, or tap it on a phone, to read the definition, as with crest. Plate 0 names the parts of a dam and gives the book's three conventions.

All drawings are simplified. Those marked "drawn to scale" are drawn to scale, from the code's own dimensions or from published ones; the rest are schematic, and say so. The soil properties are typical values for teaching, not design values: a real dam is designed from tests on its own soils.

Time, and a tutor

Plan on eight to ten hours for the whole book, more if you work the questions, plus the films. Each chapter's opening page says how long it takes to read, and the bar at the top of the screen shows where you are and opens the Contents. If you are reading on claude.ai, each live figure also has a Why did that happen? button that asks Claude about exactly what the figure is showing you; the standalone page does not have it.

Plate 0 · How to read a section

A section is what you would see if you cut a dam straight across with a very large knife, lifted one half away and looked at the cut face, the way you look at the layers in a slice of cake. Engineers draw dams this way because the things that decide whether a dam stands (how deep the water is against it, where water seeps through it, which slope might slide) all show in that one cut, and along most of a long dam one slice looks much like the next. So the book works per metre of dam: picture a slice one metre thick.

The plate names every part of a section the book uses, on the dam it keeps coming back to:

Plate 0 · How to read a sectionDrawn to scale · our dam, 24 m high
upstream downstream 1 crest 6 m 6 freeboard 3.6 m 7 reservoir 20.4 m 2 upstream face 3 downstream face 2 1 11 slope 2H:1V 9 phreatic surface 8 tailwater 24 m 4 heel 5 toe 102 m 10 foundation 0 10 20 m
  1. →Upstream and downstream. Water flows from left to right in every section in this book, so the upstream side, where the water is held, is always on the left.
  2. 1Crest. The top of the dam, here 6 m wide and 24 m above the original ground.
  3. 2Upstream face. The slope the reservoir leans on.
  4. 3Downstream face. The slope on the dry side. Where water comes out of it, the wet strip is the seepage face; here it reaches 2 m up from the toe.
  5. 4Heel. Where the upstream face meets the ground.
  6. 5Toe. Where the downstream face meets the ground. Water that seeps through or under a dam usually comes out at or near the toe, so that is where an inspector looks first.
  7. 6Freeboard. The crest's height above the water: here 24 − 20.4 = 3.6 m, the margin left for waves and floods.
  8. 7Reservoir. The water held back, 20.4 m deep at the dam.
  9. 8Tailwater. The water below the dam. Here it lies level with the ground, as low as the water can fall.
  10. 9Phreatic surface. The top of the water inside the dam, where the water pressure is zero. Below it the gaps between the grains are full of water under pressure (the blue-ruled shading); above it the soil is drier. This one is the line the seepage lab computes for this dam.
  11. 10Foundation. The ground the dam stands on, here river sand and gravel, drawn 15 m deep as the seepage lab models it.
  12. 11Slope notation. 2H:1V means 2 m across for every 1 m up, an angle of 26.6°. A larger first number means a flatter slope. Both faces of this dam are 2H:1V.
  13. ↓Abutments. The valley sides the two ends of a dam rest against, shown in the plan below. Left and right are named as you look downstream.

In plan: the same dam seen from above

The dashed line A–A marks where the section above is cut.

A A river reservoir crest left abutment right abutment

Three conventions

  • Water pressures are gauge pressures, measured above the pressure of the air (Chapter 2): still water's surface is at zero.
  • Forces and flows are per metre length of dam, as if the slice were one metre thick.
  • "El." is an elevation, a height above sea level; other heights and depths are measured from the ground or from the water's surface.

Materials: the sixteen hatches

  • Bedrock
  • Alluviumriver sand and gravel
  • Clayand clay cores
  • Filter sand
  • Transitionsandy gravel
  • Rockfill
  • Ripraparmour stone
  • Concrete
  • Fillcompacted earth
  • Drain gravel
  • Grout curtainin rock
  • Tailings sand
  • Cycloned sandtailings sand, sorted
  • Slimesfine tailings
  • Saturatedlaid over any soil: every gap full of water
  • Liquefiedgrains afloat, the soil flowing

Colours: one meaning each, all through the book

  • Blue: water, its pressure in the ground, and its head (the height it would rise to in a tube, Chapter 3).
  • Brick red: whatever pushes toward failure, and the places it shows, such as a seepage face.
  • Green: whatever resists, and a check that passes.
  • Ochre: grains pressing on grains, the effective stress.
  • Ink greys: structure, concrete, and anything neutral.

Every section in this book is read the same way: the dam cut straight across, with the water on the left and flowing right. The hatch says what each part is made of; the colour of a line or arrow says what it means, and the words that go with it say the same, so no colour has to be read alone. This is our dam, the embankment the seepage lab and the workbench share, with the water inside it where the lab computes it to be. The number to remember: 2H:1V, two across for every one up, a slope of 26.6°.

  • Measure the freeboard against the scale bar: it should come out at 3.6 m.
  • Find this dam again in Fig. 10.3, which opens on exactly this section, and watch the phreatic surface move as you lower the reservoir.
  • Find it once more in Fig. 13.1, where you paint it yourself.

The toolkit: a self-check

The book uses arithmetic, a little trigonometry and no algebra beyond turning a formula round. These eight questions cover all of it, and each uses a number you will meet later. Answer in your head or on paper first, then open the answer.

1 · Powers of ten The workbench's rockfill has a permeability of 10⁻⁴ m/s; its clay, 10⁻⁹ m/s. How many times more easily does water pass through the rockfill?

10⁻⁴ ÷ 10⁻⁹ = 10−4 + 9 = 10⁵, a hundred thousand times. Dividing powers of ten subtracts their exponents: −4 − (−9) = 5.

2 · A slope as an angle Both faces of our dam slope at 2H:1V, two metres across for every metre up. What angle does each face make with the ground?

The tangent of the angle is rise over run: tan β = 1 ÷ 2 = 0.5, so β = tan−1 0.5 = 26.6°. On a calculator, tan−1 is often marked atan or arctan, and it must be set to degrees.

3 · The area of a section Our dam is 24 m high, with a crest 6 m wide and both faces at 2H:1V. What is the area of its cross-section?

Split it into a rectangle under the crest and a triangle on each side. Each face runs 2 × 24 = 48 m across, so each triangle is ½ × 48 × 24 = 576 m². The rectangle is 6 × 24 = 144 m². In all, 1,296 m²: that many cubic metres of fill for every metre of dam, and the same 1,296 one-metre squares you will paint in the workbench at the end of the book.

4 · Mass and weight What does a 100 kg mass weigh, in kilonewtons?

Weight is mass times gravity: 100 kg × 9.81 m/s² = 981 newtons, or 0.981 kN. A tonne weighs 9.81 kN, and a cubic metre of water has a mass of one tonne, so it weighs 9.81 kN. That number turns up on almost every page.

5 · Pressure A tonne of water stands on one square metre. How hard does it press, in kilopascals?

9.81 kN spread over 1 m² is 9.81 kN/m², and a kilopascal (kPa) is exactly one kilonewton per square metre: 9.81 kPa. A column of water one metre deep presses that hard on its base; ten metres press 98.1 kPa.

6 · A log axis On an axis whose ticks read 10⁻⁶, 10⁻⁵, 10⁻⁴, what value sits exactly halfway between the first tick and the last?

10⁻⁵, the middle tick. On a logarithmic axis each step multiplies by the same factor, here ten, so halfway between 10⁻⁶ and 10⁻⁴ is the value ten times the first and a tenth of the last. The tempting answer, 5 × 10⁻⁵, sits much nearer the 10⁻⁴ end. The book needs one for permeability, which runs from 10⁻² m/s for gravel to 10⁻⁹ m/s for clay.

7 · Turning a formula round A formula says q = k · i · A. If q/A = 7.0 × 10⁻⁵ m/s and i = 0.70, what is k?

Divide both sides by i · A: k = (q/A) ÷ i = 7.0 × 10⁻⁵ ÷ 0.70 = 1.0 × 10⁻⁴ m/s. These are the numbers of the sand in Darcy's pipe in Chapter 3, and the formula is his law.

8 · Significant figures The seepage lab opens with 2,366 litres a day seeping past each metre of dam. Write that to two significant figures.

2,400. Keep the first two digits, 2 and 3, and let the next one, 6, round the 3 up. The lab works with typical soil properties, not measured ones, so two figures are as many as its answer deserves.

Missed two or more? Appendix A opens with a refresher on each of these.

Prologue

4 August 2014

How does a dam that passed its calculations fail in a night?

The dam, piece by piece so far, only the ground

1 figure · about 5 min to read

At half past ten on the night of 3 August 2014, the night shift's logbook at the Mount Polley mine, in the hills of central British Columbia, recorded three things about the tailings dam: "Good berm, good slope, no visible cracks." (A berm is a bank or ledge of earth.)Mount Polley Review Panel (2015), the timeline of the night, p. 10. The mine is about 65 km northeast of Williams Lake: Chief Inspector of Mines (2015), p. 1. At a quarter to twelve another worker drove along the top of the dam, across the stretch that was about to fail, and back again, and noticed nothing unusual.

At a quarter to one, the water in a small pond below the dam, which collects what seeps through it, began to rise. By one o'clock it was rising fast. At a quarter past one the lights went out in the electrical shop and the mill shut down. At five past two an operator found that the dam had been breached.The same timeline, p. 10, from the mine's reports and the panel's interviews.

What came out was more than water. A tailings dam holds back tailings: the rock that is left once the copper and gold have been taken out of it, ground to sand and silt and pumped in as a slurry. Over the next sixteen hours the gap kept widening, and between 21 and 25 million cubic metres of water and tailings came through it, about ten million of them water from the pond.Chief Inspector of Mines (2015), p. 13; the range is the mine owner's estimate. Spread a metre deep, that would cover 21 to 25 square kilometres.

It washed away a stand of forest and ran two ways. Some went into Polley Lake, where the debris plugged the lake's outlet and raised its level by about a metre. The rest went down Hazeltine Creek, a narrow stream, and scoured it into a mudflow 150 metres wide that ran 7 kilometres to Quesnel Lake, one of the deepest lakes in the province.Chief Inspector of Mines (2015), p. 13; on who was below the dam, p. 160. Nobody was at work below the dam that night, or out on the lake or the creek, and nobody was killed.

P.1 A dam that passed

The dam had been watched. Its first, low embankment, the starter dam, was built in 1996, and it had been raised in stages ever since, a few metres at a time, as the mine filled it; the ninth stage was being finished that summer. By that night it stood about 40 metres high where it failed, its crest 1 metre short of the height it was permitted to reach, and that evening the pond stood 2.3 metres below the crest (taken, as the review panel took it, at the top of the dam's core, the watertight zone inside it): its freeboard. Engineers designed each raise and checked it, the dam's engineers inspected it once a year, and the province's geotechnical inspectors, engineers who specialize in soil and rock, visited the site.Mount Polley Review Panel (2015), pp. 5 and i; Chief Inspector of Mines (2015), pp. 3, 5 and 47.

The checks said it was safe. When the side of the dam that later failed was checked during the raise of 2007 to 2011, its calculated factor of safety was 1.7; its designers required 1.3.Mount Polley Review Panel (2015), p. 64; for 2012, p. 79. A factor of safety is the strength holding a slope up divided by the push trying to bring it down, so 1.7 meant 70 per cent more strength than the slope needed. In 2012 its engineers judged the numbers high enough that the instruments measuring water pressure in that side needed no alarm levels.

Here is that side of the dam, cut across where it failed, drawn to scale from the panel's own section:

Fig. P.1 · The dam that failed, the evening beforeDrawn to scale · section from the Review Panel
gone by morning water and tailings, for about 16 hours 2.3 m top of core El. 969.1 m pond El. 966.8 m tailings embankment core, filters and rockfill about 40 m toe original ground 970 950 930 El. (m) 0 20 m upstream downstream gone by morning crest pond 2.3 m tailings embankment about 40 m toe 970 950 930 0 20 m
Polley Lake Quesnel Lake Hazeltine Creek, 7 km tailings breach mine N

On the evening of 3 August 2014 this section stood about 40 m high with its pond 2.3 m below the top; by morning everything inside the dashed red line was gone. The section is the Perimeter Embankment, the north side of the Mount Polley tailings dam, cut across at the place it failed, as it stood on the evening of 3 August 2014: the pond and the tailings deposited behind it on the left, the embankment in the middle, the original ground beneath. The dashed red outline marks what was gone by morning. It is traced from the panel's own section at the breach (Review Panel, p. 70), with the crest and pond levels it reports for that night (p. 6). The map is a sketch, not to scale. The number to remember: 2.3 m, the height of the top of the dam's core above the pond that evening, the margin the panel measured.

  • Use the scale bar to measure the embankment along the original ground, from where its left edge meets the ground to the toe.
  • Compare its downstream face with our dam's in Plate 0. Which is steeper?
Answers

About 77 m. Measured from the crest's edge to the toe on this drawing, its downstream face runs about 1.4 across for every 1 up, steeper than our dam's 2:1.

Notice that the pond is only a thin skin on top of the tailings, and how big the heap is that holds them: a bank of broken rock as tall as a twelve-storey building. Nothing in the drawing looks fragile.

So how does a dam that passed its calculations fail in a night? The province's Chief Inspector of Mines, in his own investigation, put the puzzle plainly: "The failure of the dam took place suddenly, without any warning signs."Chief Inspector of Mines (2015), p. 13. The province, with the Williams Lake and Soda Creek Indian Bands, set up a panel of three engineers to find out why. It was chaired by Norbert Morgenstern, whose name is on one of the standard methods of checking whether a slope will stand, and one of its members, Steven Vick, had written a standard textbook on tailings dams.Mount Polley Review Panel (2015), pp. i and 1; the textbook is Vick (1990). They had the site drilled and sampled and the samples tested, read the design reports of eighteen years, and reported in January 2015.

Their answer is in the last chapter of this book, and by then you will be able to follow it: to see what they found, and why the calculations said the dam was safe. It takes nothing more exotic than the thing every chapter between here and there is about: grains, and the water between them. We start somewhere smaller, on a beach.

Part I

Ground

By the end of Part I

Before there is a dam there is ground: grains of sand and the water between them. These five chapters start on a beach and end in a box of simulated grains. Along the way they build the one idea the rest of the book stands on.

  1. Chapter 1The beach

    Three kinds of sand, and the water that makes the difference.

  2. Chapter 2Depth, not size

    Why the depth of the water matters and the size of the lake doesn't.

  3. Chapter 3Which way the water goes

    Head, Darcy's law, and the clay that takes almost the whole drop.

  4. Chapter 4Grip

    Friction, a block on a water cushion, and Terzaghi's effective stress.

  5. Chapter 5When the water can't get out

    Loose and dense, drained and undrained, and liquefaction.

Chapter 1 · Part I · Ground

The beach

The same sand, and only the water has changed. Why does it behave three ways?

When the falling tide leaves the sand firm, as the foot falls on it the sand whitens, or appears momentarily to dry round the foot.

So far. The Prologue left a dam that had passed its calculations and failed in a night. Following the engineers who worked out why takes a handful of ideas about soil and water, and the cheapest place to collect them is a beach.

After this chapter you can

  • describe the three ways one sand behaves, and name the one thing that changed between them;
  • measure an angle of repose, and write any slope both as an angle and as a ratio, across to up;
  • say why a damp sand wall stands 30 cm tall and never 30 m.

This chapter adds a heap of sand at its angle of repose

2 figures · about 15 min to read

Walk down a beach at low tide, from the dunes to the sea, and the sand changes under you three times. Up by the dunes it is dry. Your foot sinks in, the sand pours back round your ankle, and when you kick it, it runs.

Halfway down, where the sea was an hour ago, the sand is dark and firm. It takes a crisp footprint, and as your weight comes down a pale ring spreads round your foot for a moment, as if the sand had dried where you stood. Then the colour comes back.

At the edge, where every wave lays down a fresh sheet of sand and soaks it, you can stand quite happily as long as you keep still. Wiggle your feet and the sand goes soft. You sink to the ankles, and it takes a tug to get out.

Three steps, one sand. The grains are near enough the same all the way down the beach. What changes is the water in the gaps between them. At the top there is almost none. In the middle every gap is still full, but the water is held up in the sand by the pull of its own surface, the way a sponge lifted out of the bath stays full, and the grains there are packed tight. At the edge the gaps are full too, and the grains lie loose.

1.1 Three sands

Take those three sands home in buckets. On the way the middle sand drains a little, as sand lifted out of water does, and arrives damp: air in most of its gaps, and water clinging only where the grains touch. Then the three keep their characters.

Dry sand pours. Tip it out and it spreads into a low heap whose sides settle at one slope, however carefully you pour.

Damp sand stands. Pack it into the bucket, turn it out, and you get a castle with walls that go straight up.

Loose wet sand flows. Where every gap is full of water and the grains were laid down loosely, the sand bears you while you keep still and turns to something like soup when you shake it.

Ask people what water does to sand and most of them say it makes it slippery, the way water makes a tiled floor slippery. Keep that answer in mind, because the castle is already evidence against it. Damp sand has water in it, and it stands steeper than either of the others. Whatever the water is doing in the castle, it is not greasing the grains. Chapter 4 settles the question with a block on a ramp. For now, notice only that more water does not mean weaker sand, or stronger: the soup holds the most water and the castle the next most, and they are the weakest and the strongest of the three.

The three sands also differ in a second way, which the beach hides. Sand that a wave has just dropped lies loosely packed, with wide gaps. Sand that has been walked on, rained on and settled for a tide lies densely packed, each grain nestled into the hollows of its neighbours. Wet sand that is saturated (every gap full of water, no air at all) can be either.

Below, one heap of sand is drawn four ways: dry, damp, sitting under still water, and with water draining out through its faces. A switch sets how tightly it was packed. Pick a state, then press Jiggle to shake the ground:

Fig. 1.1 · The same sand, different water Schematic · castle about 30 cm tall
Sand
Packing

Close-up: a little water gathers in rings where the grains touch.

Steepest slope it holds90°a vertical wall
In the gapsAir + waterrings of water where grains touch
After a jiggle–press Jiggle to shake it

One heap of sand, dry, damp and wet in two ways: only the water in its gaps changes, and the heap's shape follows it. The heap is a slice across a long ridge of sand, and every shape it takes holds the same amount of sand: the area of the slice stays fixed. The close-up shows what fills the gaps: air, small rings of water where grains touch, or water everywhere. The red arc marks the steepest slope each heap holds. The numbers to remember: dry sand stands at 34°, damp sand holds a vertical wall, sand under still water stands about as steep as dry, and with water draining out of its faces it slumps to about 19°. Why each happens is the business of Chapters 4 and 5.

  • Switch from Damp to Dry and watch the castle spread into a heap two-thirds as tall.
  • Pick Under still water, then Water draining out. Which one flattens the heap?
  • Under still water, press Jiggle with Loose packing, then with Dense. Picking a state or a packing again rebuilds the heap.

Notice first that the heap never gains or loses a grain. When the castle goes dry it spreads out, and the drawing keeps the area of its slice exactly the same, so a lower heap must be a wider one.

Then notice the two wet states. Sand that is wet all through sounds as if it ought to slump flatter than dry sand; wet sand, after all, is supposed to be slippery. Pick Under still water and the heap stands about as steep as dry sand, at about 34°. Still water hardly changes its slope. Now pick Water draining out. This is the state of a beach face at low tide, where the water that soaked in at high tide seeps back out through the sand, and the heap slumps to about 19°. The water is the same water in both. In one it sits still; in the other it moves, and the moving water drags the grains downhill as it leaves. Why that drag costs about fifteen degrees is a sum of a few lines, worked in Example 4.5.

Last, notice the jiggle. Dry sand and damp sand barely change. Wet sand packed loosely flows flat and swallows the flag, which is the soup at the water's edge. The same wet sand packed densely holds. So saturation alone does not turn sand to soup; it takes loose sand as well. That is the water's edge and the firm middle of the beach in one experiment, and it is also the first hint about the pale ring round your foot, which Chapter 5 explains.The pale ring was described in 1885 by Osborne Reynolds, the Manchester engineer whose name is on the Reynolds number of fluid flow, in the paper that coined the word dilatancy. The epigraph to this chapter is his description of it.

1.2 The angle of repose

The slope a dry heap settles at has a name, the angle of repose, and you can watch it form. Pour dry sand from a funnel onto a table. Grains land on the top and roll down the sides until the sides reach the steepest slope on which a grain can rest without rolling. Pour more and the heap grows, but its sides keep that one slope, because any extra steepness slides off in small avalanches. The slope of its sides is set by how hard the grains grip one another, not by how much sand you pour.Charles-Augustin Coulomb spent 1764 to 1772 building a fort on Martinique for the French army. In 1773 he read the Paris Academy a memoir on walls, arches and the earth behind them. In its worked examples, earth that, left to itself, settles at a slope of 45° is taken to have friction equal to the press on it: the slope of a loose heap measures the friction of its grains (Coulomb (1776), pp. 361–362).

Dry sands settle in the low-to-mid thirties of degrees. Angular, rough grains stand a little steeper than round, polished ones, and the lee faces of dunes, where wind-blown sand piles up until it avalanches, stand at about 32–34°.Dune faces, measured by laser scanning: Pelletier et al. (2015), who give 32–33° for the faces they surveyed in Brazil and quote earlier surveys at 32° settled and 34° just before an avalanche. Grain shape and friction set the dry angle: Hornbaker et al. (1997). The sand drawn in Fig. 1.1 settles at 34°, near the top of that range. Every slope in the figure follows from that one angle and the weights of sand and water.

The figure draws its heap as a slice. Picture the castle as a long wall of sand running away from you, like a sea wall, and the drawing as one slice cut straight across it. When the wall slumps, each slice keeps its own sand, so what stays the same is the area of the slice, which engineers call the section. Every drawing in this book is a section of something long, and every force and every flow is counted per metre of its length. A dam is drawn the same way.

That gives a sum you can do. The castle in the figure is a slice of known area. Poured dry, the same sand makes a triangular heap with 34° sides. How tall?

Example 1.1

Same sand, new shape

Given
The damp castle's section, as the figure draws it: 20.4 cm across and 29.2 cm tall, less two notches between the battlements, each 3.2 cm wide and 3.6 cm deep. Its area is A = 20.4 × 29.2 − 2 × 3.2 × 3.6 = 572.64 cm². Dry, the sand stands at β = 34°.
Find
The height and width of a dry heap holding the same sand.
The idea
The dry heap's section is a triangle with 34° sides. Write its area in terms of its height, set it equal to the castle's, and solve for the height (Fig. 1.1).
  1. half-base = H ÷ tan βeach side climbs H over a run of H/tan β, the definition of the slope's tangent
  2. A = ½ × (2H ÷ tan β) × H = H² ÷ tan βa triangle's area is half its base times its height
  3. H = √(A · tan β)rearranged for the height
  4. tan 34° = 0.67451from a calculator; the steps below keep enough digits to come out the same on yours
  5. A · tan β = 572.64 × 0.67451 = 386.25 cm²the same sand, now on a 34° slope
  6. H = √386.25 = 19.653 cmthe height of the dry heap
  7. half-base = 19.653 ÷ 0.67451 = 29.137 cm, so the base is 58.27 cmback to step 1 with the height known

The dry heap stands 19.7 cm tall on a base 58.3 cm wide: 0.67 times the castle's height and 2.9 times its width.

Check. Put the area back together: ½ × 58.27 × 19.653 = 572.6 cm², the castle's area. And the limits behave: a gentler slope (smaller tan β) gives a lower, wider heap, and a slope near vertical gives a tall thin one, more like the castle.

What it means. The castle lost no sand when it dried; it lost the grip that held its walls up, and the same sand then needed about three times the ground. Soil that cannot hold a steep face has to be laid out wide. Most of the width of an earth dam is the price of its slopes.

You can measure an angle of repose at home with a funnel, a ruler and a cup of dry sand, sugar or rice. Pour a heap, measure its height and the width of its base, and the tangent of the angle is the height divided by half the base. The questions at the end of the chapter ask you to try.

1.3 A slope is a number

Angles are awkward on a building site. Nobody holds a protractor against a hillside of fill. Engineers write a slope as a ratio instead, how far across for each unit up, and they put the across first: a slope of 2H:1V (two across for every one up) climbs one metre for every two metres of ground it covers. This book writes it 2H:1V, or 2:1 for short. The bigger the first number, the flatter the slope.Road signs turn the same ratio upside down and give it as a percentage, rise over run: a 10% hill climbs 1 m for every 10 m across, which this book would write 10:1.

The ratio and the angle are tied by one piece of trigonometry. The tangent of a slope's angle β is the rise over the run, so for a slope written n:1, tan β = 1 ÷ n. A dam drawn at 2:1 and a heap at 34° can then be compared on the same footing:

Example 1.2

A slope as a number

Given
A dam face at 2:1. The dry heap's 34° and the draining face's 19.0° from Fig. 1.1.
Find
The angle of the 2:1 face, and the two heap slopes written as ratios.
The idea
tan β = rise ÷ run = 1 ÷ n, which works in both directions.
  1. tan β = 1 ÷ 2 = 0.5one up for every two across
  2. β = arctan 0.5 = 26.6°the angle whose tangent is 0.5
  3. n = 1 ÷ tan 34° = 1 ÷ 0.6745 = 1.48across for one up, for the dry heap
  4. n = 1 ÷ tan 19.0° = 1 ÷ 0.344 = 2.9the same for the face with water seeping out

A 2:1 face slopes at 26.6°. The dry heap stands at 1.48:1, and the draining face at 2.9:1.

Check. A 1:1 slope is 45°, as steep as it is wide, and a 3:1 slope is arctan ⅓ = 18.4°. A 2:1 slope must lie between them, and 26.6° does.

What it means. A 2:1 face built of this beach's sand would stand dry, with room to spare, and would slump wherever water seeped out of it. Keeping water from coming out of the face of a dam, or catching it safely where it does, is much of Part III of this book.

Check yourself A road sign warns of a 10% hill. Write the slope as a ratio, and as an angle.

10:1, ten across for each one up. tan β = 0.1, so β = 5.7°: a hill that feels steep in a car is far flatter than any heap of sand.

1.4 Why a castle stays small

Look again at the close-up in Fig. 1.1 with the sand set to Damp. The water does not fill the gaps. It gathers in small rings where one grain touches the next, and the surface of each ring is curved and pulled tight, like the skin on a drop hanging from a tap. That skin pulls the two grains together. Every grain has several, and a handful of sand holds a great many grains, so together the rings give damp sand a grip that dry sand lacks, and that grip holds a wall up where dry grains would roll away.It takes remarkably little liquid. Films of oil less than 50 nanometres thick on plastic beads about 0.8 mm across raised the steepest slope a heap would hold sharply (Hornbaker et al. (1997)). Columns of beach sand were strongest with about 1% water by volume, and the tallest column that would stand grew as its base radius to the power ⅔ (Pakpour et al. (2012)).

But the grip is small, and it does not grow as the castle does. Each ring pulls with the same tiny force however much sand sits above it. The weight that the bottom of a wall must carry grows with every centimetre of height. A wall 30 cm tall puts a hundred times less weight on each square metre of its base than a wall 30 m tall of the same sand, and somewhere between the two the weight wins. Sand sculptors do build castles several metres high, but they build them on broad bases that lean on their own width.Sandcastles "of several meters high", and the rule that the safe height depends on the width of the base: Pakpour et al. (2012). Chapter 5 puts a number on the grip.

There is a second reason a dam engineer never counts on it. The rings exist only while the sand is damp. A week of rain fills the gaps, the rings vanish, and the grip goes with them; a drought dries them out and does the same. A castle slumps as the rising tide soaks it for the same reason, before the waves reach it. Dams are designed as if the grip were not there.

Check yourself Of the three beach sands, which holds the most water, and which stands steepest?

The soup at the water's edge holds the most, and it stands at nothing at all once shaken. The damp castle holds a little and stands vertical. Water can make sand stronger or weaker; what matters is where the water sits and what it is doing.

1.5 The beach inside a dam

Now scale the heap up to a hill, and lean a lake against one side of it. That is an earth dam. The water in the reservoir soaks into the soil and creeps slowly through it, so inside the dam there is a boundary between soil that is full of water and soil above it that is not. Engineers call that boundary the phreatic surface: the water line inside the dam. It starts at the reservoir level on the upstream face, the side the water comes from, and falls as it crosses toward the downstream face, the side the water is trying to reach. (Every section in this book is drawn with the water on the left, so it flows left to right.) Here is a dam with the three beach sands mapped onto it:

Fig. 1.2 · The beach inside an earth damSchematic · faces at 2:1
2 1 reservoir soaked dry at the surface damp water line seeps out

An earth dam holds all three beach sands at once. The dashed blue line is the water line inside the dam, the phreatic surface. Below it the soil is soaked, every gap full of water creeping from the reservoir toward the far side, like the wet sand at the water's edge. Above it the soil is damp, like the castle, and the surface of the crest dries out like the top of the beach. Where the water line meets the downstream face, the water seeps out over the last stretch of slope to the foot (red). That stretch is the state Fig. 1.1 calls Water draining out, where its heap stood at only 19° against 34° dry. Plate 0 names every part of a section like this one.

  • Set Fig. 1.1 to Water draining out and compare its face with the foot of this dam.

Notice where the water comes out. At the foot of the downstream face, the soil is in the state that slumped the heap in Fig. 1.1 from 34° to 19°, and the face there is 2:1, 26.6°. A dam of beach sand would crumble at its foot, grain by grain, and the slope above would follow. Real dams are built of better-chosen soils, and a well-designed one catches the water in drains and filters inside the dam before it can reach the face, and how they do it is the work of Chapters 9 and 11. But the danger is the beach's, at the scale of a hill.

Notice too that the upstream face, under the reservoir, is the heap under still water: soaked, and standing about as steep as it would dry. The water there is not quite still, since it is soaking slowly into the dam, and Chapter 3 follows where it goes. The damp soil above the water line has the castle's grip, and it is the grip a designer leaves out. Everything below the line is held up by grains pressing on grains, with water filling every gap between them.

1.6 One sentence for the whole book

All of this can be said in one sentence. It is printed on the title page, and every chapter after this one comes back to it:

Soil holds only as well as its grains press on one another, and the water between them can take that press away.

You can't prove it yet, and you shouldn't believe it yet. The castle seems to contradict it: there the water makes the grains press harder. By the end of Chapter 4 you will be able to prove it, castle included, with a block on a ramp and a column of soil. The sentence does not explain every way a dam can fail. Floods that pour over the top are a different story, told in Chapter 6. But it explains most of what this book asks of soil.

To get there you need two facts about water first. Its pressure grows with depth, which is Chapter 2. And it moves through the ground, slowly or very slowly depending on the ground, which is Chapter 3. Chapter 4 puts them together, and Chapter 5 comes back to this beach for the soup and the pale ring round your foot.

What to remember

  • One sand pours, stands or flows, and only the water between its grains changed.
  • Dry sand stands at its angle of repose, 34° here. Still water barely changes that; water seeping out of a face drags the grains downhill and the face slumps to about 19°.
  • Saturated sand turns to soup when shaken only if it is loosely packed.
  • A slope is written across to up, n:1, with tan β = 1 ÷ n; 2:1 is 26.6°.
  • Damp sand's grip is small and does not grow with height, so castles stay small and dams are designed without it.

Questions

  1. 1.1

    As the heap in Fig. 1.1 changes shape, one thing stays fixed. What is it, and how could you check it with the numbers in Example 1.1?

    Hint

    Work out the area of the dry heap's triangle and compare it with the castle's.

  2. 1.2

    Set the heap under still water, packed densely, and jiggle it. Then switch to loosely packed and jiggle again. What differs, and what does that tell you about the soup at the water's edge?

    Hint

    Read the "After a jiggle" readout each time.

  3. 1.3

    Take damp sand to weigh about 18 kN per cubic metre (an illustrative round number). How hard does the sand press on each square metre at the foot of a damp wall 30 cm tall, and at the foot of one 30 m tall?

    Hint

    A kilonewton (kN) is a unit of force: 18 kN is the weight of about 1.8 tonnes. A column of sand one metre square and h metres tall holds h cubic metres of sand, all resting on its square-metre base.

  4. 1.4

    At home, pour dry sand, sugar or rice through a funnel onto a flat plate until the heap stops growing taller, and measure its height and the width of its base. What is its angle of repose? Is it steeper or shallower than the figure's 34°?

    Hint

    tan β = height ÷ half the base = 2h ÷ d. Measure across the middle of the base, and pour slowly.

  5. 1.5

    Write the slopes 1.5:1, 2:1 and 3:1 as angles. Which of them could a heap of the figure's dry sand stand at?

    Hint

    tan β = 1 ÷ n, then take the arctangent.

  6. 1.6

    You must build a sand embankment of the figure's sand, and water will seep out of the lower part of its downstream face. The workbench in Chapter 13 offers faces of 1.5:1, 2:1, 2.5:1 and 3:1. By the rule of Fig. 1.1, which of these faces could the wet part stand at?

    Hint

    Compare each angle with the draining face's 19.0°.

Further reading

  • Pakpour, M., Habibi, M., Møller, P., Bonn, D. (2012). How to construct the perfect sandcastle. Scientific Reports 2: 549. doi:10.1038/srep00549 gentle Short and free to read: why a sandcastle can be taller when its base is wider, and why a splash of water does more than a bucketful.
  • Heyman, J. (1972). Coulomb's Memoir on Statics: An Essay in the History of Civil Engineering. Cambridge: Cambridge University Press. serious Coulomb's 1773 memoir in English, with a commentary that walks through its sums on walls, arches and the friction of earth.
  • Reynolds, O. (1885). On the dilatancy of media composed of rigid particles in contact. With experimental illustrations. Philosophical Magazine, Series 5, 20(127): 469–481. doi:10.1080/14786448508627791 primary source The paper that named dilatancy; the footprint is on p. 475. It is also reprinted in the second volume of Reynolds's collected Papers on Mechanical and Physical Subjects (1901), free online.

Chapter 2 · Part I · Ground

Depth, not size

Two dams hold water 50 m deep, one a long lake and one a narrow slot. Which has to be stronger?

So far. On the beach one sand behaved three ways, and only the water had changed. To say why, we need two facts about water first. This chapter is the first of them: how hard still water presses, and on what. Chapter 3 adds the second, what happens when it moves.

After this chapter you can

  • work out the water pressure at any depth, and the push of a reservoir on each metre of a dam;
  • say where that push acts, and why doubling the depth makes it four times bigger and its tipping effect eight times bigger;
  • explain why a concrete dam is a triangle, and why water on an embankment's sloping face presses it down harder than it pushes it sideways.

This chapter adds water against the left face, with the pressure triangle

4 figures · about 20 min to read · Film 2.1, 3 min 11 s

Stand on the crest of a big dam on a still day and look upstream. The reservoir runs back between the hills further than you can see, millions of tonnes of water, and it is hard not to feel all of it leaning on the wall under your feet. Somebody had to decide how strong to make that wall, so the feeling is worth testing. Make a guess before you read on:

Two dams each hold back water 50 m deep. One holds a lake 10 km long. The other holds a slot of water only 2 m wide, with solid rock behind it. Which dam has to be stronger?

lake, 10 km long 50 m deep slot, 2 m wide 50 m deep rock

To decide, start with what sets how hard water presses on anything at all. You have felt it. Swim down to the bottom of a deep pool and, a few metres down, your ears start to hurt. The water above you has weight, and the deeper you go, the more of it there is. That weight is where water pressure comes from. The rest of this chapter turns it into a number, and then into the push on a dam.

2.1 A tonne on every square metre

Picture a column of water one metre tall standing on one square metre of a pool floor. It holds one cubic metre of water, which is a tonne. On Earth a tonne weighs 9.81 kilonewtons (kN),Mass and weight are different things. A tonne is 1,000 kg of stuff; gravity pulls on it with 9.81 kN, and that pull is its weight. Turned round, a kilonewton is the weight of about 102 kg. This book uses tonnes for how much water there is and kilonewtons for how hard it pushes. so the floor under that column carries 9.81 kN on its square metre. A force of one kilonewton on each square metre is one kilopascal (kPa), the unit of pressure in this book. Stack a second metre of water on top and the floor carries two tonnes, 19.62 kPa. Every metre of depth adds the weight of one more tonne on every square metre:

p = ρ · g · h ≈ 9.81 kPa × depth in metres(2.1)

p pressure (kPa) · ρ density of water, 1,000 kg/m³ · g gravity, 9.81 m/s² · h depth below the water surface (m). With these units 1,000 × 9.81 × h comes out in pascals, newtons on each square metre; divide by 1,000 for kPa.

Notice what the rule leaves out. It names the depth and nothing else: not the width of the lake, not its length, not how much water lies beyond. Engineers call the pressure of still water hydrostatic, and in a moment we will see that it is also blind to direction.

Ten metres of water presses with about 98 kPa, roughly the pressure of the whole atmosphere above us.Galileo recorded that a suction pump could lift water no higher than about 18 braccia (Museo Galileo). A Florentine braccio, an arm's length, was a little under 60 cm (the museum gives two braccia as about 115 cm), so that is a little over 10 m; the museum's page rounds it to 11 m. In a letter of 11 June 1644 to Michelangelo Ricci, Evangelista Torricelli explained why: the air presses on the water outside the pipe, and that press can hold up a column only so tall. He showed it with mercury, which stood 1¼ braccia high, about three quarters of a metre; the museum's page rounds it to 70 cm (Museo Galileo). The standard atmosphere, 101.325 kPa, balances 10.3 m of water, or 76 cm of mercury, which is 13.6 times as dense. That raises a fair question. The air presses on the lake's surface too, so shouldn't every reading include it? A sensor that counted the air would read 396 kPa thirty metres down, not 294.Sea-level air: 101.325 kPa and 1.225 kg/m³, U.S. Standard Atmosphere (1976). But the air presses as hard on the dam's downstream face, on the rock, on everything, and equal pushes from two sides cancel. So engineers leave it out and measure pressure above the air's, as a tyre gauge does. That is gauge pressure, and this book uses it throughout: the surface of still water is at zero.

Check yourself A lake high in the mountains has thinner air above it than a lake at sea level. Does a gauge 30 m down read less there?

No. It reads 294.3 kPa in both, because a gauge measures the press of the water above the air's. The thinner air lowers the pressure on the surface and on the gauge's back alike.

Below, a sensor hangs on a cable in front of a concrete dam. Drag it up and down, or use the Sensor depth slider; the Water depth slider changes the depth of the water itself:

Fig. 2.1 · Water pressure on a dam face Depths to scale · dam shape schematic
50 m
30.0 m
Pressure at the sensor294 kPa9.81 × 30.0 m
At the foot of the dam491 kPa9.81 × 50 m
Push on each metre of dam12.3 MNas heavy as 1,250 tonnes

Water pressure grows in a straight line with depth, so the arrows on the dam's face fill a triangle. The sensor hangs h metres below the surface and reads the gauge pressure there. The brick-red arrows show the pressure every 10 m down the face, and the heavy one the pressure at the sensor's depth; the shaded triangle they fill is the whole push. The chart plots that push, per metre length of dam, against the depth of the water: the curve bends upward, and the open dot at half the depth sits at a quarter of the push. The push is in MN, meganewtons: 1 MN = 1,000 kN. The number to remember: 9.81 kPa for every metre of depth, so 294.3 kPa at 30 m.

  • Hold the sensor at 30 m and change the water depth. The reading doesn't move.
  • Set the water to 100 m, then to 50 m. The push falls to a quarter.
  • Drag the sensor to the foot of the dam and compare its reading with "At the foot of the dam".

Notice that the sensor's reading depends on how far it hangs below the surface and on nothing else. Leave it 30 m down and make the water deeper: the reading stays at 294 kPa, though there is now more water beneath it and more water behind it. Notice too that the arrows on the face are the same pressures drawn as arrows, and they grow evenly, from nothing at the surface to the most at the foot.

Example 2.1

The sensor 30 m down

Given
A sensor 30 m below the surface of a reservoir 50 m deep. Water weighs 9.81 kN per cubic metre.
Find
The pressure on the sensor, in kPa and as tonnes on each square metre.
The idea
The pressure is the weight of the column of water standing on one square metre at the sensor's depth (Eq. 2.1).
  1. Column above 1 m²: 30 m × 1 m² × 1 t/m³ = 30 ta tonne for every metre of depth
  2. Its weight: 30 × 9.81 = 294.3 kNeach tonne weighs 9.81 kN
  3. p = 294.3 kN ÷ 1 m² = 294.3 kPaa kilonewton on each square metre is a kilopascal

The sensor feels 294.3 kPa, the weight of 30 tonnes on every square metre. The 20 m of water below it play no part.

Check. Ten metres adds about 98.1 kPa, so thirty metres should add three times that: 3 × 98.1 = 294.3.

What it means. A pressure in still water is a property of one number, the depth. A gauge lowered to the foot of the lake dam and one lowered to the foot of the slot dam would read the same.

2.2 Sideways too

The column explains why a pool floor feels the water above it. It doesn't yet explain your ears, which face sideways, or a dam, which stands beside the water and not under it. A stack of bricks presses down on the floor and hardly at all on the walls beside it, because each brick holds its shape. Water can't. A liquid is a material that flows when anything tries to change its shape, however gently.

So suppose that at some point in still water the push from above were harder than the push from the sides. The water there would be squeezed out sideways, the way toothpaste leaves a tube, and it would start to flow. Still water doesn't flow; that is what still means. Therefore at any point in still water the push is the same in every direction: down, up, sideways and slantwise.A tighter version of the argument. Take a tiny wedge of still water with one face level, one upright and one sloping. The pushes on its three faces must balance its weight. Now shrink the wedge: the pushes shrink with the faces' areas, the weight much faster, with its volume. In the limit the weight drops out, and the pushes balance only if the pressure on all three faces is the same, whatever the slope. At the foot of a 50 m dam the water presses as hard on the concrete beside it as on the rock beneath it, 490.5 kPa each way.

That settles the opening question, and in 1586 the Flemish engineer Simon Stevin found a way to see it without any algebra.Stevin published it in De Beghinselen des Waterwichts (Stevin (1586)), the first systematic treatise on hydrostatics since Archimedes. He explained the paradox, as in the figure, by imagining parts of the water set solid (Minnaert (1976)). Take the lake, at rest. Imagine that all of it except a slot 2 m wide against the dam sets solid, like jelly, without changing its weight or its shape. Nothing in the lake was moving before, and setting it solid gives nothing a reason to start, so every pressure stays what it was, including every pressure on the dam. But the dam now holds back a slot of liquid water 2 m wide with a solid block behind it. The slot must push exactly as hard as the lake did. The figure below draws the argument:

Fig. 2.2 · Stevin's argument: set the lake solidDepths to scale · lake length not to scale
A · The lake ← lake runs back 10 km 50 m deep 490.5 kPa B · All but a 2 m slot set solid solid, same weight 2 m slot, still liquid 50 m deep 490.5 kPa

Setting most of the lake solid changes nothing, so a slot of water pushes on a dam exactly as hard as a lake does. In A the dam holds a lake 50 m deep; in B everything but a 2 m slot has been set solid without changing its weight, and the solid block now holds itself up on the lake bed. The brick-red arrows are the water's pressure on the face, the same in both; their length shows how hard the water presses, not how much water there is, which is why in B they reach back over the solid block. The number to remember: 490.5 kPa at the foot of either dam, and a push of 12,262.5 kN on each metre of it.

  • Cover panel B's solid block with your hand. What is left is the slot dam of the opening question, and nothing on its face has changed.

Notice that the two triangles of arrows are identical. The ten kilometres of lake behind the slot carry their own weight down into the lake bed and leave the dam alone.

Check yourself One diver is at the bottom of a flooded mine shaft 1 m across and 10 m deep. Another is at the bottom of a swimming pool 10 m deep. Whose ears hurt more?

They hurt the same. Both are 10 m down and both feel 98.1 kPa. The narrow shaft matters as little as the narrow slot.

2.3 The push on a whole dam

A sensor gives the pressure at one point. A dam has to hold the pressure over its whole face, and that pressure changes all the way down. Draw it as Fig. 2.1 does, as arrows: at the surface the arrow has no length, and it grows by 9.81 kPa for every metre down to its longest at the foot. Together the arrows fill a triangle, and the push on each metre length of dam is the area of that triangle: half its height times its width at the bottom, ½ × H × (9.81 × H). An area can be a push because of what its two sides measure. Cut one metre length of the face into strips a metre tall: each strip is one square metre, so the push on it in kilonewtons is the length of its arrow in kilopascals. Adding the strips adds up the arrows, and kilopascals times metres of height come out as kilonewtons on each metre of dam. The depth appears twice, once as the height of the wet face and once in the pressure at its foot. So the push grows with the square of the depth:

F = ½ · ρ · g · H²(2.2)

F push of the water on each metre length of dam (kN per m) · H depth of water at the dam (m)

Make the water twice as deep and the triangle is twice as tall and twice as wide at the bottom, so four copies of the first triangle fit inside it. A dam holding 100 m of water takes four times the push of one holding 50 m.

2.4 Where the push acts

A push on a dam does two things. It tries to slide the dam downstream, and it tries to tip it over about its toe, the downstream foot (Plate 0), the way you tip a bookcase by pushing near its top. For tipping, where the push acts matters as much as how big it is. Most of the triangle sits low down: the bottom third of the face carries 56% of the push.Why 56%: the push above any level grows with the square of its depth, so the top two thirds of the face carry (⅔)² = 4/9 of the whole, and the bottom third the other 5/9. Add up all the arrows and the whole push acts as though it were gathered at one point, the centre of the triangle. A triangle's area balances a third of the way in from its wide side, and the wide side of the pressure triangle is at the bottom, so the push acts H/3 up from the base. Its tipping effect about the toe, the moment, is the push times its lever arm, the distance from the pivot to the line the push acts along. For a sideways push that distance is its height above the base, H/3, wherever on the base the pivot is:

M = F · H/3 = ρ · g · H³ / 6(2.3)

M tipping moment about the toe, per metre length of dam (kN·m per m, or MN·m per m; 1 MN·m is 1,000 kN·m)

Double the depth and the push grows four times while the height it acts at doubles, so the tipping effect grows eight times. Here are all three numbers for the dams of the opening question:

Example 2.2

The push on a 50 m dam, and where it acts

Given
Still water H = 50 m deep against an upright face; each metre of depth adds 9.81 kPa.
Find
The pressure at the foot of the face; the push on each metre of dam, in kilonewtons and as tonnes; the height it acts at; its tipping moment about the toe.
The idea
The push is the area of the pressure triangle (Eq. 2.2) and acts at its centre, a third of the way up (Eq. 2.3).
  1. p at the foot = 9.81 × 50 = 490.5 kPathe width of the triangle at its base
  2. F = ½ × 490.5 × 50 = 12,262.5 kN per metrehalf the base times the height of the triangle
  3. 12,262.5 ÷ 9.81 = 1,250 tthe push, as the weight of so many tonnes
  4. Height of the push: 50 ÷ 3 = 16.67 m above the basethe triangle's centre, a third of the way up from its wide side
  5. M = 12,262.5 × 16.67 = 204.4 MN·m per metrepush times its lever arm; 1 MN·m is 1,000 kN·m
  6. At H = 100 m: F = 49,050 kN, M = 1,635 MN·mthe same steps with the depth doubled

Each metre of a 50 m dam carries a push as heavy as 1,250 tonnes, acting 16.7 m up. At 100 m the push is 4 times as big and the tipping moment 8 times.

Check. Water weighs a tonne per cubic metre, so for water the push in tonnes is ½ × H × H: ½ × 50 × 50 = 1,250, which is what the figure's readout says.

What it means. The height of the water is what a dam pays for. Each extra metre of depth costs more than the one before.

Now the slot, which looked as though it could never push that hard:

Example 2.3

A hundred tonnes that push like 1,250

Given
The slot of the opening question: water 2 m wide and 50 m deep, between a dam and solid rock. Water: a tonne per cubic metre.
Find
The weight of the water in the slot, per metre length of dam, and the push it puts on the dam.
The idea
The water's weight goes down into the slot's floor; its push on the dam depends only on the depth.
  1. Water per metre of dam: 2 m × 50 m × 1 m = 100 m³, so 100 ta tonne per cubic metre
  2. Push on the dam: 12,262.5 kN, as heavy as 1,250 tthe same as the lake's, from Example 2.2
  3. 1,250 ÷ 100 = 12.5push over the water's own weight

100 tonnes of water push on the dam with the weight of 1,250 tonnes, 12.5 times their own weight.

Check. Where can a sideways push bigger than the water's weight come from? Compare the areas the water presses on. Each metre of the slot's floor is 2 m² and carries the 100 tonnes: 50 tonnes on each square metre, the pressure at 50 m. Each metre of the dam's face is 50 m² and carries on average half that pressure, 25 tonnes on each square metre, so 50 × 25 = 1,250 tonnes. The same pressure acts on a big face and a small floor. Nothing moves, because the rock pushes back on the water as hard as the dam does, from the other side.

What it means. Pressure is not weight. A thin layer of water can carry an enormous push as long as it is deep, which is why a narrow crack under a dam, filled from the reservoir, matters so much in Chapter 7.

Check yourself A reservoir is raised from 50 m to 60 m deep. Does the tipping moment on the dam grow by more or less than the push?

By more. The push grows by (60/50)² = 1.44 times, the moment by (60/50)³ = 1.73 times.

2.5 Why a concrete dam is a triangle

Look at a concrete dam from the side and it is close to a triangle, narrow at the top and wide at the foot. One could assume the reason is that the pressure is greatest at the bottom, so the wall needs to be thickest there. That explains a thick foot, but not a thin top: a wall as thick as its foot all the way up would hold too. The full reason is about every level at once.

Cut the dam in your mind along a level line, z metres below the water's surface. Everything above the cut has to hold against the water above the cut, or it would slide or tip off along that line; so engineers check a concrete dam at levels all the way up, not only at its base. What holds it is its weight, pressing down on the concrete below (Chapter 4 shows how a press becomes a grip). The push of the water above the cut grows with z². For a triangle, the width at depth z is proportional to z, so the area above the cut, half its width times its height, grows with z² as well. Push and weight grow in step. The figure shows three cuts:

Fig. 2.3 · Weight and push above three levels of a triangular damSection to scale · bars to scale with each other

push of the water above the cut weight of concrete above the cut kN per metre of dam

z = 15 m z = 30 m z = 45 m water to the top concrete 1,104 2,115 4,415 8,460 9,933 19,035

In a triangular section the weight above any level grows with the square of the depth, exactly as the push does, so every level is equally safe. The section is 0.8 times as wide as it is deep at every level, the proportions of the concrete dam in Chapter 7. Beside each dashed cut, the brick-red bar is the push of the water above that cut and the green bar the weight of the concrete above it, both per metre length of dam. From the top cut to the bottom one the depth triples and both bars grow ninefold. The number to remember: the weight is 1.92 times the push at every level.

  • Read the two bars at z = 30 m and divide the green number by the red one.
Answers

8,460 ÷ 4,414.5 = 1.92, the same as at the other two cuts.

Notice that the green bar stays the same multiple of the red one at all three cuts, though both grow ninefold between the top cut and the bottom. Here is the arithmetic for any level:

Example 2.4

Why the triangle keeps step

Given
A concrete section with an upright upstream face, 0.8z wide at depth z, water to its top. Concrete weighs 23.5 kN/m³, water 9.81 kN/m³.
Find
At any depth z: the weight of concrete above that level, the push of water above it, and how they compare, for sliding and for tipping.
The idea
Both are areas of triangles whose sides grow with z, so both grow with z². Tipping adds one more factor of z to each, through the lever arms.
  1. Weight above z: ½ × 0.8z × z × 23.5 = 9.40 z² kNarea of the triangle above the cut, times concrete's unit weight
  2. Push above z: ½ × 9.81 × z² = 4.905 z² kNEq. 2.2 with the depth to the cut
  3. Weight ÷ push = 9.40 ÷ 4.905 = 1.92z² cancels: the same at every level
  4. Holding moment: 9.40z² × ⅔ × 0.8z = 5.013 z³a triangle balances a third of the way across from its tallest side; here that is the upright face, so the weight acts two thirds of the width from the toe
  5. Tipping moment: 4.905z² × z/3 = 1.635 z³the push acts a third of the way up, Eq. 2.3
  6. Holding ÷ tipping = 5.013 ÷ 1.635 = 3.07z³ cancels too

Above every level of the triangle the concrete weighs 1.92 times the push, and holds 3.07 times the tipping moment.

Check. At z = 30 m: weight 8,460 kN, push 4,414.5 kN, and 8,460 ÷ 4,414.5 = 1.92. At z = 45 m: 19,035 ÷ 9,932.6 = 1.92.

What it means. A triangle gives every level the same margin, so no level is weaker than the rest and none carries concrete it does not need. What this leaves out, water seeping under the dam and pushing it up, is the subject of Chapter 7.

To see what the shape buys, compare a wall with straight sides that uses the same concrete. Down to 50 m the triangle holds 1,000 m³ of concrete for every metre of dam, so the straight wall is 20 m thick. Its weight above a cut grows only in proportion to the depth, while the push grows with the depth squared. Ten metres down it weighs 9.6 times the push above, far more than it needs. At the foot its weight is the same 1.92 times the push as the triangle's, but against tipping it holds only 1.15 times the moment where the triangle holds 3.07, because its weight sits closer to the downstream toe. The triangle puts its concrete where the push is.

Dams that hold back their water by their own weight in this way are called gravity dams, and real ones are close to this shape. Revelstoke Dam on the Columbia River is 175 m high and 126 m thick at its base,Height: BC Hydro (2013). Base thickness, 126 m, and 9 m at the crest: BC Hydro (2016). a base 0.72 times its height, a little slimmer than the 0.8 used here. BC Hydro describes it as relying on its massive weight and the friction that weight produces to hold back the reservoir.

The first three minutes of the film Depth, not size retell the chapter up to this point, in pictures:

Film 2.1 · Depth, not size part one3 min 11 s · captions

Before you watch

At 2:44 the film says that most of the push on a dam is low down. Put a number on it before you watch: what share of the push does the bottom half of the face carry? The chapter worked out the bottom third.

These are the film's first three minutes. The second half, about the water under a concrete dam, plays in Film 7.1.

In this film

  1. Two dams: a lake and a slot the question
  2. Your ears, then a pressure gauge Eq. 2.1
  3. Arrows that fill a triangle Fig. 2.1
  4. Twice as deep, four triangles Eq. 2.2
  5. Why a gravity dam is a triangle Example 2.4
Read the transcript

A dam is a wall that holds back water. Here's a pair of them, each holding water fifty metres deep. Behind the top dam is a lake that runs back ten kilometres. Behind the bottom one is a slot of water two metres wide, with solid rock beyond it. Which dam has to be stronger?

It's tempting to pick the lake. There's so much more water behind it, and all of it seems to lean on the wall. But follow where the lake's weight actually goes. Cut the lake into columns, and each column just presses down on the ground beneath it. The lake bed carries the lake. Only the water right against the dam pushes sideways on it. So what decides how hard that water pushes?

If you've ever dived down off a beach, you've felt the answer. A few metres down, your ears start to hurt. The water above you presses in, and the deeper you go, the harder it presses. Water passes that press on in every direction, sideways too, so it pushes on a wall just as hard. Each metre of depth adds about a tonne on every square metre.

How wide the sea is doesn't matter, only how far you are below the surface. So lower a pressure gauge to the foot of each dam. Both read the same, about fifty tonnes on every square metre. The lake behind one dam and the slot behind the other make no difference at all. A Flemish engineer, Simon Stevin, wrote this down in fifteen eighty-six, and it still surprises people.

That's the pressure on one square metre, but the water presses on the whole face of the dam. Draw the pressure as arrows down the face. At the surface the arrow has no length at all, and it grows steadily toward the bottom. Together the arrows fill a triangle, and the area of that triangle is the whole push.

On every metre along the dam, it comes to about twelve hundred and fifty tonnes. Now make the water a hundred metres deep. The triangle is twice as tall, and twice as wide at the bottom, because the pressure there is twice as big. So it holds four copies of the first triangle. That's about five thousand tonnes on every metre of dam. Double the depth, and the push grows four times.

So the two dams feel exactly the same push. Depth sets the push, and nothing else does. The size of the lake decides something different: how much water gets out if the dam fails.

The four-times rule also explains the shape of a concrete dam. Most of the push is low down, so the wall has to be thick at the bottom. Make the wall a triangle, and its own weight grows the same way the push does. Pick any level: the weight of concrete above it and the push of water beside it grow together, so they stay in step all the way down. That's why big concrete dams that stand by their own weight are triangles in cross-section. Engineers call them gravity dams.

After you watch: the answer

Three quarters, 75%. The pressure triangle above half the depth is half as tall and half as wide as the whole one, so it holds a quarter of the push, (½)² = ¼, and the bottom half carries the other three quarters. On the film's 50 m dam that is 9,196.9 of the 12,262.5 kN on each metre, against 3,065.6 kN on the top half. The film draws the same rule at 2:05, when it doubles the depth and fits four triangles into the new one. In Fig. 2.1 the chart's open dot shows it: at half the depth, a quarter of the push.

2.6 Water on a sloping face

Most of the dams in the rest of this book are embankments, heaps of earth and rock whose upstream face leans back at a gentle slope. The still-water argument holds on any face, at any angle: the water presses equally in every direction, so on a surface it presses straight at it, at right angles. On a sloping face the arrows tilt. They point down into the slope as well as across it. Here is a 24 m embankment whose upstream face slopes at 2H:1V (two across for every one up, as in Example 1.2), holding water 20.4 m deep, the dam of the seepage lab in Fig. 10.3:

Fig. 2.4 · Water on a sloping faceSection to scale · force diagram to scale with itself
water on the face upstream face, 2H:1V embankment 20.4 m deep face runs 40.8 m across The whole push on each metre of dam, in kN, split in two: across 2,041 down 4,083 whole push 4,564, at right angles to the face

On a sloping face the water presses at right angles to the slope, so it pushes the dam down as well as across. The brick-red arrows are the water's pressure, longer with depth, each at right angles to the face. The hatched wedge is the water standing on the face, which the face holds up like a floor. Below, the whole push on each metre of dam is split in two: across, the same as on an upright wall of the same depth, and down, the weight of the hatched water. The number to remember: on a 2:1 face the water presses down 2 times as hard as it pushes across.

  • Lay a set square on the force diagram: the whole push should meet the upstream face at a right angle.

Notice that you can split the push in two. Across, it is exactly the push on an upright wall of the same depth, because a sideways push only counts the face's height. Down, it is the weight of the wedge of water standing on the face.

Example 2.5

Water pressing an embankment down

Given
An upstream face at 2:1 holding still water H = 20.4 m deep; water weighs 9.81 kN/m³.
Find
The push of the water on each metre length of dam, across and down, and the whole push.
The idea
Across: the pressure triangle on the face's height, Eq. 2.2. Down: the weight of the water standing on the face.
  1. Run of the wet face: 2 × 20.4 = 40.8 mtwo across for every one up
  2. Across: ½ × 9.81 × 20.4² = 2,041.3 kNas on an upright wall 20.4 m tall
  3. Water on the face: ½ × 40.8 × 20.4 = 416.16 m³the hatched wedge, per metre of dam
  4. Down: 416.16 × 9.81 = 4,082.5 kNthe wedge's weight, which the face holds up
  5. Whole push: √(2,041.3² + 4,082.5²) = 4,564 kNthe two parts at right angles, Pythagoras

The water presses each metre of the embankment down with 4,083 kN and across with 2,041 kN: twice as hard down as across.

Check. Every arrow is at right angles to the face, so the whole push must be too. Its angle from the vertical has a tangent of 2,041 ÷ 4,083 = 0.50, which is 26.6°, the same as the face's angle from the horizontal, 26.6°. At the other extreme, an upright face has no water standing on it and the downward part vanishes, as on the concrete dam.

What it means. On an embankment the reservoir mostly presses the dam onto its foundation, which helps it grip. The picture is exact for a watertight face, such as a concrete skin on rockfill; in an earth dam the water also soaks into the face, and Chapter 10 follows it inside. Chapter 12 shows what happens when the reservoir is let down fast: the help on the face goes, while the water inside the dam stays.

2.7 Water in the ground has pressure too

Dig a hole in the sand near the sea and it fills with water from below. The water in the hole settles at a level, and below that level every gap between the grains is full of water. That level is the water table.

The water in those gaps is joined up from gap to gap, like the water in a soaked sponge, so while it is still it obeys the same rule as the water in a lake. A metre below the water table it presses at 9.81 kPa, ten metres down at 98.1 kPa, and at every point it presses equally in every direction, on every grain around the gap. The pressure of the water in the gaps between grains is called pore pressure, written u. It is the water half of this book's one sentence: Chapter 4 shows how it takes away the press of grain on grain that gives soil its strength.

The important word there is still. The water in a lake is still. The water in the ground under a dam almost never is: it creeps through the gaps from the reservoir toward the far side of the dam, and moving water does not keep to the rule of this chapter.

What to remember

  • Still water presses equally in every direction, and how hard depends only on the depth: 9.81 kPa for every metre. The size of the lake does not enter.
  • The push on each metre of dam is the area of the pressure triangle, ½·ρ·g·H², acting a third of the way up. Double the depth and the push grows four times, its tipping effect eight.
  • A gravity dam is a triangle because the weight above any level then grows with the square of the depth, as the push does, so every level is equally safe.
  • On a sloping face the water presses at right angles to the slope; on an embankment it presses the dam down harder than it pushes it across.
  • Water in the gaps of soaked ground follows the same rule while it is still. That pore pressure carries part of the load the grains would otherwise carry between them, so they press on one another less; Chapter 4 makes this exact as σ′ = σ − u.

Questions

  1. 2.1

    A reservoir is raised from 40 m deep to 60 m. By what factor does the push on each metre of its dam grow?

    Hint

    The push grows with the depth squared, so compare 60² with 40².

  2. 2.2

    Predict the push on each metre of dam when the water is 70 m deep, then set the figure and check. How does it compare with the push at 50 m?

    Hint

    70² is close to 2 × 50².

  3. 2.3

    Your ears hurt 3 m down in a pool, but not after a 30 m ride up a lift in a tall building. Why?

    Hint

    Both are the same rule, Eq. 2.1, with a different fluid. A cubic metre of sea-level air has a mass of about 1.2 kg; a cubic metre of water, 1,000 kg.

  4. 2.4

    Where does the push on a dam holding 50 m of water act, and by what factor does its tipping moment grow if the water is twice as deep?

    Hint

    The push acts at the centre of the pressure triangle; the moment is push times height.

  5. 2.5

    Is the water on an embankment's sloping upstream face a danger or a help to the dam's stability?

    Hint

    Split its push into the part across and the part down. What does each do?

  6. 2.6

    Revelstoke Dam is 175 m high and 126 m thick at its base. Suppose water stood against it right to the top (a real reservoir is kept a little below the crest). What would the pressure be at its foot, and the push on each metre of dam? How does its base compare with the triangle of Example 2.4?

    Hint

    Use Eq. 2.1 and Eq. 2.2 with H = 175 m; in tonnes, the push is ½ × H².

Further reading

  • Ciechanowski, B. (2021). Naval Architecture. ciechanow.ski, 27 July 2021. https://ciechanow.ski/naval-architecture/ gentle An interactive essay that measures pressure with gauges you can watch, shows that it depends on height alone (including the arrangement known as Pascal's barrel), and draws pressure as arrows at right angles to a surface, before going on to ships.
  • Feynman, R.P., Leighton, R.B., Sands, M. (1964). The Feynman Lectures on Physics, Vol. II. Reading, Mass.: Addison-Wesley. https://www.feynmanlectures.caltech.edu/II_40.html serious Section 40-1, Hydrostatics, makes this chapter's argument in the language of a physics course: why a fluid at rest carries no shear, why its pressure is then the same in every direction, and how it changes with height.
  • U.S. Bureau of Reclamation (1987). Design of Small Dams, 3rd ed. Denver. https://www.usbr.gov/tsc/techreferences/mands/mands-pdfs/SmallDams.pdf serious Its chapter on concrete gravity dams shows how engineers turn the water's push into a design: the loads on a gravity section and the checks against sliding and overturning. Free from the Bureau of Reclamation.

Chapter 3 · Part I · Ground

Which way the water goes

The sea is lower than the beach. What pushes the water out of the sand, and why does it take hours?

Thus, it appears that for an identical sand, it can be assumed that the volume discharged is (directly) proportional to the head and inversely proportional to the thickness of the sand layer that the water passes through.

Henry Darcy, Les fontaines publiques de la ville de Dijon, 1856, in English as quoted by Simmons

So far. Still water presses harder the deeper you go, 9.81 kPa for every metre, and Chapter 2 turned that into the push on a dam. Water in the ground is rarely still. Under a dam it creeps through the gaps between the grains. This chapter asks which way it goes, how fast, and what it does to the grains on the way.

After this chapter you can

  • find the total head at a point from a piezometer reading, and say which way the water there will flow;
  • use Darcy's law to estimate a flow, and how long a drop of water takes to cross a layer of soil;
  • explain why a thin layer of clay takes almost the whole drop in head, and how hard moving water pushes on the grains it passes.

This chapter adds the phreatic line through the heap and one standpipe (the straw)

3 figures · about 20 min to read

Go back to the beach an hour or two after high tide, while the sea is still going out. Near the bottom of the beach, where the sand flattens toward the water, it is dark and shining, and thin streams of water run out of it and braid their way down the last few metres to the sea. They keep running for hours. Nothing is raining on the beach and nobody is squeezing it, yet the water keeps coming.

It comes out because the beach is full of it. At high tide the sea soaked into the sand and filled the gaps between the grains. As the tide falls, the water inside the sand falls too, but far more slowly, and it is left standing higher inside the beach than the sea outside. Where that buried water surface meets the sloping sand, it leaks out, and the streams are the leak.On a falling tide the water table in a beach comes apart from the sea and leaks out over a seepage face: Turner (1993), who tracked it on a beach in North Queensland.

That leaves two questions, and they are the questions of this chapter. What pushes the water out of the sand, when the sea it came from has gone? And why does it take hours?

Most people say water flows from high pressure to low pressure, the way air rushes out of a punctured tyre. It sounds right for the beach. But the rule fails in the simplest place there is. In a still lake the water at the bottom is at a far higher pressure than the water at the top: Example 2.1 found 294.3 kPa thirty metres down, and zero at the surface. If water flowed from high pressure to low, the deep water would rise toward the surface and the lake would stir itself, forever. It lies still. So pressure alone does not decide where water goes. Whatever does decide it must be the same at every point of a still lake.

3.1 Head

To tell which way the water will go, engineers use a simple measure. Imagine pushing a clear straw down into wet ground and watching how high the water climbs inside it. The level the water reaches, measured from some fixed height such as sea level, is the total head at the bottom of the straw. Water in the ground creeps through the gaps between the grains, from where its head is high to where it is low.

The level in the straw adds two things. The first is how high the bottom of the straw is, its elevation head, z, measured from the fixed level, which is called the datum. The second is how far the water pressure at the bottom of the straw can push water up it, its pressure head. A pressure u holds up a column of water u ÷ γw tall, where γw = 9.81 kN/m³ is the weight of a cubic metre of water: one metre of column for every 9.81 kPa. Together:

h = z + u / γw(3.1)

h total head (m) · z elevation head, the height of the point above the datum (m) · u water pressure at the point (kPa) · γw unit weight of water, 9.81 kN/m³. The datum can be any level you like, as long as you keep it: only differences in head move water.

Now the still lake makes sense. Go one metre deeper and you lose a metre of elevation head, but the pressure grows by 9.81 kPa, which is exactly one more metre of pressure head. The two trade evenly, so the total head is the same everywhere in still water, and with no difference in head nothing flows. The figure puts straws into a still lake, and then into the beach as the tide goes out:

Fig. 3.1 · Straws in still water and in a draining beachA: depths to scale · B: schematic
datum still lake, 8 m deep soaked sand the water in every straw stands at the lake's level pressure head, u/γw = 18 m z = 2 m total head 20 m A · A still lake datum the levels step down toward the sea water table water leaks out sea soaked sand B · The beach at low tide

Water in a straw rises to the total head at its tip: in still water every straw stands at the same level, and in a draining beach the levels fall toward the sea. In A, three straws reach 16, 8 and 2 m above a datum drawn 20 m below the lake's surface. The deeper the tip, the less elevation head it has and the more pressure head, and the dimensions on the deepest straw show the two adding to the same total. In B the soaked sand lies below the dashed water table, which stands higher than the sea; the dotted line joins the water levels in the three straws, and the arrows show the water moving toward the brick-red stretch of beach where it leaks out. The number to remember: in still water the total head is the same everywhere, 20 m above this datum at every tip.

  • Work out the pressure head at the middle straw's tip in A, and add its elevation.
  • In B, follow an arrow from the deep sand toward the leak. Does the water move toward higher pressure or lower?
Answers

The middle tip is 12 m below the surface, so its pressure head is 12 m; with its 8 m of elevation that makes 20 m. In B the water rises toward the leak, from deep sand where the pressure is high to the face where it is zero, and it also moves toward lower head, as it must.

Notice that in panel A the three straws reach very different depths, and the water in all of them stops at exactly the lake's surface. The pressures at their tips differ, but the heads don't, and nothing moves. In panel B the straws step down: the head inside the beach is highest far from the sea and lowest where the water comes out. That difference in head is what pushes the water out of the sand. The beach's water is still higher than the sea, and so it runs downhill to it, through the sand.

Two consequences surprise people. Groundwater can flow uphill, as long as it is heading toward lower head; a spring that bubbles up out of a hillside is water rising toward a point of lower head. And water can flow from low pressure to high. Follow a drop in panel B from close under the water table down and along through the sand: it goes deeper, into higher pressure, and all the while toward lower head. The rule, every time, is head.

Check yourself The middle straw in panel A has its tip 4 m into the lake bed, 12 m below the lake's surface. What water pressure does the tip feel, and why does the water in the straw stop at the lake's level?

9.81 × 12 = 117.7 kPa, a pressure head of 12 m. The tip is 8 m above the datum, so its head is 20 m, the same as the lake's surface: a straw shows the head, and in still water that is the surface everywhere.

3.2 The straws are real

Engineers sink thin pipes into dams and the ground beneath them and measure the water level inside. They are called piezometers, and reading them, along with measuring the water that seeps out, is one of the main ways engineers know what the water inside a dam is doing.A real straw takes time to fill. Water has to flow in through the ground to raise the level, and in tight ground that can be slow. Juul Hvorslev worked out the delay in 1951. For one fine-grained soil, his estimates ran from 193 days for a 2-inch cased boring with soil in its foot to 0.4 seconds for a small pressure cell set in a pocket of sand, which reads the pressure directly without waiting for a column to fill.Hvorslev (1951), his figure 14 and p. 37. Each one gives the head at a single point, a pinprick of knowledge inside a mountain of earth, so a dam carries many of them and engineers read them for as long as it stands.

Inside an earth dam they show what panel B shows. The water line in Fig. 1.2 is a water table like the beach's: it starts at the reservoir's level on the upstream face and falls as it crosses the dam. A piezometer near the upstream face stands high, one near the downstream face stands low, and the water soaking into the dam runs between them, losing head as it goes. Chapter 10 maps that water through a whole dam. For a first look in three dimensions, the Dams Academy lesson Seepage in 3D lets you colour a model dam by total head and then by pressure head; only the first picture tells you where the water goes.

3.3 Some ground is a sieve, some is a wall

Head says which way the water goes. How fast it goes depends on the ground, and above all on the size of the gaps. Gravel has gaps you can see. Clay is made of flat flakes hundreds to thousands of times smaller than a grain of sand, and the gaps between them are tiny and twisting. The same difference in head that sends water racing through gravel barely moves it through clay. To say how much, engineers need a number for the ground, and that number comes from an experiment run in a hospital courtyard in Dijon.

3.4 Darcy's law

In 1855 and 1856 Henry Darcy, the engineer who had built the water supply of Dijon in France some fifteen years before, ran water down through sand packed in an upright column 2.5 m tall and measured how much came out at the bottom.Darcy's water reached Dijon's reservoir at the Porte Guillaume on 6 September 1840, and he waived the fees the town owed him for it. The sand experiments, 35 of them, were run with the engineer Charles Ritter, using sand from the river Saône, and the mercury in their gauge shook with the water hammer of the street fountains at the hospital: Darcy's own water supply, rattling his own experiment. He published them in 1856 as Note D, an appendix to his book on the fountains of Dijon (Darcy (1856); the story is told by Simmons (2008), pp. 1025 and 1030–1031).The column, 2.50 m tall by Darcy's text (his plate marks 3.5 m) and 0.35 m across, and the dates: Simmons (2008), pp. 1025 and 1030–1031. He found a simple rule. The flow grew in proportion to the difference in head between the top and the bottom of the sand, and shrank in proportion as the sand was made thicker: double the head difference and the flow doubles; double the thickness and it halves. What mattered was how steeply the head dropped along the way: the head lost for every metre the water travels. That slope is called the hydraulic gradient, written i. The flow through each square metre of ground is the gradient times a number that belongs to the soil, and that is Darcy's law:

q = k · i · A, so q / A = k · i, with i = Δh / L(3.2)

q flow (m³/s) · A area the water flows through (m²) · q/A flow through each square metre (m/s) · k permeability (m/s) · i gradient · Δh head lost (m) · L length of the path (m)

Read the second form with i = 1, a metre of head lost for every metre travelled, and it says what k is: the flow through each square metre of ground under that push. That is why it has the units of a speed, metres per second. Engineers call it the permeability of the soil.Strictly, k in m/s is the hydraulic conductivity, and permeability is a property of the gaps alone, in m² (Freeze & Cherry (1979), section 2.3). Dam engineers say permeability for both, and so does this book. k depends on the water too: at 5 °C water is half again as viscous as at 20 °C, 1.52 against 1.00 mPa·s (IAPWS (2008)), so the same sand passes cold water about two thirds as fast. The figures in this book use 10⁻² m/s for gravel, 10⁻⁴ for sand, 10⁻⁷ for silt and 10⁻⁹ for compacted clay.

Below, Darcy's column is laid on its side: a pipe full of sand joins two tanks, and each tank's water level is the head at that end of the pipe. Drag the water levels in the tanks, or use the sliders. The thin tubes along the pipe are piezometers:

Fig. 3.2 · Darcy's experiment Schematic · flow per square metre of pipe
Material in the pipe
Clay plug in the middle
2.00 m
0.60 m
2.0 m
Head difference Δh1.40 mleft is higher
Gradient i = Δh / L0.70head lost per metre
Flow per m² q/A4.2 L/minthrough each m² of pipe
Head lost in the plug–no plug

The head falls along the pipe in a straight line, and the water flows down that line, from the higher tank to the lower. Each tank's water level is the head at its end of the pipe, measured from the pipe's centre line. The piezometer tubes show the head at five points along the pipe, and the dashed blue line through their levels is the head everywhere between. The dots are water moving through the soil; their speeds are squeezed together so that silt still visibly moves. With the clay plug in, the head line runs almost level through the soil on each side and drops almost the whole way across the plug, and the last readout says how much of the drop the plug takes. The number to remember: in sand, the default gradient of 0.70 drives 4.2 litres a minute through each square metre; a clay plug a fifth of the pipe's length takes 99.996% of the head.

  • Drag the right tank above the left one. Which way do the dots go now?
  • Set the Pipe length to 4.0 m. What happens to the Gradient and to the Flow per m²?
  • Add a clay plug, then switch the material from Sand to Silt. How much of the head does the plug take now?
Answers

The dots reverse: water flows from high head to low, whichever side that is. At 4.0 m the gradient halves to 0.35 and the flow halves to 2.1 L/min. With silt the plug takes 96.15%: silt is only a hundred times more permeable than clay, so it keeps a visible share of the drop.

Notice that the water in the tubes lines up in a straight slope from one tank to the other. The sand is the same all along the pipe, so every metre of it takes the same share of the head. Notice too that the readouts follow Darcy's rule: raise the left tank and the gradient and the flow grow together; lengthen the pipe and both shrink. Here is the arithmetic behind the readouts, and one more number the figure does not show:

Example 3.1

How fast does water move through sand?

Given
Darcy's pipe as the figure opens: sand with k = 10⁻⁴ m/s, whose gaps take up 0.35 of its volume; the left tank at 2.00 m, the right at 0.60 m; a pipe 2.0 m long.
Find
The gradient; the flow through each square metre of pipe; how fast a drop of water moves, and how long it takes to cross the pipe.
The idea
Darcy's law gives the flow through each square metre (Eq. 3.2). But the water can only travel through the gaps, so to carry that flow it has to move faster than q/A.
  1. Δh = 2.00 − 0.60 = 1.40 mthe head at each end is its tank's level
  2. i = 1.40 ÷ 2.0 = 0.70head lost for each metre of pipe
  3. q/A = 10⁻⁴ × 0.70 = 7.0 × 10⁻⁵ m/sDarcy's law: cubic metres a second through each square metre
  4. In litres: 0.070 L/s, and 0.070 × 60 = 4.2 L/mina cubic metre is 1,000 litres
  5. v = 7.0 × 10⁻⁵ ÷ 0.35 = 2.0 × 10⁻⁴ m/sthe same flow squeezed through the gaps, about a third of the area
  6. Time: 2.0 m ÷ 2.0 × 10⁻⁴ m/s = 10,000 slength of the pipe over the drop's speed
  7. 10,000 ÷ 3,600 = 2.8 hours3,600 seconds in an hour

Each square metre of pipe passes 4.2 litres a minute. A drop moves at 0.2 mm a second and takes 2.8 hours to cross the 2 m of sand.

Check. Swap the sand for silt, whose k of 10⁻⁷ m/s is a thousand times smaller. The flow should fall a thousandfold, to 7.0 × 10⁻⁸ m/s, which is 6.0 litres a day through each square metre: what the figure's readout shows with Silt.

What it means. Even through clean sand, losing seven tenths of a metre of head for every metre it travels, the water moves only about 0.7 m an hour. Seepage is slow, and the soils a dam is made of are slower still.

The last lines of that example hold a distinction worth keeping. The flow through each square metre of ground, q/A, is what a dam engineer adds up to find a leak. The speed of the water itself, the seepage velocity, is that flow divided by the gap fraction, n (engineers call it the porosity), and it is what tells you how long water takes to get anywhere.Why does head leave out the speed of the water? For a river or a pipe, engineers add a third term, the height the water's speed could lift it, v²/2g. For the drop in the sand pipe that is (2.0 × 10⁻⁴)² ÷ (2 × 9.81) = 2 × 10⁻⁹ m, two millionths of a millimetre. In the ground it is always that small, so it is dropped.

3.5 Back to the beach

That is enough to answer the second question from the beach. The sand there is like the pipe's, but the beach slopes gently, so its water table does too. Take the water table to fall 1 m over 20 m of beach, an illustrative gradient of 0.05. Then each square metre passes 10⁻⁴ × 0.05 = 5 × 10⁻⁶ m/s, and the water itself moves at that divided by the gap fraction, 1.4 × 10⁻⁵ m/s: about 5 centimetres an hour. A drop that soaked in 20 m up the beach needs about 16 days to reach the sea. The beach cannot empty in one low tide, or anything like it. It leaks steadily through the hours the tide is out, and the sea comes back and fills it again long before it is drained.

3.6 A metre of clay

Sand is the fast end of a dam's materials. The slow end is clay, and a guess first:

A one-metre block of compacted clay has water pressing on one side, with the water level 1 m higher than on the other side. How long does a drop of water take to get through the metre?

To decide, the race below runs a metre of each of four materials under the same push, with the same arithmetic as the sand pipe:

Notice that the race is lopsided in a way no ordinary clock could show. Each tick on the time axis is ten times the one before, and still the four finishes are spread across almost the whole axis. The arithmetic is the sand pipe's, done four times:

Example 3.2

The race, by hand

Given
A metre of each material (L = 1 m) under a gradient of 1. Gravel: k = 10⁻² m/s, gap fraction 0.30. Sand: 10⁻⁴, 0.35. Silt: 10⁻⁷, 0.40. Clay: 10⁻⁹, 0.35.
Find
How long a drop of water takes to cross each metre.
The idea
The drop's speed is k · i ÷ n, as in Example 3.1, so the time is the length over that: t = n · L ÷ (k · i). With L and i both 1, it is the gap fraction over k.
  1. Gravel: 0.30 ÷ 10⁻² = 30 sdividing by 10⁻² multiplies by a hundred
  2. Sand: 0.35 ÷ 10⁻⁴ = 3,500 s, and 3,500 ÷ 60 = 58 minsixty seconds to a minute
  3. Silt: 0.40 ÷ 10⁻⁷ = 4.0 × 10⁶ s, or 46 days86,400 seconds to a day
  4. Clay: 0.35 ÷ 10⁻⁹ = 3.5 × 10⁸ s, or 11.1 yearsabout 3.16 × 10⁷ seconds to a year

Thirty seconds, 58 minutes, 46 days and 11.1 years, under the same push through the same metre.

Check. Clay's time over gravel's is 3.5 × 10⁸ ÷ 30 = 1.17 × 10⁷. If gravel's 30 seconds were drawn a millimetre long, clay's bar would be that many millimetres, 11.7 km: the figure caption's twelve kilometres.

What it means. The gap fractions hardly differ; the permeabilities differ ten million times, and they decide the race. In a dam it is the choice of soil, far more than its thickness, that decides where the water is held back.

That enormous range is what makes it possible to build a dam out of soil at all. A dam needs a part that water can barely get through, and parts that let water out quickly, side by side. Clay does the first job. Sand and gravel do the second.

3.7 Layers in a line

But a dam's clay is usually a narrow strip inside a much wider bank of sand and rock. How can a strip a few metres thick hold back the head of a whole reservoir? Darcy's pipe can show it. Go back to it and put a plug of clay in its middle:

The pipe is the same length and the tanks are at the same heights, but now nearly all the head is lost across the short piece of clay. The tubes in the sand on either side read almost the same as the tank next to them. The same flow has to pass through every part of the pipe, because water can't pile up anywhere in between. The tight clay needs a far steeper gradient than the sand to push that flow through, so the clay takes almost the whole drop. Each piece of the pipe needs head in proportion to its length divided by its k, which works like a resistance:

Example 3.3

The plug as resistances in a line

Given
The 2.0 m pipe with a clay plug 0.4 m long in its middle (k = 10⁻⁹ m/s) and sand on either side of it, 1.6 m in all (10⁻⁴ m/s); head difference 1.40 m.
Find
The share of the head the plug takes, and what it does to the flow.
The idea
The same flow passes every piece, and each piece needs head in proportion to L ÷ k, its resistance. Resistances in a line add.
  1. Sand: 1.6 ÷ 10⁻⁴ = 16,000 smetres over metres per second leaves seconds
  2. Clay: 0.4 ÷ 10⁻⁹ = 4.0 × 10⁸ sa fifth of the length, but a hundred thousand times tighter
  3. Together: 4.0 × 10⁸ + 16,000 = 4.00016 × 10⁸ sresistances in a line add
  4. Clay's share: 4.0 × 10⁸ ÷ 4.00016 × 10⁸ = 99.996%each piece takes head in proportion to its resistance
  5. k of the whole pipe: 2.0 ÷ 4.00016 × 10⁸ = 5.0 × 10⁻⁹ m/sits length over its total resistance
  6. q/A = 5.0 × 10⁻⁹ × 0.70 = 3.5 × 10⁻⁹ m/s, 110 L a yearDarcy's law for the pipe as a whole; the gradient is still 0.70
  7. 7.0 × 10⁻⁵ ÷ 3.5 × 10⁻⁹ = 20,000the flow without the plug, from Example 3.1, over the flow with it

The plug takes 1.39994 m of the 1.40 m of head; all the sand together loses 0.056 mm. The flow falls 20,000-fold, to 110 litres a year through each square metre.

Check. The share depends only on the resistances, so a material nearer clay's k should take a bigger share of the drop. Silt's 1.6 m has a resistance of 1.6 ÷ 10⁻⁷ = 1.6 × 10⁷ s, a total of 4.16 × 10⁸ s, and the plug's share falls to 4.0 ÷ 4.16 = 0.9615, the figure's 96.15% with Silt.

What it means. In a line of layers, the tightest one sets the flow and takes almost all the head, however thin it is. The others are left with almost none.

This is how a dam with a clay core works. A core of clay, much narrower than the dam around it, takes almost the whole drop in head, from the reservoir level down to the downstream side. The sand and rock downstream of it carry almost none of the drop, so the water pressure in the downstream half of the dam stays low. Chapter 9 shows the zones that make this work.Water running along layers, instead of across them, meets them side by side, and then the most permeable layer carries most of the flow. Earth fill is spread and rolled in layers, and soil laid down by nature comes in beds, so both pass water more easily along their layers than across them. Hvorslev warned that because of such layering the horizontal permeability of a soil “often is many times greater than the vertical” (Hvorslev (1951), p. 37).

Check yourself Suppose the plug were moved from the middle of the pipe to its left end. Would the flow change? Would the tubes read differently?

The flow would not change: the resistances add the same way in any order, so it stays 110 litres a year. The tubes would: all the sand would now be downstream of the plug, so every tube in the sand would read almost the right tank's level.

A real core is thicker than the plug and holds back far more head, so here is a rough estimate for one:

Example 3.4

Through a real core

Given
An illustrative clay core 5 m thick, with the reservoir standing 30 m higher on its upstream side than the water on its downstream side. The figures' clay: k = 10⁻⁹ m/s, gap fraction 0.35.
Find
How long a drop takes to cross the core, and roughly how much water passes each metre of dam.
The idea
Darcy's law across the core, then the drop's speed through the gaps. For the leak, treat the whole 30 m height of the core as if it carried the full 30 m of head; nearer the top the head is less, so this overestimates.
  1. i = 30 ÷ 5 = 6steep: six metres of head lost for every metre of clay
  2. q/A = 10⁻⁹ × 6 = 6.0 × 10⁻⁹ m/sDarcy's law
  3. v = 6.0 × 10⁻⁹ ÷ 0.35 = 1.71 × 10⁻⁸ m/sthrough the gaps only
  4. t = 5 ÷ 1.71 × 10⁻⁸ = 2.92 × 10⁸ s, or 9.24 yearsthickness over speed
  5. Leak: 6.0 × 10⁻⁹ × 30 = 1.8 × 10⁻⁷ m³/s, which is 0.00018 L/sthrough a strip of core 30 m tall and 1 m long
  6. 0.00018 × 86,400 = 15.6 L a dayseconds in a day

A drop takes about 9 years to cross the core, and each metre of dam leaks about 16 litres a day through it.

Check. The race's clay took 11.09 years for a metre at a gradient of 1. Six times the gradient makes the drop six times faster, and five times the thickness makes the trip five times longer: 11.09 ÷ 6 × 5 = 9.24 years.

What it means. Even a good core leaks, about a bucket and a half a day for each metre of dam, and that water takes years to come through. Chapter 10 works out real leaks with the whole picture of the water in the dam.

Look again at the numbers for clay. A metre of it, under the push that sends water through gravel in half a minute, takes about a decade to pass a drop. The core of a dam is built on that slowness. Hold on to it. The last chapter needs it.

3.8 Water that pushes on grains

So far the water has been the thing that moves, and the grains have only been in its way. But every metre of head the water loses has gone somewhere. Water losing head is shoving its way past grains, and what it shoves, it pushes.

Take a slab of soaked sand one metre square and one metre thick, with water flowing straight up through it and losing 1 m of head on the way, a gradient of 1. If the water were still, the pressure under the slab would be higher than the pressure on top by the weight of the water between and no more, and the soaked sand would get only the lift of being under water. Flowing water needs a metre more head at the bottom to push itself through, so the pressure under the slab is higher than the still-water rule by another 9.81 kPa. That extra pressure, on one square metre, is an extra 9.81 kN pushing the slab up. The water does not rise any faster for it; it hands the push on to the grains, dragging on them as it squeezes past. In general, water that loses head at a gradient i pushes on the grains it passes, in the direction it flows, with a seepage force of

j = i · γw(3.3)

j seepage force on each cubic metre of soil (kN/m³) · i gradient · γw unit weight of water, 9.81 kN/m³

This is also where the pore pressure of Chapter 2 stops following its neat rule. Where water flows upward, the pressure grows faster with depth than 9.81 kPa per metre; where it flows downward, slower. The difference is the push on the grains. How big is it, set against what holds the grains down?

Example 3.5

Nearly enough to float it

Given
Soaked sand weighing 20 kN per cubic metre, grains and water together (the sand of Fig. 4.2); water at 9.81 kN/m³, flowing straight up through it at a gradient of 1.
Find
The seepage force on each cubic metre, and how it compares with the weight of the grains that is left once the water around them holds part of it up.
The idea
Under still water, soaked sand is buoyed up by the weight of the water it displaces, so the grains press down with the difference. Water flowing up pushes them up by i · γw on top of that (Eq. 3.3).
  1. j = 1 × 9.81 = 9.81 kN/m³upward, the way the water flows
  2. Grains' weight, under water: 20 − 9.81 = 10.19 kN/m³the soaked weight less the water's lift
  3. 9.81 ÷ 10.19 = 0.963the push up as a share of the weight down

Water flowing up through sand at a gradient of 1 pushes the grains up with 96% of the weight they have left under water. It nearly floats them.

Check. At a gradient of zero the water is still and the push is zero: the grains press down with their full 10.19 kN/m³. The push grows in step with the gradient, so it would match their weight at a gradient of 10.19 ÷ 9.81 = 1.04, only a little steeper than 1.

What it means. Grains that are pushed up nearly as hard as they weigh hardly press on one another at all. Chapter 4 shows why that matters: a grain's grip on its neighbours comes from that press.

The same push works in any direction. Water seeping down through the upstream part of a dam presses its grains together harder. Water seeping out of a slope drags its grains outward and down, and that is what the moving water does at the foot of the beach, where the streams come out. In Fig. 1.1 the heap with water draining out of its faces stood at 19° instead of 34°, and the drag is why. Example 4.5 does the sum.

Check yourself Water seeps straight down through a layer of sand at a gradient of 0.5. Does the seepage force make the grains press on one another harder or less, and by how much on each cubic metre?

Harder. The force acts the way the water flows, here downward, adding 0.5 × 9.81 = 4.9 kN to each cubic metre's press. Water flowing down presses soil together; water flowing up lifts it.

What to remember

  • Water flows from high total head to low, where head is elevation plus pressure head, h = z + u/γw. In still water the head is the same everywhere, so high pressure at depth moves nothing.
  • Darcy's law: the flow through each square metre is k · i. Permeability runs from 10⁻² m/s for gravel to 10⁻⁹ for compacted clay, and a drop moves faster than k · i, because only the gaps carry it.
  • In a line of layers the tightest takes almost all the head: a clay plug a fifth of the pipe's length takes 99.996%. A core holds back a reservoir the same way, and a metre of clay takes about a decade to pass a drop.
  • Moving water pushes on the grains it passes, i · γw on each cubic metre. Flowing up at a gradient of 1 it carries 96% of the weight the grains have left under water, so they hardly press on one another: the water has taken their press away.

Questions

  1. 3.1

    Piezometer A has its tip 10 m above the datum, and the water in it stands 25 m above the datum. Piezometer B has its tip 2 m above the datum, with water to 22 m. Which way does the water between them flow? Which tip feels the higher pressure?

    Hint

    The water level in each pipe is the head at its tip. The pressure head is the height of that level above the tip.

  2. 3.2

    The pressure on the bed of a deep lake is far higher than at its surface. Why doesn't water flow up out of the lake bed toward the surface?

    Hint

    Push a straw into the lake bed. Where does its water stand?

  3. 3.3

    With sand and the clay plug in Darcy's pipe, raise the right tank from 0.60 m to 1.80 m. What happens to the share of the head the plug takes, and to the flow?

    Hint

    The share depends only on the resistances of the pieces. The flow depends on the head difference too.

  4. 3.4

    How long does a drop of water take to cross 2 m of the figures' silt under a gradient of 1?

    Hint

    Time is the gap fraction times the length, over k times the gradient. Silt: k = 10⁻⁷ m/s, gap fraction 0.40.

  5. 3.5

    The core of Example 3.4 is built twice as thick, 10 m, holding back the same 30 m of head. How do the leak and the time a drop takes to cross it change?

    Hint

    Doubling the thickness halves the gradient. What does that do to the flow, and to the speed of a drop that also has twice as far to go?

Further reading

  • Simmons, C.T. (2008). Henry Darcy (1803–1858): immortalised by his scientific legacy. Hydrogeology Journal 16: 1023–1038. doi:10.1007/s10040-008-0304-3 gentle A short, well-illustrated life of Henry Darcy: the Dijon water supply, the sand column and its rattling gauges, and how one appendix to a book about fountains became the law that every groundwater calculation starts from.
  • Freeze, R.A., Cherry, J.A. (1979). Groundwater. Englewood Cliffs: Prentice-Hall. https://gw-project.org/books/groundwater/ serious The classic groundwater textbook. Its second chapter sets out Darcy's law, hydraulic head and the difference between conductivity and permeability, with piezometers drawn exactly as this chapter's straws. Free to download from the Groundwater Project.
  • Cedergren, H.R. (1989). Seepage, Drainage, and Flow Nets, 3rd ed. New York: Wiley. serious The engineer's book on water in and under dams: how seepage is estimated and how drains and filters control it. Best read alongside Chapter 10.

Chapter 4 · Part I · Ground

Grip

If water doesn't grease the grains, what does it do to them?

All the measurable effects of a change of the stress, such as compression, distortion and a change of the shearing resistance are exclusively due to changes in the effective stresses.

So far. Still water presses harder the deeper you go, and water in the ground flows from high head to low, dragging on the grains it passes. Chapter 3 ended on a cubic metre of soaked sand that flowing water could nearly float. This chapter puts the water and the grains together.

After this chapter you can

  • work out the factor of safety of a block on a ramp, with and without water under it;
  • apply σ′ = σ − u to a column of soil, and say why a lake on top changes nothing;
  • derive the critical gradient, and the steepest face a sand can hold dry, under still water and with water seeping out of it.

This chapter adds small paired arrows under the slope: grains pressing (σ′)

4 figures · about 20 min to read

A wet floor is slippery. So is a wet road, a wet rock in a stream, a wet bar of soap. Everyone learns from the soles of their feet that water greases things. Most people say the same of sand: wet sand is slippery sand, and a bank that slumps after rain has been greased.

Chapter 1 asked you to keep that answer in mind, and it is already in trouble. The damp castle has water in it and stands with upright walls, steeper than the dry heap beside it. If water greased the grains, the castle would be the first to fall.

There is also a direct test. In 1962 two engineers, H. M. Horn and Don Deere, slid polished pieces of minerals over one another, first oven-dried and then soaked in water. Clear quartz, the commonest mineral in sand, gripped with a friction of 0.10 dry and 0.23 wet: more than twice as hard with water on it.Horn & Deere (1962), their friction coefficients for three surface-moisture conditions: clear quartz 0.10 oven-dried, 0.10 air-dried, 0.23 saturated.Not every mineral behaves like quartz. In the same tests the flaky sheet minerals went the other way: muscovite mica fell from about 0.43 dry to 0.23 soaked, and chlorite from 0.53 to 0.22 (Horn & Deere (1962)). Clay is made of flakes of this family, which is one reason clay comes back later in this book as a different kind of soil.

So water isn't greasing the grains. What it does instead was worked out by Karl Terzaghi, the engineer who founded soil mechanics in the 1920s, and the rest of this book stands on it. By the end of this chapter it explains why water can make the castle stronger, why a heap under still water keeps its slope, the 19° face that Chapter 1 left unexplained, and the number every later chapter uses to say whether a slope will stand. To see it, start with something simpler than sand.

4.1 Friction comes from pressing

Put a book on a plank and slowly tilt the plank. The book stays put until the slope gets steep, then it slides. Friction holds it, and friction comes from the book pressing on the plank. Press twice as hard and it takes twice the force to slide it. The friction on offer is always a fixed fraction of the press, and that fraction is what Horn and Deere measured.

The fraction sets a steepest angle for each pair of surfaces. Tilt beyond it and the book slides, whatever it weighs. Engineers call it the friction angle, written φ, and the fraction is its tangent, tan φ. For sand as a whole it is about 34°, the same angle as the dry heap in Fig. 1.1, because the surface of a heap is a ramp of sand resting on sand. Polished quartz grips far less than that: at Horn and Deere's dry 0.10, a polished block would slide off a polished ramp tilted past about 6°.Horn & Deere (1962), clear quartz oven-dried; the angle whose tangent is 0.10 is 5.7°.That dry 0.10 says as much about the surface as about the mineral. Later tests at MIT, in a vacuum chamber, found smooth quartz gripping with anything from about 0.1 to 1.0 depending on how clean its surface was, while rough surfaces varied much less (Bromwell (1966)). None of the tests found water making quartz slippery. Real grains are rougher than a polished block, and they also interlock: to move, they have to roll up and over one another. Both lift sand to its 34°.

The weight drops out of that angle, and a block on a ramp shows why. Put a block of weight W on a ramp tilted at an angle a. Gravity pulls it straight down, and the ramp feels that pull in two parts. The part pressing straight into the ramp is W cos a, and the part pulling the block down the slope is W sin a. Friction can offer up to the press times tan φ. So the block holds as long as W cos a · tan φ is at least W sin a. The weight is on both sides and cancels: the block holds while tan a is less than tan φ, and slides when the ramp passes φ, heavy or light. That is why the dry heap of Chapter 1 keeps one slope however much sand you pour on it.

A designer needs more than "holds" or "slides". The dam in the Prologue had been checked and found to hold; how comfortably it held was a number, and that number is the factor of safety, the friction available divided by the friction needed:

FS = N′ · tan φ ÷ (W · sin a) (4.1)

N′ the press left between the block and the ramp (kN) · tan φ the friction fraction · W sin a the pull down the ramp (kN). For a dry block N′ = W cos a, and the factor of safety is tan φ ÷ tan a.

Above 1 the block holds. Below 1 it slides. At 1.5 the ground could lose a third of its grip before anything moved, and about 1.5 is what dams are designed to, to leave room for everything nobody can measure exactly.1.5 is a convention, not a law of nature. The US Army Corps of Engineers asks a new embankment dam for 1.5 under steady seepage with the reservoir full, 1.3 at the end of construction, and 1.1 to 1.3 right after a fast drawdown of the reservoir; it says such values "continue to be based on experience" (USACE EM 1110-2-1902, p. 3-1 and its table of minimum factors on p. 3-2). The lower values are for conditions that last a short time. Every later chapter of this book uses the same definition, applied to a dam's slope instead of a block.

Now slip a thin cushion of water under the block's base. Make a guess before you see it:

A block rests on a ramp tilted to 25°, holding comfortably. A thin water cushion sits under its base. You pump up the water pressure in the cushion. The block's weight, shape and surface stay exactly the same. What happens?

To decide, tilt the ramp and pump up the cushion in the figure below:

Start with the dry block, before any water. Notice that its two bars stand in the ratio of the factor of safety: 12.2 kN of friction available against 8.5 kN needed, so 1.45. Then notice what the water touches. As the cushion's pressure rises, the press left over falls and the friction bar shrinks with it, kilonewton for kilonewton times tan 34°. The bar for the friction needed doesn't move at all, because the water pushes straight up from the ramp and has no part along it. The block slides when the two bars meet, though its weight, its shape and its surface never changed. Here are the readouts by hand:

Example 4.1

How much water slides the block?

Given
A block weighing W = 20 kN on a base of 1 m², on a ramp at a = 25°. Friction angle φ = 34°, so tan φ = 0.6745.
Find
The factor of safety with no water, and the cushion pressure that slides the block.
The idea
Friction works only with the press left over; the water takes its pressure, times the base area, off the press and leaves the pull alone (Eq. 4.1).
  1. N = 20 × cos 25° = 20 × 0.9063 = 18.13 kNthe part of the weight pressing into the ramp
  2. Pull = 20 × sin 25° = 20 × 0.4226 = 8.45 kNthe part pulling it down the slope
  3. Friction available = 18.13 × 0.6745 = 12.23 kNthe press times the friction fraction
  4. FS = 12.23 ÷ 8.45 = 1.45available over needed, with no water
  5. Press that barely holds: 8.45 ÷ 0.6745 = 12.53 kNat FS = 1 the friction available equals the pull
  6. Water must carry 18.13 − 12.53 = 5.60 kN, so u = 5.60 kPaspread over the 1 m² base, a kilonewton per square metre is a kilopascal
  7. 5.60 ÷ 18.13 = 0.309that pressure as a share of the press

Dry, the block holds with FS = 1.45. It slides when the cushion reaches 5.6 kPa, when the water carries 30.9% of its press: a little under a third.

Check. There is a shortcut. The block slides when the press left is the pull divided by tan 34°, so the share the water must take is 1 − tan 25° ÷ tan 34° = 1 − 0.4663 ÷ 0.6745 = 0.309, the same. On a gentler ramp the water needs more: at 10° the share is 1 − 0.1763 ÷ 0.6745 = 0.7386 of a press of 20 × 0.9848 = 19.70 kN, and 0.7386 × 19.70 = 14.55 kPa.

What it means. Nothing about the block got weaker. The water took a third of its press away and left the pull alone, and a third was all the margin the block had. The steeper the ramp, the less water it takes.

Check yourself Two dry blocks sit on the same 25° ramp, one twice as heavy as the other. Which has the larger factor of safety?

Neither. Doubling the weight doubles the press and the pull alike, and a dry block's factor of safety, tan 34° ÷ tan 25° = 1.45, has no weight in it.

4.2 Soil is thousands of little blocks

Now shrink the block down to a grain of sand. A handful of soil is a crowd of grains, each resting on its neighbours at a few points of contact. Every contact is a tiny block on a tiny ramp, held by friction, and that friction comes from how hard the grains press on each other. The strength of soil is the friction of all those contacts added up.The press is not shared evenly. In experiments with a layer of plastic disks that glow between polarizing filters where they are squeezed, the load travels along branching chains of disks while their neighbours carry much less, and when the pile is sheared the chains line up with the push (Majmudar & Behringer (2005)). Soil mechanics works with the average over many grains, which is what σ′ is.

But fill the gaps with water and something has to carry the water's pressure. To see where it goes, picture a level surface one metre square, deep in soaked sand, that threads its way through the gaps and crosses the grains only where they touch. Everything above that surface, grains and water together, presses down across it. That is the total stress, written σ. The contacts between grains are tiny, so the water in the gaps meets almost all of the surface, and it presses on all of it with its own pore pressure, u. The water therefore carries u of each square metre's σ, and only the rest is left for the grains to carry, through their contacts, by pressing on one another. Terzaghi called that rest the effective stress:Karl Terzaghi, an Austrian engineer, worked the idea out while teaching at Robert College in Constantinople (now Istanbul), in a laboratory he built partly from cigar boxes and kitchenware, and published it in Erdbaumechanik in 1925 (Bennett (2023); Terzaghi (1925)). At the first international conference on soil mechanics, at Harvard in 1936, he put it in the sentence that heads this chapter (Terzaghi (1936), p. 54).

σ′ = σ − u(4.2)

Friction comes from pressing, so the most shear stress the grains can resist, their strength, is that press times the friction fraction:

τ = σ′ · tan φ′(4.3)

σ total stress, the weight of everything above, per square metre (kPa) · u pore pressure (kPa) · σ′ effective stress, how hard the grains press on one another (kPa) · τ shear strength, the most shear stress the soil can carry on a plane (kPa) · φ′ the soil's friction angle, with a prime because it goes with the effective stress; 34° for this sand, tan φ′ = 0.6745

The water does to the grains exactly what the cushion did to the block: it takes over part of the load they would otherwise carry by pressing on one another, and it adds no grip of its own.

To use the rule you need σ at a depth, the weight of everything above it, and for that you need the soil's unit weight, the weight of one cubic metre. Damp sand weighs about 18 kN per cubic metre. Soaked, it weighs about 20, because water has filled the gaps that air held. Below the water table the water in the gaps joins up like a lake, so in still groundwater u grows by 9.81 kPa for every metre below the water table, the rule of Eq. 2.1. The figure below stacks ten metres of this sand and lets you move the water table, from below the column to three metres above the ground:

Fig. 4.2 · Stresses in 10 m of sand as the water table rises Drawn to scale · 18 kN/m³ damp, 20 kN/m³ soaked
4.0 m down
Total stress σ192 kPaat 10 m depth
Pore pressure u59 kPawater 6.0 m above this point
Effective stress σ′133 kPaσ − u
Strength left74%against the water table below the column
Total stress σ Pore pressure u Effective stress σ′

As the water table rises the sand gets heavier, and its grains press on one another less. On the left is a column of sand 10 m deep, damp above the water table and soaked, shaded, below it; drag the water line or use the Water table slider. On the right, stress is plotted against depth: the thin total-stress line σ, the blue pore-pressure line u, and the heavy effective-stress line σ′, which plots the shaded gap between the other two. The readouts are at the column's foot, 10 m down; Strength left compares its effective stress with the column's when the water table is below it, 180 kPa. The number to remember: flooded to the surface, the column is heavier than it was damp, yet its grains at 10 m press with 102 of their 180 kPa, 43% less.

  • Drag the water table up to the ground. Read the three stresses and Strength left.
  • Keep going, three metres above the ground. Which readouts change, and which one doesn't?
  • Drag the water table all the way down, below the column. What does the pore-pressure line do?
Answers

At the ground: 200, 98.1 and 101.9 kPa, with 57% of the strength left. With 3 m of water on top, σ and u both rise to 229 and 128 kPa, and σ′ stays at 101.9. Below the column there is no pore pressure at 10 m, and all 180 kPa of the weight is pressed grain on grain.

Notice first what happens as the water table rises from below the column to the ground. The sand gets heavier, because water is replacing air in its gaps, yet the effective stress at 10 m falls from 180 to about 102 kPa. That is 43% of the soil's strength gone, and no grain had to move. Below the water table each extra metre of sand adds 20 kPa to the total stress and 9.81 to the pore pressure, so the grains feel only the difference, 10.19 kPa a metre: the sand's buoyant unit weight, written γ′. Above the water table they feel all 18.

Notice second what happens once the water rises above the ground. Every extra metre of lake adds the same 9.81 kPa to the total stress and to the pore pressure, so the effective-stress line stops moving. A lake sitting on soil doesn't press its grains together any harder, however deep it is. Here are the numbers:

Example 4.2

Ten metres of sand, three water tables

Given
A column of sand 10 m deep: 18 kN/m³ above the water table, 20 kN/m³ below it; water 9.81 kN/m³.
Find
σ, u and σ′ at the foot with the water table 4 m down, at the ground, and under 3 m of lake.
The idea
σ is the weight of everything above; u is the depth below the water surface times 9.81; the grains get the difference (Eq. 4.2).
  1. 4 m down: σ = 18 × 4 + 20 × 6 = 192 kPa4 m of damp sand on 6 m of soaked
  2. u = 9.81 × 6 = 58.86 kPathe foot is 6 m below the water table
  3. σ′ = 192 − 58.86 = 133.14 kPawhat the grains press with
  4. At the ground: σ = 20 × 10 = 200, u = 9.81 × 10 = 98.1, σ′ = 101.9 kPaall ten metres soaked
  5. 101.9 ÷ 180 = 0.566against 18 × 10 = 180 kPa with the water table below the column
  6. 3 m of lake: σ = 9.81 × 3 + 200 = 229.43, u = 9.81 × 13 = 127.53 kPathe lake's weight is added to both; the foot is now 13 m below the water's surface
  7. σ′ = 229.43 − 127.53 = 101.9 kPaunchanged by the lake

With the water table 4 m down the grains at the foot press with 133.1 kPa. Flooded to the ground they press with 101.9 kPa, 43% less than with the water table below the column, and 3 m of lake on top leaves that at 101.9.

Check. Count the grains' press metre by metre: 18 kPa for each damp metre and the buoyant 10.19 for each soaked one. With the water table 4 m down, 72 + 6 × 10.19 = 133.14 kPa, the same as step 3.

What it means. Water in the ground weakens soil by taking over part of the grains' press. Raising a reservoir over a sand bed adds to σ and u alike and leaves the bed's grains exactly as they were; what does change the press is water that flows, which comes next.

Check yourself Wells pump the water table down from the ground surface to 4 m below it. Is the sand at 10 m now heavier or lighter, and stronger or weaker?

Lighter and stronger. σ falls from 200 to 192 kPa, but u falls further, from 98.1 to 58.9, so the grains press with 133.1 kPa instead of 101.9.

4.3 One line for the strength

Every strength in this chapter so far has been a press times the same fraction. So if you plot strength against effective stress, every state of this sand, wet or dry, deep or shallow, lands on one straight line through zero, rising at the friction angle:

Fig. 4.3 · The strength of one sand at any pressBoth axes in kPa, to the same scale
50 100 150 200 50 100 0 effective stress σ′ (kPa) shear strength τ (kPa) τ = σ′ tan 34° out of reach below the line: it holds the block at 25° (18.1, 12.2) column flooded (101.9, 68.7) water table 4 m down (133.1, 89.8) water below the column (180, 121.4) flooding

Every state of this sand, wet or dry, lies on one straight line through zero: strength is the grains' press times tan 34°. Across is the effective stress σ′, how hard the grains press; up is the shear strength τ, the most shear stress they can carry. Both axes have the same scale, so the line rises at the friction angle itself. Below the line the soil holds; the shaded region above it is out of reach, because the grains of this sand can never carry more shear than their press allows. The open dots are the block of Fig. 4.1, given sand's friction angle, and the foot of the column of Fig. 4.2 at three water tables; the blue arrow is the column being flooded. The number to remember: flooding takes the column's foot from 121.4 to 68.7 kPa of strength, without changing the line.

  • Divide the strength of any dot by its effective stress. What do you get each time?
Answers

The same fraction every time, tan 34°: 121.41 ÷ 180 = 0.6745, and 68.73 ÷ 101.9 = 0.6745.

Notice that the water never moves the line. Flooding slides the column's point down along it, to a smaller press and so a smaller strength; the sand's friction angle is the same throughout. That is the whole effect of still water on sand. Engineers call this line the strength envelope. Some soils, clays among them, keep a little strength even when the grains are barely pressed, called cohesion and written c′; their line starts at c′ instead of zero, τ = c′ + σ′ tan φ′. Clean sand has none.

Example 4.3

Reading strength off the line

Given
The effective stress at the column's foot from Example 4.2: 180 kPa with the water table below the column, 133.14 kPa with it 4 m down, 101.9 kPa flooded. tan φ′ = 0.6745.
Find
The shear strength of the sand at the foot in each case.
The idea
Strength is the press times the friction fraction (Eq. 4.3).
  1. Below the column: 180 × 0.6745 = 121.41 kPathe most strength the foot has
  2. 4 m down: 133.14 × 0.6745 = 89.80 kPathe figure's opening state
  3. Flooded: 101.9 × 0.6745 = 68.73 kPathe same line, further down it

The foot can resist 121.4, 89.8 and 68.7 kPa of shear in the three cases.

Check. The flooded strength over the dry one should be the flooded press over the dry press, since the fraction is the same: 68.73 ÷ 121.41 = 0.566, as in Example 4.2.

What it means. These are strengths, not loads: the grains at the foot of a level column are not being sheared at all. A slope shears them, and Chapter 12 sets the shear a slope asks for against this line, all along a curved slip surface.

4.4 Water that flows up

Still water, however deep, only takes away the press that its own pressure accounts for. But Chapter 3 ended with water that flows, and flowing water drags on the grains it passes: water losing i metres of head for every metre it travels pushes on the grains in each cubic metre with i × 9.81 kN, in the direction it flows.

Turn that flow upward, as it does where water that has passed under a dam comes up again on the far side. Soaked sand presses down with its buoyant unit weight, 10.19 kN per cubic metre. The water flowing up through it pushes up with i × 9.81. Raise the gradient and the upward drag eats into the downward press, until at one gradient they cancel. There the grains still touch but no longer press on one another, the effective stress is zero, and the sand has no grip at all. That gradient is the critical gradient:

ic = γ′ ÷ γw = (γsat − γw) ÷ γw (4.4)

γsat unit weight of the soaked soil (kN/m³) · γw unit weight of water, 9.81 kN/m³ · γ′ buoyant unit weight, γsat − γw

Example 4.4

The gradient that floats the grains

Given
Soaked sand of 20 kN/m³, the column's; water 9.81 kN/m³, flowing straight up through it. For the second route, typical quartz sand: grains 2.65 times as heavy as water, and 0.65 m³ of gap for every cubic metre of grain.
Find
The upward gradient at which the grains stop pressing on one another.
The idea
The water's upward drag, i × 9.81 per cubic metre, against the grains' buoyant weight (Eq. 4.4).
  1. γ′ = 20 − 9.81 = 10.19 kN/m³the press of each soaked cubic metre on the grains below
  2. ic = 10.19 ÷ 9.81 = 1.04upward drag equals that press
  3. From the grains: γsat = 3.30 ÷ 1.65 × 9.81 = 19.62 kN/m³a cubic metre of grain weighs as much as 2.65 m³ of water, and its gaps hold 0.65 m³ more: 2.65 + 0.65 = 3.30 cubic metres' worth of water's weight in 1.65 m³
  4. ic = (2.65 − 1) ÷ (1 + 0.65) = 1.65 ÷ 1.65 = 1.00the same rule, γ′ ÷ γw, written with the grains' numbers

Sand floats when the water flowing up through it loses about a metre of head for every metre it climbs: 1.04 for the column's sand, 1.00 for the typical one.

Check. The two routes differ only because the second sand is a little lighter, 19.62 kN/m³ against 20. A soaked sand weighs about twice as much as water, so its buoyant weight is about one water's worth, and a gradient of about 1 lifts it.

What it means. No sand anywhere is much stronger than this against water rising through it. Where seepage comes up at the downstream foot of a dam, a gradient near 1 makes the ground heave or boil, and Chapter 11 follows what happens next.

4.5 The beach face, worked out

Everything so far has been level ground. A slope adds a pull, and that brings back the two wet heaps of Fig. 1.1. A heap under still water stands about as steep as a dry one. The water carries part of each grain's weight, so the grains press on each other less, but the same buoyancy lightens their pull down the slope by the same share, and the angle depends only on the ratio of the two. The heap slumps to about 19° only where water drains out through its face, because the flowing water drags the grains downhill as well, the drag of Chapter 3. Chapter 1 promised the sum that turns that drag into degrees.

It needs one more idea. A beach face, or the long face of a dam, is so long and even that every patch of it is like every other. Engineers call it an infinite slope, and its great convenience is that one grain on the face, or one cubic metre of it, stands for the whole. Take that grain as a block on a ramp of sand and draw its forces in the three states:

Fig. 4.4 · One grain on the face of a heap, three waysSchematic · arrows to scale within each panel
34° weight press pull friction A · Dry 34° weight in water press pull friction B · Under still water 19.0° weight in water press pull drag of the seeping water friction C · Water seeping out

Water seeping out of a face adds to the pull on its grains and not to their press, so the face can stand only about half as steep, measured as rise over run. Each panel is a section through the face of a long heap of the same sand, at the steepest angle it holds, with one grain drawn large. The weight (straight down) splits into a press into the face (ochre) and a pull down it (brick red); the press makes the friction available (green), which exactly matches the pull in all three. Under water the grain weighs less, so every arrow in B is shorter than in A by the same share. In C the blue arrows inside the face are the water seeping out parallel to it, and the blue arrow added to the pull is their drag. The numbers to remember: 34° dry, 34° under still water, 19.0° with water seeping out.

  • In panel C, lay a ruler along the pull and the drag together and compare them with the pull alone.
  • Set Fig. 1.1 to Water draining out and compare its face with panel C.
Answers

The drag is nearly as long as the pull, so together they come to about twice it: the drag is 49% of the whole pull, γw ÷ γsat. The heap in Fig. 1.1 stands at the same 19°.

Notice that panels A and B have the same shape. Under still water every arrow shrinks by the same share, because buoyancy takes the same fraction off the grain's weight wherever it points, and so the angle stays at 34°. Notice then what is new in C. The grain weighs what it did in B, and its press into the face is no bigger, but a second arrow has joined the pull: the drag of the water flowing past it. Friction comes only from the press, so the face must flatten until the press's friction catches up. Here is the sum Chapter 1 promised:

tan β = (γ′ ÷ γsat) · tan φ′ (4.5)

β the steepest angle of a face with water seeping out of it, parallel to the face, in sand with no cohesion · γ′ buoyant unit weight · γsat soaked unit weight · φ′ the sand's friction angle. Dry or under still water, tan β = tan φ′.

Example 4.5

Why the draining face stands at 19°

Given
The beach sand of Fig. 1.1: soaked, γsat = 20 kN/m³; water 9.81 kN/m³; φ′ = 34°, tan φ′ = 0.6745. Water seeps out through the face, flowing parallel to it.
Find
The steepest angle β the face can stand at.
The idea
Treat one cubic metre of the face as the block on its ramp (Example 4.1), and add the drag of the flowing water to its pull.
  1. Press into the face: 10.19 cos β kNthe grains' weight in water, 20 − 9.81, split as on the ramp
  2. Gradient along the flow: i = sin βflowing parallel to the face, the water stays at one depth and one pressure, so the head it loses equals the height it drops: sin β per metre
  3. Pull: 10.19 sin β + 9.81 sin β = 20 sin β kNthe weight's pull plus the drag, i × 9.81
  4. On the point of sliding: 20 sin β = 10.19 cos β × tan 34°friction available equals the pull, FS = 1; dividing by cos β gives Eq. 4.5
  5. 10.19 ÷ 20 = 0.5095γ′ ÷ γsat
  6. tan β = 0.5095 × 0.6745 = 0.3437a little over half of tan 34°
  7. β = arctan 0.3437 = 19.0°the angle whose tangent that is

The face stands at 19.0°, 15.0° flatter than the dry heap's 34°.

Check. Take the drag away and the pull is 10.19 sin β, the same weight as the press: the 10.19 cancels and tan β = tan 34°, the heap under still water. And at 19.0° the face is in balance: the press is 9.637 kN, which gives 9.637 × 0.6745 = 6.500 kN of friction, against a pull of 3.312 + 3.188 = 6.500 kN.

What it means. The 19° in Fig. 1.1 was right; Chapter 1 only couldn't say why yet. Soaked sand weighs about twice as much as water, so water seeping out of a face roughly doubles the pull on its grains and roughly halves the tangent of the steepest slope. The downstream face of an earth dam, where seepage comes out, is this face at the scale of a hill, and the drains of Chapter 9 are there to keep the water from reaching it.

Check yourself Water seeps out of the face of a sand with heavier grains, soaked at 22 kN/m³ instead of 20, with the same friction angle. Does its face stand steeper or flatter than 19°?

Steeper. Its buoyant weight is a larger share of its soaked weight: γ′ ÷ γsat = 12.19 ÷ 22 = 0.554, so tan β = 0.554 × 0.6745 = 0.374 and β = 20.5°. The heavier the grains against the water, the less the water's drag counts.

4.6 The sentence, proved

Turn back to the title page and read the sentence there. You can now prove it:

Soil holds only as well as its grains press on one another, and the water between them can take that press away.

The first half is Eq. 4.3: strength is the grains' press times a fixed fraction, and nothing else. The second half is Eq. 4.2: whatever pressure the water has, it subtracts from that press. Every result in this chapter is one of the two halves at work. Water in the cushion took 31% of the block's press and slid it. Flooding the column took 43% of the grains' press. Water flowing up at a gradient of about 1 takes all of it. Water seeping out of a face adds a pull and no press, and the face falls from 34° to 19°.

The castle of Chapter 1 seemed to contradict the sentence, because there the water makes the sand stronger. It obeys it too. In damp sand the water hangs in small rings between the grains, and the curved skin of each ring pulls, so the water's pressure is below the air's: u is negative. Subtract a negative and σ′ = σ − u is larger than σ. The water adds to the press instead of taking it away, by the same rule with the sign turned over (in damp sand only part of that pull counts, because the water touches only part of each grain). Chapter 5 puts a number on that pull and shows why it can hold up a wall a few tens of centimetres tall and never a 30 m one.

The sentence doesn't cover everything. Water that pours over the top of a dam washes it away whatever its grains are doing, and Chapter 6 takes up that story. But for what soil does with the water inside it, this is the rule the rest of the book applies.

What to remember

  • Friction comes from pressing: the grains' strength is their press times tan φ′, and the factor of safety is the friction available over the friction needed. A dry block, or a dry heap, slides at φ whatever it weighs.
  • Water pressure takes part of the grains' press and none of the pull: σ′ = σ − u. Under the 25° block, 31% was enough to slide it.
  • A rising water table makes soil heavier and its grains press less; a lake on top adds to σ and u alike and changes nothing.
  • Water flowing up at the critical gradient, about 1, takes the grains' whole press away.
  • Water seeping out of a face adds a pull and no press, so the steepest face falls to tan β = (γ′/γsat) tan φ′: 19.0° for the beach.

Questions

  1. 4.1

    Set the ramp in Fig. 4.1 to 30°, with no water. What is the block's factor of safety, and what cushion pressure slides it?

    Hint

    Use the shortcut of Example 4.1: the share of the press the water must carry is 1 − tan 30° ÷ tan 34°.

  2. 4.2

    A block twice as heavy as the one in Fig. 4.1, with the same base, sits on the same 25° ramp with the same 5 kPa of water in its cushion. Is it more or less likely to slide than the lighter block?

    Hint

    The water carries the same number of kilonewtons under each block. What share of each block's press is that?

  3. 4.3

    Put the water table in Fig. 4.2 2 m below the ground. What are σ, u and σ′ at 10 m, and how much of the strength is left?

    Hint

    Two metres of damp sand on eight of soaked; the foot is 8 m below the water table.

  4. 4.4

    A reservoir is filled 20 m deeper over a bed of sand that it has covered for years, and nothing flows through the bed. Is the sand under the lake weaker than before?

    Hint

    What do 20 m of extra water add to σ, and what do they add to u?

  5. 4.5

    Below the downstream foot of a dam, water rises through the sand of the column (20 kN/m³), losing 0.6 m of head for every metre it climbs. What share of the grains' press is left?

    Hint

    Each cubic metre presses down with its buoyant weight and is pushed up by i × 9.81.

  6. 4.6

    A beach of sand with φ′ = 30° and the same unit weights has water draining out of its face. What is the steepest face it can hold?

    Hint

    Use Eq. 4.5 with tan 30° in place of tan 34°.

Further reading

  • Bennett, M. (2023). Review of Modern Conceptions Concerning Foundation Engineering (Terzaghi, 1925). Geo-Institute of the American Society of Civil Engineers. https://www.geoinstitute.org/news/review-modern-conceptions-concerning-foundation-engineering-terzaghi-1925 gentle A short, free account of Terzaghi's years in Constantinople and the book of 1925 that came out of them.
  • Holtz, R.D., Kovacs, W.D., Sheahan, T.C. (2011). An Introduction to Geotechnical Engineering, 2nd ed. Upper Saddle River: Pearson. serious A standard first course in soil mechanics; its chapters on effective stress and on shear strength take this chapter's two equations further, with laboratory tests and worked problems.
  • Terzaghi, K. (1936). The shearing resistance of saturated soils and the angle between the planes of shear. Proceedings of the First International Conference on Soil Mechanics and Foundation Engineering, Cambridge, Massachusetts, vol. 1: 54–56. https://www.issmge.org/uploads/publications/1/44/1936_01_0017.pdf primary source Three pages, free from the ISSMGE library. The first paragraph states the principle of effective stress, under the name Terzaghi used then, the "neutral stress" for u.

Chapter 5 · Part I · Ground

When the water can't get out

Why does firm wet sand go pale under one foot, while the sand at the water's edge swallows the other?

So far. A soil grips only as hard as its grains press on one another, and water pressure in the gaps takes part of that press away: σ′ = σ − u. A lake standing on top changes nothing, because it adds to the weight and to the water pressure alike. All of that assumed the water could come and go as it pleased.

After this chapter you can

  • put a number on the grip of damp sand, and say why a sand wall stands a quarter of a metre and never thirty;
  • predict from how sand is packed whether shearing makes it swell or shrink, and so explain the pale footprint, the sinking foot and liquefaction;
  • estimate how long water takes to leave a layer of sand or clay, and so tell a drained loading from an undrained one.

This chapter adds the soaked zone below the phreatic line shaded (water carrying load)

3 figures · about 20 min to read · Film 5.1, 8 min 45 s

Go back to the beach at low tide and walk down to the firm sand in the middle, where the sea was an hour ago. Put your weight on one foot. A pale ring spreads round it, as if the sand had dried where you stood, and fades in a second or two. Lift the foot and the print turns dark and shiny for a moment, wetter than the sand around it.

Now step down to the edge, where each wave lays a fresh sheet of sand. Stand still and it holds you. Wiggle your feet and it goes soft and swallows them to the ankle.

Both sands are saturated: every gap between the grains is full of water. Chapter 4 showed that water pressure in the gaps takes part of the grains' press away, and that still water standing on top changes nothing. By that account two wet sands pressed by the same foot ought to behave alike. They behave as opposites. Most people, asked what a foot does to wet sand, say it squeezes the water out, so the sand round the foot should look wetter. On the firm beach it looks drier.

Chapter 1 left a clue. When its heap of wet sand was jiggled, the densely packed heap held and the loosely packed one flowed, so packing matters. Time matters too. Every sum in Chapter 4 let the water flow in or out as fast as it liked, and under a moving foot it can't, not quickly enough. This chapter is about what water does when it is trapped. It starts with the simplest case, the damp castle, where the water pulls.

5.1 Water that pulls

The damp castle has only a little water in it, and that water gathers in tiny rings where the grains touch (the close-up in Fig. 1.1 draws them). The surface of each ring is curved and stretched, like the skin of a soap bubble, and it pulls the two grains toward each other. So the water in a ring is at less than the pressure of the air around it: a negative pore pressure, which engineers call suction.

Put a negative u into Eq. 4.2 and the effective stress goes up: σ − u is more than σ when u is below zero. The grains press together harder than their own weight could make them, and by Eq. 4.3 more press means more grip. That extra grip does not depend on how much sand lies above, so it can hold grains on a vertical face, where weight alone would roll them off.

How much grip? Nobody can put a pressure gauge inside a ring of water a fraction of a millimetre across, but the grip can be measured whole. Shear damp sand in a box with nothing pressing on it and it still resists, with about 0.6 kPa for a sand of fine, angular grains.Richefeu, El Youssoufi & Radjaï (2006): sand of angular grains 0.1–0.4 mm across, sheared in a box; its cohesion rose with the water content and levelled off at about 600 Pa once the water was about 3% of the weight of the grains. Grip that is there with no press at all is called cohesion, written c. In damp sand it is borrowed from the water, so engineers call it apparent cohesion. It is the grip a suction of 0.6 ÷ tan 34° = 0.89 kPa would give if it pressed on every grain, the weight of about 5 cm of damp sand. Whether that matters depends on what it is set against.

A wall of soil with cohesion fails when a wedge of it slides off along a sloping crack: the block on the ramp of Fig. 4.1, with a little glue on the ramp. Try every angle of crack and every shape of wedge, and the worst one gives the tallest vertical wall the soil can hold:The factor 3.83, the best upper bound for a vertical cut whatever the friction angle: Salençon (2024), p. 273. Chapter 12 does the same kind of search for slopes.

H ≈ 3.83 · (c ⁄ γ) · tan(45° + φ′⁄2) (5.1)

H tallest vertical wall (m) · c cohesion (kPa) · γ unit weight of the soil (kN/m³) · φ′ its friction angle

Notice that the height grows with the cohesion and shrinks with the weight. The ratio c/γ is the cohesion written as a height of soil, and the wall can stand a few times taller than that.

Example 5.1

How tall can a damp wall stand?

Given
Damp sand weighing γ = 18 kN/m³ (the stress column's damp sand, Fig. 4.2), friction angle φ′ = 34° (the dry heap's), and the measured apparent cohesion c = 0.6 kPa. The castle in Fig. 1.1 is about 0.3 m tall.
Find
The press at the foot of the castle from its own weight, the tallest wall this sand can hold, and the cohesion a 30 m wall would need.
The idea
The weight presses harder the taller the wall grows; the cohesion stays the same. Eq. 5.1 says where the two balance.
  1. σ = 18 × 0.3 = 5.4 kPathe weight on each square metre at the foot of a 30 cm wall
  2. tan(45° + 17°) = tan 62° = 1.8807half the friction angle, added to 45°
  3. H = 3.83 × 0.6 ÷ 18 × 1.8807 = 0.24 mEq. 5.1: the tallest wall
  4. σ = 18 × 30 = 540 kPathe press at the foot of a 30 m wall
  5. 3.83 × 1.8807 = 7.203the two numbers in the formula that don't change
  6. c = 540 ÷ 7.203 = 75.0 kPathe same formula turned round: c = γH ÷ (3.83 tan 62°)
  7. 75 ÷ 0.6 = 125how many times the grip the water rings supply

This sand holds a vertical wall about 0.24 m tall. A wall 30 m tall would need 75 kPa of cohesion, 125 times what the water supplies.

Check. The answer falls a little short of the 0.3 m castle, and the gap is fair: 0.6 kPa was measured on one sand, prepared one way. Finer sand has more contacts in each square metre, each with its own ring, so it grips harder, and a round tower is braced by its own curve. What the sum settles is the size: tens of centimetres, never tens of metres. And with no cohesion at all, dry sand, H = 0: no vertical face, only the 34° heap.

What it means. At the foot of the castle the weight already presses with 5.4 kPa, six times the 0.89 kPa the suction adds, so suction matters most near the top, where the grains carry almost no weight. At the base of a dam the grains press with hundreds of kilopascals. Designers leave suction out, as Chapter 1 said, and a week of rain or drought would take it away in any case.

Dry the sand and the rings evaporate; flood it and they merge into one body of water at ordinary positive pressure. Either way the suction goes and the wall slumps.Treating suction as a negative u slightly overstates it. The rings wet only part of each grain, so only part of the suction presses the grains together. Alan Bishop wrote the effective stress of a damp soil with a factor χ between 0 (dry) and 1 (saturated) in front of the suction (Bishop (1959)). The measured cohesion used here already includes that factor. Suction needs air in the gaps. But the firm sand under your foot has no air in its gaps. It is saturated, and it stiffens anyway.

5.2 Loose and dense

The answer lies in how the grains are packed. Push on a heap of grains so that they must slide past one another, and what happens next depends on how they sat. Densely packed grains sit in each other's hollows. To slide, each must ride up over the one in front, so the heap swells as it shears. Loosely packed grains have hollows below them, and shearing drops them in, so the heap shrinks. Engineers call the first kind dilative and the second contractive.Osborne Reynolds named the swelling dilatancy in 1885. He filled a rubber bag with lead shot and water and sealed it, and the bag went rigid in whatever shape it had: to change shape the shot had to spread apart, and there was no more water to fill the room. Connected to a measuring vessel, the bag drew water in as he squeezed it out of shape. He explained the pale footprint in the same paper (Reynolds (1885), p. 475).

Here are both, in close-up, with every gap full of water:

Fig. 5.1 · Two wet sands under a footSchematic · grains drawn far larger than life
Firm sand: densely packed foot before sheared: it swells the water is drawn down between the grains, and the top grains are left dry: the pale ring Soft sand at the edge: loosely packed foot sinks before wiggled: it can't shrink yet the water can't get out, so it takes the load: the grains no longer press on one another

Sheared under a foot, dense sand must swell and loose sand must shrink, and the water in the gaps cannot keep up with either. Each panel shows a few grains before and after, with the gaps full of water. The ochre lines join grains that press on one another, thicker where they press harder (the sandbox later in this chapter draws them the same way). Left: the dense stack rides up into a looser one; the water it holds no longer fills the gaps, so the water surface (the dashed line) sinks below the top grains, which are left pale. Right: the loose stack starts to drop toward a tighter one, but it can shrink only by pushing water out, and the water can't leave in time, so the box keeps its size; the water's pressure rises (drawn darker) and it carries the load the grains were carrying. The number to remember: the water can follow only as fast as it drains, and leaving 2 m of sand takes it about 20 s (Example 5.2), while a footstep lasts about one.

  • Count the grains in each box. What changes between before and after, and what doesn't?
  • Lay a ruler against the left panel. How much taller is the sheared stack than the one before? Then measure the two right-hand boxes.
Answers

Thirty grains in every box; only the space between them changes. The sheared dense stack is about an eighth taller: where the first measures 40 mm, the second measures about 45 mm (the drawing exaggerates the swelling so you can see it). The two right-hand boxes are the same size: the loose sand has not shrunk at all, because shrinking means pushing water out.

Notice that in both drawings the water is the wrong amount for the new arrangement. The dense stack has grown more gap than it has water to fill. More water can only flow in through the narrow gaps from the sand around, which takes time. Until it arrives, the water already there is stretched to fill the larger space, and its pressure falls below what still water at that depth would have. Engineers call the difference from the still-water pressure the excess pore pressure, Δu; here it is negative. A negative Δu works like suction: by Eq. 4.2 the grains press harder, and the sand under your foot stiffens.

The water the sheared sand pulls in has to come from somewhere, and the nearest supply is the water among the top grains. So the water surface is drawn down below the top grains and leaves them looking dry: that is the pale ring. Reynolds saw both supplies, the surface and the sand around, and saw why the surface wins at first: drawing water through the sand "requires time to accomplish".Reynolds (1885), p. 475 (p. 209 of the 1901 reprint). A second or two later the water from below arrives, the surface rises, and the ring darkens again.

Keep the words apart. Suction is the pull in damp sand, where air fills most of each gap. Negative excess pore pressure is the pull in saturated sand that is trying to swell faster than water can follow. Both make the grains press harder.

Check yourself You lift your foot out of the pale ring and the print turns dark and shiny for a moment. Why?

With the foot gone, the swollen sand settles back into its dense arrangement and shrinks. It now holds more water than its gaps need, and the extra is pushed up to the surface until it drains away.

The loose sand at the water's edge does the opposite. Each wiggle of your foot shears it, and its grains start to drop into the hollows below them. For the pack to shrink, water has to leave it, and water leaves sand only through the same narrow gaps. So the pack can't shrink yet. The water, squeezed between grains that are trying to close on it, rises in pressure: a positive Δu. It now carries part of the weight the grains were carrying, so they press on one another less, and grip less, and your foot sinks. Stand still for a few seconds and the water seeps away upward, the grains settle and press again, and the sand holds you once more.

A house on the same sand is a much bigger foot, and an earthquake shakes it for longer than you can wiggle. Make a prediction before reading on:

A house stands on loose sand that is full of water. An earthquake shakes the ground for a few seconds, too fast for the water to drain away. What happens to the house?

The question turns on its last words: too fast for the water to drain away. How fast is that?

5.3 A matter of time

Whether the water keeps up is a race between two clocks: how long the load takes to come on, and how long the water needs to move. If the water wins, its pressure stays at the still-water value while the grains rearrange, and engineers call the loading drained. If the load wins, the water is trapped, the soil can't change its volume, and the water pressure changes instead; that is undrained. The same sand can be either; what decides is how fast.

The water's clock can be worked out from things you already have. Picture a saturated layer L thick, with free-draining gravel at its top, and a load put on it suddenly. At first the water carries the whole of the new load as an excess pressure Δu. To hand it on to the grains, the layer must squeeze out enough water to let it shorten, and how far it shortens depends on its stiffness against squeezing, M, in kPa: a press of Δu shortens it by the fraction Δu/M of its thickness. That is a volume of water Δu·L/M under each square metre. Darcy's law, Eq. 3.2, says how fast it leaves: the excess is a head of Δu/γw spread over a path of length L, so each square metre passes k·Δu/(γw·L) cubic metres a second. Divide the volume by the rate. Δu cancels, and what is left is a time:Karl Terzaghi published the theory of how excess water pressure drains out of a layer of clay in 1923 (Terzaghi (1923)), two years before the book Chapter 4 credits with effective stress. The rough sum here gives the form of his answer; his full solution adds a factor close to 1 for most of the pressure to drain.

t ≈ L² ⁄ cv (5.2)
cv = k · M ⁄ γw (5.3)

t time for most of the excess pressure to drain (s) · L the longest path the water must travel to a free-draining face (m) · cv the coefficient of consolidation (m²/s) · k permeability (m/s) · M the soil's stiffness against squeezing (kPa) · γw = 9.81 kN/m³

Notice the square. A layer twice as thick has twice the water to lose along a path twice as long, so it takes four times as long. And notice the k. Sand lets water through a hundred thousand times more easily than clay, so for the same stiffness its water leaves a hundred thousand times sooner. The slow squeezing out of water under a new load is called consolidation.

Example 5.2

How long does a layer take to drain?

Given
Sand with k = 10⁻⁴ m/s and clay with k = 10⁻⁹ m/s, the permeabilities of Fig. 3.3. Stiffnesses, illustrative round numbers (real soils vary widely, and stiffen with depth): M = 20,000 kPa for the sand and 10,000 kPa for the clay. Layers 2 m and 10 m thick, each draining through one face.
Find
How long each layer takes to drain.
The idea
Work out cv from Eq. 5.3, then the time from Eq. 5.2.
  1. M ÷ γw = 20,000 ÷ 9.81 = 2,038.7 mthe sand's stiffness written as a height of water
  2. cv = 10⁻⁴ × 2,038.7 = 0.20387 m²/spermeability times that height
  3. t = 2² ÷ 0.20387 = 19.6 s2 m of sand
  4. t = 10² ÷ 0.20387 = 490.5 s10 m of sand: about 8.2 minutes
  5. M ÷ γw = 10,000 ÷ 9.81 = 1,019.4 mthe clay's stiffness as a height of water
  6. cv = 10⁻⁹ × 1,019.4 = 1.0194 × 10⁻⁶ m²/stimes 86,400 seconds in a day, 0.08807 m² a day
  7. t = 2² ÷ 0.08807 = 45.4 days2 m of clay
  8. t = 10² ÷ 0.08807 = 1,135 days10 m of clay
  9. 1,135 ÷ 365.25 = 3.11 yearsin years

The sand drains in 20 seconds (2 m) to about 8 minutes (10 m). The clay takes 45 days to 3.1 years.

Check. For the same 2 m layer the clay should take longer than the sand by the factor of 100,000 in k times the factor of 2 in M, so 200,000 times. In seconds the clay takes 45.4 × 86,400, and that divided by 19.6 s is about 200,000. The 10 m layers take 25 times as long as the 2 m ones, as the square says.

What it means. Sand is drained under anything slower than a few minutes and undrained under anything quicker: a footstep, an earthquake. Clay is undrained under almost everything except the slow work of months and years.

Laid on one line beside the loads a soil meets, those four times sort themselves. The strong shaking of the 1971 earthquake at Lower San Fernando Dam, which comes back below, lasted about 15 seconds:Strong shaking at the dam for about 15 s: ASDSO, Lower San Fernando Dam.

Fig. 5.2 · Two clocks: the load and the waterTime on a ten-fold scale · illustrative soils
a footstep an earthquake's strong shaking a year of building 2 m of sand · 19.6 s 10 m of sand · 8.2 min 2 m of clay · 45 days 10 m of clay · 3.1 years 1 s 1 min 1 h 1 day 1 yr 10 yr time: every tenfold step is the same length

Sand gives up its water in seconds to minutes and clay in months to years, so the same load can be drained for one and undrained for the other. Time runs left to right on a scale where each equal step is ten times longer. The dashed brick-red lines mark how long three loads last. Each blue bar runs to the time water needs to leave a layer, from Example 5.2. Where a bar reaches past a load's line, the water is still trapped when the load is done: that load is undrained for that layer. The number to remember: 2 m of sand drains in about 20 s, 2 m of clay in about 45 days.

  • Find the load that is undrained for every layer, and the one layer whose water is still trapped after a year of building.
  • Where would 4 m of sand end? Its time is four times that of 2 m.
Answers

A footstep: every bar runs past its line. After a year of building only the 10 m of clay still holds its water. Four metres of sand takes 4 × 19.62 = 78.5 s, a little past one minute on the line.

Notice where the bar for 2 m of sand ends: a little past the earthquake's line. The sand's water has barely started to leave when the shaking stops, and each swing of the ground, a small fraction of the whole, is far quicker still: the shaking is undrained. The clay bars run past the earthquake and the footstep by a long way, and the thickest runs past a year. A dam raised over clay a few metres at a time loads it undrained at first, and the clay hands the load from its water to its grains over the months that follow.

Check yourself A 4 m layer of clay has sand above it and sand below it, and water can leave through both faces. Does it drain faster or slower than a 2 m layer that drains through one face?

The same. Water in the middle of the 4 m layer has 2 m to go either way, so L is 2 m and t = 2² ÷ 0.08807 = 45.4 days. Drained through one face only, it would take four times as long.

5.4 The sandbox as a laboratory

The two clocks are hard to watch in real sand: the grains are small and the water is hidden. The sandbox below shows both. Every grain in it is simulated as a disc that pushes, rubs and rolls on its neighbours, and the water in the gaps is a pressure that rises when the pack is squeezed and falls when it swells. To shear the grains is to push the top sideways so that they slide past one another. The Drain switch sets the race: Open lets water flow in and out of the box freely, which is drained; Shut traps it, which is undrained.

The four experiments are this chapter's four cases: the damp castle, dense and loose sand sheared, the house on loose wet sand shaken, and dense wet sand sheared with its water trapped, which is the firm beach under your foot. Pick one, or set the box up yourself:

Fig. 5.3 · Grain sandbox · a live discrete-element model 2D teaching model · stresses in scaled model units (kPa*)
Experiments

Applied load σ–the load on top; still water left out
Extra water pressure Δu–above the still-water pressure
Grain stress σ′ = σ − Δu–platen force from grain contacts
Void ratio e–2D discs
Shear stress τ–τ/σ′
Packing
Water
Drain
On top
40 kPa*

A box of about four hundred simulated grains, packed loose or dense, dry, damp or saturated, that you can shear, shake or cut, with nothing about the outcome scripted. The lines between grains are force chains: the thicker the line, the harder those two grains press together. The blue tint in the gaps is the water, darker where its pressure is higher. The readouts leave out the still-water pressure and the buoyancy of the grains: σ is the load on top and Δu the water pressure above its still-water value, so they are not the σ and u of Fig. 4.2, and the grain stress is σ′ = σ − Δu, measured from the forces between grains. The chart under the readouts follows all three in time. Every reading is measured from the model as it runs. It is a 2D teaching model: discs pack tighter than real sand, so void ratios (space between grains ÷ space taken by grains) come out lower than real ones (about 0.5 to 0.9), and the water is one pressure field on a coarse grid. Stresses are scaled to read like kPa (marked kPa*); they are not lab results. Read the trends, not the numbers. The numbers to remember are the model's two fixed settings: friction between grains in contact of 0.5 (an angle of about 27°), and the lowest water pressure it allows, −100 kPa*, where the water would boil.

  • Experiment 1: the damp cut stands and the dry one slumps.
  • Experiment 2: watch the platen. Dense sand lifts it as it shears (it dilates), and its strength peaks then drops; loose sand lets it sink (it compacts).
  • Experiment 3: with the drain shut, follow the extra water pressure on the chart as it climbs toward the load, and the force chains under the house as they fade.
  • Experiment 4: dense sand pulls the water pressure below its still-water value (a negative Δu, which engineers call negative excess pore pressure) and gets stronger. Then run it again with the drain open.

Notice two things the sandbox does that the rest of this chapter has only described. In Experiment 2 the dense sample's strength rises to a peak and then falls as it keeps swelling: once its grains have climbed out of each other's hollows, the pack is no longer dense, and it grips like loose sand. And in Experiment 4 the extra water pressure falls and then stops falling, near −100 kPa*. Water can be pulled only so hard before it boils into bubbles of vapour, which is called cavitation, and the model stops it there.Why about −100? Pressures in this book are gauge pressures, counted from the air's 101 kPa (U.S. Standard Atmosphere (1976)). Lower the water's pressure to near zero absolute, which is about −100 kPa gauge (strictly, to its vapour pressure, a little above zero absolute), and dissolved air and vapour come out of it as bubbles. So a dense sand sheared with its water trapped can gain only so much strength. Laboratory tests meet the same ceiling, which is why they are run with the water under raised pressure, and the sandbox copies it.

5.5 Liquefaction

Push Experiment 3 to its end and the loose sand's trapped water carries nearly all the load. The grains still touch, but they hardly press, the effective stress falls toward zero, and the grip goes with it. The sand flows like a heavy liquid. That is liquefaction, and your wiggling feet do a small version of it at the water's edge. It needs three things at once: loose grains, every gap full of water, and a load that comes faster than the water can leave. An earthquake supplies the speed, but any load that comes on fast enough will do, even a small slip that sets the grains moving. Liquefaction with no earthquake is called static liquefaction, and Chapter 8 meets it.

Water dams are not immune. Lower San Fernando Dam, north of Los Angeles, was begun in 1912 largely as hydraulic fill: soil sluiced into place with water, which settles loose and stays saturated. Early on 9 February 1971 an earthquake of magnitude 6.6 shook it hard for about 15 seconds. Within about 12 of them, the water pressure in the sandy fill near the base of the upstream slope had climbed so high that much of that fill had liquefied, and most of the upstream slope slid into the reservoir. What was left of the embankment stood about 1.5 m above the water, and that was with the water kept, since 1966, at least 2.9 m below its old maximum level. Some 80,000 people below the dam were evacuated for four days while the reservoir was pumped down, and nobody was killed at the dam.The slide and its cause: Seed et al. (1975). The date, magnitude, about 15 s of strong shaking with high water pressures after about 12 s, the 5 ft (1.5 m) of freeboard left, the 80,000 evacuated for four days and the 9.6 ft (2.9 m) cut in the maximum operating level in 1966: ASDSO, Lower San Fernando Dam.

Dense sand behaves differently. Once its grains start to slide they must climb over one another, the pack must swell, its water is pulled thin, and the grains press harder. That is why the densely packed heap of Fig. 1.1 held when it was jiggled. Engineers who want an embankment that will not liquefy compact its fill densely, layer by layer, as Chapter 9 describes.In 1936 Arthur Casagrande sheared loose and dense samples of sand side by side and found that both ended at the same density, which he called the critical void ratio: sand looser than that shrinks as it shears, and sand denser swells (Casagrande (1936)). Two years later the upstream slope of Fort Peck Dam in Montana, a hydraulic fill dam, slid during construction and killed eight men. The board that investigated blamed weak shale and bentonite under the dam, and thought liquefaction of the fill may have carried the slide further (ASDSO, Fort Peck Dam).

This is why mine waste worries engineers so much. The ground-up rock a mine pumps out as slurry, called tailings, often settles loose and stays saturated for decades, and Chapters 6 and 8 follow what that means for a dam that holds it back, or stands on it.

5.6 Clay, the slow soil

Everything so far was sand. Clay, the soil of the cores that Chapter 3 made watertight, follows the same rule, σ′ = σ − u, on a far slower clock. Its flakes are so small that its water needs about a decade to cross a metre of it (11 years in Fig. 3.3), and by Example 5.2 a 2 m layer takes weeks to drain and a 10 m layer years. So almost any load on clay starts undrained.

Clay also grips less than sand. Its flat flakes slide on one another more easily than rounded quartz grains climb over each other: the workbench of Chapter 13 gives its clay a friction angle of 26°, against 34° for the beach sand. Loaded slowly, so that its water keeps pace, a clay grips as Eq. 4.3 says, with σ′ · tan φ′.

Loaded fast, it is another matter, and the reason is the loose sand's. A soft clay that has never carried more load than it carries now is loosely packed for its load, and when it is sheared it tries to contract. Its water can't leave in time, so the water pressure rises, σ′ falls, and the grip falls with it. The strength a clay shows when it is sheared too fast for its water to move is its undrained strength, su. Engineers who worked back from failures of soft clay under embankments, footings and excavations found the average strength the clay had actually supplied along the slip, and Gholamreza Mesri showed in 1989 that laboratory tests give the same answer:Mesri (1989): su = 0.22 times the largest effective stress the clay has ever carried, which for a clay never loaded more than now is its present σ′; first from field failures, then confirmed from laboratory tests.

su ≈ 0.22 · σ′ (5.4)

su undrained strength of a soft clay never loaded more than now (kPa) · σ′ its effective stress before the fast loading began (kPa)

Example 5.3

The same clay, slow and fast

Given
A soft clay whose grains press with σ′ = 150 kPa (an illustrative value, about what 15 m of soil and water above might give) and friction angle φ′ = 26°, with no cohesion.
Find
The grip it offers when sheared slowly, and when sheared fast.
The idea
Slowly, the water keeps pace and the grip is σ′ · tan φ′ (Eq. 4.3). Fast, the clay tries to contract, its trapped water takes part of the press, and the grip is su (Eq. 5.4).
  1. tan 26° = 0.4877the clay's grip per kilopascal of press
  2. grip, slow = 150 × 0.4877 = 73.2 kPadrained: the water keeps up, σ′ stays at 150
  3. grip, fast = 0.22 × 150 = 33 kPaundrained: the trapped water takes part of the press
  4. 33 ÷ 73.2 = 0.45the fast grip as a share of the slow

Sheared slowly the clay grips with about 73 kPa; sheared fast, with about 33 kPa, less than half.

Check. The undrained grip is what the slow formula gives for a smaller press: 33 ÷ 0.4877 = 67.7 kPa. So the fast shear behaves as if the trapped water had taken about 80 of the grains' 150 kPa of press. That is a rough reading: the two strengths act on surfaces at different angles, and real clays scatter around 0.22, but the size of the gap is real.

What it means. The same clay, in the same place, has two strengths, and which one counts depends on how fast it is asked for.

Clay, loaded fast, can be weaker than the same clay loaded slowly. Of everything in this chapter, that is the sentence the last chapter of the book needs most.

Undrained isn't always weaker. A clay that was once pressed harder than it is now, by ice or by soil since worn away, is densely packed for its load, and behaves like dense sand: sheared fast, it tries to swell, its water pressure falls, and for a while it is stronger. The danger then comes later, as water seeps in and the extra strength seeps away. And any clay that has slid a long way along one surface loses more still, because its flakes line up along the slip; the grip left is its residual strength.

Check yourself A dam is raised over a 2 m layer of the example's clay, with sand beneath it, by one metre every month. Is each raise drained or undrained for the clay?

Each raise goes on in days, far quicker than the 45 days the layer needs, so each starts undrained. But a month is not much shorter than 45 days, so most of each raise's water pressure has seeped away before the next. Over the life of the dam the clay is largely drained; for any one fast event, such as a slip, it is undrained.

The film The same sand, three ways retells Part I, from the beach to this chapter's sandbox, in under nine minutes:

Film 5.1 · The same sand, three ways8 min 45 s · captions

Before you watch

Near the end, at 7:53, the film shakes its box of soaked grains again, packed densely this time, with the same little house on top. This chapter has shown that dense sand presses harder when it is sheared, and the dense heap of Fig. 1.1 held when it was jiggled. Does the house on the dense box stay exactly where it is? Watch the orange lines, and how far the house goes compared with the loose box.

The film was made first of the series, so it ends on the question that opens Chapter 2, which you have answered; this segment stops before it.

In this film

  1. A house on soaked sand the question
  2. Three kinds of beach sand Fig. 1.1
  3. A block on a tilting plank Fig. 4.1
  4. A cushion of water under the block your guess
  5. Ten metres of sand, soaked Fig. 4.2
  6. The castle's pull §5.1
  7. Loose grains, shaken Fig. 5.3
  8. Dense grains, shaken Fig. 5.1
Read the transcript

These are grains of sand, and each orange line joins two of them that are pressing together. The thicker the line, the harder they press, and together they hold up this little house. Every gap between the grains is full of water. Now shake the box, the way an earthquake shakes wet ground. No grain breaks, and nothing gets slippery. The house goes down anyway. So what let go?

You already know three kinds of sand from the beach. Dry sand pours, and the steepest heap you can make settles at about thirty-four degrees, however you pour it. Damp sand, packed into a bucket and turned out, stands as a castle with walls that go straight up. And at the water's edge, where every gap is full of water, the sand feels firm while you stand still, then swallows your feet when you wiggle them.

It's the same sand every time. Only the water between the grains changes. An earth dam is a huge heap of sand and soil, with a lake soaking into one side. Which of these three it acts like decides whether it stands. So how can water make the same sand stronger, and then weaker?

The first answer most people reach for is that water is slippery. Wet the grains and they slide past each other, the way a wet floor makes you slip. The castle says otherwise. It has water in it, and it's the strongest of the three. Sand grains are mostly quartz. When engineers slid polished quartz against itself, wet surfaces gripped more than twice as hard as dry ones. And on real, rough sand grains, water barely changes the grip at all. So whatever water does to sand, it isn't greasing the grains. To find out what it does, start with something simpler than sand.

Put a block on a plank and slowly tilt the plank. The block stays put until the slope gets steep, and then it slides. Friction holds it, and friction comes from pressing. Press the block onto the plank twice as hard, and it takes twice the push to move it. The grip on offer is always the same fraction of the press. That fraction sets a steepest angle: tilt past it and the block slides, whatever it weighs.

A heavier block presses harder, but gravity also pulls it harder down the slope, by the same share. For sand, that steepest angle is the thirty-four degrees of the dry heap. Now tilt our ramp to twenty-five degrees. Two things are pulling against each other here. Gravity pulls the block down the slope, and friction grips back. The grip on offer is about one and a half times what the block needs to stay put. That ratio is called the factor of safety. Above one, the block holds, and below one, it slides.

Now slip a thin cushion of water under the block, and slowly pump up the pressure inside it. The block and the ramp stay exactly as they were. Does the block hold, slide, or hold even better?

It slides. The water pushes up on the block, so it carries part of the block's weight. Every bit of weight the water carries comes off the block's press on the ramp. The grip shrinks with the press, while gravity pulls down the slope as hard as ever. The block started with about half again the grip it needed. So once the water carries just under a third of the press, there's no spare grip left, and the block goes. The block and the ramp never changed.

Now shrink the block down to a grain of sand. A handful of sand is a crowd of grains, each resting on its neighbours at a few points of contact. Every contact is a tiny block on a tiny ramp, held by friction, and that friction comes from how hard the two grains press. Locked together, so that to move they have to climb over each other, those contacts give the sand its strength. When the gaps are full of water, the water in them pushes neighbouring grains apart.

Like the cushion, it holds up part of the weight above, so the grains press on each other less. About a century ago, an engineer named Karl Terzaghi put this into one rule, and it founded soil mechanics. The strength of soil follows only the press left over between the grains. That press is the weight of everything above, minus the water pressure in the gaps. Engineers call it the effective stress.

Here's what that rule does in real ground. Take a column of sand ten metres deep. The water table is the level below which every gap is full of water. Keep it low, and the sand above it is only damp. Then the grains at the bottom press on each other with about eighteen tonnes on every square metre. Now raise the water table all the way to the surface. Soaking makes the sand heavier, by about two tonnes on every square metre, because water now fills the gaps.

But the water in those gaps pushes back up with about ten. So the grains now press with only about ten tonnes. More than forty percent of the sand's strength is gone, and not one grain had to move. Keep raising the water, above the ground, like a lake sitting on top. The lake presses down from above, and pushes up inside every gap by the same amount. So the press between the grains doesn't change at all. So here is what water does to sand. It takes weight off the grains, and their grip goes with it.

That explains why soaked sand is weaker. The castle needs one more piece. Our friction rule said nothing stands steeper than thirty-four degrees, yet the castle's walls go straight up. In damp sand there's only a little water, and it gathers in tiny rings where the grains touch. The surface of each ring is stretched, like the skin of a soap bubble, and it pulls the two grains toward each other. The water in those rings is at less than the pressure of the air around it. That's called suction. So this water pulls the grains together, and the press between them goes up.

That press doesn't come from weight, so the thirty-four degree limit no longer holds. The pull is small, a few hundred kilograms on a square metre. Even at the bottom of a thirty-centimetre castle, the grains press with only about half a tonne, and near the top, with almost nothing. So up there, the pull is most of what holds the grains together. Here's the castle and the dry heap again, in a simple model of grains, drawn flat. Cut a wall in damp grains, and it stands. Cut it in dry grains, and it slumps into a heap. Dry the sand, and the rings evaporate. Flood it, and they merge into ordinary water. Either way, the castle falls.

That leaves the sand at the water's edge, and the box we started with. The grains in our box are loosely packed, with big holes between them, and every hole is full of water. Each shake shoves the loose grains, and they try to settle into the holes below them, like a shaken jar of rice. But the water is in the way, and it can't get out fast enough. Trapped, its pressure climbs, until it nearly matches the weight above. For a moment, the water holds up the house, and the grains don't. Weight minus water pressure is almost nothing, so the grains hardly press on each other at all.

The orange lines fade, and the grip goes with them. The house tips and sinks. That is liquefaction. Now the same shaking, on densely packed grains. These grains have no holes to drop into; to move at all, they have to ride up over each other. That opens the gaps, and bigger gaps need more water than can flow in so quickly. The water is pulled thin, a pull like the castle's, and the grains grip harder. The house sinks only a fraction as far. Loose, soaked sand like this is what swallows your feet when you wiggle them. Stand still, and the water seeps away, and the grains press again.

So that's the idea under everything that follows. Sand and soil are strong because their grains press on each other. Water pressure in the gaps takes some of that press away, and suction can add some back. Every dam in this series stands or falls on that one idea.

After you watch: the answer

The house sinks, but only a fraction as far as on the loose box. Dense grains shaken back and forth still shuffle a little, so the house settles some way. But every shuffle that would open the pack needs water that can't arrive in time; the water is pulled thin, the grains grip harder, and the orange lines stay thick. Dense sand is far safer than loose, and not immune. Two of the film's phrasings are looser than this book's. At 4:08 it says the water pushes neighbouring grains apart; Chapter 4 put it more exactly: the water presses equally on every grain round a gap, and takes part of the load the grains carried between them. At 6:30 its "few hundred kilograms on a square metre" is the suction in the castle's rings; the grip that suction adds is smaller, the 0.6 kPa of Example 5.1, because the rings wet only part of each grain and grip is only a fraction, tan 34°, of a press. The film ends Part I with the sentence on the title page, adding the castle's part: water pressure in the gaps takes some of the grains' press away, and suction can add some back. Run Experiment 3 in Fig. 5.3, then press the hint's "Same, dense sand" to see both boxes; Fig. 4.1 is the same rule with one contact in place of thousands.

5.7 Where Part I comes back

Soil holds as well as its grains press on one another, and water in the gaps, pushing or pulling, changes that press. Every later chapter leans on that pair:

  • Chapter 6 · A lake can't lose strength. Tailings can. A water dam holds a liquid; a tailings dam holds a loose, saturated deposit that can liquefy.
  • Chapter 7 · A concrete dam is the block on the cushion. Water under its base carries part of its weight, and the friction that stops it sliding goes with it.
  • Chapter 8 · An upstream raise stands on tailings. If the tailings under it liquefy, the raise has nothing left to stand on.
  • Chapter 9 · The core takes the head so the downstream shell keeps its grip. Low water pressure behind the core means a high effective stress and a strong downstream slope. The upstream shell stays soaked, so it relies on draining freely.
  • Chapter 10 · Drains put the press back. Lowering the water inside a dam lowers the pore pressure, so the grains press harder.
  • Chapter 11 · Sand boils: water flowing up can lift the grains. When its push on them equals their weight under water, the effective stress is zero and the sand boils.
  • Chapter 12 · A slope's water can't always leave as fast as the reservoir does. Let a reservoir down quickly and the upstream shell's water pressure lags behind, with nothing outside to balance it.
  • Chapter 13 · Your design is judged on this. The stability check weighs the friction the grains can supply against what's needed, using the water pressures from the seepage calculation.

What to remember

  • Water can pull as well as push. In damp sand its rings press the grains together, but only with a fraction of a kilopascal: enough for a wall a quarter of a metre tall, never one of thirty.
  • Sheared, densely packed grains must swell and loosely packed grains must shrink. Where the water can't follow, the dense sand's water is pulled thin and its grains press harder; the loose sand's water takes their weight and they press less.
  • Whether the water follows is a race between the load and the drainage time, t ≈ L²/cv: seconds to minutes for sand, months to years for clay. Drained if the water wins, undrained if the load does.
  • Liquefaction is the loose case taken to its end: loose, saturated and loaded faster than it can drain, the grains stop pressing on one another and the soil flows.
  • Clay, loaded fast, can be weaker than the same clay loaded slowly.

Questions

  1. 5.1

    Explain, in the terms of this chapter, why the firm sand goes pale under one foot while the sand at the water's edge swallows the other.

    Hint

    Which sand is dense and which loose? What must each do when it is sheared, and what does the water do meanwhile?

  2. 5.2

    Run Experiment 4: dense, saturated sand, drain shut, sheared. Why does the grain stress rise with the drain shut, and stay level when you run it again with the drain open? And why does the extra water pressure stop falling where it does?

    Hint

    Dense sand must swell to shear. Where does the water to fill the new space come from in each case? Then read the model's lowest allowed water pressure in the caption.

  3. 5.3

    Run Experiment 3, the house on loose saturated sand with the drain shut. On the chart, does the grain stress fall all at once or a little with each shake? Then open the drain and shake again: what changes, and does the house still settle?

    Hint

    Each shake lets a few grains drop toward the gaps. With the drain shut, where does the water they displace go? With it open?

  4. 5.4

    Using the illustrative soils of Example 5.2: is a 3 m layer of sand, draining through one face, drained during 15 seconds of earthquake shaking? Is a 5 m layer of clay drained under a dam raised 5 m in a year?

    Hint

    t = L² ÷ cv, with cv = 0.20387 m²/s for the sand and 0.08807 m² a day for the clay.

  5. 5.5

    Lower San Fernando was a water dam, yet in 1971 its upstream slope failed the way loose mine waste fails. What did it have in common with tailings, and why did 15 seconds of shaking matter?

    Hint

    How was its fill placed, and how would that fill have been packed? Compare 15 s with the drainage time of a few metres of sand.

Further reading

  • Johansson, J. (2000). Soil Liquefaction Web Site. Department of Civil Engineering, University of Washington. https://depts.washington.edu/liquefy/html/main.html gentle A plain-language tour of liquefaction: what it is, the earthquakes that taught engineers about it, why loose saturated sand is the culprit (with Casagrande's critical void ratio), and what can be done about it.
  • Terzaghi, K., Peck, R.B., Mesri, G. (1996). Soil Mechanics in Engineering Practice, 3rd ed. New York: Wiley. serious The working engineer's textbook, begun by the man who founded the subject. Read its treatment of consolidation, and of the strength of sands and clays under drained and undrained loading, for the full versions of this chapter's two rough sums.
  • Reynolds, O. (1885). On the dilatancy of media composed of rigid particles in contact. With experimental illustrations. Philosophical Magazine, Series 5, 20(127): 469–481. doi:10.1080/14786448508627791 primary source Thirteen pages that named dilatancy: the sealed rubber bag of shot and water that goes rigid, and the pale footprint on p. 475 (p. 209 of the 1901 reprint, free online), where "the sand is full of water".
Part II

Two kinds of dam

By the end of Part II

Some dams hold water that is worth money. Others hold mine waste that nobody wants back and that has to stay put long after the mine has closed. Part II looks at both: at the flood every dam must pass, at a block of concrete that stands by its weight alone, and at dams that keep growing for as long as a mine runs.

  1. Chapter 6Two dams, two jobs

    A dam whose water is the product, a dam whose waste must stay put, and the flood both must pass.

  2. Chapter 7A block on a cushion of water

    The concrete gravity dam, and the water that lifts it from below.

  3. Chapter 8Never finished

    Upstream, downstream and centreline: three ways to raise a tailings dam, and a water dam built like one.

Chapter 6 · Part II · Two kinds of dam

Two dams, two jobs

What changes when the thing behind the dam can lose its own strength?

So far. Part I built one idea from a beach, a bucket and a block on a ramp: soil holds only as well as its grains press on one another, and the water between them can take that press away, all at once when it cannot get out in time (Chapter 5). Part II puts the idea to work on real dams, starting with two that look alike in section and do opposite jobs.

After this chapter you can

  • set a water dam beside a tailings dam and say how they differ in what they hold, how long they must last and how they have failed;
  • work out the power of a hydro plant and the load on its dam, and tell the depth that loads a dam from the head that drives its turbines;
  • explain why a tailings beach sorts its own grains, and why every dam must pass its flood with metres to spare.

This chapter adds freeboard mark and a spillway notch at the crest

2 figures · about 20 min to read

At 12:28 in the afternoon on 25 January 2019, a video camera above an iron mine near Brumadinho, in the Brazilian state of Minas Gerais, was looking across the valley at the face of a dam. Dam I held about 12.7 million cubic metres of waste from the mine, the Córrego do Feijão, and nothing had been added to it for two and a half years. A drone had filmed it seven days before and seen nothing wrong. Its instruments had recorded nothing that mattered either: survey markers on the crest, tubes that measure movement deep inside it, radar watching the face, and gauges reading the water pressure within it.Robertson et al. (2019), pp. i–ii (the time, the drone, the instruments, deposition ending July 2016), p. 5 (the camera) and p. 12 (the volume stored).

In the footage the crest sags first, along most of the dam's length. A fifth of a second later the face bulges outward, low down, just above the toe. Five seconds in, water spurts out near the toe. In under ten seconds the whole slope has collapsed, and what was a slope is a heavy liquid, moving. Within five minutes 9.7 million m³ of tailings, three quarters of everything the dam held, had poured out. The flow ran through the mine's canteen and offices, through houses and farms, and on down the valley to the Paraopeba River. It killed 270 people; two years later 11 of them had still not been found.The sequence, the volume and the flow's path: Robertson et al. (2019), pp. i, 3, 5–6. The toll, 259 deaths and 11 people missing: Franks et al. (2021), p. 1.

When a water dam fails, water escapes. Here the dam itself went liquid. The ground-up rock behind its face had stood as a slope for decades, firm enough to walk on, and in seconds it had no grip left. A lake has no grip to lose; it is a liquid already. That difference is this chapter's question: what changes when the thing behind the dam can lose its own strength?

6.1 Two jobs

A hydroelectric dam holds back water, and the water is the product. Its height above the turbines is stored energy, released on demand. A tailings dam holds back mine waste, called tailings: rock ground to sand, silt and finer, pumped out of the mill as a slurry with the water used to process it. Nobody wants it back. The job is to keep it in place, forever.

There is a great deal of it. In April 2019, after Brumadinho, a group of investors asked the world's listed mining companies to describe their tailings facilities. The 107 companies that replied listed 1,743 of them, holding 44.5 billion m³; spread evenly over Manhattan, the authors worked out, that would stand 750 m deep. The same companies expected to add about 2.5 billion m³ a year.Franks et al. (2021), pp. 1–2.

Both kinds of dam face the physics of Chapters 2 and 3. Water pressure grows by 9.81 kPa for every metre of depth, which comes to nearly 1,000 kPa, or 1 MPa, at the foot of a 100 m reservoir, and water seeps through anything that isn't perfectly tight. Below, one of each is drawn in section, side by side:

Fig. 6.1 · Water dam and tailings dam in section Schematic · not to scale

Concrete gravity dam with powerhouse hydroelectric

H = 85 m
150 m³/s
Power, P = ηρgQH112 MWη = 0.90 turbine–generator
Pressure at the heel0.87 MPa89 m deep at the heel · water load 39 MN per metre of dam

Tailings storage facility mine waste

A water dam turns the drop of its water into power; a tailings dam holds a deposit that sorts itself by grain size as it settles. Left: a concrete dam with its powerhouse at the toe. Water enters the intake, falls through the penstock to the turbine and leaves by the tailrace. The head H, dimensioned on the right, is the drop from the reservoir to the tailwater, and it drives the turbines. The brick-red triangle is the water's pressure on the upstream face; it comes from the full depth of water at the heel, a few metres more than H. Moving dashes show the water in the penstock and the tailrace, whose speed follows the flow, and the power leaving along the line, whose speed follows the flow times the head. Right: a tailings dam raised in stages from a starter dam. Slurry pours from pipes on the crest, the spigots. Watch the grains: the coarse sand drops out first and builds a sloping tailings beach, while the finest grains, the slimes, ride the water out to the pond, where a barge pumps water back to the mill. The dashed line labelled Phreatic surface is the top of the water inside the deposit, the phreatic surface; the finer dashes in the embankment mark the stages it was raised in, and the underdrain at its base lets water out. The sorting decides how strong, and how wet, the tailings next to the dam will be. The numbers to remember: at its opening settings the plant makes 112 MW from 150 m³/s falling 84.8 m, and the water presses on the heel with 0.87 MPa.

  • Set Reservoir level to its lowest. Read H, then the depth at the heel under Pressure at the heel. What is the difference?
  • Halve the Turbine flow. Which readouts change, and which stay where they were?
  • Watch where the smallest grains on the right come to rest.
Answers

H = 56 m and the heel is 60 m deep: the 4 m of tailwater over the base, the same at every level. Halving the flow to 75 m³/s halves the power, to 56.2 MW; the pressure and the load at the heel do not move, because they come from the depth of the water, not from how much of it is flowing. The smallest grains settle in the pond, far from the dam.

Notice that on the left the dimension H stops at the tailwater, while the pressure triangle on the dam's face runs on down to the heel. Notice too, on the right, that the pond lies at the far end of the beach, as far from the dam as the deposit allows. The rest of the chapter takes those two details in turn.

6.2 Head drives the turbines, depth loads the dam

Start with the water dam, because its sums are short. A cubic metre of water weighs 9.81 kN. Let it fall H metres and gravity does 9.81 × H kilojoules of work on it, since a kilojoule is a kilonewton moved through a metre. The turbine catches as much of that as it can. Send Q cubic metres through every second and the power is the energy each cubic metre gives up, times the cubic metres per second, less what the machines lose:

P = η · ρ · g · Q · H(6.1)

P power (W, watts; a megawatt, MW, is a million watts) · η efficiency, the share of the falling water's energy that the turbine and generator turn into electricity · ρ 1,000 kg/m³ · g 9.81 m/s² · Q flow through the turbines (m³/s) · H head, the drop from the reservoir surface to the tailwater, strictly the gross head (m)

The figure takes η = 0.90 for the turbine and generator together.H here is the gross head, surface to surface. Part of it is spent pushing the water through the intake and the penstock, so the turbine itself works on a little less, the net head. The figure uses the gross head, and folds the losses on the way into one efficiency. At its opening settings:

Example 6.1

The power of a drop

Given
A turbine flow Q = 150 m³/s falling a head H = 84.8 m (the figure rounds it to 85); efficiency η = 0.90; water weighs 9.81 kN/m³. At the highest reservoir level the head is 92 m.
Find
The power, and the flow that would give 200 MW at the highest level.
The idea
Energy per cubic metre, times cubic metres per second, times the share the machines keep (Eq. 6.1).
  1. Each m³: 9.81 × 84.8 = 831.888 kJits weight times the drop; a kilonewton through a metre is a kilojoule
  2. Each second: 831.888 × 150 = 124,783 kW150 cubic metres every second; a kilojoule each second is a kilowatt
  3. Kept: 0.9 × 124,783 = 112,305 kW = 112.3 MWthe turbine and generator keep 90%; a megawatt is 1,000 kW
  4. At 92 m, per m³/s: 0.9 × 9.81 × 92 = 812.268 kWthe same three steps with one cubic metre a second
  5. Q = 200,000 ÷ 812.268 = 246.2 m³/s200 MW is 200,000 kW

The plant makes 112 MW, the figure's readout. With the reservoir at its highest, 200 MW takes 246 m³/s, inside the Turbine flow slider's range.

Check. Power is proportional to flow, so each cubic metre a second is worth 112,305 ÷ 150 = 748.7 kW at this head, and doubling the flow to 300 m³/s should double the power: the figure gives 224.6 MW.

What it means. A hydro plant sells head and flow in equal measure. The height of its water is what it earns from, and, as the next example shows, it is also what the dam has to hold.

The turbine in the figure turns at the same speed whatever the flow, and that is deliberate.Every generator on a grid has to keep step with the grid's alternating current, 60 cycles a second in North America, so it turns at one fixed speed; the rotors at Mica and Revelstoke turn at 112.5 revolutions a minute (BC Hydro (2016)). Open the gates wider and the water pushes the runner harder, so the generator delivers more power at the same speed. More flow means more push on the runner at the same speed.

One could assume that the H that drives the turbines is also the depth that loads the dam. But look at the left drawing again: H ends at the tailwater, and the dam's upstream face goes on below it, down to the heel on the rock. The water against the dam is deeper than the drop it makes past it. The figure's readouts use the depth, and a reader who checks them with H gets the wrong answer:

Example 6.2

Depth, not head

Given
The figure at its opening settings: the water is 88.8 m deep at the heel and the head is 84.8 m. Water weighs 9.81 kN/m³.
Find
The pressure at the heel, the push of the water on each metre of dam and the height it acts at; then the same two numbers worked, wrongly, from H.
The idea
The load is the pressure triangle on the whole wet face (Eqs. 2.1 and 2.2), and the face is wet all the way down to the heel.
  1. p = 9.81 × 88.8 = 871.1 kPa = 0.87 MPathe pressure at the bottom of the wet face
  2. F = ½ × 9.81 × 88.8² = 38,678 kNthe area of the triangle, per metre length of dam
  3. Height: 88.8 ÷ 3 = 29.6 m above the basethe triangle's centre, a third of the way up
  4. From H: 9.81 × 84.8 = 831.9 kPaa pressure taken at the tailwater, 4 m too high
  5. ½ × 9.81 × 84.8² = 35,272 kNthe push on a face that stopped at the tailwater

The heel feels 0.87 MPa and each metre of dam carries 38.7 MN, acting 29.6 m up: the figure's 0.87 MPa and 39 MN. Worked from H, the load comes out 8.8% short.

Check. The two depths differ by 4 m, the tailwater over the base, so the two pressures should differ by 9.81 × 4 = 39.24 kPa, and 871.1 − 831.9 = 39.2.

What it means. The depth of water at the dam sets its load; the drop to the tailwater sets its power. The tailwater also pushes back on the downstream face, a push this figure leaves out; Chapter 7 adds it, and the water underneath the dam.

Check yourself With the reservoir held where it is, the turbines are opened from 150 to 300 m³/s. What happens to the water load on the dam?

Nothing. The load comes from the depth at the heel, still 88.8 m. Only the power changes: it doubles, to 224.6 MW.

The drawing shows a concrete dam, but a water dam need not be concrete, and in British Columbia most of the big ones are not. Mica, on the Columbia, is an earthfill dam 244 m high, the tallest in the province, and W.A.C. Bennett on the Peace, the dam Chapter 9 opens with, is earthfill 183 m high. Revelstoke, downstream of Mica, has the concrete dam Chapter 2 measured.Mica and Revelstoke: BC Hydro (2016). Bennett: BC Hydro (2020). A dam is a water dam because of what it holds back.

6.3 A beach that sorts itself

The other half of the figure works differently. The mill sends its tailings out as slurry, and from the spigots along the crest it pours onto the deposit and runs downhill toward the pond. As the flow spreads and slows it drops what it carries, and it drops the biggest grains first. The coarse sand settles close to the spigots and builds the tailings beach, sloping gently away from the dam. The finest grains stay in suspension far longer and settle in the pond at the far end. A mine builds a beach, as the sea does, and sorts it the way a river sorts its bed. Why the sizes part so cleanly takes one new fact, about how grains sink; why it matters to the dam takes Darcy's law from Chapter 3. The numbers below are illustrative, chosen to be typical:

Example 6.3

Why the beach sorts, and why that helps

Given
Illustrative grains: fine sand 0.1 mm across, slime 0.01 mm. For grains this small, the speed at which a grain sinks through still water grows with the square of its size. Permeability: tailings sand 10⁻⁵ m/s (the workbench's, in Chapter 13), slimes an illustrative 10⁻⁸ m/s; gap fraction 0.4 for both.
Find
How much faster the sand settles out, and how long water takes to cross a metre of each.
The idea
Settling sorts the grains; Darcy's law then says how fast each kind drains. The crossing time is Fig. 3.3's: gap fraction ÷ (k × gradient), here with a gradient of 1.
  1. Size: 0.1 ÷ 0.01 = 10the sand grain is ten times wider
  2. Sinking speed: 10 × 10 = 100speed grows with the square of the size
  3. Permeability: 10⁻⁵ ÷ 10⁻⁸ = 1,000how much more easily water passes the sand
  4. Sand: 0.4 ÷ 10⁻⁵ = 40,000 s = 11.1 ha metre of sand, at a gradient of 1
  5. Slimes: 0.4 ÷ 10⁻⁸ = 4 × 10⁷ s = 463 daysthe same metre of slimes

The sand sinks out of the slurry about 100 times faster than the slimes, so it settles by the dam, and it passes water 1,000 times as easily: water crosses a metre of it in about half a day, and a metre of slimes in more than a year.

Check. Fig. 3.3 gave 46.3 days for silt with k = 10⁻⁷ m/s and the same gap fraction. A k ten times smaller should take ten times as long: 10 × 46.3 = 463 days.

What it means. The sorting does the dam a favour. It puts coarse sand that drains in hours next to the dam, where the deposit must be firm, and keeps the wet, slow slimes far away. The favour holds only while the pond stays at the far end of the beach.

Check yourself A mine lets its pond spread until the water lies against the crest. What now sits against the dam?

Water, and the slimes that settle out of it: the finest, wettest, slowest-draining tailings, where the dam most needs firm, drained ground. That is why tailings rules keep the pond far from the crest (Chapter 8).

6.4 What differs is almost everything else

The physics of the two dams is the same: pressure that grows with depth, and water that seeps from high head to low. Almost everything else about them differs:

Table 6.1 · A water dam and a tailings dam, side by side
Hydroelectric (water) damTailings dam
Holds backWater, released through turbines on demandGround rock (silt to fine sand size) and process water, settling into a soft deposit
BuiltUsually to full height in one campaign, before the reservoir fillsRaised in stages for the life of the mine, often for decades, while tailings keep arriving
Made ofConcrete, or earth and rockfill from quarries and borrow pits, specified and testedOften the tailings themselves (sand separated from the slurry), mine waste rock and soil dug nearby
FloodsA spillway passes the design flood past the damOften no spillway: the pond stores the flood, so the dam must keep room for it; pond water is pumped back to the mill
Design lifeDecades to a century of operation; it can be taken out of service and removedMust stay stable indefinitely, long after the mine has closed
If it failsA flood wave of waterA flow of liquefied tailings: denser than water, sometimes toxic, able to run tens to hundreds of kilometres
RulesNational dam-safety regulators; guidance from the International Commission on Large Dams (ICOLD)Mining regulators and, since 2020, the Global Industry Standard on Tailings Management (GISTM)
In British ColumbiaThe Dam Safety Regulation under the Water Sustainability Act, and the Canadian Dam Association's Dam Safety GuidelinesPart 10 of the province's mines code, which requires an Engineer of Record for every tailings facility and an Independent Tailings Review Board at every mine with one; the Canadian Dam Association's bulletin on mining dams

Sources for the last two rows: GISTM (2020); Dam Safety Regulation (2016); Mines Code, Part 10, sections 10.4.1 and 10.4.3; CDA (2013); CDA (2019).

Notice the two rows that decide most of what follows. A water dam is built once and then watched; a tailings dam is raised for as long as the mine runs and then has to stand indefinitely after it, earning nothing.The Global Industry Standard on Tailings Management was published in August 2020 by a review that the International Council on Mining and Metals, the United Nations Environment Programme and the Principles for Responsible Investment set up after Brumadinho (GISTM (2020)). It is the first standard meant to apply to tailings facilities anywhere, new or old. And when a water dam fails it lets out water, while a tailings dam lets out a deposit that may itself have turned to liquid.

6.5 Why tailings dams fail differently

Water has no strength to lose. A deposit of loose, saturated tailings has some, and can lose nearly all of it at once: it is the liquid beach from Chapter 5 at the scale of a mine. Sheared, the loose grains try to pack tighter, the trapped water takes over their load, and the deposit flows like a heavy fluid. That is liquefaction. An earthquake can set it off. So can something much quieter, with no shaking at all (static liquefaction): a rising water table, slow creep, or a small movement of the slope. It does not lose quite all: a liquefied deposit keeps a small residual strength, and what is left of its grip decides how far it runs before it stops.

Six failures show the range. For each, the table gives what the mine produced, how the dam was raised, what set it off, and what happened:

Table 6.2 · Six tailings-dam failures: how each dam was raised, what set it off, and what failed
YearDamOreRaisedTriggerWhat happenedKilledReleased
1965El Cobre, ChileCopperUpstreamAn earthquakeThe tailings liquefied and ran 12 km down the valley, burying the village of El CobreMore than 2001.9 million m³
1985Stava, ItalyFluoriteUpstream, and still being raisedNone from outside: no earthquake, no floodThe upper of two dams burst; its mud broke the lower one and ran more than 4.2 km through two villages268180,000 m³
1994Merriespruit, South AfricaGoldUpstreamAbout 50 mm of rain; water then flowed over the crest for one to two hoursThe water washed away the outer wall; the loose tailings behind it flowed about 3 km into a suburb, by static liquefaction in a later study's analysis17About 600,000 m³
2014Mount Polley, British ColumbiaCopper and goldModified centrelineNone seen that nightNot a liquefaction of the tailings; see Chapter 13None21 to 25 million m³ of tailings and water
2015Fundão, Mariana, BrazilIronUpstreamThree small earthquakes 90 minutes earlier hastened a failure already well under wayA liquefaction flowslide; the tailings travelled down three rivers, more than 660 km, to the Atlantic19, two of them listed as missing32 million m³
2019Feijão Dam I, Brumadinho, BrazilIronUpstreamNone from outside: slow creep, and suction lost after heavy rainStatic liquefaction; the slope collapsed in under 10 seconds270; 11 still missing in 20219.7 million m³

Sources: El Cobre, Fondazione Stava 1985, El Cobre; Stava, Luino & De Graff (2012); Merriespruit, Fourie, Blight & Papageorgiou (2001), pp. 707–708; Mount Polley, Mount Polley Review Panel (2015), p. 8 (the method) and Chief Inspector of Mines (2015), p. 13 (the volume, the mine owner's estimate); Fundão, Fundão Review Panel (2016), pp. i–ii and 18, and Fondazione Stava 1985, Fundão (the toll and the distance); Brumadinho, Robertson et al. (2019), pp. i–ii, and Franks et al. (2021), p. 1 (the toll).

Notice that every dam in the table that killed people was raised the same way. Upstream means that each new wall was built out over the tailings already in place, so that the dam stands partly on its own loose, wet waste; Chapter 8 shows what that means and why some countries now forbid it. Notice too the trigger column. At Stava and at Brumadinho nothing came from outside at all, and at Fundão three small tremors only hastened a failure already well under way. A deposit that is loose and saturated can be brought down by a small push, or by none that anyone saw.

Two rows do not fit. Mount Polley, in British Columbia, the dam of the Prologue, was raised by a variant of the centreline method that Chapter 8 describes, and nobody died. It was not a liquefaction of the tailings. Its cause is the subject of Chapter 13, and the chapters between here and there collect what you need to follow it.

Merriespruit is the other. Its tailings probably did liquefy, as a later study argued, but only after water had poured over the crest for an hour or two and washed away the wall in front of them. What failed first was not the grip between grains at all. That failure belongs to a different story from the one this book mostly tells, and it is the commonest single way an embankment dam dies.

6.6 Every dam needs a door

Most people picture a dam failing because the water pushes it over or bursts through it. For embankment dams the commonest single way is simpler: the water goes over the top, which engineers call overtopping. Foster, Fell and Spannagle went through the record of large embankment dams built up to 1986 and sorted the failures that happened in service by how each one began:

Table 6.3 · How large embankment dams failed in service, up to 1986
How the failure beganFailuresShare of those with a known cause
Water over the crest (overtopping)4034.2%
A spillway or its gates failed1512.8%
Over the top or past the spillway48.4%
Internal erosion through the embankment3832.5%
Internal erosion through the foundation1815.4%
Internal erosion from embankment into foundation21.7%
Water inside carrying soil away46.9%
Slides54.3%
Earthquake and liquefaction21.7%
Not known7

124 failures in all. Source: Foster, Fell & Spannagle (2000); shares as in the UNSW table reproduced by Reclamation, Internal Erosion (2010), slide 4, counts from Adamo et al. (2020).

Notice that the shares add to more than 100%. Of the 124 failures, 117 have a known cause, and the main rows add to 102.6%, because some failures were filed under more than one heading. The two totals are the source's own, and for the same reason they are not quite the sums of the rows above them. Statistics like these are counts of messy events sorted into boxes that overlap, and nobody should read the decimals as exact. Two things are plain all the same. About a third of these failures began with water over the crest, and about half did once the failed spillways and gates are added. Most of the other half is water inside the dam carrying its soil away, which Chapter 11 takes up.Foster, Fell & Spannagle (2000), failures in operation; the counts as reproduced in Adamo et al. (2020), which also gives the size of the record, 11,192 dams.

The sentence on this book's title page does not reach those overtopping failures. A flood pouring over an embankment does not need water in the gaps to take the grains' press away. It runs down the outer face fast, drags the grains off the surface and carries them away, and each grain it takes uncovers the next. A heap of earth can be cut through in a few hours; at Merriespruit an hour or two of water over the crest was enough. A concrete gravity dam can be built with its own crest as the spillway; an embankment cannot.Merriespruit: Fourie, Blight & Papageorgiou (2001), p. 708. An overflow crest "may make up an entire gravity dam": USBR (1987), §9.8(c). So every dam needs a door, a way to let a flood past that is not over the crest.

The door is a spillway: a channel past the dam, usually at one end of it, with a crest, the sill the water pours over, set lower than the crest of the dam. In ordinary times the lake is kept at or below the spillway's crest. When a flood comes in faster than the turbines and outlets can pass it, the lake rises and begins to pour over the sill, and the higher it rises, the faster it pours. It goes on rising until the spillway passes water as fast as the flood brings it. The designer's job is to make sure that happens with the lake still below the dam's crest, by a margin called the freeboard. Here are both, to scale, for an illustrative embankment 30 m high:

Fig. 6.2 · The levels of a reservoir in flood, and the spillway that sets themEach section to scale · illustrative dam
A · Through the dam freeboard 2 m flood rise 3 m Dam crest, 30 m Design-flood level, 28 m crest level Normal level, 25 m: the spillway's crest earth embankment, faces 2:1 10 m B · Through the spillway dam crest, beside the spillway, 30 m h = 3 m freeboard 2 m reservoir at the design flood, 28 m spillway crest, 25 m: the normal level flood flow 520 m³/s over 50 m of spillway crest rock 5 m

A flood lifts the lake above its normal level until the spillway passes the flood as fast as it arrives, and the freeboard is what is left above that. In A, a section through the dam, the darker water is the lake at its normal level, the same height as the spillway's crest; the paler band above it is the lake in the design flood; the dotted line carries the dam's crest level out over the water. In B, a section through the spillway at a larger scale, the same flood stands h metres over the rounded concrete crest, draws down as it pours over it, and runs away down the chute; the shape of the falling sheet is sketched. The dimensions in each panel are measured from those levels, and each panel has its own scale bar. The levels and the flow are illustrative, worked out in Example 6.4. The numbers to remember: a flood 3 m deep over 50 m of crest passes 520 m³/s, with 2 m of freeboard left.

  • With panel A's scale bar, measure the height of the dam's crest above the normal level, then above the flood level.
  • With panel B's scale bar, check that the lake stands 3 m over the spillway's crest.
Answers

5 m above the normal level, 2 m above the flood: the flood has used 3 m of the margin. In B, the bar is 5 m long and the dimension h is three fifths of it.

Notice that the freeboard is measured from the flood level, not from the normal one. The 5 m between the lake's everyday level and the crest is not all margin: the flood spends 3 m of it just to make the spillway pass the water.

How much a spillway passes depends on how deep the lake stands over its crest. The sheet of water pouring over is a fixed share of that height, h, deep (about two thirds of it, over a broad crest), and it moves at a speed that grows with the square root of h, so the flow over each metre of crest grows as h × √h, which is h to the power 3/2:

Q = C · L · h3/2(6.2)

Q flow over the spillway (m³/s) · L length of its crest, across the flow (m) · h height of the lake above the crest, measured upstream where the water is still level (m) · C a coefficient set by the shape of the crest (m½/s)

For a flat, broad crest, theory gives C = (2/3)3/2 × √g = 1.70 m½/s, which is the 3.087 that the Bureau of Reclamation's Design of Small Dams gives in feet. A rounded crest shaped to fit the falling water, over a deep approach, does better: the same book takes 4.0 in feet, 2.21 m½/s.USBR (1987), §9.10, equation (3) (p. 365), §9.12(a) for 3.087 (p. 369) and §9.12(d) for 4.0 (p. 370). A coefficient in feet becomes one in metres on multiplying by √0.3048: 3.087 gives 1.70. The example takes a round value between the two:

Example 6.4

The flood a spillway passes

Given
The illustrative dam of Fig. 6.2: spillway crest, and normal level, at 25 m; dam crest at 30 m. A design flood raises the lake to 28 m, so h = 3 m. The spillway's crest is L = 50 m long with C = 2.0 m½/s.
Find
The flow the spillway passes at that level, and the freeboard left.
The idea
Eq. 6.2, then the crest minus the flood level.
  1. √3 = 1.732the square root of the height over the crest
  2. h3/2 = 3 × 1.732 = 5.196depth times the square root of depth
  3. Q = 2.0 × 50 × 5.196 = 519.6 m³/stimes the coefficient and the length of crest
  4. Freeboard: 30 − 28 = 2 mwhat is left between the flood and the crest

The spillway passes about 520 m³/s with the lake 3 m over its crest, leaving 2 m of freeboard.

Check. The units come out right: m½/s × m × m3/2 is m³/s. And each metre of crest carries 519.6 ÷ 50 = 10.4 m³/s, about what a sheet of water a couple of metres deep carries at five metres a second.

What it means. The margin is counted in metres of water, and every metre of flood rise is a metre of freeboard spent. Because the flow grows faster than the rise, a bigger flood costs less extra height than you might fear; a flood the spillway was never designed for can still spend the rest.

Check yourself The flood turns out 10% bigger than the one the spillway was designed for. Does the lake rise 10% higher over the crest?

Less. With Q growing as h3/2, h grows as Q2/3: 1.12/3 = 1.066, so h goes from 3 m to 3.20 m, only 6.6% higher, and 1.80 m of freeboard is left.

Spillways can fail too, and when one does, the freeboard is tested in earnest.On 7 February 2017 the concrete chute of the main spillway at Oroville Dam, California, broke up while carrying a flood: water forced through cracks and joints lifted a section of the slab, and the weak rock beneath it washed out. With the main spillway damaged, the lake rose over the emergency spillway's weir for the first time in the dam's history, and the ground below the weir eroded so fast that about 188,000 people downstream were ordered to leave. The dam itself, an embankment 770 feet (235 m) high and the tallest in the United States, held (Oroville Independent Forensic Team (2018), pp. S-1, 7 and 26). Why water under a slab lifts it is a subject of Chapter 7. A tailings dam usually has no spillway at all. It stores the flood in its pond and pumps the water back to the mill, which works only if the pond has room. Suppose, for illustration, that a design flood brings 3 million m³ into a pond of 2 km²: the pond rises 3,000,000 ÷ 2,000,000 = 1.5 m. That 1.5 m, plus room for waves, has to be kept empty below the crest all year round, and kept while the dam is raised and the deposit climbs. Merriespruit's history records too little freeboard and often poor control of its pond.Fourie, Blight & Papageorgiou (2001), p. 707: "insufficient freeboard provision and often poor pool control". Water over the top of a tailings dam does what it does to any embankment, and then it uncovers what lies behind the wall: tailings that may be loose, saturated, and ready to flow.

What to remember

  • A water dam holds a liquid, which has no grip to lose. A tailings dam holds loose, saturated ground rock whose grains grip only while they press on one another, and the water in its gaps can take that press away in seconds.
  • The head H, reservoir to tailwater, drives the turbines: P = η·ρ·g·Q·H, 112 MW in the figure. The depth of water at the heel, a few metres more, sets the load on the dam: 38.7 MN on each metre.
  • Slurry sorts itself: coarse sand settles by the dam and drains fast, so its grains press firmly on one another; the slimes settle by the pond and drain about a thousand times more slowly. The sorting helps only while the pond is kept away from the dam.
  • Every deadly failure in the case table was of a dam raised upstream, standing partly on its own loose, wet waste, and twice nothing from outside set it off at all.
  • Water over the crest, with failed spillways and gates, began about half the failures of large embankment dams, and it washes grains away however hard they press. Every dam needs a spillway, and freeboard above the flood.

Questions

  1. 6.1

    At a turbine flow of 150 m³/s, drag Reservoir level from its lowest to its highest. By what factor does the power grow, and by what factor the water load on the dam? Why are they different?

    Hint

    The power follows the head H; the load follows the square of the depth at the heel.

  2. 6.2

    The figure's drawing says H = 85 m, and 9.81 × 85 is about 834 kPa, yet the heel readout says 0.87 MPa. Which is wrong?

    Hint

    Where does H end, and where does the dam's wet face end?

  3. 6.3

    Which failures in the case table of this chapter did not begin as a liquefaction of the stored tailings?

    Hint

    Read the Trigger column for water that came over the top, and the What happened column for a row that says what the failure was not.

  4. 6.4

    "Water has no strength to lose." A liquefied deposit has lost nearly all of its strength. What does it keep, and why does that matter once it is moving?

    Hint

    Nearly all is not all. What stops a flow?

  5. 6.5

    With the illustrative spillway of Example 6.4, how high over its crest must the lake rise to pass twice the design flow, and how much freeboard is left?

    Hint

    Twice the flow needs twice h3/2. Then h is that number to the power 2/3.

Further reading

  • Franks, D.M., Stringer, M., Torres-Cruz, L.A., Baker, E., Valenta, R., Thygesen, K., Matthews, A., Howchin, J., Barrie, S. (2021). Tailings facility disclosures reveal stability risks. Scientific Reports 11: 5353. doi:10.1038/s41598-021-84897-0 gentle The survey behind this chapter's 1,743 facilities, open to read: how the world's tailings dams are built, how much they hold, and which kinds most often report stability problems (upstream dams, by a clear margin).
  • U.S. Bureau of Reclamation (1987). Design of Small Dams, 3rd ed. Denver. https://www.usbr.gov/tsc/techreferences/mands/mands-pdfs/SmallDams.pdf serious Its chapter on spillways sets out how a crest is shaped, where the coefficients of Eq. 6.2 come from, and how a chute carries the flood away. Free from the Bureau of Reclamation.
  • Robertson, P.K., de Melo, L., Williams, D.J., Wilson, G.W. (2019). Report of the Expert Panel on the Technical Causes of the Failure of Feijão Dam I. 12 December 2019. https://damfailures.org/sites/default/files/wp-pdf/IR_Feijao-Dam.pdf primary source The expert panel's report on Dam I. Its executive summary, a few pages long, tells the story of the cold open in the panel's own words and explains how loose, wet tailings with no sign of distress could collapse in seconds.

Chapter 7 · Part II · Two kinds of dam

A block on a cushion of water

A concrete dam weighs more than twice the push of the water on it. Why isn't that the end of the story?

The danger peculiar to high dams, as they have been built until now, abroad as much as in France, comes from the uplift, or upward pressure, that water produces in them when it gets into the masonry through cracks over a large enough extent.

So far. Chapter 2 worked out the push of a lake on a dam, and why a concrete dam is a triangle. Chapter 4 put a block on a ramp and showed that water under it takes away the press its grip depends on. Chapter 6 set the water dam, whose water is the product, beside the tailings dam. This chapter takes the simplest water dam of all, a block of concrete that holds its lake by weight alone, and finds the block's weakness in it.

After this chapter you can

  • work out the weight, the push, the uplift and the factor of safety against sliding of a concrete gravity dam, with its drains working and clogged;
  • explain how a grout curtain and a line of drains spend the water's head before it reaches the base, and why the grip of the rock matters as much as the weight;
  • find where the resultant crosses the base, apply the middle-third rule, and say what Lévy asked for after Bouzey and why.

This chapter adds a small concrete block beside the dam, with uplift arrows under it

3 figures · about 20 min to read · Film 7.1, 5 min 13 s

On 14 March 1884 the reservoir behind the new dam at Bouzey, in the Vosges hills of eastern France, was filling for the first time. The dam was a wall of stone masonry 22 m high, built to store water for the Canal de l'Est, and it stood on sandstone laid down in flat beds.Height, foundation and the events of 1884 and 1895: USBR (1998), pp. 16–17. The canal: Bordes (2010), p. 104. With the water a little over 3 m below the crest, a 135 m length of the wall slid as much as 34 cm downstream on its foundation. It sheared off the key that anchored its heel in the rock, and the leaks below the dam jumped to 230 litres a second.

The wall moved no further. In 1888 and 1889 the engineers thickened its lower third with a buttress of new masonry on the downstream side and sealed the crack along the heel; a drainage gallery was proposed and turned down.The buttress: USBR (1998), p. 16; the 4 m crest, its Fig. BF-1. The gallery that was proposed and set aside: Bordes (2010), p. 104. The upper two thirds kept their thin original section, 4 m thick at the crest.

On 27 April 1895, with the reservoir 55 cm below the crest, the upper wall broke away along 170 m of its length. The break ran almost level through the masonry about 10.5 m below the crest, well above the foundation, which this time did not move. The water went down the valley of the Avière through Bouzey and the villages below it. Accounts of the dead differ; the figure most often given is 87.The break: USBR (1998), pp. 16–17. The toll of 87: Bordes (2010), p. 76; the same article gives 70 on p. 104, and USBR (1998), p. 17, "more than 100".

Engineers argued over the cause: the mortar, the thin upper wall. But masonry cannot lose its strength the way the loose wet sand of Chapter 1 did, and nothing had shaken it. So what pushed the wall along its foundation in 1884, and what broke it in two in 1895? The answer that came out of Bouzey is the cushion of water from Chapter 4, at the base of a dam and in the cracks inside one.

7.1 The block, grown into a dam

Take the triangle of Fig. 2.3, 50 m tall and 40 m wide at its base, and lean it against a wall of concrete 8 m thick and 60 m tall, whose top carries a road. That makes the gravity dam of this chapter, 48 m wide at its base, drawn to scale in Fig. 7.1. Fill its reservoir to 57 m, 3 m below the crest, and let the river below it, the tailwater, run 5 m deep.Not every big water dam is concrete. British Columbia's tallest, Mica on the Columbia, is earthfill, 244 m high and nearly a kilometre thick at its base. Revelstoke, downstream, is a concrete gravity dam 175 m high (BC Hydro (2016), slides 7 and 8; BC Hydro (2013)).

Each metre of the dam is a slab of concrete 1,480 m² in section, and concrete weighs 23.5 kN/m³, so each metre of dam weighs 23.5 × 1,480 = 34,780 kN, or 34.78 MN. The reservoir pushes on its upright face with ½ × 9.81 × 57² = 15,936 kN (Eq. 2.2), and the river pushes back on the toe with 123 kN. The net push, F, is 15,814 kN, or 15.81 MN.

Weight holds a gravity dam the way it held the block in Fig. 4.1, by pressing on what is underneath: the dam is that block on level ground. The push F tries to slide it downstream; the rock grips back with the press times tan φ, where φ is the friction angle between concrete and rock. Take φ = 45°, so tan φ = 1, and count no glue. The factor of safety against sliding, the grip on offer divided by the grip needed, is then 34,780 ÷ 15,814 = 2.20: the opening question's dam, more than twice as heavy as the push on it.

But the block in Fig. 4.1 also had a thin cushion of water under its base, and the question about the cushion showed what that does to a grip. The dam stands on rock, and rock is cracked.

Check yourself With only 30 m of water behind it, is the same dam about twice as safe against sliding?

Far safer than that. The push grows with the square of the depth, so at 30 m it is only ½ × 9.81 × (30² − 5²) = 4,292 kN, and the weight is 8.10 times that. A gravity dam is in most danger in its last few metres of water.

7.2 Water under the base

One could assume a dam this heavy would hardly notice water under it. Concrete is 2.4 times as heavy as water, and the water in a crack is a film millimetres thick. To test the idea, go back to the hole in the sand of §2.7, which fills from below because the water in the gaps is joined to the sea. The rock under a dam is cracked by joints and by the partings between its beds, and the reservoir fills them in the same way. At the heel, where the base meets the reservoir, the water in the cracks has the full pressure of the 57 m above it, 9.81 × 57 = 559 kPa. The film holds almost no water, but Example 2.3 showed that a thin layer of water presses as hard as the depth it hangs from.

At the toe the water in the cracks is at the river's pressure, 49 kPa. In between it creeps from high head to low, and if the rock is the same all the way across, it loses its head evenly with distance, as the water did in the uniform sand of Fig. 3.2. So the pressure under the base falls in a straight line from heel to toe, pushing up on every square metre of the base. That push from below is called uplift, U. Before the numbers, a guess:

Nothing is done to stop the water: no drains, no seal. What share of the 60 m dam's weight does the uplift take away?

To decide, compare the two per square metre of base: how hard the concrete presses down on the rock, and how hard the water presses up. Here is the dam with its forces, drawn to scale:

Notice that the uplift arrow is nearly as long as the push, although the cracks hold very little water: the uplift is pressure spread over 48 m of base. The wedge under the base does in the dam what the cushion did in Fig. 4.1. It carries part of the weight, so the dam presses on the rock with less, and the grip shrinks with the press. The factor of safety becomes

FS = (W − U) · tan φ ÷ F (7.1)

W weight of the dam · U uplift on its base · F net push of the water, all per metre of dam (kN or MN) · φ friction angle between dam and rock. This is Eq. 4.1 on level ground, with W − U the press left.

Example 7.1

Water under the base, with nothing done about it

Given
The 60 m dam: weight 34,780 kN and net push 15,814 kN per metre, base 48 m, reservoir 57 m and river 5 m above the base, tan φ = 1. Water weighs 9.81 kN/m³.
Find
The uplift with no drains, its share of the weight, and the factor of safety against sliding.
The idea
The pressure under the base falls in a straight line from the reservoir's at the heel to the river's at the toe, and a straight-line wedge pushes as hard as its average all the way across.
  1. Mean head under the base: halfway between 57 and 5 m = 31 ma wedge that falls evenly averages its two ends
  2. U = 9.81 × 31 × 48 = 14,597 kNthe mean pressure times the base, a metre long
  3. 14,597 ÷ 34,780 = 0.420the share of the weight the water carries
  4. Press left: 34,780 − 14,597 = 20,183 kNW − U, what the grip works with
  5. FS = 20,183 ÷ 15,814 = 1.276Eq. 7.1 with tan 45° = 1

With no drains the water lifts 14.60 MN off each metre of dam, 42% of its weight, and the factor of safety falls from 2.20 to 1.28.

Check. Compare pressures instead of forces. Spread over its base, the section is 30.833 m of concrete deep on average, pressing with 23.5 × 30.833 = 724.6 kPa. The water under it averages 9.81 × 31 = 304.1 kPa, and 304.1 ÷ 724.6 = 0.420, the same share.

What it means. Nothing about the concrete changed. Water in the rock took away two fifths of the press that gives the dam its grip, and with it most of the margin the weight seemed to promise.

7.3 Drains, and what happens when they clog

The weight is fixed, and so is the lake. What an engineer can change is the water under the base, with two rows of holes. Near the heel runs a grout curtain: holes drilled deep into the rock in a line along the dam and pumped full of cement grout, to seal as many cracks as they can reach. A few metres downstream runs a line of drains, drilled into the rock from a gallery, a tunnel along the dam inside it, close to the base. Water that reaches a drain rises into the gallery and is led away, so the pressure at the drain line can be little more than the gallery's own.Foundation drains as 3-inch (76 mm) holes on 10-foot (3 m) centres, the rule below taken as conservative when they reach 40 to 50 per cent of the dam's height: USBR (1987), §8.10, p. 321.

How much do drains take off? Nobody can see into the rock, and the answer depends on which cracks the holes happen to cross. Designers start from a rule. The US Bureau of Reclamation's rule assumes that the head at the drain line is the tailwater's, plus a third of the difference between the reservoir and the tailwater:

hd = ht + (hr − ht) ÷ 3 (7.2)

hd head at the drain line · ht head of the tailwater · hr head of the reservoir, all in metres above the base. On either side of the drain line the pressure is taken to fall in a straight line.

In the language of Chapter 3, the water reaches the heel with 52 m of head to lose before it gets out at the toe. Uniform rock would spend it evenly; the curtain and the drains make it spend two thirds of it in the first 6 m.The one-third rule: USBR (1987), §8.10, pp. 320–321. The US Army Corps of Engineers is more cautious: its drains remove 25 to 50 per cent of the straight-line pressure above the tailwater's at the drain line, up to 67 per cent with tests to justify it (USACE EM 1110-2-2200, para. 3-3.d(1)(c)). The figure below draws the dam with its curtain, gallery and drains. Switch Foundation drains between Working and Clogged, and move the Reservoir slider:

Fig. 7.2 · Uplift and sliding of a 60 m concrete gravity dam Drawn to scale · per metre of dam

Drawing scrolls sideways →

Foundation drains
57.0 m
Weight W–concrete at 23.5 kN/m³
Water push F–net of tailwater
Uplift U–
Sliding safety–(W − U)·tan 45° / F
Resultant at base–

Working drains nearly halve the uplift under a gravity dam, and give back much of the grip the water took from it. The reservoir is on the left, with the brick-red triangle of its pressure on the upright face. Under the base the brick-red wedge is the uplift, in metres of head: the reservoir's at the heel, the river's at the toe, and with the drains working a sharp drop at the drain line, the dashed line below the gallery; clogged drains are crossed out and the wedge returns to a straight line. The heavy arrows are the weight W, the push F and the uplift U, to scale, in MN per metre of dam. The small triangle on the base marks where the resultant crosses it: green inside the middle third, marked by the double arrow, red outside. The sliding check assumes a friction angle of 45° between concrete and rock (tan 45° = 1) and no bond. The number to remember: with the reservoir at 57 m the dam's factor of safety is 1.70 with its drains working and 1.28 with them clogged.

  • Clog the drains and watch the wedge fill out to a straight line. Which readouts change, and which stay put?
  • With the drains clogged, lower the reservoir until the sliding safety reaches 1.5.
  • With the drains clogged, raise the reservoir to 59 m. Does the resultant leave the middle third?
Answers

Only the uplift, the sliding safety and the resultant change; the weight and the push do not. The safety crosses 1.5 at about 53.6 m: the slider reads 1.51 at 53.5 m. At 59 m the resultant sits at 0.66 B, inside the middle third but close to its end at 0.67 B.

Notice that the pressure at the heel never changes when you switch the drains: the heel touches the reservoir, which sets it. What the drains change comes after the heel. The wedge drops steeply to the drain line and runs gently on to the toe, and its area, the uplift, shrinks to a little over half:

Example 7.2

The same dam with its drains working

Given
The dam of Example 7.1, with a line of drains 6 m from the heel. Reservoir 57 m, river 5 m.
Find
The head at the drain line, the uplift, and the factor of safety against sliding.
The idea
Eq. 7.2 sets the head at the drains; the wedge is then two straight-line pieces, each pushing as hard as its own average.
  1. hd = 5 + 52 ÷ 3 = 22.33 mthe river's head plus a third of the 52 m difference
  2. Heel to drains: 39.667 × 6 = 238.0 m²the mean of 57 and 22.33 m of head, over 6 m
  3. Drains to toe: 13.667 × 42 = 574.0 m²the mean of 22.33 and 5 m, over the other 42 m
  4. U = 9.81 × 812.0 = 7,966 kNthe two pieces together, a metre of head being 9.81 kPa
  5. Press left: 34,780 − 7,966 = 26,814 kNW − U
  6. FS = 26,814 ÷ 15,814 = 1.696Eq. 7.1

With its drains working the dam carries 7.97 MN of uplift, 23% of its weight, and its factor of safety is 1.70. Clog the drains and the uplift grows 14,597 ÷ 7,966 = 1.83 times.

Check. The uplift is the area of the head wedge times 9.81, so the two states should compare as their areas: 812 m² against the straight line's 31 × 48 = 1,488 m², and 1,488 ÷ 812 = 1.83.

What it means. The drains act on one line, 6 m from the heel, but the saving is felt over the 42 m beyond it, where the head now falls from 22.3 m to the river's 5 instead of from 50.5 m.

With no glue counted, a common target for this check is a factor of safety of 1.5 under the worst ordinary loads, a convention that leaves room for what nobody can measure, such as the real head in the cracks.1.5 for the worst static case when no cohesion is counted, 1.3 for the probable maximum flood: FERC (2016), p. 3-27, the table of factors for a no-cohesion analysis. The dam passes with its drains working and fails with them clogged.

But drains clog. Water seeping through rock and grout dissolves calcium out of them, and in the drain holes the calcium comes out again as a crust, like the scale in a kettle; silt and bacterial slime add to it.Drains are usually "plugged with calcium deposits, sediment, or biological growth": Campbell, Rivett & Walpole (2022). Watching drain efficiency for clogging: FERC (2016), §3-2.4.3.2. A drain that passes no water lowers no pressure, and the wedge creeps back toward the straight line. That is what the gallery is for: engineers walk it to measure what each drain gives, read gauges set in some of the holes, and clean or redrill the holes that have gone quiet. A falling flow from a drain is as much a warning as a rising one.Uplift lifts more than dams. On 7 February 2017 the fast flow down the concrete chute of Oroville Dam's main spillway, in California, drove water through cracks and joints into the ground under a slab, probably faster than the drains there could carry it away. The forensic team concluded that the uplift most likely beat what the slab's weight, its anchors and the water on top could hold; the slab lifted out, and the flow tore into the poor rock beneath (Oroville Independent Forensic Team (2018), pp. S-1 and 33).

Check yourself Drains clog and a grout curtain needs no cleaning. Why not seal the rock with grout and leave out the drains?

Grout seals only the cracks it reaches, and a curtain ages. Water that finds a way past it arrives under the base with nearly the reservoir's head, and with no drains the wedge is the straight line again. The US federal regulator of hydropower dams allows no reduction in uplift for grouting alone.FERC (2016), §3-2.4.3.2.

Clogged drains matter most when the reservoir is high:

Example 7.3

How fast the margin goes as the lake rises

Given
The dam with its drains clogged, the reservoir at 50 m and at 57 m, the river at 5 m. Weight 34,780 kN per metre.
Find
The factor of safety at 50 m, and the level at which it falls to 1.5.
The idea
The push grows with the square of the depth (Eq. 2.2). The uplift grows only with the depth itself, through the head at the heel.
  1. F = ½ × 9.81 × 2,475 = 12,140 kN2,475 is 50² − 5², the reservoir's push less the river's
  2. U = 9.81 × 27.5 × 48 = 12,949 kNthe mean of 50 and 5 m of head, over the base
  3. FS = 21,831 ÷ 12,140 = 1.8034,780 − 12,949 = 21,831 kN left pressing
  4. FS = 1.5 at h = 53.6 mread from Fig. 7.2, and checked below

With its drains clogged the dam has a factor of safety of 1.80 with 50 m of water, 1.5 at 53.6 m and 1.28 at 57 m. The last 7 m take away more than half of its margin above 1.

Check. For a reservoir h metres deep, U = 235.44(h + 5) and F = 4.905(h² − 25), in kN. Setting W − U = 1.5F gives 7.3575h² + 235.44h − 33,787 = 0, whose positive root is h = 53.6 m.

What it means. One more metre of water at 57 m adds about 559 kN of push and only 235 kN of uplift. So a flood that lifts a full reservoir a few metres costs far more margin than the same rise lower down.

7.4 How hard concrete grips rock

Every number so far leaned on tan φ = 1, as the figure's readout says. The Bureau of Reclamation's guide gives a friction coefficient of 1 for concrete in a first design, and a federal regulator's worked example of a gravity dam uses 45°.Coefficient of internal friction 1.0 for concrete, for preliminary designs: USBR (1987), §8.6, p. 316. A worked example with φ = 45°: FERC (2016), Appendix C. But a dam can also slide on a joint or a bedding plane a few metres down in the rock, which is only as rough as the rock's own partings, and those may be polished smooth or filled with soft material.

With its drains working and a grip of 45°, the 60 m dam has a factor of safety of 1.70. If the rock under it grips at only 35°, the dam:

To decide, look at where tan φ sits in Eq. 7.1. It multiplies everything else.

Example 7.4

The friction angle decides it

Given
The factors of safety at tan φ = 1: 1.696 with the drains working, 1.276 with them clogged (Examples 7.1 and 7.2).
Find
Both at φ = 35°, and the friction angle at which each would slide.
The idea
The factor of safety is proportional to tan φ; the dam slides where it reaches 1.
  1. tan 35° = 0.7002from a calculator
  2. Drains working: 1.696 × 0.7002 = 1.19the factor scales with tan φ
  3. Drains clogged: 1.276 × 0.7002 = 0.89below 1: it slides
  4. Working, slides at tan φ = 1 ÷ 1.696 = 0.590, φ = 30.5°the tan φ that brings the factor to 1
  5. Clogged, slides at tan φ = 1 ÷ 1.276 = 0.784, φ = 38.1°the same, with the bigger uplift

At 35° the dam holds with its drains working, at 1.19, and slides with them clogged, at 0.89. With working drains it is safe down to 30.5°; with clogged drains, only down to 38.1°.

Check. Work the clogged case from Eq. 7.1 directly: 20,183 × 0.7002 ÷ 15,814 = 0.894, as step 3 found by scaling.

What it means. Weight is easy to know and friction is hard. That is why the rock under a gravity dam is dug down to sound material, cleaned and grouted before any concrete is placed on it.

Site C, the newest dam on the Peace River in British Columbia, stands on bedded shale.BC Hydro's 1978 preliminary design for Site C was already an earthfill dam, because a concrete gravity dam was not feasible on that rock. The powerhouse and spillways sit on a concrete buttress keyed down through one bedding plane; in 2018 a deeper plane moved about 5 mm. The remedy chosen in 2020 sets up to 125 piles through that plane and adds drainage to lower the water pressure in the rock, the two levers of Eq. 7.1 (BC Hydro (2021), pp. 3, 4 and 6). During its construction a bedding plane under the concrete that carries its powerhouse and spillways turned out to grip at about 11°, below the 16° its contingency design had allowed for.BC Hydro (2021), Appendix A (Technical Advisory Board), pp. 3–4: "≈11º" against the contingency case's 16º.

7.5 The middle third

Sliding is one way for the push to win. The other is tipping: the push turns the dam about its toe with a moment, as Chapter 2 showed. Engineers check it by adding the weight, the push and the uplift into one force, the resultant, and asking where its line crosses the base.

Where the resultant crosses the middle of the base, the rock is pressed evenly. Move it downstream and the toe is pressed harder and the heel less, until the heel is not pressed at all: the pressure under the base, taken to change in a straight line, is then a triangle with its point at the heel, which balances a third of the way in from its tallest side (§2.4). So while the resultant stays between a third and two thirds of the base from the heel, in the middle third, the whole base stays pressed onto the rock. For the 60 m dam that is 16 to 32 m from the heel; the readout of Fig. 7.2 puts the resultant at 0.56 B, 26.7 m, with the drains working and 0.62 B with them clogged. The same straight-line picture gives the pressure at the heel:

pheel = (V ÷ B) · (4 − 6x ÷ B) (7.3)

pheel pressure of the dam on the rock at its heel (kPa) · V the downward force on the base (kN per metre) · B base width (m) · x where the resultant crosses the base, from the heel (m). At x = B/2 it gives an even press, V/B; at x = 2B/3, zero.

The middle third was the rule of the day at Bouzey. Lévy pointed out that it only keeps the joints from tending to open at their upstream ends, which is not enough: a crack can open anyway, from heat or a bad joint, and the reservoir fills it. A crack no thicker than a sheet of paper carries the full pressure of its depth (Example 2.3), which pushes its faces apart so that it grows.Lévy, an engineer and member of the Academy of Sciences, read his note on 5 August 1895. He compared the danger to a boiler without a safety valve, and proposed a valve: a guard wall on the upstream face with wells behind it, so that any crack opened into a well and drained instead of holding pressure. The wells would empty into a drain along the dam, whose flow would warn of what was happening inside (Lévy (1895), pp. 288–289). Drains in the body of the dam later did the same job (Bordes (2010), p. 104). So Lévy asked for more: at the upstream end of every joint, the masonry must press harder than the water presses there, and then water that finds a crack is squeezed back out.The usual condition and Lévy's new one: Lévy (1895), p. 289. Here is his test at the base of the 60 m dam, with no water inside the joint, as he posed it:

Example 7.5

Does the heel press harder than the water?

Given
The 60 m dam, reservoir at 57 m: weight 34,780 kN per metre, acting 15.71 m from the heel; pushes of 15,936 kN from the reservoir, 19 m up, and 122.6 kN from the river, 5/3 m up; base 48 m.
Find
Where weight and push together cross the base, and the pressure at the heel against the water's.
The idea
The push moves the resultant downstream of the weight by its moment ÷ the weight; Eq. 7.3 does the rest.
  1. Moment of the push: 15,936.345 × 19 = 302,790.6 kN·mthe reservoir's push times its height, H/3
  2. Less the river's: 302,790.6 − 204.4 = 302,586.2 kN·mthe river pushes the other way, 122.6 kN at 5/3 m
  3. Shift: 302,586.2 ÷ 34,780 = 8.700 mhow far downstream of the weight the resultant crosses the base
  4. x = 15.712 + 8.700 = 24.412 ma little past the middle, 24 m; well inside the middle third
  5. 6 × 24.412 ÷ 48 = 3.0515, and 4 − 3.0515 = 0.9485the bracket of Eq. 7.3
  6. pheel = 724.58 × 0.9485 = 687.3 kPaV/B = 34,780 ÷ 48, times the bracket
  7. Water at the heel: 9.81 × 57 = 559.2 kPa, and 687.3 ÷ 559.2 = 1.23Lévy's comparison

At its base the dam presses the rock at the heel with 687 kPa, 1.23 times the water's 559 kPa, so it passes Lévy's test there.

Check. With no push the resultant would sit under the weight, 15.71 m from the heel, pressing the heel harder than the toe. The push moves it 8.7 m downstream, almost to the middle, so the heel pressure should be close to the even press, 725 kPa, and it is.

What it means. The section puts its weight upstream of the middle, where the push needs it. A thinner wall, or one with its weight further downstream, leaves less press at the heel for the water to beat.

Bouzey's upper wall had no such margin. Here is its section as it stood in 1895, from the published drawing:

Fig. 7.3 · Case file: Bouzey Dam, 1884 and 1895Section to scale from the published drawing · curves traced
reservoir, 27 April 1895 crest 371.95 m break, 361.45 m upper wall swept away buttress, 1888–89 1884 slip key bedded sandstone reservoir 5 m
Dam
Bouzey, on the Avière near Épinal, Vosges, France
Built
Finished 1880–81 (accounts differ), to store water for the Canal de l'Est
Height
22 m; crest 4 m thick; base about 11 m before the buttress
Type
Masonry gravity dam on horizontally bedded sandstone
What happened
14 March 1884, first filling: 135 m of the wall slid up to 34 cm on its foundation. 27 April 1895, reservoir 55 cm below the crest: the upper 10.5 m broke away along 170 m.
Toll
Most often given as 87; accounts range from 70 to more than 100
Sources
Section after ICOLD (1974), redrawn in USBR (1998), Fig. BF-1; events, pp. 16–17; Bordes (2010), pp. 76, 77, 104

Bouzey's buttress strengthened the bottom of the wall and left the thin top as it was, and the top is where it broke. The section is the dam in 1895, to scale, its upright face on the left against the reservoir. The original wall is solid; the buttress added after 1884 is the paler, dashed shape on the downstream side; the hatched part above the dashed break line is what the water took. In 1884 the wall slid on its sandstone and sheared the key under its heel (the short dashed line). In 1895 it broke through the masonry, level from the upstream face for about 3.5 m and then dipping steeply downstream to the top of the buttress. Back-calculating that break, the US Bureau of Reclamation found the upper wall's resultant outside its middle third even with no water in the joint, and its upstream face in tension; with water in the joint, the face was in about 0.23 MPa of tension, and a crack that opened and filled would run right through.USBR (1998), p. 17 and Fig. BF-4: at the break, the resultant lay 4.65 ft from the centre of the 16.24 ft joint with no uplift, where the middle third ends 2.71 ft out. The number to remember: 10.5 m, the height of wall above the break, about half the dam.

What this chapter explains. In 1884 the foundation's cracks put the cushion of Fig. 7.1 under the whole base with nothing to drain it. In 1895 a thin wall with tension at its upstream face let water into a joint, and the joint carried the reservoir's full pressure, the failure Lévy's condition is written to prevent.

  • Measure the wall's thickness at the break with the scale bar, and its height above the break. Compare their ratio with the 60 m dam's base and height.
Answers

About 5 m thick under 10.5 m of wall, a ratio near 0.5, against 0.8 for the 60 m dam. Bouzey's upper wall was much slimmer than its own base, and slimmer still than the triangle of Fig. 2.3.

Notice how high the swept-away part sits. The lower wall, thickened and keyed, held; the upper wall took the push of a full reservoir on a section hardly thicker than a road.

After Bouzey, a ministerial decision of 10 March 1896 had every masonry dam in France checked again,Bordes (2010), p. 104. and gravity dams since have been built thicker, with drains in their bodies and their foundations. Lévy's condition survives in design manuals in a gentler form, with an allowance for drains.In the Bureau of Reclamation's version, the face must be pressed harder than four tenths of the water's pressure there when drains work and harder than all of it when they do not, less an allowance for the concrete's strength in tension (USBR (1987), sec. 8.19(b)(2), p. 329).

The second half of the film Depth, not size retells this chapter in five minutes and answers a question it left open:

Film 7.1 · Depth, not size part two5 min 13 s · captions

Before you watch

This chapter checked Lévy's condition only at the base of the 60 m dam, where it passed. Bouzey broke 10.5 m below its crest, through its thin upper wall, and higher up the 60 m dam is thinner too. Does it pass Lévy's test all the way to the top?

This is the second half of the film that starts in Film 2.1, and it opens with a short recap of the block.

In this film

  1. The block on its cushion, again Fig. 4.1
  2. Water under the base: uplift Fig. 7.1
  3. Drain holes from a tunnel Eq. 7.2
  4. Does it slide? Example 7.2
  5. When the drains clog Fig. 7.2
  6. Bouzey, 1895 Fig. 7.3
  7. The middle third, and Lévy's test Example 7.5
  8. A dam that water can soak through Chapter 9
Read the transcript

But a concrete triangle can't count on being glued to the rock beneath it. Weight and friction hold it in place, like a block resting on a ramp. Friction comes from pressing: the harder the block presses down, the more grip it has. Now slip a thin cushion of water under the block, and pump up its pressure. The water carries part of the block's weight, so the block presses less, and its grip shrinks. Nothing about the block changed, and yet it slides. The grip on offer, divided by the grip the block needs, is called the factor of safety. Above one, the block holds, and below one, it slides.

Is there a cushion like that under a concrete dam?

Go back to the beach and dig a hole near the water's edge. It fills with water, from below. The water between the sand grains is connected to the sea, so at each depth it has the pressure the sea would have there. The rock under a dam is cracked, and water gets into the cracks the same way. Here's a concrete dam drawn to scale, sixty metres tall, with the water nearly to its top. Water seeps in under its base.

At the upstream edge it pushes up with the full pressure of the reservoir. At the downstream edge it pushes up only as hard as the shallow river below the dam. In between, the push falls in a straight line, so it makes a wedge under the base. That push from below is called uplift. It's tempting to think a dam this heavy wouldn't notice. But concrete is less than two and a half times as heavy as water, and the water pushes on the whole base.

So designers bleed it off. From a tunnel inside the dam, they drill a line of holes down into the rock. Each hole is under ten centimetres across, and they're spaced a few metres apart along the dam. Engineers call them drains. Water that reaches the holes escapes up them, so the pressure under the base drops sharply there. A rule designers often start from is that, at the drains, the pressure falls two thirds of the way from the reservoir's down to the river's.

With the drains working, the uplift cancels about a quarter of the dam's weight. Now check whether it slides. The grip is the weight still pressing on the rock, times a friction factor for concrete on rock. Divide the grip by the water's push, and you get the dam's factor of safety, just like the block's. For this dam it's about one point seven. A common target for this check is one and a half, so the dam passes with room to spare.

But drains clog over the years, and they have to be watched and cleaned. Block them, and the pressure under the base runs straight from the reservoir's down to the river's. The uplift nearly doubles. Now it cancels about two fifths of the dam's weight. The factor of safety falls to about one point three. The dam still stands, but it has lost the margin its designers asked for. The concrete didn't change. The water under it did.

A concrete dam is the block on the cushion, and its grip is the weight still pressing on the rock. It's the same idea that holds sand together: the grip comes from pressing, and water pressure in between takes it away.

Some of these lessons were learned from dams that failed. At Bouzey, in eastern France, a masonry dam held water for a canal. On its first filling, part of the wall slid on its leaky sandstone. It was propped up with extra masonry and filled again. Then, in April eighteen ninety-five, it broke, through the masonry itself, well above its foundation. More than eighty people drowned.

Experts disagreed about why it broke. That summer, without naming Bouzey, the engineer Maurice Lévy told the French Academy of Sciences what he saw as the special danger of high dams. Water forced into cracks pushes from inside, and he compared it to a boiler without a safety valve. The rule of the day said the weight and the water's push, taken together, must land in the middle third of the wall. That had to hold at every level, so the upstream edge stays pressed down.

Lévy said that wasn't enough. At every level, the masonry at the upstream face must press down harder than the water pushes there, so any water that gets into a crack is squeezed back out. Our dam passes his test all the way up. A test like his, with an allowance for drains, is in the US Bureau of Reclamation's design manual for gravity dams. Failures like Bouzey led engineers to build gravity dams thicker, and to drain their foundations deeper.

Slot or lake, the push is the same, and so is the uplift. Water can't soak through the concrete itself. But most of the world's big dams aren't concrete at all. They're heaps of earth and rock, and water passes through soil. So how can a wall of dirt hold back a lake?

After you watch: the answer

Yes, all the way up; at 7:43 the film says so. The base, which Example 7.5 checked, is the closest call: 687 kPa at the heel against the water's 559, 1.23 times. Higher up the wall above each joint is thinner and lighter, but the water there is shallower, and its push, which tips the joint, falls with the square of the depth. Thirty metres above the base the upstream end of the joint is pressed with 474 kPa where the water presses 265, 1.79 times, and from a metre up to the water line the margin never falls below 1.24. The film also retells the drains, the sliding check and Bouzey, and ends by asking how a wall of earth, which water can soak through, can hold back a lake. Chapter 9 answers that.

What to remember

  • A gravity dam is the block on the ramp, on level ground: its weight presses it onto the rock, and the rock's grip, (W − U) · tan φ, holds it against the water's push.
  • Water in the cracks under the base presses up with the reservoir's pressure at the heel and the river's at the toe. With nothing to stop it, it carried 42% of the 60 m dam's weight and cut its factor of safety from 2.20 to 1.28.
  • A grout curtain and a line of drains make the water spend most of its head near the heel; working drains raised the factor to 1.70. Drains clog, and the gallery is there so they can be measured and cleaned.
  • The friction angle multiplies everything: at 35° the same dam fell to 1.19.
  • The resultant must cross the base inside the middle third, or the heel stops being pressed. Lévy asked for more after Bouzey: the masonry at every joint's upstream end must press harder than the water there, so that water is squeezed out of a crack instead of into it.

Questions

  1. 7.1

    In one sentence, say how the uplift under a gravity dam is the water cushion under the block of Fig. 4.1.

    Hint

    What did the cushion take away from the block, and what does the friction depend on?

  2. 7.2

    By hand: the reservoir behind the 60 m dam stands at 50 m and the drains are working. What are the head at the drain line, the uplift and the factor of safety against sliding?

    Hint

    Use Eq. 7.2 with 50 and 5 m, then the two straight-line pieces of Example 7.2. The push at 50 m is in Example 7.3.

  3. 7.3

    With the drains working, slide the reservoir from 30 m up to 59 m, a metre below the crest. Does the factor of safety ever fall below 1.5? Why does it fall faster near the top?

    Hint

    Compare how the push and the uplift grow with depth (Example 7.3).

  4. 7.4

    Clog the drains and raise the reservoir to 59 m. Where does the resultant cross the base, and is the heel still pressed?

    Hint

    The middle third runs from 0.33 B to 0.67 B. What happens to the heel pressure in Eq. 7.3 as x approaches 2B/3?

  5. 7.5

    The US Army Corps of Engineers assumes that drains remove only 25 to 50 per cent of the straight-line pressure above the tailwater's at the drain line. With the reservoir at 57 m, what does the 60 m dam's factor of safety come to at each end of that range, and would you accept the Bureau of Reclamation's one-third rule for this dam without measurements?

    Hint

    With no drains the straight line stands at 50.5 m at the drain line; take 50 or 25 per cent of its height above the river's 5 m away.

  6. 7.6

    Lévy said that the masonry at the upstream end of every joint must press harder than the reservoir's water presses there. What does that prevent, and why is the middle-third rule alone not enough to prevent it?

    Hint

    Think about a crack that has already opened, and what pressure the water in it carries (Example 2.3).

Further reading

  • U.S. Bureau of Reclamation (1987). Design of Small Dams, 3rd ed. Denver. https://www.usbr.gov/tsc/techreferences/mands/mands-pdfs/SmallDams.pdf serious Its chapter on concrete gravity dams sets out the loads of this chapter (uplift with and without drains, the one-third rule), the stress check at the upstream face that descends from Lévy's, and the sliding check. Free from the Bureau of Reclamation.
  • Anderson, C., Mohorovic, C., Mogck, L., Cohen, B., Scott, G. (1998). Concrete Dams: Case Histories of Failures and Nonfailures with Back Calculations. Report DSO-98-05. Denver: U.S. Bureau of Reclamation, Dam Safety Office, December 1998. https://www.usbr.gov/ssle/damsafety/TechDev/DSOTechDev/DSO-98-05.pdf serious Case histories of concrete and masonry dams that slid, cracked or failed, Bouzey among them, each with the Bureau's own back-calculation of the forces. A good place to see this chapter's checks applied to real failures.
  • Lévy, M. (1895). Quelques considérations sur la construction des grands barrages. Comptes rendus hebdomadaires des séances de l'Académie des sciences 121(6): 288–300 (séance du 5 août 1895). https://archive.org/details/mobot31753002064894 primary source Lévy's note, in French. The first two pages hold the boiler, the drained wells and the new condition; the rest works out the stresses inside a dam.

Chapter 8 · Part II · Two kinds of dam

Never finished

Why would a country write the word “upstream” into its laws?

So far. Chapter 6 set a tailings dam beside a water dam: what it holds can lose its strength, and the dam keeps growing for as long as the mine runs. Chapter 7 looked at a dam that is built once, of a material that never goes soft. This chapter watches a tailings dam grow, and asks why the direction it grows in matters so much.

After this chapter you can

  • sketch the three ways to raise a tailings dam, and work out the fill each one needs and where its crest ends up;
  • explain why an upstream raise can fail with no earthquake, years after the last raise was built;
  • estimate how fast a tailings deposit rises, and name the levers that keep it safe (rate of rise, a wide beach, a pond kept away from the crest) and the alternatives to a wet deposit.

This chapter adds a raise stepped onto the crest

3 figures · about 20 min to read · Film 8.1, 9 min 6 s

On 30 September 2020 Brazil changed its national law on dam safety. Most laws about building say what a structure must achieve: how strong, how safe, how often inspected. This one did something rarer. It named a way of building and forbade it: no mining dam in Brazil may be built or raised by the upstream method, a montante in Portuguese. Then, so that nobody could argue about what the words meant, the next paragraph defined the method, in translation: a way of building a dam "in which the containment dykes rest on the tailings themselves, or on sediment previously discharged and deposited."Lei 14.066 (2020), Art. 3, adding Art. 2-A to Law 12.334 of 2010. §1 reads: "Entende-se por alteamento a montante a metodologia construtiva de barragem em que os diques de contenção se apoiam sobre o próprio rejeito ou sedimento previamente lançado e depositado." The translation is ours.

The law gave every mine with such a dam until 25 February 2022 to take it out of service as a dam.The Portuguese word is descaracterização, "decharacterization": draining, reshaping or removing the deposit until it no longer works, or fails, as a dam. The next paragraph lets the regulator extend the deadline where the work cannot technically be finished in time. It was not the first word on the subject. Less than a month after the collapse at Brumadinho in January 2019, Brazil's mining agency had already banned the method by resolution, with the same definition.The deadline: Lei 14.066 (2020), Art. 2-A §2. The resolution: ANM Resolution 4 of February 2019, as summarized by Freire (2019). Chile had turned against it half a century earlier.

A lawmaker does not usually care whether a wall steps one way or the other as it grows. So why would a country write the word "upstream" into its laws? The answer is in what the wall stands on, and to see it we have to watch one grow.

8.1 A dam that grows with the mine

A water dam is finished the day it is built. A tailings dam never is while the mine runs. The mill sends out more ground rock and water every day, so the dam has to rise with the deposit behind it. It begins as a starter dam of soil and rock dug from pits nearby, and the tailings are piped out over its crest as a slurry. As Chapter 6 showed, the slurry sorts itself on the way: the coarse sand settles near the pipes and builds a gently sloping beach, while the fine grains and the water run on to a pond. When the deposit nears the top, the dam is raised, and then raised again. Each raise can step toward the pond, away from it, or straight up, and that direction names the method:

  • Upstream: the crest steps toward the pond. Each raise is a small dyke built on the tailings beach, often from the beach's own sand.
  • Downstream: the crest steps away from the pond. Each raise is built over the previous outer slope, on compacted fill and natural ground.
  • Centreline: the crest rises straight up. Each raise straddles the top of the beach and the outer slope.

Below, three dams are raised side by side from the same starter dam, 20 m tall, in four raises of 12 m each, to the same final height. Press Play, or drag the Stage slider:

Fig. 8.1 · Staged construction of three tailings dams, same height, same starter dam Schematic · vertical scale ×1.5 · beach and pond not to scale

Drawing scrolls sideways →

Raise 4 complete · 68 m
Fill resting on tailings Path of the crest, stage by stage Water table inside the dam (phreatic surface); hatched below it is saturated Drains Low-permeability zone

Raised to the same height from the same starter, the upstream dam needs the least fill and puts most of it on the tailings. Each panel is a section through one dam, pond on the left. The starter is 20 m tall with an 8 m crest and faces of 2H:1V, the notation of Example 1.2; each raise adds 12 m. The tailings fill the space behind the dam, and the dashed lines in them are old beach surfaces. Over each panel the readouts give the fill (its cross-section area, which is the volume per metre length of dam), how far the crest has moved, and the share of the fill that rests on tailings, which a brick-red line under a raise marks. The dots and the dashed brick-red line trace the crest from stage to stage. The blue dashed curve is the phreatic surface, the water table inside the deposit; below it the deposit is hatched as saturated. The downstream dam carries a low-permeability zone and a drain up its inner face, the centreline dam a chimney drain, and both a drain along the base. The beach is drawn at 8% so it can be seen; real beaches slope about 1% where the slurry comes out and flatten toward the pond (Jewell (2012), p. 86). Centreline raises put only a small upstream wedge on the beach (9% of the fill here), which must be sandy, drained and compacted. The Earthquake button is a cartoon of a loose deposit; a dense, drained beach would hold. The numbers to remember: 2,496 m³ of fill per metre for upstream, 9,792 for downstream.

  • Press Play and stop at "Raise 2 · placing fill". Compare how far each crest has moved so far.
  • Drag the Stage slider to the end and divide the downstream fill by the upstream fill.
  • Press Earthquake. Which of the three dams has the brick-red line under every raise?
Answers

At the end, 9,792 ÷ 2,496 = 3.92: almost four times the fill. The upstream dam is the one whose every raise rests on tailings, and it is the one that fails.

Notice two things. The upstream crest walks a long way toward the pond, −112 m by the end, while the downstream crest walks 96 m the other way. And notice what the fill costs: the downstream dam's numbers climb faster with every raise. You can work out that bill from the shapes alone. A dam section h metres tall with an 8 m crest and faces at 2:1 is a band 8 m wide down the middle and a triangle on each side, each ½ × h × 2h = h². So its area, which is also its volume of fill per metre length of dam, is

A = 8h + 2h² (8.1)

A area of the section, m² (m³ of fill per metre of dam) · h height, m · the 8 is the crest width in metres, and the 2 comes from the two 2:1 faces

The first term grows in step with the height; the second with its square. That square is the whole story of the next example:

Example 8.1

The fill bill

Given
The dams of Fig. 8.1: a starter 20 m tall, four raises of 12 m to 68 m, a crest 8 m wide, every face at 2:1.
Find
The fill in each finished dam, per metre of its length, and how fast the downstream bill grows with height.
The idea
Every piece of fill is a whole dam section or a part of one, so Eq. 8.1 prices each piece.
  1. Starter: 8 × 20 + 2 × 20² = 960 m³Eq. 8.1 with h = 20 m; every number here is per metre of dam
  2. One upstream dyke: 8 × 12 + 2 × 12² = 384 m³a dyke is a small dam of the same shape, 12 m tall
  3. Upstream: 960 + 4 × 384 = 2,496 m³the starter and four dykes
  4. Downstream: 8 × 68 + 2 × 68² = 9,792 m³each raise covers the old outer face, so the finished dam is one whole section 68 m tall
  5. Centreline: 68² + 8 × 68 + 20² + 4 × 144 = 6,144 m³the outer triangle and the crest band full height; on the pond side, the starter's own triangle and four wedges on the beach, each ½ × 24 × 12
  6. Last downstream raise: 8 × 56 + 2 × 56² = 6,720, and 9,792 − 6,720 = 3,072 m³the dam at 68 m less the dam at 56 m
  7. 3,072 ÷ 384 = 8 dykesthat raise alone, counted in upstream dykes
  8. At half the height: 8 × 34 + 2 × 34² = 2,584, and 9,792 ÷ 2,584 = 3.79a downstream dam twice as tall

Upstream 2,496, centreline 6,144, downstream 9,792 m³ per metre of dam. The last downstream raise takes as much fill as 8 upstream dykes, and doubling the height of a downstream dam multiplies its fill by 3.79.

Check. Raising a downstream dam by a little adds a strip of fill as wide as the whole base, because the outer face moves out from top to bottom. Halfway up the last raise, at 62 m, the base is 8 + 4 × 62 = 256 m wide, and 256 × 12 = 3,072 m³, the same as step 6.

What it means. Fill means trucks, time and money, and a downstream dam pays for each new metre across its whole width while an upstream dam pays for one small dyke. That is the pull toward upstream, and it grows as the dam does.

The cheap dam has a price of its own, and it shows in where each crest ends up and in what each raise stands on:

Example 8.2

Where the crest goes, and what it stands on

Given
The finished dams of Fig. 8.1. Each upstream dyke is 384 m³ per metre, all of it on tailings; each centreline raise puts a wedge of 144 m³ on the beach.
Find
How far each crest moves, and what share of each dam's fill rests on tailings.
The idea
Each upstream dyke stands with its outer foot on the middle of the crest below, and leans back toward the pond at 2:1.
  1. Upstream, per raise: 2 × 12 + 4 = 28 ma 2:1 face climbing 12 m runs 24 m back; the foot on the middle of the old crest adds half the crest
  2. Four raises: 4 × 28 = 112 m toward the pondthe readout's −112
  3. Downstream: 2 × 48 = 96 m away from itthe crest moves 2 m out for every metre of the four raises
  4. Upstream on tailings: 4 × 384 = 1,536 m³, and 1,536 ÷ 2,496 = 0.615four dykes, out of the whole dam
  5. Centreline on the beach: 4 × 144 = 576 m³, and 576 ÷ 6,144 = 0.094four wedges, out of the whole dam

The upstream crest ends 112 m toward the pond with 62% of the dam on tailings; the centreline crest stays put with 9% on the beach; the downstream crest moves 96 m away with none.

Check. The starter's base reaches 2 × 20 + 4 = 44 m upstream of its crest, so the final upstream crest stands 112 − 44 = 68 m out beyond the starter's inner toe. The whole top of that dam has nothing under it but tailings.

What it means. Of the three, only the upstream dam has to trust the tailings to carry it. Whether it holds is a question about the beach, and no amount of care with the dykes can answer it.

8.2 What the wall stands on

Most people say a dam fails when its wall breaks, so the way to make it safe is to build the wall strong. For an upstream dam that answer looks in the wrong place. Each dyke is compacted fill, small and sound. Under it lies the beach, and the beach was never built at all. It was laid down from a slurry, grain by grain, the way a wave lays down the loose sand at the water's edge in Chapter 1, and nobody rolled it afterward. Near its surface, in the sun and wind, it drains and can look as firm as the damp sand of a castle. A few metres down it sits below the water table of the deposit, the phreatic surface, and there every gap is full of water.

How high that water stands under the dykes is not fixed. It depends on where the pond is. Here is one upstream dam drawn twice, with the pond kept far from the crest and with the pond close to it:

Fig. 8.2 · What an upstream raise stands onSchematic · not to scale
A · Pond kept far from the crest pond long beach starter dykes on the beach tailings, loose and soaked phreatic surface B · Pond near the crest pond short beach starter tailings, loose and soaked phreatic surface

Where the pond sits sets how much of the beach under the dykes is soaked. Both panels show the same upstream dam: a starter on the right, a drain at its base, and three dykes stepping toward the pond, each with a brick-red line where it rests on the tailings. The blue dashed curve is the phreatic surface; below it the tailings and the fill are hatched as saturated. In A a long beach keeps the water's edge far from the crest, and the phreatic surface falls well below the dykes before it reaches them. In B the pond reaches almost to the crest, and the phreatic surface runs through the dykes themselves. The thing to remember: the deep tailings are soaked in both panels; the pond decides how close the soaked zone comes to the dykes.

  • In each panel, follow the brick-red line under the top dyke and say whether the beach right beneath it is above or below the phreatic surface.
Answers

In A it is above: the top dyke rests on beach that has drained. In B it is below: the top dyke rests on soaked tailings. The lowest dyke stands on soaked tailings in both.

Notice that the only difference between the two panels is the water. The dykes, the beach and the tailings are the same. In B the dykes stand on loose grains with water in every gap, and as Chapter 5 showed, that is a material whose strength depends on what happens to that water when something pushes on it.

Something always does. Each new dyke adds its weight to the beach beneath it. The weight of the sloping dam presses the tailings under it outward, toward its open face. Nobody needs to shake anything. So the question to settle is what loose, soaked grains do under a push that is slow and steady:

A new dyke is placed, slowly, on a beach of loose tailings with water in every gap. The water can only leave slowly. What happens to the grains under the dyke?

The film Dams that are never finished puts the same push on a box of grains, then follows it out into the world:

Film 8.1 · Dams that are never finished9 min 6 s · captions

Before you watch

Brazil and Chile each banned one way of building a tailings dam. The film gives its reason with three boxes of grains, pushed slowly sideways, and only one lets go. You have predicted the first box: loose grains, drain shut. Before you watch, predict the other two, the same loose grains with the drain open and dense grains with the drain shut, and say which way of raising a dam each box stands for.

The film counts fill in round numbers ("about four times"); Example 8.1 has the exact ones.

In this film

  1. Grains under a dam, fading Fig. 5.3
  2. Never finished: three ways up Fig. 8.1
  3. What an upstream wall stands on Fig. 8.2
  4. A slow push with the drain shut your prediction
  5. El Cobre, Chile, 1965 §8.3
  6. Brumadinho, 2019 §8.3
  7. Coquitlam, a water dam Fig. 8.3
  8. Mount Polley: what gave way? Chapter 13
Read the transcript

These grains are holding up a dam. Something is pushing them sideways, very slowly. Nobody is shaking them. Watch the lines between the grains: they show how hard the grains press on each other. They're fading. In twenty nineteen, in Brazil, a dam standing on grains like these collapsed. It took under ten seconds, and there was no earthquake. Within weeks, Brazil banned the way it was built. Chile had turned against the same way of building half a century before. Why that one way?

A mine grinds rock to powder to get the metal out. What's left is ground rock and water, called tailings. It's piped out as a slurry, like runny grey mud, and kept behind a dam for good. To see why that's hard, go back to the beach. Dry sand pours. Damp sand stands up as a castle wall. But at the water's edge, soaked sand turns to soup: wiggle your feet, and they sink. That's the idea behind this whole series. Soil is strong because its grains press on each other. Water in the gaps pushes them apart, and takes that strength away. Remember the soup.

A water dam is finished the day it's built. A tailings dam never is. The mill sends more every day, for as long as the mine runs, so the dam has to keep growing. It starts as a small starter dam. Pipes along its top, the crest, pour the slurry out. The coarse grains settle near the crest, and build a beach. The fine grains and the water run on, to a pond.

When it's nearly full, the dam is raised, and then raised again. There are three ways to do it, from the same starter dam. Upstream: each new wall steps back toward the pond, onto the beach. It's often built from the beach's own sand. Centreline: each one goes straight up. Downstream: each one steps away from the pond, over the dam's own outer slope.

Here's a drawing from the course. The starter dam is twenty metres tall, and it's raised four times. Upstream, every raise is the same small wall. Downstream, every raise has to cover the whole outer face, so each one is bigger than the last. In all, downstream takes about four times the fill of upstream. Its last raise alone takes eight times as much as an upstream one. Make a dam like this twice as tall, and it takes nearly four times the fill. Centreline sits in between. Fill means trucks, time and money. So upstream is the cheapest way, and the quickest. Why would anyone build any other way?

Here's the idea most people start with: a dam fails when its wall breaks, so build the wall strong. But look at what an upstream wall stands on. It isn't on rock, or on the old dam. Its crest has walked toward the pond, and most of the dam now sits on the beach: tailings laid down by the pipes, years before. On top, that beach can look like the damp sand of the castle. But inside the dam, there's a water level. And below it, the tailings are loose, and soaked. That's the soup. So what does the soup do, when something pushes on it?

This is the course's grain model, a teaching model in two dimensions. Each disc is a grain. These are loose, with water in every gap. On top, a weight, standing in for the wall. At the bottom, a drain, shut, so the water can't get out. Now push it sideways, slowly. No shaking at all. Will the grains pack down and hold, like sand under a footstep?

They don't. The box hardly gets any shorter, and the water takes over. The bar is the weight on the box: amber is the share the grains carry, blue the share the water carries. Before long, the water is carrying about nine tenths of it. The grains hardly press on each other, and most of the lines fade. Nothing broke. The grains just stopped pressing. That's liquefaction: the grains lose their press, and the soil flows like a heavy liquid.

So why does the water take over? Open the drain, and give the same push. This time the sand settles, just a little, and the grains keep pressing. Pushed, loose grains shuffle into the gaps between them, and try to pack tighter. With the drain open, the water squeezed out of those gaps simply leaves. With the drain shut, it has nowhere to go. So the grains can't move closer, and the weight passes from the grains to the water. Now try dense grains, with the drain shut.

Pushed, they have to climb over each other, so the gaps grow, and the water in them is pulled tight. That pull draws the grains together, and they press more than three times as hard. It has to be loose enough, though. In the model, only the loosest packings let go. Pack them a little tighter, and they hold. That's why a downstream raise is built of compacted fill, rolled dense in layers, with drains inside, and stands on it. A centreline raise stands mostly on it too.

But isn't liquefaction an earthquake thing? In nineteen sixty-five, an earthquake shook Chile. The El Cobre tailings dams, raised the upstream way, liquefied, and buried the village below. More than two hundred people were killed. In nineteen seventy, Chile's first rule for tailings dams allowed only the other two ways, where people live below. Shake our box, and it does the same: the grains stop pressing.

But Brumadinho needed no earthquake. Dam One, at an iron ore mine near the town, was raised the upstream way, ten times. On the twenty-fifth of January, twenty nineteen, no earthquake was recorded in the region. Cameras were watching it, and until the moment it went, nothing looked wrong. Its slope collapsed in under ten seconds. Within five minutes, about three quarters of its tailings had flowed out.

Two hundred and seventy people were killed. And no one had raised it for more than five years. The growing had stopped, but the soup was still inside. A tailings dam has to stand long after the mine has closed, for good. Investigations differ on what set it off. They agree on what happened: the tailings in the dam itself liquefied. The wall didn't break first. It stood on the soup, and the soup let go.

So here's why that one way was banned. An upstream raise stands on the soup. And a slow push is enough to make the soup let go.

At the beach, the soup lets your feet sink. Under an upstream raise, it can let the whole dam go. The next year, Brazil wrote the ban into law. It defines the banned method just like that: walls resting on the tailings themselves.

Brumadinho wasn't the first. Stava, in Italy, where an upper dam fell onto the one below it: two hundred and sixty-eight killed. Merriespruit, in South Africa, where a dam overtopped in a storm, and its tailings flowed through a suburb: seventeen. Fundão, in Brazil: nineteen. Fundão's tailings ran down the rivers for more than six hundred kilometres, all the way to the Atlantic. Each had its own trigger. All of them were raised the upstream way.

So is the trouble mine waste? Coquitlam Dam, in British Columbia, holds water, some of it Vancouver's drinking water. It was built more than a century ago by hydraulic fill: soil washed into place with water. That leaves it loose and wet, like an upstream raise. Engineers found a zone in it that could liquefy in a big earthquake. So BC Hydro built a new dam of compacted earth and rock, against the old one's outer toe. What matters isn't water or tailings. It's loose and wet, or dense and drained. It's the series' idea once more: soil holds while its grains press on each other, and water in the gaps can take that away.

Mount Polley, also in British Columbia, was a tailings dam with a core, a filter and rockfill. It wasn't raised the upstream way. In twenty fourteen, it failed anyway. What gave way?

After you watch: the answer

Only the loose box with its drain shut lets go, and it stands for an upstream raise on a soaked beach. With the drain open, the water squeezed out of the gaps leaves, so the loose grains settle a little and keep pressing: that is a raise with drains in it. The dense grains, pushed, must climb over one another; the gaps grow, the water in them is pulled tight, and the grains press harder than before: that is compacted fill. A downstream raise is both, dense fill with drains, and stands on it; a centreline raise stands mostly on it. One line in the film needs care: at 1:14 it says water in the gaps "pushes them apart". More exactly, the water carries part of the load the grains would otherwise carry between them, so they press on one another less (Chapter 4). The film's last question, what gave way at Mount Polley, stays open until Chapter 13.

8.3 A slow push is enough

Put the box back into the dam. Below the phreatic surface, the beach under an upstream raise is loose sand and silt with water in every gap; it is contractive, so a push makes it try to pack tighter. Whether the water can get out of the way in time depends on how far it has to go and how easily it moves, and Chapter 5 put a clock on that: the time grows with the square of the distance (Eq. 5.2). A sample in a box drains in seconds. A deposit tens of metres thick, laced with fine silt from the pond, does not drain in the seconds or minutes it takes a slope to start moving. For that moment the tailings are undrained, the water takes the load, and the grains stop gripping. The deposit flows as a heavy liquid, and the dykes on it go with it. That is static liquefaction: liquefaction with no earthquake at all.

The trigger can be small, and it can come late. Dam I at the Córrego do Feijão mine near Brumadinho was raised upstream in ten raises over 37 years, to 86 m. Its last raise was built in 2013, and tailings stopped arriving in July 2016. On 25 January 2019, with no sign of trouble on any of its instruments, its slope collapsed in under ten seconds, and about three quarters of its tailings flowed out within five minutes. The expert panel found flow liquefaction of the tailings inside the dam itself. They were loose, saturated and brittle, and the water level in them had stayed high near the toe after deposition stopped. What tipped them was a combination of slow creep and the loss of suction in the drier tailings above the water table after heavy rain at the end of 2018.Robertson et al. (2019), Executive summary, pp. i–ii, and p. 16 (raises, height, dates). The toll is in Chapter 6. The growing had stopped. The loose, soaked tailings were still there.

An earthquake is a faster push of the same kind, and Chile, where the ground shakes often, learned it first. After the El Cobre tailings dams failed in an earthquake in 1965 and buried the village below, Chile wrote its first rules for tailings dams, in 1970. They covered dams with people living within reach downstream; among other things they kept the pond as far from the sand wall as possible and had the wall's sand raised outward, over the crest, and they have been read ever since as shutting out the upstream method for those dams. Chile's rules of 2007 say it in one line, which translates as "the use of the upstream method is prohibited."El Cobre in 1965: Dobry & Alvarez (1967). The rules: Chile DS 86 (1970), Arts. 30, 31 and 44; Chile DS 248 (2007), Art. 14(h). Stava, in Italy, failed with no earthquake at all.On 19 July 1985 the upper of two tailings dams above the village of Stava, in the Italian Alps, burst. Its mud poured into the lower basin and on down the valley, about 180,000 m³ of it, and 268 people died. The upper dam was still being raised, by the upstream method. Luino and De Graff put first among the causes the mistaken assumption that the tailings would consolidate fairly quickly.

That answers the question this chapter opened with. A country writes "upstream" into its law because no rule about the wall can make an upstream dam safe. The danger is in what the wall stands on: loose, saturated tailings that a slow push can turn to liquid, and that stay that way after the raising stops. So the law names the one thing it can name, the place the wall is allowed to stand.

Check yourself A mine proposes to build its upstream dykes of crushed rock instead of beach sand, so that the dykes themselves are strong. Does that make the dam safe?

No. Stronger dykes still stand on the same loose, soaked beach. If the tailings under them liquefy, rock dykes go down with them as surely as sand ones.

8.4 How fast it may rise

Upstream dams are still built where the law allows them; most South African tailings dams, for one, are raised this way. The ones that stand are kept standing by keeping the tailings under them drained. A long beach keeps the pond, and with it the high water, away from the crest, as in panel A of Fig. 8.2. The pond itself is kept small: its water is drawn off to a separate dam as soon as it can be, not stored against the tailings.SANCOLD (2020), §4.8, p. 18. The third lever is time: the rate of rise, how many metres a year the surface of the deposit climbs.

Time matters because every metre of new tailings is a load on the tailings below. Below the phreatic surface that load is carried at first by the water, and it passes to the grains only as the water seeps away and the grains settle closer, the slow process Chapter 5 called consolidation. Rise slowly and each metre has drained and settled before the next arrives. Rise fast and the water pressure builds up from one season to the next, the tailings stay loose, and the phreatic surface climbs with them.

So the deposit runs a race between two clocks. One is the time it takes to add a metre. The other is the time the water below takes to get out, and by Eq. 5.2 that grows with the square of the distance the water must travel, so it grows as the deposit thickens; the fine, silty tailings laid down toward the pond are the slowest of all. Nobody can work out the second clock exactly for a real deposit, layered as it is, so the industry limits the first one instead. South African guidance puts the limit for most upstream dams at about 2 to 2½ m a year, up to about 6 m a year for dams whose outer wall is built of coarse, free-draining sand, and says plainly that the figure is a judgement based on experience.SANCOLD (2020), §4.6, p. 15. At 2 m a year, each metre has half a year to drain before the next one covers it.

The rate of rise itself is plain arithmetic: the volume the mill sends out in a year, spread over the area of the deposit,

r = M ⁄ (ρd S) (8.2)

r rate of rise, m a year · M dry tailings sent out, tonnes a year · ρd tonnes of solids in each cubic metre of settled deposit · S area of the deposit's surface, m²

A mine of ordinary size shows how fast that can be:

Example 8.3

How fast the deposit climbs

Given
An illustrative mine that sends out 50,000 tonnes of dry tailings a day. Settled in the deposit they hold 1.4 tonnes of solids in each cubic metre (also illustrative). The deposit covers 2 km², and the dam is raised 12 m at a time.
Find
How fast the deposit rises, and how often the dam must be raised.
The idea
Eq. 8.2, one step at a time: tonnes to cubic metres, a day to a year, then spread over the area. A square kilometre is a million square metres, so millions of cubic metres over square kilometres gives metres.
  1. 50,000 ÷ 1.4 = 35,714 m³ a daytonnes over tonnes per cubic metre
  2. 35,714 × 365 ÷ 1,000,000 = 13.04 million m³ a yeara year of days, counted in millions
  3. 13.04 ÷ 2 = 6.52 m a yearmillions of cubic metres over millions of square metres
  4. 12 ÷ 6.52 = 1.84 yearsbetween one 12 m raise and the next
  5. 13.04 ÷ 2.5 = 5.2 km²the area that would hold the rise to 2½ m a year

The deposit climbs 6.5 m a year, a new raise every 1.8 years: two to three times the usual limit for an upstream dam. To stay under 2½ m a year it would need about 5.2 km².

Check. The order of magnitude: 13 million m³ is a block 1 km by 1 km by 13 m, and spread over two such squares it stands 6½ m deep.

What it means. The rate of rise is set by the mine's output and the size of the valley, not by the dam engineer. A mine that grows its output, or a deposit that narrows as it climbs, can outrun its own tailings. The engineer's choices are a bigger footprint, a slower mine, or a dam that does not stand on the tailings at all.

Check yourself The mine of Example 8.3 doubles its output. What happens to the rate of rise, and to the time between raises?

The rate doubles, to 13.04 m a year, and a 12 m raise falls due every 11 months. An upstream dam on that deposit would be far past any rule of thumb.

8.5 Raising away from the water

The other two methods do not ask the tailings to carry the dam. A downstream raise stands on compacted fill and on the ground beyond the old toe. Because each raise covers the whole outer face, a sloping core and drains can be carried up with every lift, as in a water dam, and the downstream panel of Fig. 8.1 draws them. Such a dam can hold back a pond, and it rides out earthquakes best. Its costs are the ones Example 8.1 counted, and land: its toe marches out 4 m for every metre of rise, 192 m over the four raises of the figure, and that ground has to be set aside from the day the mine opens.

A centreline raise splits the difference. Its crest stays put, so pipelines and roads along it never move, and a vertical drain can rise with every lift. A small wedge of each raise rests on the top of the beach, 9% of the fill in the figure, so that part of the beach must be sandy, drained and compacted. It can be built of cyclone sand: the slurry is spun in a hydrocyclone, which flings the coarse grains to its wall and out at the bottom while the fines and most of the water leave at the top.SANCOLD (2020), §4.4.4, p. 12. The coarse sand drains freely and can be compacted, which is what a wall needs.

Check yourself Tailings behind a downstream dam rise 6 m a year. Is that rate a danger to the dam?

Far less than it would be to an upstream dam. The limit exists because each upstream raise stands on the last few years of tailings. A downstream raise stands on compacted fill and the ground beyond the toe, and the tailings behind it only lean on it, as a heavy liquid would; that is a load it is designed to carry. South African guidance says the rate of rise is not relevant behind a fully engineered earthfill or rockfill wall (SANCOLD (2020)). The wall must still be raised fast enough to stay ahead of the tailings.

Between centreline and upstream lies the modified centreline raise. Each raise steps its crest a little toward the pond, so more of it rests on the beach than in a centreline raise, and the core leans slightly upstream, supported on that side by a zone of tailings. Mount Polley, in British Columbia, was designed to be raised this way.Mount Polley Review Panel (2015), pp. 8 and 53. The panel adds that as built, one early raise was entirely upstream construction and later ones crept upstream (pp. 55 and 69). Hold that thought; it comes back in Chapter 13.

The table gathers the three methods:

Table 8.1 · Three ways to raise a tailings dam, for and against
QuestionUpstreamDownstreamCentreline
Each raise rests onThe tailings beach. Each raise is a small dyke, often built from the tailings themselves.Compacted fill and natural ground.Fill and the starter, with a small upstream wedge on the beach.
Fill to 68 mLeast by far: 2,496 m³ per metre. Cheapest and fastest.Most: both faces move out with every raise, so the section grows fastest (9,792 m³ per metre here, against 6,144 for centreline).A middle path on fill and on footprint.
The crestWalks toward the pond, onto the tailings.Walks away from the pond; the toe marches out, so land downstream has to be reserved from the start.Stays put, so pipelines and roads don't move.
Drains and coreCan't be extended upward.Raised with every lift, as in a water dam.A vertical chimney drain raised with every lift.
Water and earthquakesCan't store much water: the pond must stay far from the crest. Rests on loose, saturated tailings that can liquefy, with or without an earthquake.Can store water. Best in earthquakes.Good, if the beach next to the crest is sandy, drained and compacted.
RulesNew upstream dams have been barred in Chile since 1970, at first only where people live within reach downstream and under its 2007 rules everywhere, and banned outright in Brazil (since 2019).None special.None special.

Fill from Fig. 8.1 and Example 8.1. Rules: Chile DS 86 (1970), Chile DS 248 (2007), Freire (2019), Lei 14.066 (2020).

8.6 Less water in the deposit

Every method so far stores the tailings wet and builds a wall to hold them. The other road is to take the water out. Thickened tailings leave the mill with less water and stand on steeper beaches than a thin slurry does (Jewell (2012)). Filtered tailings, sometimes called dry stack, go further: filters take out enough water that the tailings can be carried by truck or conveyor and spread and compacted like earthworks. The review panel that investigated the Mount Polley breach set out the case for them from first principles of soil mechanics: take away the surface water, keep the tailings unsaturated with drainage, and "achieve dilatant conditions throughout the tailings deposit by compaction".Mount Polley Review Panel (2015), §9.3.1–9.3.2, pp. 121–123, which also gives cost as the main reason filtered tailings are not used more widely. Each of those is a way of keeping the grains pressing on one another: no pond to soak them, no water in the gaps to take the load, and grains packed dense enough to swell, not collapse, when they are sheared. Even then the compaction is not optional; a loose, dry stack that later gets wet can flow too. The catch is cost: filtering is more expensive than piping a slurry, and most filtered operations so far have been small.

8.7 A water dam built like a tailings dam

All of this might suggest the trouble is mine waste. Coquitlam Dam, in British Columbia, shows that it is not. It holds water, for BC Hydro's power plants and for about a third of Metro Vancouver's drinking water, and it was built in 1913Drinking water: BC Hydro (2026). The date and the way it was built: BC Hydro (2008); Water Power (2001). by hydraulic fill: soil washed into place with water and left to settle, which leaves it loose and saturated, much like the beach under an upstream raise. Here it is as a case file:

Fig. 8.3 · Case file: Coquitlam Dam · 1913, backed up by a new dam in 2007Schematic · heights to scale; slopes not published, drawn for illustration
10 m steps both crests about 30 m above the valley floor Coquitlam Lake 1913 dam hydraulic fill, loose, wet below new dam, 2007 compacted earth core, rockfill shells finer core valley floor; the new dam stands mostly on dense glacial soil
Dam
Coquitlam Dam, on the Coquitlam River, British Columbia; owned by BC Hydro.
Built
1913, replacing a dam 4.5 m high. The new dam was finished in 2007, and BC Hydro declared the whole seismic upgrade complete in October 2008, at a cost of C$65 million.
Height
About 30 m, the old dam and the new alike; the new dam is 300 m long.
Type
Hydraulic fill. The new dam: compacted earth core and rockfill, founded mostly on dense glacial soil.
What happened
Investigations found a loose zone in the fill that could liquefy in a moderate or large earthquake. The reservoir was held below a safe level until the new dam was built against the old one's downstream toe.
Toll
None. Nothing failed; the work was done before an earthquake could test the fill.
Sources
BC Hydro (2008); Renewable Energy World (2008); Water Power (2001); Yan, Sy & Thavaraj (2008); drinking water: BC Hydro (2026).

A water dam of loose, saturated fill carries the same risk as an upstream tailings dam, and was made safe the same way: with a dense, drained dam that does not depend on it. Coquitlam Dam was built in 1913 by hydraulic fill, washed into place, so its sandy shells and finer core settled loose and stayed wet. Investigations found that a zone of it could liquefy in a large earthquake. Rather than dig out or repair the old fill, BC Hydro built a new compacted earth-core rockfill dam against its downstream toe, founded mostly on dense glacial soil. The new dam was finished in 2007, and BC Hydro declared the whole seismic upgrade complete in 2008. In the drawing the old fill is hatched as saturated below its phreatic surface (blue dashed), and the new dam's core sits between two filters inside rockfill shells; only the heights are to scale.

What this chapter explains. Whether fill holds in a shake or a push depends on its state: loose and wet, or dense and drained. Clean water or tailings behind it, the fill behaves the same.

  • Look for the line in the drawing that tells you the old fill is loose. Why is there none?
Answers

Density does not show in a section: a loose fill and a dense one look the same on paper and from the crest. It took investigations of the fill itself to find the loose zone.

Notice that the old dam is still there, and still holds the lake. The new one stands behind it for the earthquake that might make the old fill flow. Brazil asked the same of its upstream tailings dams in 2019: before taking one out of service, a mine had to reinforce it from downstream or build a new containment structure below it.ANM Resolution 4 of 2019, as summarized by Freire (2019).

What to remember

  • A tailings dam grows with the mine, by upstream, centreline or downstream raises. To the same height, upstream needs the least fill (2,496 m³ per metre in Fig. 8.1, against 9,792 downstream) because it builds each raise on the beach.
  • An upstream raise stands on loose tailings with water in every gap. Pushed, even slowly, the grains try to pack tighter, the water cannot leave in time and takes the load, and the grains stop gripping. No earthquake is needed, and the danger lasts after the raising stops.
  • A slow rate of rise gives the water time to leave the tailings; a long beach and a small pond keep it away from the dykes. How fast the deposit rises is set by the mine's output and the area the deposit covers.
  • Downstream and centreline raises stand on dense, drained fill, so the grains under the dam keep pressing; filtered, compacted tailings take the water out of the deposit itself.
  • What decides the danger is the state of the fill, loose and wet or dense and drained, not what it holds back: Coquitlam is a water dam that needed the same cure.

Questions

  1. 8.1

    Example 8.1 found that a downstream dam twice as tall takes 3.79 times the fill. Do the same for the upstream and centreline dams of Fig. 8.1, from 34 m to 68 m, treating the raising as smooth.

    Hint

    Upstream: the 960 m³ starter, then one 384 m³ dyke for each 12 m, which is 32 m³ for each metre above 20 m, so the fill is 32h + 320. Centreline: h² + 8h + 20², plus a 144 m³ wedge for each 12 m above 20 m, which comes to h² + 20h + 160.

  2. 8.2

    Set the figure to "Raise 2 · depositing tailings". What share of the centreline dam rests on the beach at that stage, and why can it hold in an earthquake when the upstream dam does not?

    Hint

    Look at where the brick-red lines are in each panel, and at which part of each dam holds up the crest.

  3. 8.3

    Why could the dam at Brumadinho fail with no earthquake, more than five years after its last raise?

    Hint

    Ask what the end of raising changed, and what it did not change, about the tailings inside the dam and the water in them.

  4. 8.4

    In the grain sandbox, run the experiment with loose, saturated sand, its drain shut and a house on top, then set Drain to Open and shake again. Which way of raising a tailings dam is each run like?

    Hint

    Which raise stands on loose, soaked grains that can't drain, and which stands on fill whose water can get out?

  5. 8.5

    An illustrative mine sends out 20,000 tonnes of dry tailings a day, which settle at 1.4 tonnes per cubic metre over 1.5 km². How fast does the deposit rise? Is that within the usual limit for an upstream dam, and what area would keep it to 2½ m a year?

    Hint

    Follow Example 8.3 line by line with the new numbers.

  6. 8.6

    Coquitlam Dam holds drinking water, not mine waste. Why did BC Hydro treat it like an upstream tailings dam?

    Hint

    How was it built, and what state does that leave the fill in?

Further reading

  • Grant-Stuart, D.J., for the South African National Committee on Large Dams (2020). Your Tailings Dam. Pretoria: SANCOLD, June 2020. ISBN 978-1-990981-73-9. https://sancold.org.za/wp-content/uploads/2024/05/SANCOLD-Your-Tailings-Dam-2020.pdf gentle A short, plain guide written for mine owners: how tailings are deposited (spigots, paddocks, cyclones), why the rate of rise limits an upstream dam, and what to watch. Free to download.
  • Robertson, P.K., de Melo, L., Williams, D.J., Wilson, G.W. (2019). Report of the Expert Panel on the Technical Causes of the Failure of Feijão Dam I. 12 December 2019. https://damfailures.org/sites/default/files/wp-pdf/IR_Feijao-Dam.pdf serious The expert panel's report on Brumadinho. Its executive summary is the clearest short account there is of how an upstream dam that had stopped growing could still liquefy, and its photographs of the construction history are worth the download.
  • Vick, S.G. (1990). Planning, Design, and Analysis of Tailings Dams. Vancouver: BiTech. serious The standard textbook on tailings dams: the raising methods, beaches, seepage and stability, by a member of the Mount Polley review panel.
Part III

Inside the embankment

By the end of Part III

Most of the world's large dams are heaps of earth and rock. Part III opens one up: why it is built from several soils, where the water inside it goes, and what happens when that water starts to carry the soil away.

  1. Chapter 9Every zone has one job

    Core, filters, transitions and shells, and how loose earth is made into dense fill.

  2. Chapter 10Where the water goes

    Seepage paths, flow nets drawn by hand, and a seepage lab you can run.

  3. Chapter 11When water carries soil

    Boils, pipes, the four kinds of internal erosion, and the filter that stops them.

Chapter 9 · Part III · Inside the embankment

Every zone has one job

Why build a dam out of six different soils?

The voids are the channels of flow for percolating water.

So far. Chapter 8 ended inside a tailings dam raised downstream, which can carry a core and drains up with it, like a dam built to hold water. Part III opens that kind of dam properly: the embankment, a heap of earth and rock, and the different soils it is built from.

After this chapter you can

  • name each zone of an embankment and its job, and say what goes wrong if it is missing;
  • explain why the upstream shell can be soaked and still strong, and what one event makes it dangerous;
  • say how loose earth is made into dense fill, and check a rolled layer with a field test.

This chapter adds a core and a filter inside the heap

2 figures · about 20 min to read · Film 9.1, 6 min 51 s

Between 1964 and 1967, in the canyon of the Peace River in northern British Columbia, crews built a hill one layer at a time. Each layer of its middle was spread no more than 10 inches thick (6 inches from late August 1965), scratched up on top so the next would bond to it, and rolled four times by a 90-ton compactor on rubber tyres. Each layer was meant to come out at least 98 per cent as dense as the same soil packed in a standard laboratory test. When they stopped, the W.A.C. Bennett Dam stood 183 m high and ran 2.0 km across the canyon and the terraces on either side, about 44 million cubic metres of earth and rock.BC Hydro expert panel (2012), pp. 1–3: 600 ft high, 6,700 ft long, 57.5 million cubic yards of fill, built 1964 to 1967; the rules for placing the core are in its table of zones, p. 3.Bennett is far from the tallest dam of its kind. Shuangjiangkou, in China, a rockfill dam built round a core of gravelly soil, is 315 m high. It began storing water in May 2025 and connected its first generator to the grid in June 2026 (International Water Power (2025); China Power News Network (2026)). It is reported as the world's tallest dam, a title it is expected to hold until Rogun, in Tajikistan, reaches its planned 335 m, which was projected for 2033.

From outside, a dam like this is one long slope of earth. Cut it across and it is seven materials. The panel of engineers that reviewed Bennett in 2012 lists them as zones, from the middle outward. Zone 1 is the core, there to hold the reservoir back. Zones 2 and 3, a transition and a filter, are each described as a filter to the core. Zone 4 is a drain. Zone 5 is a bedding layer for the riprap, the armour of large rocks on the upstream face. Zone 6, the random shell, "provides the weight for the stability of the dam". Each went in by its own rules: the core in the thinnest layers at a water content held close to one chosen value, the transition and filter in thicker layers under a vibrating roller, the shell with coarser streaks allowed because they help water out.BC Hydro expert panel (2012), the table of zones, pp. 3–4, quoted for Zone 6.

Keep an eye on Zone 2, the transition. It sits between the core and everything coarser, and of all the zones its job is the hardest to see in a drawing. Chapter 11 is where it earns its keep.

Why go to that trouble? The canyon had plenty of soil. Why not build the whole hill out of the tightest soil in it?

9.1 One soil, or several?

Chapter 3 gave half an answer: a short plug of clay in a pipe of sand takes nearly the whole drop in head. The film at the end of Chapter 7 asked the other half: how can a wall of dirt hold back a lake at all, when water passes through dirt? Most people say the wall must be made watertight. Build it of clay, the tightest soil there is, and the water stays out. Test that before you read on:

The same 24 m dam is built twice on solid rock: once all of silt, and once all of clay, a hundred times tighter. In the silt dam, water seeping through breaks out of the back slope well above the ground. Once the water has had time to soak into the clay dam, where does it stand inside it?

The film below runs that experiment, and then a third dam built of both soils.

Its first few minutes retell the race and Darcy's pipe from Chapter 3, as a reminder; then it looks inside a dam:

Film 9.1 · A wall made of dirt6 min 51 s · captions

Before you watch

The film's third dam puts a core of clay, about a fifth of the dam's width, inside the silt. Before you watch, decide two things. Does it leak more or less than the dam built all of clay? And which of the two would you rather live below?

The film was made for a series that had not yet come to filters. The job it promises to "come to later" is the subject of Chapter 11.

In this film

  1. The permeability race Fig. 3.3
  2. A straw in the beach: head Chapter 3
  3. A dam of one soil
  4. The same dam, all of clay the question above
  5. Darcy's pipe Fig. 3.2
  6. A plug of clay in the pipe Fig. 3.2
  7. The plug inside the dam Example 9.1
  8. Real zoned dams Fig. 9.1
Read the transcript

This is a dam made of dirt, holding back a lake twenty metres deep. Water soaks through dirt. So why is the back half of this dam dry? These clear tubes show how high the water stands inside it: high on the lake side, and nothing at all at the back. How can a heap of soil hold back a lake, when water passes through soil?

Start on a beach, at low tide. For hours after the sea has pulled back, little streams keep trickling out of the sand at the foot of the beach. That's water that soaked in at high tide, still finding its way out. This is the beach's soaked sand, with every gap between the grains full of water. And water moves through those gaps. The question is how fast.

Take a metre of each of four soils, and put the same push of water on top of each. Gravel has big gaps, and water gets through it in about thirty seconds. Sand takes about an hour. Silt, which is finer still, takes about a month and a half. Compacted clay takes about eleven years. These are typical values, and real soils vary, but the spread is enormous.

Water passes through gravel about ten million times as easily as through clay. How easily water passes through a soil is called its permeability. So here's the obvious idea: a dam must be watertight, so build it from the tightest soil there is. But clay isn't watertight, only slow. To judge a dam, we need to see what the water inside it is doing.

Push a clear straw down into the wet sand of the beach. Water climbs up the straw and settles at a level. That level is called the head. Water always flows from high head to low head. Up the beach, the water in the sand still stands high, but the sea has dropped. So the water flows down through the sand, and seeps out at the foot of the beach. Engineers sink thin tubes like this into dams, and read the levels in them. They're called piezometers.

Now look inside a dam. This one is twenty-four metres tall, drawn to scale, standing on solid rock. First, build it all from one ordinary soil, the silt from the race. Fill the lake, and the water soaks in and slowly finds its level inside the dam. The tube nearest the lake reads almost the lake's level.

The one in the middle reads lower, and the one at the back lower still. And the back slope weeps: water seeps out of its face, up to ten metres above the ground. That's the foot of the beach at low tide, on the side of a dam. Through each metre along the dam, about three buckets of water leak out every day. The obvious fix is to make it tighter.

So build the same dam all of clay, a hundred times tighter. The leak falls a hundredfold, to about a cup a day. Now let the water soak in. A clay dam takes far longer to get there, but it gets there in the end. Once it has, the water inside stands exactly where it stood before.

A hundred times tighter, and not one bit drier. Why?

In eighteen fifty-five, in a hospital courtyard in Dijon, an engineer named Henry Darcy ran water down a column of sand. The column stood taller than a person, and he measured the water pressure at its top and at its bottom. He had built the city's water supply, and he wanted to size the sand filters that clean it. He found that the flow grows with the drop in head, and shrinks with the distance the water has to travel.

Here's his test laid on its side: a pipe of sand between two tanks, with tubes along it. The levels in the tubes fall in a straight line, from one tank to the other. Swap the sand for gravel, or silt, or clay. The flow changes enormously, but the line doesn't move at all. Through any one material, the head falls in a straight line, however tight the material is. That's why the all-clay dam was no drier.

But what if the pipe isn't all one material? Fill the middle fifth of the pipe with clay, and leave sand on both sides. Now practically the whole drop happens inside the plug. The tubes on the left read the same as the left tank, and the tubes on the right read the same as the right tank, to within a tenth of a millimetre. The same flow has to pass through every part of the pipe. Sand lets it through with almost no push, but clay needs a steep drop to force it through. So the clay takes nearly all of it.

Now put the plug inside the dam. The dam becomes Darcy's pipe: sand, a plug of clay, then sand again. A core of clay runs up the middle, about a fifth of the dam's width, with sand on both sides. Of the twenty metres of head, the core takes all but half a metre. The back half of the dam stays dry. It does leak more than the all-clay dam, a little over a litre a day through each metre. That's about four times as much, and yet it's the better dam.

Here's why. Soil is strong because its grains press on each other, and water pressure in the gaps takes that press away. At the foot of the one-soil dam's back slope, the water carries about two fifths of the weight. Behind the clay core it carries almost none, so the grains there press hard, and grip. A dam doesn't have to be watertight. It has to keep the water pressure out of the part that holds it up.

Real earth dams are built this way. Here's a sixty-metre dam from the course, drawn to scale. Read it from the middle out: a tight core, then zones of coarser soil and rock toward the outside. The core holds the water back, and the coarse outer zones give the dam its weight and let any water that gets past run out freely. The sand on each side of the core has a job of its own, and we'll come to it later in the series. Most of the world's big dams are heaps of earth and rock like this.

Shuangjiangkou in China, set to become the tallest dam in the world, is built around a core of tight soil.

But every dam leaks, even this one. Ours lets a little over a litre a day through every metre. Where does that water go, and what does it do on the way?

After you watch: the answer

It leaks more, and it is the one to live below. A little over a litre a day passes each metre of the cored dam, about four times as much as the all-clay dam lets through, because its clay is only a fifth as wide. But the core takes all but half a metre of the twenty metres of head, so the back half of the dam is left with almost no water in it. At the foot of the one-soil dam's back slope the water carries about two fifths of the soil's weight; behind the core it carries almost none, and the grains there press on one another with nearly their whole weight. The film's own summary is the line to keep: "A dam doesn't have to be watertight. It has to keep the water pressure out of the part that holds it up." Example 9.1 measures the same thing in our dam. The film's last question, where the leak goes, opens Chapter 10.

The film puts its answer in one line: a dam does not have to be watertight; it has to keep the water pressure out of the part that holds it up. Every zone in this chapter serves that idea. Chapter 4 showed that soil holds only as well as its grains press on one another, and that water in the gaps takes part of that press away. The part of an embankment that holds it up against the lake is the downstream half. So the first question about any design is where the water will stand inside it, and only the second is how much of it will get through.

To put numbers on it, take the one dam this book keeps coming back to, our dam: the 24 m embankment with 2:1 faces (the notation of Example 1.2) drawn in Plate 0, whose water the seepage lab computes in Chapter 10 and whose slopes the workbench tests in Chapters 12 and 13. The lab can build it of a single fill or with a clay core between shells of sand and gravel, on a foundation of gravel or of clay. Here it is both ways, on clay:

Example 9.1

The plug inside our dam

Given
Our dam holding water 20.4 m deep, on the seepage lab's clay foundation, built two ways: all of one fill (k = 10⁻⁷ m/s), or with a clay core (10⁻⁹ m/s, 3 m wide at the top and 21.4 m at the base) between shells of sand and gravel (10⁻⁴ m/s). The lab computes where the water stands.
Find
How high the water stands in the downstream half each way, how much of the head the core takes, and how much each version leaks.
The idea
The same flow passes every zone in turn, so the tightest one takes nearly the whole drop in head, as the clay plug did in Darcy's pipe (Fig. 3.2).
  1. One fill: water breaks out of the downstream face 8 m above the groundthe highest point where water seeps out of the face, read from the lab
  2. With the core: the water behind it stands 0.65 m above the groundthe phreatic surface close behind the core
  3. Head the core takes: 20.4 − 0.65 = 19.75 mfrom the lake's level down to the water behind the core
  4. 19.75 ÷ 20.4 = 0.968, so 96.8%the core's share of the drop
  5. Leak: 44.4 L/day per metre with one fill, 37.2 with the corethe lab's Seepage readout, both ways

The core takes 96.8% of the head. The water in the downstream half drops from 8 m up the face to 0.65 m above the ground, while the leak falls by only about a sixth.

Check. Estimate the flow through the core by hand with Darcy's law. Ten metres up, the core is 13.4 m wide, so the water crossing it loses head at about 19.75 ÷ 13.4 = 1.47 metres per metre. At 10⁻⁹ m/s over a wet height of about 20 m, that is about 2.5 litres a day through each metre of dam. The lab with a full cutoff under the core, so that the core is the only way through, gives 2.1: the hand sum is within a fifth of it. The rest of the 37.2 goes under the core, through the clay foundation, where Chapter 10 follows it.

What it means. A core works by where it puts the water, far more than by how much water it stops. The leak changed by a sixth; the water level in the downstream half fell by more than seven metres, and with it the water pressure under the grains that hold the dam up.

9.2 Four jobs

A core alone does not make a dam, because the water it lets through has to go somewhere and the core has to be held in place. Every embankment with a core hands out the same four jobs to different soils.

Hold the water back. That is the core's job: compacted clay, or glacial till, the unsorted mix of clay, silt, sand and stones that a glacier leaves behind. Bennett's core is glacial soil of that kind, a non-plastic mix of silt, sand and gravel.BC Hydro expert panel (2012), p. 31 ("non-plastic silt sand gravel", and a glacial soil tested as similar to it) and p. 48 (the moraine materials used in the core). Where the ground beneath is gravel or cracked rock, the barrier has to go down into it too. A cutoff trench carries the core down through loose river deposits to rock, and under the trench a grout curtain, lines of holes drilled into the rock and pumped full of cement grout, closes the rock's cracks.Rock is tested for grouting by pumping water into a section of a drill hole. One Lugeon is an inflow of one litre a minute into each metre of hole at a pressure of 1,000 kPa, equivalent to a rock permeability of about 1.3 × 10⁻⁷ m/s (BC Hydro expert panel (2012), p. 5).

Let the water out. Whatever gets through the core must leave the dam without building up pressure in the part that holds it up. Drains of clean sand and gravel do that: a chimney drain standing against the downstream side of the core, a blanket along the base, and a toe drain that carries the water out to where it can be measured.

Stop the soil. But water leaving fine grains for coarse ones carries the fine grains with it, so a drain laid straight against a core would slowly wash the core away. Filters and transitions sit between them and step the grain size up gradually, each zone fine enough that the grains of the zone inside it cannot wash into its gaps, and coarse enough to let the water through. This is the thinnest zone in the dam, and its job is the subtlest; Chapter 11 is about it.

Add weight. The shells of rockfill or sand and gravel are most of the dam. They hold the core in place against the push of the reservoir and hold themselves up by friction, which is to say by the press of their grains on one another. Over the upstream shell goes riprap, against waves and ice; along the top, the crest carries a road and stands high enough above the highest flood to leave some freeboard.

Read a cored dam from the middle outward and you meet the jobs in the same order every time: tight in the centre, coarser and freer-draining toward the faces. Each zone is coarser than the one inside it, so the water can get out, and fine enough to hold that zone's grains in. Engineers number the zones in that order too. Zone 1 is the barrier, Zone 2 the filters and transitions under or beside it, Zone 3 the main rockfill, though each dam adapts the scheme: Bennett's run from 1 to 6, core outward.The convention, as Cooke and Sherard set it out and Cruz quotes it: Cruz (2008), p. 57.

Check yourself Rockfill drains freely and is strong. Why not pile it directly against the core?

Water leaving the core would carry its fine grains into the big gaps between the rocks, and the core would wash away from the inside. The filter and the transition step the grain size up so that no zone can wash into the next.

Below, a 60 m earth–rockfill dam is drawn to scale with all its zones in place, beside three other kinds of dam. Click a zone, or its chip, to read what it is made of, how easily water passes through it, its job, and what happens if it fails. Build shows the order in which the dam goes up, and Explode pulls the zones apart:

Fig. 9.1 · Zoned earth–rockfill dam with central core Drawn to scale · 60 m high

Drawing scrolls sideways →

An earth–rockfill dam is tight in the middle and coarser toward its faces, and each zone has its own material and its own job. The section opens finished and drawn to scale; each hatch is a material, as in Plate 0, and each code a zone: 1 the core, 2A and 2B the filters and transitions either side of it, 3U and 3D the upstream and downstream shells, 4 the riprap and CR the crest; beneath them the cutoff trench (1T) cuts through the river deposits (AL) to the rock (BR), with the grout curtain (GC) below it, and behind them lie the filter blanket (BF) and the toe drain (TD). The line labelled Phreatic surface is the top of the water inside the dam: it holds the reservoir's level through the upstream shell, falls through the core and runs out low along the base. Build replays the construction stage by stage. The other tabs show a concrete-face rockfill dam, a concrete gravity dam and a tailings dam raised on its centreline, each to its own scale. The number to remember: the core passes water about a million times less easily than the shells beside it, about 10⁻⁹ m/s against 10⁻³ m/s or more.

  • On the Earth–rockfill tab, press Build. Which zones rise together, in the same stage?
  • Press Explode and count the zones between the core and the downstream shell.
  • Open the Concrete-face rockfill tab. Where is its barrier, and where does its phreatic surface run?
  • Open the Tailings tab and click CS. What is this dam built of?
Answers

The core, both filters, both transitions, both shells, the blanket and the toe drain rise together in compacted layers: none of them can be added afterwards. Two zones lie between the core and the downstream shell, the fine filter 2A and the transition 2B. The concrete-face dam's barrier is the concrete slab on its upstream face, and its phreatic surface stays at the base. The tailings dam is built of cycloned sand, the coarse part of the tailings themselves (Chapter 8).

Notice that the phreatic surface stays at the reservoir's level all the way through the upstream shell and only drops inside the core: the upstream shell is soaked. Notice too how thin the filters and transitions are, a few metres wide in a dam more than 250 m across at its base, and that Build places them in the same stage as the core, layer by layer, since they could never be slipped in later. The cards are easier to compare side by side, so here they are for the earth–rockfill section:

Table 9.1 · The zones of an earth–rockfill dam, from the middle outward
ZoneMade ofPermeability, m/sJobIf it fails
1 CoreClay or glacial till, compacted slightly wet of its optimum water contentabout 10⁻⁹The barrier: almost all of the head is lost across itA crack or a wet seam opens a leak; the filters decide whether a leak becomes a failure
1T Cutoff trenchCore soil compacted in a trench dug through the river deposits to rockabout 10⁻⁹Carries the barrier down to rockA badly cleaned rock surface lets core soil wash into the rock's cracks
GC Grout curtainCement grout pumped under pressure into rows of drill holesbrings rock toward 10⁻⁷Seals cracks in the rock beneath the coreGaps between holes leave windows for water
2A Fine filterProcessed clean sand, graded to suit the coreabout 10⁻⁵Lets water out of the core and keeps every core grain in; downstream, it is also the chimney drainA cracked core with no filter behind it can erode and fail in hours
2B TransitionSandy gravel or crushed rock, a few metres wideabout 10⁻³Steps the grain size up from filter sand to rockfillThe filter sand washes into the rockfill and the core is left unprotected
3U Upstream shellRockfill or sand and gravel, in layers 1–2 m thick10⁻¹ to 10⁻³Weight; soaked by the reservoir, it holds the core in placeIf it drains slowly when the reservoir is lowered fast, the water left in it can slide the upstream slope
3D Downstream shellRockfill, in layers 1–2 m thick10⁻¹ to 10⁻³Weight; kept drained, so it is strong and can stand steeperIf the drains clog, it fills with water and loses much of its margin against sliding
BF Filter blanketSand filter spread over the river deposits under the downstream shellabout 10⁻⁵Lets water rising from the foundation into the shell without its sandRising water carries foundation sand into the rockfill and undermines the toe
TD Toe drainCoarse gravel round a perforated pipe, wrapped in filterdrains freelyCollects the seepage and leads it to a measuring weirFlow that rises with no rise in the reservoir, or turns cloudy, warns that soil is moving
4 RiprapLarge, durable rock, 0.5–1.5 m, on a gravel bedding layerdrains freelyArmours the upstream face against waves and iceWaves pluck out the shell beneath and steepen the face
CR CrestA road on a capping layer that protects the top of the core–Access, freeboard above the highest flood, and extra height for settlementToo little freeboard invites overtopping

Condensed from the zone cards of Fig. 9.1; the full cards also cover the river deposits and the rock beneath. Permeabilities are typical values for each material, not measurements of one dam.

Tailings dams can be zoned the same way. Mount Polley's, in British Columbia, was: the review panel's section through it shows a core of till, a filter and a transition downstream of the core, rockfill supporting them, and a zone of tailings or rockfill on the pond side. The panel found the filter thin and much of it more poorly graded than a filter should be, and still found no evidence that the breach was caused by water eroding the core.Mount Polley Review Panel (2015), pp. 7–8 (the zones and their duties) and pp. 11–12 (the filter, and the finding on erosion of the core). Its zones are not where its answer lies.

9.3 Two shells, one soaked

The two shells look alike in a drawing, and they lead different lives. The downstream shell is the part the core protects. Behind our dam's core the water stood 0.65 m deep, so above that the gaps in the downstream shell hold air, the water pressure in them is zero, and the grains press on one another with their whole weight.

The upstream shell gets no such protection. It sits under the reservoir, and every gap in it is full of water at whatever pressure the lake's depth puts there. One could assume that a soaked shell must be a weak one, since water in the gaps takes part of the grains' press away. It does take part of it, and the grains grip one another less for that; but as long as the water is still, the shell is no less safe. Buoyancy lightens each grain's press on the grains below it, and it lightens the pull that tries to slide the grains down the slope by exactly the same share. Chapter 4 showed the result for a heap of sand: under still water it stands as steeply as it does dry. Here are the numbers for the shells of Fig. 9.1, made of the workbench's rockfill:

Example 9.2

Soaked, and still the safer face

Given
Rockfill weighing 22 kN/m³ soaked and 20 kN/m³ moist, friction angle φ′ = 42°, no cohesion; water 9.81 kN/m³. The upstream face slopes at 2.2:1, the downstream face at 1.9:1. The upstream shell lies under still water; the downstream one is dry.
Find
How hard the grains press 20 m below the surface of each shell, and the factor of safety of each face treated as a long, straight slope; then the upstream face's if the water in it were trapped and seeping out.
The idea
Under still water the grains press with their buoyant weight, soaked weight less water. For a long slope of cohesionless grains, FS = tan φ′ ÷ tan β, the same dry or under still water; with water seeping out parallel to the face it falls in the ratio of buoyant to soaked weight (Chapter 4).
  1. Buoyant weight: 22 − 9.81 = 12.19 kN/m³each cubic metre's weight less the water it pushes aside
  2. Upstream, 20 m down: 12.19 × 20 = 243.8 kPathe press of the grains on the grains below
  3. Downstream, 20 m down: 20 × 20 = 400 kPano water in the gaps, so the whole weight presses
  4. 243.8 ÷ 400 = 0.61the soaked grains press 61% as hard
  5. Upstream face: tan 42° × 2.2 = 0.9004 × 2.2 = 1.98tan β = 1 ÷ 2.2; the still water lightens the press and the pull alike
  6. Downstream face: 0.9004 × 1.9 = 1.71steeper, at tan β = 1 ÷ 1.9
  7. Upstream, water trapped: 1.98 × 12.19 ÷ 22 = 1.10water seeping out through the face, as at the foot of a beach at low tide

Soaked, the upstream shell's grains press only 0.61 times as hard as the dry shell's, and yet its face is the safer of the two, 1.98 against 1.71, as long as its water is still. Trap the water and the upstream face falls to 1.10.

Check. The two faces are the same rockfill at different slopes, so their factors should differ only by the ratio of the slopes: 1.98 × 1.9 ÷ 2.2 = 1.71. And at a 1:1 face, 45°, the dry rockfill would have FS = tan 42° = 0.90: no heap of this rockfill stands that steep.

What it means. A soaked shell is safe while its water can come and go with the lake. Rockfill, at 10⁻¹ to 10⁻³ m/s, drains as fast as any reservoir can be lowered; a shell of fine fill would not, and its face would fall toward the trapped case. These are long, straight slopes; real slips are curved and run deeper, and Chapter 12 checks them on our dam.

That trapped case has a name. Lowering a reservoir faster than its water can leave the upstream shell is rapid drawdown: the water that pressed on the face from outside is gone, the water inside is still there, and it pushes outward through the slope. It is the one event that turns a soaked shell from safe to dangerous. It is why designers prize an upstream shell that drains freely, and one reason the earth–rockfill section's upstream face is the flatter of the two.

9.4 Where the rule breaks

"Tight in the middle, coarse outside" is the rule for a dam with a core. The other three tabs of Fig. 9.1 break it, each for a reason.

The concrete-face rockfill dam has no core. Its barrier is a slab of reinforced concrete laid over the upstream face, joined to the rock around its edges by a concrete plinth, with the grout curtain drilled from the plinth. Behind the slab the whole dam is rockfill, rolled hard so that it settles little and the slab does not crack. Its finest zones are directly under the slab, not in the middle: a cushion of crushed rock and sand gives the slab a firm, smooth bed and limits the leak if a crack opens. Water that passes the slab drops through the rockfill to the base and runs out, and the rockfill needs no filter against the slab because it has no fine grains to lose. The leaks it can take are large. At two big dams of this kind in Brazil, Barra Grande and Campos Novos, leaks of 1,000 to 3,000 litres a second were measured with no sign of trouble in the downstream slope.Cruz (2008), pp. 60–61, reviewing Cooke & Sherard (1987).

The concrete gravity dam has zones only in a looser sense: a richer, frost-resistant concrete on its faces, a lean mass concrete inside so that the heat of setting stays low, and drains and a gallery near the upstream face. None of them is there to stop water carrying grains away, because concrete has no loose grains. Its danger, as Chapter 7 showed, is the water underneath it.

The tailings dam is built of its own waste. In the centreline dam on the last tab, sand split from the tailings by cyclones is spread and compacted on the downstream slope, a chimney drain keeps that sand from filling with water, and the finest waste settles out by the pond, on the upstream side, as far from the slope as it can be (Chapter 8 showed how such a dam rises). The materials are the mine's, and the aim is the cored dam's: keep the water away from the part that holds the dam up.

So "tight in the middle" is one way of doing something every dam on the four tabs does. Each decides where the water will stand, and keeps it out of the part that carries the load.

Check yourself A crack opens in a concrete-face rockfill dam's slab and the leak through the dam rises tenfold. Why would its engineers worry less than about a crack in a clay core?

The water that passes the slab falls into rockfill with no fine grains to lose, and drains out at the base without raising the water pressure in the dam. Water through a crack in a core can carry the core itself away, which is why a core needs filters and a slab does not.

9.5 How a heap becomes fill

So far every zone has been described by what it is made of. How it is put in place matters as much. Soil dug from a borrow pit and tipped into a heap is loose, the loose sand of Chapter 5: wide gaps, ready to settle when it is wetted or shaken, and inclined to shrink when it is sheared, handing its load to the water in its gaps. A dam of loose soil would settle unevenly and crack, and a soaked shell of it could lose its grip altogether. So every zone goes in as thin layers, each squeezed dense by rollers before the next is spread. That squeezing is compaction, and it is why Bennett's core went in no more than 10 inches at a time.

How dense can a soil be made? In 1933 R. R. Proctor, a field engineer with the water department of Los Angeles, published a way to find out.Proctor's first article (Proctor (1933)) rested on laboratory tests of more than 200 soils, checked in the field on four dams his bureau had built in the two years before. He explained the dry side much as this chapter does below: in a very dry soil, films of water where the grains touch draw them firmly together, and a little more water releases that pull. Past that point he wrote of the water as having "an effect similar to that of lubrication". Pack a sample of the soil into a mould with a set number of blows from a standard hammer. Weigh it, dry it in an oven, and weigh it again. Then add a little water to the soil and repeat. A cubic metre of the packed sample weighs its bulk unit weight γ, grains and water together; its water content w is the weight of the water as a fraction of the weight of the grains. Take the water out and what is left is the weight of grain in each cubic metre, the dry unit weight:

γd = γ / (1 + w)(9.1)

γd dry unit weight, the grains in each cubic metre (kN/m³) · γ bulk unit weight, grains and water (kN/m³) · w water content, weight of water ÷ weight of grains

The more grain packed into each cubic metre, the denser the fill. Plotted against the water content, Proctor's dry unit weights climb, reach a peak and fall again, here for an illustrative soil:

Fig. 9.2 · Packing the same soil with more and more waterIllustrative soil · grains of specific gravity 2.65
68101214161820 17181920 water content, w (%) dry unit weight (kN/m³) zero air voids: every gap full of water 98% of the peak optimum: 19.0 kN/m³ at 12% water field test: 18.4 dry of optimum wet of optimum

Rolled with the same effort, a soil packs densest at one water content, its optimum; drier or wetter, it ends up looser. The solid curve is a compaction test of one illustrative soil: each point is a sample packed the same way at a different water content, plotted by the weight of grain in each cubic metre. The dashed line falling across the top is zero air voids, the weight the soil would reach with every gap full of water and no air at all; no compacted sample can sit above it. The dotted level line is 98 per cent of the peak, the share Bennett's specification asked of its core. The filled dot is the peak, with a thin dashed drop line to its water content, and the open circle is the field test of Example 9.3. The number to remember: the peak, 19.0 kN/m³ at 12% water.

  • Read the curve at 10% and at 14% water. Which side of the optimum loses more for two points of water?
  • Divide the field test's 18.393 by the peak's 19.0. Does the open circle pass the dotted line?
Answers

18.70 at 10% and 18.60 at 14%: on this soil the wet side drops a little faster. 18.393 ÷ 19.0 = 0.968, short of 0.98, so the circle sits below the line and the layer fails a 98% specification.

Notice that the curve has a single peak and falls away on both sides. On the dry side, the castle's grip from Chapter 1 works against the roller: rings of water where the grains touch bind them into crumbs that resist being pushed into one another's gaps, and a little more water weakens the rings and lets the grains pack closer. On the wet side, water fills nearly every gap, and in the moment a roller takes to pass it cannot get out: the water carries the roller's press, as it carried the load in the undrained tests of Chapter 5, and the grains are pushed no closer. The peak is the optimum water content, and the peak's weight is the most this effort can achieve with this soil. A heavier roller, or more blows, gives a higher peak at a little less water.

A specification asks for a share of that peak. Its relative compaction is the dry unit weight reached on site divided by the laboratory maximum:

RC = γd / γd,max(9.2)

RC relative compaction, usually given as a percentage · γd,max the peak of the laboratory curve for the same soil (kN/m³)

Bennett's core had to reach 98 per cent, at a water content between 2 per cent below the optimum and 1 per cent above it.BC Hydro expert panel (2012), the table of zones, p. 3 (the specification) and p. 9 (permeability against the water content at placing). The water content matters for more than density. The panel found that the same core soil, placed drier, with less of its gaps filled with water, let water through 60 to 140 times more easily once soaked than when it was placed wetter. The core's card in Fig. 9.1 gives a second reason to keep a core near its optimum or a little wet of it: there the soil bends as the dam settles, where a drier, stiffer one would crack.

On site an inspector checks each layer: dig a small hole, weigh what comes out, measure the hole's volume (by filling it with sand of known density, for instance), and dry a sample for its water content. Those two numbers give γ and w, and with them the gaps. A cubic metre of solid grain would weigh Gs γw, where Gs, about 2.65 for quartz sand, is how many times denser the grain is than water (Chapter 5). So the volume that one unit of grain takes up, gaps included, is Gs γw / γd, and the gaps alone, the void ratio, are that less one:

e = Gs · γw / γd − 1(9.3)

e void ratio, volume of gaps ÷ volume of grains · Gs specific gravity of the grains · γw unit weight of water, 9.81 kN/m³

Example 9.3

Does the layer pass?

Given
An illustrative field test on a freshly rolled layer: bulk unit weight 20.6 kN/m³, water content 12%. The laboratory peak for this soil is 19.0 kN/m³, and its grains have Gs = 2.65. The specification asks for 98%, as Bennett's did.
Find
The dry unit weight, the relative compaction, the void ratio and the gap fraction; and whether the layer passes.
The idea
Take out the water's weight (Eq. 9.1), compare with the laboratory peak (Eq. 9.2), then count the gaps from the room the grains take (Eq. 9.3).
  1. γd = 20.6 ÷ 1.12 = 18.393 kN/m³1 + w = 1.12: each kilonewton of grain carries 0.12 kN of water
  2. RC = 18.393 ÷ 19.0 = 0.968, so 96.8%the share of the laboratory's best
  3. 2.65 × 9.81 ÷ 18.393 = 1.413room taken by one unit of grain, gaps included
  4. e = 1.413 − 1 = 0.413the gaps alone
  5. 0.413 ÷ 1.413 = 0.29gaps as a share of the whole volume

The layer reaches 96.8% of the laboratory peak, with a void ratio of 0.413. It is dense, and it fails a 98% specification: it falls short of what Bennett's core was meant to reach.

Check. The water must fit in the gaps. The share of the gaps it fills is w × Gs ÷ e = 0.12 × 2.65 ÷ 0.413 = 0.77, less than 1, so air fills the rest. The open circle in Fig. 9.2 must then sit below the zero-air-voids line at 12%, which is 2.65 × 9.81 ÷ 1.318 = 19.72; at 18.393, it does.

What it means. Gaps that fill 0.29 of the volume are fewer than the third of the volume that Chapter 3 allowed soaked soil. Sand packed this dense swells when it is sheared and draws water in (Chapter 5), so it holds. Compaction turns a loose heap that would hand its load to the water into fill that keeps it on the grains.

Check yourself A layer tests at 95% and well wet of the optimum. The contractor offers to roll it a few more times. Will that help much?

Not much. Wet of the optimum the gaps are nearly full of water, which cannot get out while a roller passes, so the water takes the extra press and the grains move little closer. Drying the soil toward the optimum, or digging the layer out, does more.

Each zone, then, is a soil chosen for one job, packed by rules written for that job, and set where its designers expected the water to be. The core expects the water to stop at it, the filter expects to meet whatever gets through, and the drain expects to carry it off.

What to remember

  • A dam need not be watertight. It has to keep the water pressure out of the part that holds it up, so that the grains there press on one another with nearly their whole weight. In our dam a clay core takes 96.8% of the head and the leak barely changes.
  • Four jobs, middle outward: hold the water back (core, cutoff, grout curtain), let it out (drains), stop the soil moving with it (filters and transitions), and add weight (shells). Each zone is coarser than the one inside it.
  • The downstream shell is kept nearly dry. The upstream shell is soaked and still strong, because still water lightens the grains' press and the pull on them alike, and rockfill drains as fast as the lake falls. Rapid drawdown of a shell that drains slowly is its danger.
  • "Tight in the middle" is the cored dam's way. A concrete-face rockfill dam puts its barrier on the face, a gravity dam is its own barrier, and a tailings dam builds itself from its waste, all to keep water away from what carries the load.
  • Compaction packs loose fill dense, so that it keeps its load on its grains. Each soil packs densest at its optimum water content, and relative compaction measures a layer against that laboratory best.

Questions

  1. 9.1

    Why can the upstream shell be soaked and still strong, and what one event makes it dangerous?

    Hint

    Compare what still water does to the press of a grain on the grains below it and to the pull that tries to slide it downhill. Then ask what happens if the water cannot leave.

  2. 9.2

    A concrete-face rockfill dam has no core. Where is its barrier, and why does its main rockfill need no filter against the slab?

    Hint

    Ask what a filter is for, and whether the rockfill has anything a leak could carry away.

  3. 9.3

    Press Build on each tab of Fig. 9.1. Which dam puts its watertight element in only after the body of the dam is built, and what does that mean for inspecting it?

    Hint

    Watch the stage in which the barrier appears, and whether anything is built over it afterwards.

  4. 9.4

    The earth–rockfill dam has filters on both sides of its core. Why does the downstream one matter more?

    Hint

    Which way does the water cross the core, most of the time?

  5. 9.5

    An illustrative field test on a rolled layer gives a bulk unit weight of 21.0 kN/m³ at 14% water. The laboratory peak for the soil is 18.9 kN/m³. What is the relative compaction, and does the layer pass a 98% specification?

    Hint

    Take the water out first with Eq. 9.1, then divide by the peak.

  6. 9.6

    Bennett's core had to reach 98% of the laboratory peak and be placed between 2% below and 1% above its optimum water content (BC Hydro expert panel (2012), p. 3). Why specify the water content as well as the density?

    Hint

    Two layers with the same density can still differ in how they behave once the reservoir soaks them.

Further reading

  • U.S. Bureau of Reclamation (1987). Design of Small Dams, 3rd ed. Denver. https://www.usbr.gov/tsc/techreferences/mands/mands-pdfs/SmallDams.pdf serious Its chapter on earthfill dams explains how a section is zoned and why, with many drawn examples; the next chapter does the same for rockfill dams, and an appendix covers how embankments are built. Free from the Bureau of Reclamation.
  • Hoeg, K., Fell, R., Bridle, R. (2012). WAC Bennett Dam: Expert Engineering Panel Report, Volume 1. BC Hydro, Report No. N3405, 13 August 2012. https://www.bchydro.com/content/dam/BCHydro/customer-portal/documents/projects/gm-shrum/EEP%20report_V1.pdf primary source Its table of zones lists the rules each was placed by, and its section on the filter system judges how well the transition, filter and drain do their job. Chapter 11 comes back to this dam.
  • Proctor, R.R. (1933). Fundamental principles of soil compaction. Engineering News-Record 111(9), 31 August 1933, beginning p. 245. primary source The first of four short articles in which Proctor set out the test later named after him, with the first published compaction curves. Readable in an afternoon.

Chapter 10 · Part III · Inside the embankment

Where the water goes

If every dam leaks, what makes a leak safe?

So far. Chapter 9 built our dam, a 24 m embankment with slopes of 2:1 (two across for every one up, as in Example 1.2), and in its zoned form a clay core between free-draining shells. Every zone went where it did on a guess about where the water would go. This chapter follows the water itself: where it gets in, which way it goes, and where it comes out.

After this chapter you can

  • trace the five ways water gets past a dam, and read a flow net to work out how much flows and how steeply the head falls where it leaves;
  • estimate a dam's leak by hand with Darcy's law, and check the estimate against the seepage lab's solution;
  • explain why a drain makes a dam leak more and yet makes it safer, and choose a defence from a picture of the seepage.

This chapter adds two flow lines and three equipotentials

4 figures · about 25 min to read

Walk along the downstream toe of an earth dam and sooner or later you come to a small concrete box set into the ground. Across one end stands a steel plate with a V cut into its top edge, and through the V slides a thin, clear sheet of water. It drops a few centimetres into a channel and runs off toward the river. On a regular round someone comes by, reads how deep the water stands behind the plate, and writes it down. That depth says how many litres a second are coming out.

The water in the box is the dam's leak, its seepage. It went into the ground on the floor of the reservoir, or soaked into the dam's upstream face, and crept through to this side. Nobody is alarmed by it. The US Army Corps of Engineers' manual on seepage states it as a fact before it says anything else: "All earth and rock-fill dams are subject to seepage through the embankment, foundation, and abutments."USACE EM 1110-2-1901, para. 1-5, and again in para. 8-1; para. 13-3 describes the V-notch weirs. Every dam leaks. The useful question is what makes a leak safe, and here is a first test of your feel for leaks:

Our dam stands on 15 m of sand and gravel, with nothing under it to stop water passing beneath, and it is built of one kind of earth fill. Build it again with a narrow core of clay, a hundred times tighter than that fill, between shells of free-draining sand and gravel. The ground underneath stays as it was. What happens to the leak?

To decide, follow the water: where it enters, which way it goes and how far, and where it comes out. That takes a map of the ways past a dam, a way of drawing the water's paths, and in the end a solver.

10.1 Five ways past a dam

You met the rule in Chapter 3: water flows from high head to low, as fast as the soil lets it, and Darcy's law, Eq. 3.2, says how fast. The race in Fig. 3.3 showed how wide the range of permeability is: gravel, at 10⁻² m/s, passes water ten million times as easily as compacted clay, at 10⁻⁹ m/s. Zoning depends on that spread. A clay core much narrower than the dam can take almost the whole drop in head while the free-draining shells beside it take almost none.

A dam is more than one pipe, though. Water has five main ways past it, and the drawing numbers them:

Fig. 10.1 · The five main seepage paths Schematic

Drawing scrolls sideways →

Water gets past a dam five ways, and only two of them pass through the embankment: through its fill, and along a conduit buried in it. The section cuts across the valley; the plan beside it looks down on the dam from above. Dotted blue lines with moving dots are seepage paths, numbered in blue circles. 1 Through the embankment: slowed by the clay core, caught by the chimney filter and led along the base to the toe drain. 2 Under it: through the alluvium, the sand and gravel a river leaves on its valley floor, rising at the toe, where it can boil sand. 3 Through rock: along open cracks (joints) below the grout curtain, a screen of cement grout injected into the rock. 4 Along the conduit: soil packed around a pipe or culvert is hard to compact well, and a gap along it is a classic concentrated leak. 5 Around the abutments: through the valley sides the dam is built against, best seen in plan. The number to remember: three of the five paths never enter the embankment.

  • Follow path 2 from the reservoir floor to the sand boil. Which parts of the dam does it pass through?
Answers

None. It stays in the alluvium the whole way, dipping under the short cutoff below the core, and comes up beyond the toe.

Notice that the core and its cutoff, which take most of the drawing's care, stand in the way of path 1 and of path 2 only as far down as the cutoff reaches. Paths 2, 3 and 5 run through ground that was there before the dam. Its designers chose neither its permeability nor its layering; they could only find them out, and the water uses whatever they missed.

10.2 Drawing the water: the flow net

To say how much water takes path 2, and how hard it pushes where it comes out, you need the head everywhere in the ground. Fig. 3.2 found it for a pipe, where the water has one way to go and the head falls along a straight line. Under a dam the water spreads out and bends. The rule in every small piece of ground is still the same, though: in steady flow nothing piles up, so each piece passes on exactly the water it receives, and it passes it toward lower head by Darcy's law.In symbols the two facts are one equation, ∇·(k∇h) = 0. Read it from the inside out: ∇h is the slope of the head, k∇h the flow it drives by Darcy's law, and ∇· (the divergence) asks how much more flows out of a tiny box than flows in. Zero means nothing piles up and nothing runs dry. In uniform soil k comes outside and it becomes Laplace's equation, the same one that governs the steady flow of heat through a metal plate. Those two facts fix the head everywhere once you know it along the edges: the reservoir floor, the ground beyond the toe, the underside of the structure.

For most of the twentieth century engineers drew the answer by hand, as a flow net: two families of lines. Flow lines are the paths the water follows. Equipotentials join points of equal total head: push one of the straws of Fig. 3.1 into the ground anywhere along one, and the water in it rises to the same level. In soil that passes water equally well in every direction, the water crosses every equipotential at right angles, because straight across them is the way the head falls fastest.Philipp Forchheimer, an Austrian engineer, recognized that steady groundwater flow obeys Laplace's equation and drew some of the first flow nets (Houben et al. (2025)). Arthur Casagrande, at Harvard, made the net a working tool for dam designers in a paper of 1937, with rules for beginners that still hold: start with four or five flow channels, and adjust the whole net before any of its details (Casagrande (1937)).

What makes a net useful is a choice of spacing. Space the flow lines so that each neighbouring pair carries the same share of the flow; the strip between them is a flow channel. Space the equipotentials so that the head falls by the same amount, Δh, between each pair; that is a head drop. Now look at one cell of the net, b wide across the flow and l long along it. By Darcy's law the flow through it, per metre of dam, is k × (Δh/l) × b. Draw the net so that every cell is a square, b = l, and the flow through each cell is k·Δh, whatever its size. With Nd drops sharing the total head H, Δh = H/Nd, and Nf channels each carrying k·Δh give the whole flow:

q = k · H · Nf / Nd(10.1)

q flow per metre length of dam (m³/s per m) · k permeability (m/s) · H head lost between where the water enters and where it leaves (m) · Nf number of flow channels · Nd number of head drops

The ratio Nf/Nd depends only on the shape of the ground and the structure. The rules for sketching a net are short: lines cross at right angles, every cell looks square, an edge that passes no water (a concrete base, a rock surface) is itself a flow line, and ground under still water is an equipotential. Getting every cell square at once takes practice and an eraser. Here is a finished net under a small concrete weir (a low dam across a river, not the measuring weir of the toe), an illustrative example sunk into a bed of sand:

Fig. 10.2 · A flow net under a weir on sandIllustrative · drawn to scale

flow line equipotential the smallest square, at the exit

1 2 3 4 5 6 7 8 9 10 11 12 H = 10 m upstream water downstream water, level with the sand weir 1.4 m smallest square flow line sand, k = 10⁻⁴ m/s rock: no water passes

Under a weir on sand, four flow channels and twelve head drops divide the ground into curved squares, smallest where the water leaves beside the weir. The weir is 13 m wide and sunk 2 m into sand 10 m deep, on rock that passes no water. Upstream the water stands 10 m above the sand; downstream it is level with the sand. The solid lines are flow lines, the paths the water takes; the dashed lines are equipotentials, each a line of equal total head. The weir's underside and the rock are flow lines too, and the sand surfaces on either side are the first and last equipotentials. The twelve drops are numbered along the bottom channel, from upstream. The shaded cell is the smallest, where the water turns upward and leaves beside the weir's downstream face. This net was computed rather than sketched, so it shows what a careful hand sketch aims for. The number to remember: Nf/Nd = 4/12, a third.

  • Count the drops a drop of water crosses in the channel nearest the weir, then in the bottom channel. Why must they be the same?
  • Lay a ruler across a few cells: the distance across each one, between its flow lines, should be about the distance along it, between its equipotentials.
Answers

Twelve in each. Every flow line starts on the upstream sand, at the full head, and ends on the downstream sand, at none, so it crosses every equipotential once, however short or long its path.

Notice that the squares are big where the water has room, out under the reservoir and deep under the weir, and small where it is squeezed, around the weir's buried corners. A small square means the same drop in head over a short distance: a steep gradient, and fast water. Crowding in a flow net is a picture of the gradient. Notice too where the smallest square is: at the exit, where the water turns upward beside the weir and leaves the sand. That is where the water pushes hardest on the grains it is leaving.

Example 10.1

Reading the weir's net

Given
The net of Fig. 10.2 (illustrative): Nf = 4 channels, Nd = 12 drops, head H = 10 m, sand with k = 10⁻⁴ m/s. The smallest square, at the exit, is 1.4 m deep along the weir's downstream face.
Find
The flow under each metre of weir, and the gradient where the water leaves.
The idea
Every square passes k·Δh, so the flow is a count (Eq. 10.1). In the last square the water loses one drop of head over the square's length.
  1. Δh = 10 ÷ 12 = 0.833 mthe head lost in each drop
  2. Each channel: 10⁻⁴ × 0.833 = 8.3 × 10⁻⁵ m³/sa square passes permeability times its drop in head
  3. q = 4 × 8.33 × 10⁻⁵ = 3.3 × 10⁻⁴ m³/sfour channels, each carrying the same
  4. In a day: 0.000333 × 86,400 = 28.8 m³86,400 seconds in a day
  5. i at the exit = 0.833 ÷ 1.4 = 0.595the last drop, spent over the depth of the last square

About 29 m³ of water a day passes under each metre of the weir, and it leaves the sand beside the weir at a gradient of 0.60.

Check. Make the sand ten times as permeable. The squares are the same squares, because nothing in the drawing knows about k, so the flow rises tenfold, to 3.3 × 10⁻³ m³/s, and the exit gradient stays 0.60. A gradient is a drop in head over a distance, and both come from the drawing alone.

What it means. The exit gradient of 0.60 is below the 1 at which upward-flowing water floats sand (Chapter 4), but not by much, and it sits exactly where the net is most crowded. The weir's weak spot is a square metre or two of sand beside its downstream face.

The Check holds a lesson for every seepage calculation. The permeability of a natural sand is rarely known to better than a factor of a few, so the flow from a net is only as good as that guess; the pattern of heads, and with it every gradient and every water pressure, comes from the geometry, which the engineer can draw. That is why engineers sketch nets carefully where they want pressures and roughly where they want quantities."For calculation of seepage quantity only a crude flow net is required. Accurate flow nets are required to determine pressure distribution": USACE EM 1110-2-1901, para. 4-6. Real ground adds a complication the drawing leaves out: soil laid down in layers passes water more easily along its layers than across them, a property called anisotropy.Casagrande warned that where layering is visible the permeability along the layers "may easily be ten times greater" than across them, and much more in distinctly layered soils; even a rolled earth dam, compacted layer by layer, keeps some of it (Casagrande (1937), pp. 151–152). Engineers draw the net on a section stretched by the square root of the ratio, then squash it back (USACE EM 1110-2-1901, para. 4-7). A net drawn for uniform soil is the start of the analysis, not the end.

Check yourself The water upstream of the weir rises from 10 m to 15 m above the sand. Which lines of the net move?

None of them. The net is fixed by the shape of the weir and the sand. Every drop in head, the flow and the exit gradient grow by half again, in proportion to H.

10.3 Seepage lab

A net through a whole earth dam is harder to sketch than one under a weir, because the top of the wet part of the dam is unknown until the net is finished. So the page computes it instead. Everything in the next drawing is calculated. The page divides our 24 m embankment and the 15 m of ground beneath it into 1 m cells, applies Darcy's law between every pair of neighbours, and solves for the head in all of them at once (the steady seepage equation, ∇·(k ∇h) = 0). Along the way it finds the phreatic surface, the water table inside the dam, using the same saturated–unsaturated approach as commercial seepage software. Lines of equal head are drawn every 10% of the reservoir head. Flow lines are drawn in two sets: four start on the dam's upstream face and five on the reservoir floor. Within each set, each neighbouring pair carries the same flow, as in a hand-drawn flow net. Between the sets it differs: on the starting settings a gap between two foundation lines carries about 8 times the flow of a gap between two dam lines, so compare line spacing only within a set. The lab's lines are therefore no square net, and Eq. 10.1 can't count them: read the heads from its equipotentials and the flow from its readout.

It opens on our dam built of one fill, 10⁻⁷ m/s, standing on sand and gravel a hundred times more permeable, 10⁻⁵ m/s, with no drain and no cutoff, and the reservoir 20.4 m deep:

Fig. 10.3 · Seepage lab · finite-volume solution Drawn to scale · 24 m embankment, 2H:1V slopes

Drawing scrolls sideways →

Dam
Drain
Cutoff
Foundation
20.4 m
Seepage–per metre of dam
Leaves via–
Max exit gradient–unfiltered exits only
Safety vs piping–ic ≈ 1.0
Phreatic surface–

The solver finds the head in every cell of our dam and the ground beneath it; on the starting settings almost all the water goes under the dam. The shading is total head, darkest at the reservoir and palest where the water leaves. The thin lines labelled in per cent are equipotentials, every 10% of the reservoir head; the fine blue lines with moving dots are flow lines, their dots' speeds compressed so that slow paths stay visible. The dashed blue line is the phreatic surface. Where it reaches the downstream slope, water seeps out over a seepage face, the thick brick-red stroke, softening the soil until pieces of the slope can slump away. The ringed dot marks the steepest exit not covered by a drain. Seepage is in litres a day for each metre length of dam. Safety against piping compares that steepest exit gradient with the 1.0 that floats sand out of level ground; where the water leaves the downstream slope it drags grains out of the face rather than lifting them, so there it is a measure of how hard the water drags. Exit gradients right at the toe depend on the 1 m grid, so read them as indicative. The number to remember: 2,366 L/day per metre, nearly all of it out through the ground beyond the toe.

  • With the homogeneous dam and no drain, the seepage face on the starting sand & gravel foundation is only about 2 m high, because most of the water goes under the dam; switch the foundation to Clay to see a face about 8 m high.
  • Add the chimney and blanket drain: the phreatic surface dives into it and the downstream half stays dry. Read the seepage.
  • Pick Clay core + shells, then add the Full cutoff. Read the seepage each time.
Answers

With the chimney drain 4,052 L/day per metre. With the core 5,434; with the core and a full cutoff, 2.2.

Notice that the equipotentials under the dam stand almost upright and nearly evenly spaced, like the head along the pipe of Fig. 3.2. The gravel passes water 100 times as easily as the fill, so to the water the dam is nearly a lid and the gravel a pipe 15 m deep and as long as the dam's base. Notice too how little goes in through the dam's own face: 221 of the 2,366 litres a day. That picture can be checked by hand.

Example 10.2

Our dam's leak, by hand

Given
The lab on its starting settings: reservoir 20.4 m deep, tailwater at the ground; sand and gravel 15 m deep with k = 10⁻⁵ m/s; the dam's base, toe to toe, 102 m.
Find
The leak under each metre of dam, in litres a day, and how it compares with the lab's.
The idea
Treat the gravel as the pipe of Fig. 3.2: the whole head lost along the base, through an area 15 m deep for each metre of dam (Eq. 3.2).
  1. i = 20.4 ÷ 102 = 0.200head lost per metre of path
  2. q = 0.00001 × 0.200 × 15 = 0.00003 m³/sDarcy's law through 15 m² for each metre of dam
  3. 0.00003 × 86,400 × 1,000 = 2,592 L/dayseconds in a day, litres in a cubic metre
  4. 2,592 ÷ 2,366 = 1.096the hand number against the lab's
  5. A kilometre of dam: 2,366 × 1,000 ÷ 86,400 = 27.4 L/sa thousand metres of dam, litres a day turned into litres a second

About 2,600 litres a day under each metre of dam by hand, 10% more than the lab's 2,366. A kilometre of this dam would leak about 27 litres a second.

Check. Should the hand number be high or low? Darcy's pipe sends the water straight along 102 m. The real water has to dive into the gravel at the upstream toe and climb out again beyond the downstream one, and the lab's flow lines bow downward to do it. A longer path passes less, so the straight pipe should be a little high, and it is. The pipe also ignores the water that goes through the fill, but that is small: 99% of the lab's leak crosses the dam's centre line under the dam.

What it means. Ten per cent is better than the permeability deserves. A natural gravel's k is rarely known within a factor of a few, and the Bureau of Reclamation, which teaches this very calculation, says its result should be read only as the order of magnitude of the seepage (USBR (1987), sec. 6.9(b)).

10.4 What the lab knows that intuition doesn't

A hand estimate is only as good as the picture of the flow behind it, and the lab earns its place where that picture is wrong. The first case is the question this chapter opened with. Give our dam a clay core a hundred times tighter than its fill, 10⁻⁹ m/s, between shells of sand and gravel, 10⁻⁴ m/s, ten times more permeable than the gravel beneath them:

Example 10.3

The core that leaks more

Given
Both dams on the sand and gravel, no drain, no cutoff, reservoir 20.4 m. The dam of one fill leaks 2,366 L/day per metre; the cored dam, 5,434. The core is 21.4 m wide at its base.
Find
Why the cored dam leaks more, and what a full cutoff under its core does.
The idea
Almost all the water goes under both dams. What changes is how much of the base the water must get under, which the head the solver finds at the top of the gravel shows.
  1. 5,434 ÷ 2,366 = 2.30the core more than doubles the leak
  2. In through the upstream shell: 5,343; through the reservoir floor: 91 L/daythe shell is the easy way in, so the water reaches the gravel beside the core
  3. Head at the top of the gravel, 12 m upstream of the centre line: 19.1 m (94%); with no core, 12.6 m (62%)the shell spends almost none of the head, so the water arrives at the core with nearly all of it
  4. Length to get under: 102 ÷ 21.4 = 4.77the whole base under a dam of fill, which is a lid to the gravel; only the core's base under the cored dam
  5. Full cutoff under the core: 5,434 ÷ 2.177 = 2,496a wall from the core's base to the rock: the leak falls to 2.2 L/day

The core more than doubles the leak, because the shells deliver the water to the gravel right beside it and the water only has to pass under the core's 21.4 m base. Taken down to rock with a cutoff, the same core cuts the leak about 2,500 times.

Check. If the length under the dam were the whole story, a length 4.8 times shorter would pass 4.8 times the water; the lab gives 2.3. The rest of the trip does not shrink with the core: the water still has to climb out of the gravel and cross some 40 m of it to the toe, and 6.0 m of head is still left at the top of the gravel a metre downstream of the core. So the factor should fall short of the ratio of lengths, and it does.

What it means. A barrier works only if the water can't go round it. A core that stops on top of a pervious foundation is a wall with a gap under it, and its shells hurry the water to the gap. The same core joined to rock by a cutoff makes the dam almost dry.

Check yourself A full cutoff cuts the cored dam's leak about 2,500 times. Would it do as much for the dam of one fill?

No: 2,366 falls only to 166, about 14 times less. With the gravel cut off, the water goes through the fill instead, and the fill passes water a hundred times more easily than the core.

The second surprise runs the other way. A drain inside a dam is a strip of clean sand and gravel, graded so that the soil around it cannot follow the water in (Chapter 11 explains how), and open to the air at its outlet, so the water in it is at zero pressure. Every bit of water in the dam or the ground that can reach it heads for it. The Corps of Engineers' manual notes the consequence for a horizontal drain under the downstream half of a dam: it lowers the pressure under the dam and increases the quantity of seepage.USACE EM 1110-2-1901, para. 9-3 and its Figure 9-1. The lab shows by how much:

Example 10.4

A drain that leaks more and is safer

Given
The dam of one fill on the sand and gravel, reservoir 20.4 m: 2,366 L/day per metre and a safety against piping of 1.27. The chimney drain stands 4 m downstream of the centre line, and its blanket runs from there along the base to the toe. The toe drain fills the dam's toe.
Find
The leak and the safety against piping with each drain, and why the chimney raises the leak.
The idea
A drain holds everything it touches at zero pressure. It ends the water's trip early, and it gives the water a way out that holds the soil back.
  1. Chimney and blanket: 4,052 ÷ 2,366 = 1.7171% more water
  2. Path under the dam: 55 m instead of 102; 102 ÷ 55 = 1.85the gravel's water now leaves through the blanket, which starts 4 m past the centre line, 55 m from the upstream toe
  3. Steepest exit not covered by the drain: i = 0.0029; safety 343the water leaves through the drain, not through bare ground or the face
  4. Toe drain: 2,647 L/day, 12% more; safety 12.8a shorter cut at the end of the trip

The chimney drain makes our dam leak 71% more and raises its safety against piping from 1.27 to over 300. The toe drain costs 12% more water and raises the safety to 12.8.

Check. Darcy's law predicts a path 1.85 times shorter should pass 1.85 times the water; the lab gives 1.71, a little less, because the water still has to dive into the gravel and rise into the blanket. In the gravel immediately under the blanket's start the solver finds a head of 0.54 m: the drain has taken almost all of the 20.4 m by the time the water reaches it.

What it means. The extra water costs nothing but water. In return the drain takes the water pressure out of the downstream half of the dam, so the grains there press on one another with nearly their full weight and grip accordingly, which is what Chapter 5 meant by drains putting strength back. And it gives the water an exit that holds the soil back.

Check yourself Where would you put a piezometer to show that the chimney drain works, and what should it read?

In the fill a few metres downstream of the chimney, near the base. With the drain working, the water in it should stand barely above its tip, because the fill there is drained to about zero pressure; a reading well above that means the drain is clogged or cut off.

10.5 A dam on clay

Now take the gravel away. Switch the foundation to Clay, 5 × 10⁻⁸ m/s, half as permeable as the fill. The leak collapses to 44.4 L/day per metre, and something else appears. On gravel the phreatic surface reached the downstream slope only 2 m up; on clay it meets the slope 8 m up, and 65% of the water leaves through the slope's face. With no easy way underneath, the water comes through the dam, and that needs a different hand estimate: one for water moving through the fill with a free surface on top.

The estimate engineers use is due to the French engineer Jules Dupuit, in 1863.Dupuit (1863). The entry correction: Casagrande (1937), p. 150. Pretend the water moves level. Cut the dam into upright slices. In a slice where the saturated fill is h deep, the water flows through h square metres for each metre of dam, and the gradient is the slope of the water table there, Δh over Δx. By Darcy's law q = k × (Δh/Δx) × h, and the same q passes every slice, so where the wet part is thin the water table must be steep. Rearranged, q · Δx = k · h · Δh. Add up the slices from the downstream end to the reservoir. On the left the Δx add up to the length L. On the right the products h · Δh, as h rises from nothing to H, add up to the area of a triangle with height H and base H, ½H², the same sum that gave the push of water in Eq. 2.2:

q = k · H² / (2L)(10.2)

q flow through the dam per metre length (m³/s per m) · k permeability of the fill (m/s) · H depth of the reservoir (m) · L length the water travels, from where it enters to the downstream toe (m)

Where does the water enter? The upstream wedge of the dam stands in the reservoir and its water is at nearly the reservoir's head, so measuring from the upstream toe would make the trip too long. Arthur Casagrande showed in 1937 that a simple rule does well: start the trip upstream of where the water surface meets the face, by 0.3 times the horizontal run of the wetted face.

Example 10.5

Our dam on clay, by hand

Given
The dam of one fill, k = 10⁻⁷ m/s, slopes 2:1, toes 102 m apart, on 15 m of clay with k = 5 × 10⁻⁸ m/s; reservoir H = 20.4 m, meeting the upstream face 10.2 m upstream of the centre line. A flow of 10⁻⁷ m³/s is 8.64 litres a day.
Find
The leak by hand, through the fill and under it, and where the water breaks out on the downstream slope.
The idea
Two flows side by side: through the fill by Eq. 10.2, and under it through the clay by Darcy's pipe, as in Example 10.2.
  1. Run of the wetted upstream face: 2 × 20.4 = 40.8 mtwo across for every one up
  2. Entry correction: 0.3 × 40.8 = 12.24 mCasagrande's rule
  3. L = 12.24 + 10.2 + 51 = 73.44 mfrom the entry point to the centre line, then half the base to the downstream toe
  4. H² ÷ 2L: 416.16 ÷ 146.88 = 2.833 m20.4² = 416.16, and 2 × 73.44 = 146.88
  5. Through the fill: 2.833 × 8.64 = 24.48 L/dayEq. 10.2, with k = 10⁻⁷ m/s turned into litres a day
  6. Under it: 4.32 × 0.200 × 15 = 12.96 L/dayDarcy's pipe through the clay, its k in litres a day
  7. Total: 24.48 + 12.96 = 37.44 L/daythe lab: 44.4
  8. Break-out: a = √(L² + H²) − √(L² − 4H²) = 15.16 m up the slopeCasagrande's construction for a slope this gentle; 4H² because the slope runs 2 across for 1 up
  9. Its height: 15.16 × 0.447 = 6.78 m0.447 is the sine of the 2:1 slope's angle, 1 ÷ √5

By hand, about 37 L/day per metre, 16% below the lab's 44.4, with the water breaking out about 6.8 m up the downstream slope against the lab's 8 m.

Check. The lab can also say where its water crosses the dam's centre line: 31.8 L/day inside the fill, 12.5 in the clay beneath. Darcy's pipe gets the clay's share almost exactly; Dupuit's slices come out low, because the hand sum keeps the fill's water and the clay's in separate pipes and the solver lets them trade. The Bureau of Reclamation's rule of thumb for a dam of one fill puts the break-out at roughly a third of the reservoir's depth (USBR (1987), sec. 6.3(c)): 20.4 ÷ 3 = 6.8 m. The lab resolves its face only to its 1 m cells.

What it means. A clay foundation half as permeable as the fill is far from watertight: more than a quarter of the water still goes under it. Under the cored dam of Example 9.1 it is most of the water: a cutoff beneath its core takes its leak on clay from 37.2 to 2.1 litres a day. Every hand method here gets the flow within a factor of two and the break-out within a couple of metres, enough to tell a sound design from a poor one. What the break-out does to the slope it softens is Chapter 12's question.

10.6 Five ways to manage seepage, and one to watch it

The chapter's question now has an answer. How much a dam leaks matters to whoever pays for the water, and to the neighbours of a tailings pond; on its own it seldom threatens the dam. A leak is safe when it leaves where the designer meant it to, through a drain and not out of a slope; when it leaves gently, at a gradient well below the one that floats grains, through a filter that holds them back; and when it leaves the downstream part of the dam at low water pressure, so that the grains there press on one another with their full weight. Engineers have six kinds of defence for it, drawn here at one scale on our dam:

Fig. 10.4 · Six defences against seepage, on our damSections to one scale · schematic
cutoff to rock

A · Cut it off. A core, a cutoff trench or slurry wall to rock, a grout curtain, a concrete face or a membrane: a barrier the water cannot cross or go round.

blanket

B · Lengthen the path. An upstream blanket of clay on the reservoir floor, joined to the dam's watertight part, makes the water enter further away.

C · Drain it. Chimney, blanket and toe drains inside the dam; relief wells in the ground beyond the toe; drains in a concrete dam's galleries. They decide where the phreatic surface sits.

D · Filter every exit. Wherever water leaves soil, it passes through a filter, so it leaves without the soil.

E · Weigh it down. A berm, a low bank of earth beyond the toe, adds weight where the water in the ground pushes up.

weir

F · Watch it. Piezometers read the head inside the dam; a weir measures the leak; cloudy water means soil is moving.

Six defences on our dam, all at one scale: the first stops the water, four control how it arrives and leaves, and the last measures it. Each panel is our 24 m dam on 15 m of sand and gravel over rock, with the reservoir on the left; hatches are the materials of Plate 0. Blue arrows are water; in E the brick-red arrows are the water pushing up from the ground and the green one the berm's weight pressing down. The dashed line in C is the phreatic surface; the short blue ticks in F are the water levels in the two standpipes. The number to remember: our cored dam leaks 5,434 L/day per metre on gravel, and 2.2 with a cutoff to rock.

  • In which panels does water still pass under the dam? Pick one and try its nearest equivalent in the lab.
Answers

In all but A. B makes the water's road longer, C, D and E change how it arrives and leaves, F measures it. In the lab, the chimney drain is C: the leak rises to 4,052 and the safety against piping to over 300.

Notice that only the first panel stops the water. The other five accept that it will arrive and manage the arrival. Lengthening the path is the oldest of them. Before flow nets were in everyday use, weirs on sandy riverbeds were sized by a rule of thumb: make the water's path along the underside of the structure long enough for the soil.The rule of thumb was W. G. Bligh's "line of creep": the head a structure could hold grew with the length of its contact with the soil. E. W. Lane, at the Bureau of Reclamation, proposed counting the level parts of that path at a third of the steep ones. A Bureau study of 1934, which checked the rules against pressures measured under real dams, recommended his weighting over Bligh's (Hebert (1934)), and Lane published it in 1935 (Lane (1935)). The lab's half-depth cutoff is a lesson in that rule's limits: it adds 15 m to a 102 m path, and the leak barely moves.

Watching is the defence that says whether the others work. Piezometers are the straws of Fig. 3.1 made permanent: a tube, or a pressure sensor, sealed at a point inside the dam or its foundation, reporting the head there. Read against the reservoir level for years, their trend shows whether the drains are clear and whether the pressure in the downstream slope is where the design put it. The leak itself is measured at weirs like the one this chapter began with.A 90° V-notch converts depth into flow. The Bureau of Reclamation's form of Cone's formula, Q = 2.49 h2.48 with Q in cubic feet a second and h in feet, becomes Q = 1.342 h2.48 in cubic metres a second and metres (USBR Water Measurement Manual, eq. 7-6). Because the flow grows faster than the depth's square, a weir reads small flows finely. Engineers judge the leak against the reservoir. A leak that rises and falls with the reservoir is the ground obeying Darcy's law. A leak that grows while the reservoir does not, or water that comes out cloudy, is a warning: turbidity means soil is moving, and even clear seepage is sampled, because it can carry fine soil the eye can't see.USACE EM 1110-2-1901, paras. 13-3 and 13-4.

The Prologue's dam had such a system: a pond below it caught what came through, and the first sign of trouble that night was the water in that pond rising. So it is fair to ask whether the leak brought it down, and the review panel asked. Piping and cracking, water opening a path through a dam and carrying soil out along it, is one of the commonest ways earth dams fail, and the panel found details that worried it. But apart from one brief spike in the flow, in April 2013, the records showed no abnormal seepage, and the panel found no evidence that piping or cracking caused the breach. Mount Polley's seepage was not what failed it.Mount Polley Review Panel (2015), section 4.3.3, pp. 11–12; the pond that night, p. 10. What did fail it is the business of the last chapter.

What to remember

  • Every dam leaks. A leak is safe when the water leaves where the designer chose, gently and through a filter, and leaves the downstream slope at low water pressure, so its grains keep pressing on one another.
  • A flow net of squares turns Darcy's law into counting, q = k·H·Nf/Nd. Crowded squares mean a steep gradient; the smallest sit at exits and corners, where the water pushes hardest on the grains it leaves.
  • Head over path, times k, times depth, estimates the leak under our dam to within 10%, and Dupuit's slices estimate the leak through it. A hand estimate is an order-of-magnitude tool, because k is rarely known better than that.
  • A barrier works only if it reaches something tight. A core that stops on gravel can leak more than no core at all; joined to rock by a cutoff, it cuts the leak about 2,500 times.
  • A drain makes a dam leak more and stand safer: it takes the water pressure out of the downstream half, so the grains there grip with their full weight, and it lets the water out through a filter.

Questions

  1. 10.1

    A flow net sketched under a steel sheet-pile wall has 4 flow channels and 8 head drops. The water stands 6 m higher on one side of the wall than the other, and the sand passes 2 × 10⁻⁵ m/s. How much water passes under each metre of wall in a day? How much head is lost in each drop?

    Hint

    Use Eq. 10.1; there are 86,400 seconds in a day.

  2. 10.2

    Suppose our dam on gravel, as the lab opens, were 400 m long, and all its leak were collected at one 90° V-notch weir. How many litres a second would reach the weir, and how deep would the water stand over the notch? For the weir use Q = 1.342 h2.48, with Q in m³/s and h in metres.

    Hint

    Multiply the lab's litres a day per metre by 400 and divide by 86,400. Then h = (Q ÷ 1.342)1/2.48, with Q in cubic metres a second.

  3. 10.3

    Set the lab to the dam of one fill on gravel and add the half-depth cutoff. The leak falls from 2,366 to 2,160 L/day per metre, and the safety against piping rises only from 1.27 to 1.47. Why does a wall halfway down the gravel do so little?

    Hint

    How much longer does the water's path get, if it goes down one side of the wall and up the other?

  4. 10.4

    In the lab, lower the reservoir to 60% and read the seepage, then raise it to 95%. Does the leak grow in proportion to the depth of water? What happens to the safety against piping?

    Hint

    Darcy's law makes the flow proportional to the head. Compare the two ratios: depth at 95% over depth at 60%, and leak over leak.

  5. 10.5

    Our dam of one fill stands on the clay foundation. A toe drain and a chimney drain both remove its seepage face. The toe drain leaks 49.9 L/day per metre with a safety against piping of 6.7; the chimney leaks 110, more than twice as much, with a safety against piping of 183. Which would you build, and why?

    Hint

    The lab's fill is the same in every direction. Is a real fill, compacted in layers, likely to be?

Further reading

  • Casagrande, A. (1937). Seepage through dams. Journal of the New England Water Works Association 51(2): 131–172. primary source The paper that taught dam designers to sketch flow nets, with Casagrande's advice for beginners, the seepage line through an earth dam, the 0.3 entry correction used in Example 10.5, and an early warning about layered fills.
  • US Army Corps of Engineers (1986). EM 1110-2-1901, Seepage Analysis and Control for Dams. Washington, DC: Department of the Army, 30 September 1986; Change 1 (a new Appendix D on filters and drains), 30 April 1993. https://damtoolbox.org/wiki/Seepage_Analysis_and_Control_for_Dams_(EM_1110-2-1901) serious The Corps of Engineers' manual. Its fourth chapter teaches flow nets, with drawing guidelines and the counting formula used here; the eighth and ninth cover seepage control in embankments and foundations, and the thirteenth the weirs and piezometers that watch it.
  • U.S. Bureau of Reclamation (1987). Design of Small Dams, 3rd ed. Denver. https://www.usbr.gov/tsc/techreferences/mands/mands-pdfs/SmallDams.pdf serious The chapter on earthfill dams works the hand estimate of Example 10.2 for a real foundation and lays out the treatments for sand and gravel foundations: cutoffs, blankets, toe drains and relief wells. Free from the Bureau of Reclamation.

Chapter 11 · Part III · Inside the embankment

When water carries soil

Bennett and Teton both leaked through their cores. Why does one of them still stand?

The Panel considers that too much was expected of the grout curtain, and that the design should have provided measures to render the inevitable leakage harmless.

So far. Every dam leaks, and Chapter 10 followed the water through ours: where it gets in, which way it goes, where it comes out, and how drains and cutoffs decide all three. At the toe of our dam on sand and gravel, the water left the ground with a gradient of 0.79. This chapter asks what that water does to the grains it leaves behind.

After this chapter you can

  • explain a sand boil as the block of Chapter 4 floating, and judge an exit gradient against the margin engineers want;
  • name the four ways water carries soil out of a dam, place a real failure in one of them, and say why a pipe, once started, speeds up;
  • size a sand filter for a sand, and say why a clay core needs a different rule.

This chapter adds a filter and drain at the toe

7 figures · about 20 min to read · Film 11.1, 9 min

On the morning of Saturday, 5 June 1976, the reservoir behind Teton Dam in eastern Idaho was nearly full for the first time. The dam was new: an earth embankment 93 m high (305 ft above the riverbed) across a canyon of hard volcanic rock, finished the November before and filling since October. Two days earlier, springs of clear water had appeared in the rock of the canyon wall, about 400 m downstream. The engineer in charge took them for a good sign: the ground around the reservoir was filling up as it should.Teton Panel (1976): the dam, p. v and p. 1-1; the springs of 3 June, 1,300 and 1,500 ft downstream, p. 12-1, and the project engineer's testimony, p. 2-1.

Shortly after seven o'clock, the first people to arrive for work found muddy water running out of the rubble at the foot of the right abutment, where the dam met the canyon wall, at 20 to 30 cubic feet a second (0.57 to 0.85 m³ a second), and a trickle of cloudy water higher up the same junction. Both grew all morning. At about half past ten, with a loud burst, a large new leak broke out of the downstream face, from a tunnel about 1.8 m across, and the water began cutting a gully back up the face. At eleven a whirlpool opened in the reservoir, above the upstream slope. At 11:55 the crest began to fall in between the whirlpool and the head of the gully, and the reservoir went through. Eleven people died in the flood, by the count of the Bureau of Reclamation, which built the dam; the panel that investigated the failure associated fourteen deaths with it, directly or indirectly.The morning: Teton Panel (1976), p. 12-1, pieced together by the panel from sworn testimony and photographs (pp. 2-1 to 2-15). Eleven deaths: Stene (1996), p. 1. Fourteen: Teton Panel (1976), p. v and p. 1-1.

Teton leaked through its core. So, twenty years later, did the W.A.C. Bennett Dam on the Peace River in British Columbia, the dam whose zones opened Chapter 9. In June 1996 a hole opened in its crest, and the investigation that followed, reviewed by a panel of experts in 2012, found that seepage had carried part of its core away. Bennett is still holding back its reservoir. In both dams water found a way through the core and picked up soil on the way. The difference lay in where it could take that soil, and that difference is what this chapter is about.

11.1 Sand that floats

Start at the beach. Chapter 1 met the soft sand at the water's edge as loose sand that a wiggle shakes into soup. As the tide goes out there is a second reason for it. The water held in the beach seeps back out at the foot of the face, and where it flows upward through the sand it takes part of the grains' press away. Push it up harder and it takes all of it.

Chapter 4 put numbers on that. Water that loses i metres of head for every metre it travels drags on the grains in each cubic metre with i × 9.81 kN, in the direction it flows. A cubic metre of soaked sand presses down on the grains below it with its buoyant weight, about one water's worth. Turn the flow upward, raise the gradient, and the drag eats into that press until, at the critical gradient, the two cancel. The grains still touch, but they no longer press on one another, the effective stress is zero, and the sand has no grip at all. It is the block on the ramp of Fig. 4.1 with so much water under it that it floats. For a typical quartz sand, grains 2.65 times as heavy as water with 0.65 m³ of gap for every cubic metre of grain, Example 4.4 worked it out from the grains: (2.65 − 1) ÷ (1 + 0.65) = 1.00.

That is easy to see in a tube. Fill it with a metre of that sand, feed it water from below out of a tank, and let the water spill out of a spout at the top of the sand. Then raise the tank:

Fig. 11.1 · A metre of sand, fed with water from belowSchematic · arrows to scale with each other
tank level with the outlet 1 m of sand A · i = 0 grains press with9.81 kN per m³ tank 0.5 m higher B · i = 0.5 grains press with4.9 kN per m³ tank 1.0 m higher C · i = 1.0 grains press with0 kN per m³ the bed boils

Raise the tank until the water loses a metre of head through each metre of sand, and the sand floats. Each tube holds the same metre of soaked sand on a gravel screen, with a spout at the top of the sand that sets the level the water leaves at. The tank beside it stands level with the spout in A, half a metre above it in B and a metre above it in C. Inside the sand, the ochre arrow is the weight of each cubic metre of grains under water, 9.81 kN, and the blue arrow the upward drag of the water flowing through it; the press the grains have left is the difference, written under each tube. The number to remember: this sand boils at a gradient of 1.00.

  • Measure the two arrows in B with a ruler. What share of the grains' press has the water taken?
Answers

Half: the blue arrow is half the length of the ochre one, so the grains press with 4.9 kN in each cubic metre instead of 9.81, 50% of their press.

Notice that nothing about the sand changes from one tube to the next. The grains, the gaps and the water in them are the same; only the tank moves. In A the water stands still and the grains carry their whole weight under water. In B the water climbs through the sand, losing half a metre of head over its metre, and its drag takes half the grains' press. In C the drag matches the grains' weight, and the top of the bed lifts and churns. That is a sand boil, and it is what an engineer fears to find beyond the toe of a dam, where water that has passed under it rises out of the ground.

Check yourself A looser sand of the same grains has 0.8 m³ of gap for each cubic metre of grain. Does it boil at a lower or a higher gradient?

Lower. Its critical gradient is (2.65 − 1) ÷ (1 + 0.8) = 0.92: more of each cubic metre is water, so the grains weigh less under it and take less drag to float.

The gradient where seepage leaves the ground is the exit gradient, and engineers judge it against the critical gradient by a ratio, their margin against heave:

FSheave = ic / iexit (11.1)

ic critical gradient, about 1 for sand · iexit the upward gradient where seepage leaves the ground

A slope in Chapter 4 was judged at a factor of safety of 1.5. For an exit, the Bureau of Reclamation's design standard asks for at least 3 at an unfiltered exit of an existing dam, and 4 for a new dam or a repair, or about 2 to 2.5 where the soil and the water pressures are well measured.USBR DS-13 ch. 8, section 8.2.2.2, pp. 8-5 to 8-7, which also gives the warning about internal erosion quoted in the next paragraph. The margin is high because the number it divides by is shaky. Right at the toe the gradient has no settled value: the closer you look, the steeper it gets, so a computed exit gradient depends on how finely the ground was divided.Water flowing under a dam's base turns a sharp corner to come up beside the toe, and in the mathematics of a flow net the gradient at such a corner grows without limit. A grid of cells smooths the corner over one cell, so its answer depends on the cell's size. The seepage lab reads its exit gradient over the last half metre of ground. Natural ground is never as even as a model of it, and a lens of fine sand can sit where the model has gravel. And in fine sand under a roof, grains can be carried along at far gentler gradients than the one that floats them, as the next section shows.

The standard also says what the ratio does not measure: it flags sand boils, not every kind of internal erosion. Its own words are that it "is not a factor of safety against the initiation of internal erosion". A dam can boil without failing, and fail with no boil at all, as Teton did. For now, here is the ratio at work on our dam:

Example 11.1

The margin at our dam's toe

Given
Our dam of one fill on sand and gravel, as the seepage lab of Fig. 10.3 opens: reservoir 20.4 m deep, no drain, no cutoff. The lab's largest exit gradient is 0.786, in the ground right beyond the toe, and it takes ic = 1.0, a round value for sand.
Find
The margin against heave at the toe; the exit gradient a new dam would need; what a cutoff halfway down the gravel does.
The idea
Divide the critical gradient by the exit gradient (Eq. 11.1), and turn the target margin round to find the gradient it allows.
  1. FS = 1.0 ÷ 0.786 = 1.27the margin as the lab opens; its readout says the same
  2. For a margin of 3: iexit ≤ 1.0 ÷ 3 = 0.33the most an existing dam's exit may have
  3. For a margin of 4: iexit ≤ 1.0 ÷ 4 = 0.25the most a new dam's exit may have
  4. 0.786 ÷ 0.25 = 3.1how many times gentler the exit would have to become
  5. Half-depth cutoff: FS = 1.0 ÷ 0.682 = 1.47the lab's exit gradient with the wall in place

The toe has a margin of 1.27 against heave. A new dam would need its exit gradient cut about 3.1 times, to 0.25; a cutoff halfway down the gravel takes the margin only to 1.47.

Check. The margin times the exit gradient should give back the critical gradient: 1.27 × 0.786 = 1.0. And the answer sits where Fig. 11.1 says it should: an exit gradient of 0.79 leaves the grains at the toe about a fifth of their press, 21%, against the half that tube B keeps.

What it means. The sand at our toe is not boiling, but its margin is a third of what a designer would accept. Stopping the water is hard: the wall halfway down barely helped, because the water goes under it. The defences of Fig. 10.4 offer another way, at the exit itself, and the rest of this chapter comes to it.

11.2 A pipe that feeds itself

A boil is a warning, and grains are only part of what it carries. Each grain that comes up with the water leaves a hole the size of a grain behind it. The question is what that hole does next. Make a guess before you read on:

Water rising beyond the toe of a dam on sand has started to carry grains out, and a small hollow has opened under the edge of the dam. As the hollow works its way back under the dam, does it get easier or harder for the water to carry more sand away?

To decide, count the head the water has to lose, and the distance it has to lose it over. Under the intact dam the water loses its whole head along the full path from the reservoir floor to the toe. What happens next depends on what holds the hollow open. The dam above it holds up its roof, so it stays open; water in it moves freely, and the hollow can grow back toward the reservoir, against the flow. Engineers call this backward erosion piping: it starts at the exit and travels upstream. The figure plays it from the first boil to the breach:

Fig. 11.2 · Backward erosion piping, from boil to breach Schematic · hours to decades compressed

Drawing scrolls sideways →

1 · Steady seepage under the dam
Toe

Once grains wash out at the toe, the hole they leave works back under the dam, and each metre it gains steepens the push at its tip. The section is a dam on sand, with the reservoir on the left; the stage line under the slider names each step, from steady seepage through the boil, the pipe working back, the pipe reaching the reservoir and widening, to the breach. The Toe switch puts a filter and a weighting berm, a mound of extra fill, where the water comes out: the water still leaves, the sand cannot, and the chain never starts. That is why dam inspectors walk the toe looking for boils, cloudy (turbid) seepage and sinkholes. The number to remember: our dam spends 20.4 m of head over a 102 m path, an average gradient of 0.20; with half the path eaten it is 0.39.

  • Press Play with the toe Unprotected and watch where the pipe starts and which way it grows.
  • Switch the Toe to Filter + weighting berm and play it again. At which stage does it stop?
Answers

Unprotected, the pipe starts at the toe and grows toward the reservoir, against the flow. With the filter and berm the sequence stops at its first stage, steady seepage: the water comes out and the sand stays.

Notice which way the pipe grows. The water flows from the reservoir to the toe, and the pipe grows from the toe to the reservoir. It does its damage out of sight, under the dam, and the only sign above ground is at the exit: a boil, a cone of sand round a spring, water that runs cloudy. By the time the pipe reaches the reservoir, the water has a tunnel straight through, and the tunnel widens until its roof falls in. The arithmetic of a pipe that feeds itself is short:

Example 11.2

A head budget for a growing pipe

Given
Our dam on sand and gravel: 20.4 m of head between the reservoir and the toe, spent along a path 102 m long under the dam's base. The lab's exit gradient at the toe, 0.786. An open pipe that loses no head along its length (illustrative: a real pipe loses a little).
Find
The average gradient on the path the water has left to cross, before the pipe and as it grows; and roughly what the water does at the pipe's tip.
The idea
The head is fixed by the reservoir. Whatever length of sand is left has to lose all of it (gradient = head lost ÷ distance, as in Chapter 3).
  1. i = 20.4 ÷ 102 = 0.20the average gradient under the intact dam
  2. 0.786 ÷ 0.20 = 3.9how much steeper the water runs where its paths crowd together at the exit
  3. Pipe 50 m back: 20.4 ÷ 52 = 0.39the same head over the sand that is left
  4. At its tip: 0.39 × 3.9 = 1.5if the water crowds into the tip as it crowded into the toe: past the critical gradient
  5. Pipe 90 m back: 20.4 ÷ 12 = 1.7now even the average floats the sand

Halfway under the dam the average gradient has doubled, to 0.39, and the push at the pipe's tip is past the critical gradient. With 12 m of sand left, the average alone is 1.7.

Check. The average gradient reaches 1 when the sand left is as long as the head is high, 20.4 m: a pipe 81.6 m back, four fifths of the way under. Roughly halving the path roughly doubles the gradient, as step 3 found.

What it means. These numbers sketch a head budget and leave out how fast sand erodes, but the direction is certain: every grain that leaves makes it easier for the next one to leave. A pipe that has started will not stop by itself. The place to stop it is the exit, before it starts.

Check yourself A pipe has eaten 70 m back under our dam. What average gradient is the water spending on the sand that is left?

20.4 ÷ 32 = 0.64, more than three times the 0.20 under the intact dam.

How fast this happens depends on the sand. Fine, uniform sand pipes most easily: in laboratory tests on fine sands under a roof, grains began to move at average gradients as low as 0.08.Schmertmann's tests, and a gradient of about 0.08 back-calculated from piping at A. V. Watkins Dam: FEMA (2011), pp. 109–110. A well-graded sand, with a spread of sizes, resists, because its larger grains hold the smaller ones back. The figure's scale note says hours to decades, and both ends are real.Teton went from the first muddy water to the breach in about five hours. At the other end, the panel that reviewed Bennett Dam notes that a sinkhole forming over erosion deep in a dam is often slow, and may take years to show at the crest (BC Hydro expert panel (2012), p. 62).

11.3 Four ways water carries soil

A boil at the toe is one way in. Teton showed nothing at its toe until its last morning; its trouble started deep inside, in the core. Engineers now sort the ways that erosion inside a dam can begin into four. The International Commission on Large Dams names them, and the panel that reviewed Bennett used the same four: concentrated leak, backward erosion, suffusion and contact erosion.ICOLD Bulletin 164; the four, with their definitions, in the terminology of BC Hydro expert panel (2012), pp. 60–62. Here they are, each drawn on our dam at one scale:

Fig. 11.3 · Four ways erosion starts inside a damSections to one scale · circles magnify grains · schematic
A · Concentrated leaka crack through the coreTeton, 1976 · Bennett, 1996 B · Backward erosiona pipe works back from a boil at the toeour dam on sand and gravel C · Suffusioncircle: fine grains wash out between coarse onesparts of Bennett’s transition D · Contact erosioncircle: the underside of the fine layerfine soil laid on open gravel

Water can start carrying soil out of a dam through a crack, from an exit, between coarse grains, or along a boundary between two soils. Each panel is our 24 m dam with the reservoir on the left. In A a crack crosses the clay core and the water in it scours its walls; in B a pipe works back from a boil at the toe; in C, inside a coarse zone, fine grains wash out through the gaps between coarse ones, which stay put; in D water running along open gravel under a layer of fine soil scours the underside of the layer. Blue arrows are flowing water and the brick-red lines are where soil is being lost; the circles in C and D are grain-scale close-ups. The line under each panel says where this book meets that way in. The number to remember: four ways in, and in every one of them the soil has to have somewhere to go.

  • For each panel, put your finger where the lost soil ends up. Which panels give it an exit you could see from outside the dam?
Answers

A sends the core's grains into the downstream fill, B up through the boil at the toe, C downstream through the coarse zone and D along the gravel and out beyond the toe. Only B shows at the surface from the start; the others show only when enough soil has gone for the ground above to sink, or the seepage to turn cloudy.

Notice that only one of the four needs an exit at the ground surface. A concentrated leak (A) is water running through an opening that is already there: a crack, a gap beside a pipe through the dam, a loose layer, or open joints in the rock under a core. Its walls are scoured from the inside, and the opening widens. Backward erosion (B) is the pipe of the last section. Suffusion (C) happens in a soil whose grain sizes have a gap, coarse grains and fine ones with few in between: the coarse grains carry the load, the fine ones sit loose in their gaps, and seepage can wash the fine ones out and leave the skeleton standing. Contact erosion (D) happens where a fine soil lies against a coarse one and water runs along the boundary, as in a dam built on a layer of silt over open gravel.

A concentrated leak needs a crack, and a core is compacted precisely so that it has none. But a crack can open after the core is built. A core that settles more over a deep part of the valley than over a shallow one stretches, and the stretched part can crack. And water can open one by itself. Where something holds part of the core up, as the steep walls of a narrow trench can, the soil below is pressed together less than its depth would suggest. Teton's core was carried down into the rock of its abutments in such a trench, a key trench. If the water pressure in the soil rises above the least of those presses, the water can split the soil apart, the way a wedge splits wood. That is hydraulic fracture, and the panels that investigated both Teton and Bennett found it a likely way their cores first opened.

All four end the same way if nothing stops them: soil carried out of the dam, and a channel that grows. The statistics count them together, under an older name. Engineers often call every kind of internal erosion piping, though strictly the word means the backward erosion of the last section. Foster, Fell and Spannagle's 2000 study of 124 embankment-dam failures up to 1986 counted every kind as piping, through the embankment, through the foundation, or from one into the other.Foster, Fell & Spannagle (2000), their table 3: failures to 1986, excluding failures during construction. Internal erosion of all kinds caused about as many failures as overtopping: 47% against 48% once spillway and gate failures are counted with overtopping. That is more than overtopping alone (34%). Some failures had more than one cause, so the shares add up to a little over 100%. Water pouring over the top of a dam is the other half of the story, the one Chapter 6 told. The rest of this chapter is about the half that starts inside.Most clays resist erosion far better than sand, because their tiny flakes cling together: in laboratory tests a sand moves at an upward gradient of about 1, a clay often not until tens or hundreds (USBR DS-13 ch. 8, p. 8-6). A few clays come apart into separate flakes in fresh water instead (FEMA (2011), p. 78), and such a dispersive clay loses that defence. The pinhole test finds them: water is forced through a fine hole punched through a sample, and a dispersive clay turns it cloudy (Sherard et al. (1976)).

11.4 Why a filter works

The first cure anyone reaches for is to stop the water. Example 11.1 showed how hard that is under our dam, and Teton's designers tried it with a grout curtain, cement grout pumped into the joints of the rock under the core. Karl Terzaghi had proposed the opposite half a century earlier: let the water out, and keep the soil in.Teaching in Istanbul in the early 1920s, Terzaghi showed in simple laboratory tests that a layer of coarser sand with weight on top of it stops a fine sand from boiling, and patented the idea. He wrote about piping under dams and how to prevent it in 1922 (Terzaghi (1922)). The grain-size rules grew out of his consulting on small weirs in Europe; Fannin (2008) traces the story. A layer of the right sand where the water leaves the soil passes the water and holds the grains. A filter stops soil, not water.

How can anything let water through and not sand? The figure shows it at the scale of grains: on the left the grains of a core meet rockfill, on the right a filter of sand:

Fig. 11.4 · Core grains meeting rockfill vs a filter Grain scale · schematic

Against rockfill the core's grains wash into the big voids; against a filter of sand the coarsest of them lodge at its face and hold back the finer ones behind. In each panel the core is on the left of the dashed line and seepage, the blue streaks, runs left to right. On the left the voids between the rockfill's blocks are many times the size of the core's grains, and the counter of grains lost keeps climbing. On the right the filter's gaps are smaller than the core's coarser grains: those stop at the face, and the finer grains pile up behind them, so the water passes and the counter stays at zero. The drawing simplifies: it stops every grain at the filter's face by rule. In a real filter only grains bigger than its openings stop there, and the layer they build holds back the rest. The number to remember: a graded sand filter passes particles smaller than about 0.11 of its own 15% size, and holds anything larger.

  • Watch the two counters for half a minute. Then find, on the right, the layer of core grains building up at the filter's face.

Notice that the filter does not need gaps smaller than every grain of the core. It needs gaps smaller than the core's coarsest grains. Those lodge in the filter's face, their own gaps are smaller still and catch the next size down, and so on: the soil builds its own filter against the real one, and the water goes on through.Three equal balls touching leave a gap through which a ball only 0.155 of their diameter can pass, about a sixth. A real filter has grains of many sizes, the small ones sit in the gaps of the large, and laboratory tests by Sherard and his colleagues found it acts like a sieve whose openings are about 0.11 times its 15% size (FEMA (2011), pp. 275–276). A filter that does this is designed by grain size, and grain size is read off a curve.

Pour a soil through a stack of sieves, finest at the bottom, and weigh what each sieve holds. The grain-size curve plots, for each size, the share of the soil by weight that is finer. Engineers then name the soil by points on that curve: D15 is the size that 15% of the soil is finer than, D85 the size that 85% is finer than. D85 stands for the coarse grains that have to stop at the filter's face; D15, for the fine grains that decide how easily water passes. Here are a fine sand and a filter for it:

Fig. 11.5 · A sand and its filter, as grain-size curvesIllustrative soils · sizes on a ten-fold scale
0.01 0.1 1 10 100 0% 15% 50% 85% 100% grain size, mm (each step to the right is ten times coarser) % by weight finer silt and clay sand 0.20 mm 0.45 mm 1.2 mm × 4 × 4 0.8 1.8 the sand its filter

A filter is chosen by its fine grains: its 15% size must fall between four times the sand's 15% size and four times the sand's 85% size. Each curve rises from the soil's finest grains on the left to its coarsest on the right; a step to the right is ten times coarser. The ochre curve is the sand to be protected, the curve in ink its filter; the dashed lines mark 15% and 85%, and the dotted one the 0.075 mm sieve that divides sand from silt and clay. The green arrows multiply the sand's two sizes by four and land on the ends of the green band where the filter's 15% size is allowed. The number to remember: for this sand the filter's 15% size must lie between 0.8 and 1.8 mm.

  • Read the filter's 85% size off its curve, and the sand's 50% size off its curve.
Answers

The filter's 85% size is 4.0 mm; the sand's 50% size is 0.30 mm, a fine-to-medium sand.

Notice that the filter's curve lies almost wholly to the right of the sand's: nearly all its grains are coarser than nearly all of the sand. That looks as if the sand could slip through, and it would, if the filter's openings were as wide as its grains. They are about a ninth of its D15. Terzaghi's two rules put the filter in its band:

D15F ≤ 4 · D85B and D15F ≥ 4 · D15B (11.2)

D15F the filter's 15% size · D85B, D15B the base soil's 85% and 15% sizes (mm); the base is the soil the filter protects

The first rule keeps the soil in. A filter whose D15 is at most four times the sand's D85 has openings at most 0.11 × 4 = 0.44 of the sand's D85, so at least the coarsest 15% of the sand cannot get in, and the layer those grains build holds back the rest. The second rule lets the water out. Water passes a sand roughly in proportion to the square of its fine grains' size (its D15), so a filter whose fine grains are several times coarser passes water many times more easily, and the water leaves the soil without building up pressure in the filter. The agencies that design dams use these rules for sands and gravels, with 4 in the first and 3 to 5 in the second.FEMA (2011), table 5-2 (p. 84) for the retention rule by soil category; p. 84 for the permeability rule as the Bureau of Reclamation, the Corps of Engineers and the Natural Resources Conservation Service each write it; p. 88 for the square of D15.

Example 11.3

A filter for a fine sand

Given
The sand of Fig. 11.5 (illustrative): D15 = 0.20 mm, D85 = 0.45 mm. A graded filter's openings are about 0.11 of its D15.
Find
The range of D15 a filter may have; how its openings compare with the sand's coarse grains; how much more easily it passes water.
The idea
Apply both rules of Eq. 11.2, choose inside the band, then check the openings against the grains they must stop.
  1. Largest D15F: 4 × 0.45 = 1.8 mmkeep the soil in
  2. Smallest D15F: 4 × 0.20 = 0.8 mmlet the water out
  3. Choose D15F = 1.2 mminside the band, clear of both ends, since no pit delivers a curve exactly
  4. Openings: 0.11 × 1.2 = 0.13 mmthe smallest particle the filter stops
  5. 0.13 ÷ 0.45 = 0.29openings against the sand's coarse grains
  6. 1.2 ÷ 0.20 = 6the filter's fine grains against the sand's
  7. 6 × 6 = 36water passes a sand roughly as the square of its fine grains' size

Any filter with D15 between 0.8 and 1.8 mm will do. One at 1.2 mm has openings under a third of the sand's coarse grains and passes water about 36 times as easily as the sand.

Check. Even at the coarse end of the band the openings are 0.11 × 1.8 = 0.20 mm, under half the sand's 0.45 mm: the rule keeps a margin. And the band is 1.8 ÷ 0.8 = 2.25 times wide, which is the sand's own D85 over its D15: a sand with a wider spread of sizes leaves a wider choice, and a uniform sand a narrow one.

What it means. A filter works by grain size alone. Nothing in it is tight, and water passes it more easily than the soil it protects; it holds the grains because its coarse grains bridge the gaps and the soil's own coarse grains bridge the rest. Where water leaves a soil through such a layer, the grains at the exit keep pressing on one another.

Check yourself Would rockfill with a D15 of 30 mm hold the sand of Example 11.3?

No. Its openings are about 0.11 × 30 = 3.3 mm, 7.3 times the sand's coarse grains, so the sand washes straight in: the left-hand panel of Fig. 11.4. A dam steps the size up through a sand filter and a coarser transition before the rockfill.

A clay core needs a different rule. Its D85 may be a few hundredths of a millimetre, and four times that would ask for a filter of very fine sand or silt, which would drain poorly. But clay does not wash out grain by grain from a sound core; it erodes from the walls of a crack, and what the filter must do is catch what comes out of the crack and seal it. James Sherard and Lorn Dunnigan tested filters in the laboratory against many core soils with a leak forced through them, in the 1980s, and found that ordinary filters of clean sand reliably stop erosion from concentrated leaks in fine silts and clays, while broadly graded glacial soils need a relatively fine one.Sherard & Dunnigan (1989); the rules as the agencies now write them in FEMA (2011), table 5-2, p. 84, and the cake on the filter's face, pp. 275–276. The rules the agencies use today come from that work, sorted by how much of the core is finer than 0.075 mm: for a core more than 85% that fine, the filter's D15 may be up to nine times the core's D85; for one 40 to 85% that fine, it must be no more than 0.7 mm whatever the core. The eroded clay builds a thin cake on the face of such a filter, finer as it builds, and the crack stops leaking soil.

11.5 Two cores, two endings

Now Teton and Bennett can be set side by side. Here is Teton's core where it failed, on the right abutment, drawn to scale from the review panel's own section:

Fig. 11.6 · Case file: Teton Dam, 1976Section to scale · traced from the review panel
reservoir at failure, El. 5,301.7 ft Zone 1: windblown silt Zone 2 Zone 3 Zone 5 rockfill jointed rhyolite erosion tunnel grout curtain leak under the grout cap out along the rock crest El. 5,332 ft 02040 m
Dam
Teton Dam, on the Teton River near Newdale, Idaho; built by the U.S. Bureau of Reclamation.
Built
February 1972 to November 1975; filling from 3 October 1975.
Height
305 ft (93 m) above the riverbed, 405 ft (123 m) above the lowest point of its foundation; crest about 3,100 ft (945 m) long.
Type
Zoned earthfill with a wide central core of windblown silt (Zone 1). On the abutments the core was carried down a key trench 70 ft (21 m) deep, with steep sides and a grout curtain beneath it.
What happened
On first filling, 5 June 1976, with the reservoir 3.3 ft below the spillway sill: leaks from about 7 a.m., a tunnel through the downstream face at 10:30, the crest collapsing from 11:55. The panel found that water through open joints under the grout cap eroded a tunnel across the base of the key-trench fill.
Toll
Eleven deaths in the Bureau of Reclamation's count; the panel associated fourteen with the failure, directly or indirectly. Damage estimated at $400 million to $1 billion.
Sources
Teton Panel (1976), pp. iii, v, viii–ix, 1-1, 8-3 and 12-1; the section at station 15+00, p. 12-11; the zones from the contract drawing, p. 1-3. Stene (1996), p. 1.

Teton's core of easily eroded silt was packed into a trench cut in open-jointed rock, with nothing between the silt and the joints. The section crosses the dam at station 15+00 on the right abutment, near where the first leaks appeared, with the reservoir at its level on the morning of the failure. The core, Zone 1, fills a key trench cut down through the most broken rock to rock that could be grouted; below the trench a grout curtain was meant to seal the deeper joints. The panel's tests found the rock immediately under the grout cap poorly sealed. Water reached the base of the trench through it, the dashed blue line; in the silt it eroded a tunnel across the trench, the brick-red bar; and the eroded silt found its way out along the contact between the core and the open-jointed rock downstream, the dashed brick-red line. The panel also found that the trench's steep walls held up part of the fill's weight, so that the silt at its base was pressed together weakly enough for the water to split it. Only the main zones are drawn; the rock's joints are schematic.

What this chapter explains. A concentrated leak (Fig. 11.3, A), possibly opened by hydraulic fracture, in a soil that erodes easily, with an open way out for its grains: no filter stood between the core and the rock.

  • With the scale bar, measure how far the water had to travel through the core, across the bottom of the key trench, compared with across the whole core higher up.
Answers

About 10 m across the floor of the trench, against well over 100 m through the core a little above the original ground. Water in the joints of the rock had only the width of the trench's floor of silt to cross, the narrowest place in the whole core.

Notice how little stands between the core and the rock. The silt of Zone 1 lies straight against the walls and floor of the trench, and the rock there is full of open joints, gaps wide enough to take silt. The panel called this choice of material a major factor in the failure: wind-deposited silt, it found, is highly erodible, and putting it against heavily jointed rock gave the water both something to carry and somewhere to carry it.Teton Panel (1976), pp. viii–ix, conclusions 3, 4, 7, 8, 9 and 13. Its closing judgment is the epigraph of this chapter: too much was expected of the grout curtain, and the design should have made the inevitable leakage harmless. A filter between the silt and the rock would have done that.

Here is Bennett, at its highest, drawn to scale from the dimensions its designers published:

Fig. 11.7 · Case file: W.A.C. Bennett Dam, 1996Section to scale · from the design dimensions
reservoir at its normal maximum shell shell core transition filter drain sinkhole 1, about 80 m deep 183 m 0100200 m
Dam
W.A.C. Bennett Dam, on the Peace River in northern British Columbia; owned by BC Hydro.
Built
1964 to 1967; filling from November 1967, full by the summer of 1970.
Height
600 ft (183 m); 6,700 ft (2,042 m) long.
Type
Zoned earthfill. A core of glacial soil as wide as the water is deep; downstream of it a transition 80 ft (24 m) wide at the base and 33 ft (10 m) near the top, a filter 12 ft (3.7 m) wide and a drain 24 ft (7.3 m) wide. Slopes 2.5:1 upstream and 2:1 downstream, steeper near the crest.
What happened
On 14 June 1996 a pothole on the crest turned out to sit over a survey benchmark casing left off the design drawings; drilling around it on 17 June set off a collapse of 21 ft (6.4 m). Below lay a column of badly disturbed core 2.5 to 3 m wide, in a disturbed zone 6 to 8 m across, about 80 m deep. A second, smaller sinkhole over another casing was found in September 1996. Both were repaired by compaction grouting.
Toll
None. The dam stayed in service.
Sources
BC Hydro expert panel (2012), pp. 1–2, 15, 35–37, 47–48; BC Hydro expert panel (2012), vol. 2, pp. C-4, C-5, C-7 and C-10; the collapse of 17 June: Stewart (1998). The section takes the foundation as level at the dam's greatest height; the drain's blanket is drawn as thick as the drain is wide.

Bennett's core leaked into a transition built to catch what the leak carried, and the transition did. The section is the dam at its full height, with the reservoir at its normal maximum on the left. Downstream of the wide core stand the transition, the filter and the drain, each coarser than the one before, the drain turning into a blanket along the base; the shells carry the weight. The brick-red column below the crest shows how deep the loosened core under the first sinkhole went, about 80 m. That sinkhole lay in a lower part of the dam, near the right canyon wall; it is drawn here at the dam's axis, at its own depth, for scale. The number to remember: about 80 m of loosened core, and no breach.

What this chapter explains. A concentrated leak (Fig. 11.3, A), opened most likely by hydraulic fracture where the core was pressed together weakly beside a casing and near the canyon wall, which washed some of the core into the transition until the transition stopped it.

  • With the scale bar, measure the width of the transition near the base and near the top, and compare them with the filter's width.
Answers

About 24 m near the base and about 10 m near the top, against a filter about 3.7 m wide throughout: the transition is the wide first line of defence, the filter a narrow second one.

Notice the transition, the zone Chapter 9 asked you to keep an eye on. It is many times wider than the filter and sits right against the core. The panel that reviewed Bennett in 2012 concluded that the settling of loosely packed core around the casing, as it got wet, could not by itself explain how far the first sinkhole had sunk: some of the core must have been carried away. The likeliest route was concentrated leaks through cracks opened by hydraulic fracture, where the casing and the nearby canyon wall left the core weakly pressed. The eroded core went into the transition, and the erosion probably went on only for a short time before the transition stopped it. The panel judged the transition, filter and drain together a good filter system, one that might let a little core erode at their boundary but would prevent ongoing erosion.BC Hydro expert panel (2012), pp. 35–37 (the two mechanisms and the cracks), 47 (the sinkholes) and 48 (the filter system). The holes were filled by compaction grouting, and the dam went on working.

That answers this chapter's question. Teton and Bennett both leaked through their cores, and in both the water carried core soil with it. At Teton the soil went into open joints in the rock, nothing stopped it, and in five hours the tunnel had taken the dam. At Bennett the soil went into a transition graded to hold it, and the leak sealed itself against the transition's grains. The water at Bennett went on through, as water in every dam does; the soil stopped, and the grains of the core went on pressing on one another.

The film The sand that boils retells the chapter, from the tube of sand to Bennett:

Film 11.1 · The sand that boils9 min · captions

Before you watch

At 7:29 the film gives Terzaghi's second rule, that a filter's small grains must be at least four times the size of the sand's small grains, and a few seconds later it says the water then passes through the filter at least sixteen times as easily as through the sand. Where does sixteen come from? Example 11.3 has everything you need.

This film was made after Film 12.1, which closes the next chapter. When the narrator says “the last episode”, that's the one.

In this film

  1. A metre of sand in a tube Fig. 11.1
  2. Our dam on sand and gravel Example 11.1
  3. A pipe that feeds itself Fig. 11.2
  4. Teton, 1976 Fig. 11.6
  5. A wall halfway down Fig. 10.3
  6. A filter at the toe Fig. 10.4
  7. How a filter holds sand back Fig. 11.4
  8. Bennett Dam, 1996 Fig. 11.7
Read the transcript

This is a dam made of packed dirt, twenty-four metres tall, standing on sand and gravel. Every dam like this leaks, through it and under it. The water comes out near its foot, which engineers call the toe. Clear water at the toe is just a leak. Cloudy water means the dam itself is coming out with it. In one big study of earth dam failures, nearly half went this way: water carried the soil out of them. So how does a trickle at the toe bring down a whole dam?

First, why soil stands at all. Its grains press on each other, and friction needs that press. Water in the gaps between the grains pushes them apart, and takes part of the press away. You've seen it at the beach. As the tide goes out, there's a glassy strip of sand near the water, where water held in the beach seeps back out. Stand there and wiggle your feet, and they sink.

Under a dam, the water coming up is pushed far harder. To measure how hard, engineers use head: the height water would climb to in a thin tube pushed into the ground. At the lake, the head is about twenty metres. At the toe, it's down to ground level. All of that head is lost on the way under the dam. So how hard can that lost head push?

Say we fill a tube with a metre of this sand, and feed it water from below. The tank starts level with the top of the sand, and nothing moves. The grains at the bottom press on each other with the weight of the sand above them. Now raise the tank. Water climbs through the sand, and as it squeezes through the gaps, it loses head. That lost head is a push on the grains, upward. The higher the tank, the harder the push, and the less the grains press on each other. Raise the tank about a metre above the sand, and the press is gone. The bed boils. Here's why it's a metre.

Soaked sand weighs about twice as much as water. Under water, the water around it holds up half of that weight, so it presses down only about as hard as the same amount of water. So if the water loses a metre of head for every metre it climbs, it carries the sand's whole weight. Head lost per metre is called the gradient, and the one that floats the sand is the critical gradient. For most sands, it's about one. That's the beach strip, with the water pushed hard enough to lift the sand. And it's the idea behind this whole series, at its limit: the grains stop pressing on each other at all.

Now our dam. This is the same dam as in the last episode, now standing on the course's sand and gravel. The course's seepage solver works out where the water goes. These lines are the paths the water takes. Nine tenths of the leak dives into the sand and gravel, runs under the dam, and comes up beyond the toe. On average, under the dam, the water loses only a fifth of a metre of head for each metre it travels. But look where the paths crowd together, as they turn up at the toe.

There, the push is four times the average. That's about zero point eight. That's under one, so the sand isn't boiling. Engineers call soil washing out of a dam piping. The margin against it is one divided by that push: here, about one point three. Designers want at least three. For a new dam, they want four. The course warns that its number right at the toe is a guide, not a measurement. Either way, this margin is thin.

So say a boil starts anyway, at a weak spot near the toe. The water carries sand away, grain by grain, and a hole opens under the dam. The dam above holds up its roof, so the hole stays open, and grows into a channel. And the channel is a shortcut. Water in it flows freely, so all the head is now lost in the shorter stretch of sand ahead of it. Here's a teaching model on the same solver: an open channel along the base of the dam, growing ten metres at a time. Watch the push at its tip. It climbs with every step.

By the time the channel is halfway under, the push is past one, and the leak has nearly doubled. Near the lake end, the push reaches three. And the leak is five times what it was. A shorter path means a steeper push. A steeper push means faster erosion. And faster erosion means a longer channel, and a shorter path. The pipe feeds itself, until the roof falls in and the lake pours through. In some sands, a pipe keeps growing at a far gentler push than one. So inspectors walk the toe looking for boils, cloudy water and sinkholes.

This isn't just a model. In June nineteen seventy-six, Teton Dam in Idaho was being filled for the first time. It was an earth dam ninety-three metres tall, with a core of windblown silt, built against heavily cracked volcanic rock. Both official investigations found that water carried the silt of the core into open cracks in the rock, with nothing in the way.

It wasn't a boil at the toe. But it was the same trouble: soil that washes out easily, water moving through it, and an open way out for the grains. It took about five hours from the first seepage seen near the dam to the breach. Eleven people died. Across earth dam failures, water carrying soil out has brought down more dams than floods pouring over the top.

So the obvious fix is to stop the water. Back to our dam: try a wall in the ground under it, halfway down through the sand. The water just goes under the wall. The leak falls by less than a tenth, and the margin stays far short of three. Take the wall all the way down to rock and it works, where rock is close enough to reach.

Now try the opposite. Build the toe of the dam from a filter: clean, carefully graded sand, right where the water comes out. More water comes out now, about a tenth more. But where water still leaves the ground unfiltered, the push drops to under a tenth. The margin against piping jumps to thirteen. A filter doesn't stop water. It stops soil.

In the last episode, sand like this dried out a slope. Here, it does its other job: it keeps the grains in place.

But how can anything let water through, and not sand? Look at three grains, touching. The gap between them is about a sixth of their size. In a real filter the grains come in many sizes, and its openings end up about a fifth the size of its small grains. Say the sand at the toe is like beach sand, most of its grains about a third of a millimetre. The engineer Karl Terzaghi gave two rules for the filter it needs. First, the filter's small grains can be at most four times the size of the sand's big grains.

Then its openings are smaller than the sand's big grains. So those big grains jam at the filter's face, and hold back the small ones behind them. Second, the filter's small grains must be at least four times the size of the sand's small grains. Then water passes through the filter at least sixteen times as easily as through the sand, so no push builds up inside it. Terzaghi's very first filter was this episode's fix: sand with weight on top, at the toe of a dam built on sand.

It works inside real dams too. In June nineteen ninety-six, crews checked a small pothole on the crest of the Bennett Dam, in British Columbia. The dam is a hundred and eighty-three metres tall. Under the pothole, they found a column of loosened core, about eighty metres deep. It ran down a pipe for a survey marker, put in while the dam was built, and left off the drawings.

An expert panel later found that seepage had washed part of that core away. The fine grains that washed out got into the filter layer beyond the core, and filled its gaps until it held the rest. Since then, the panel wrote, it has worked as a filter should. The holes were grouted, and the dam is still in service.

Every dam leaks. After Teton, the independent panel wrote that the design should have provided measures to render the inevitable leakage harmless. That's what a filter does. It's the series' idea once more: soil is strong when its grains press on each other, and water pushing up through it can take that away. A filter lets the water out, and keeps the grains pressing.

After you watch: the answer

Four times four. Water passes a sand roughly in proportion to the square of its small grains' size, so a filter whose 15% size is four times the sand's passes water about 4 × 4 = 16 times as easily, and one at five times about 25 times (FEMA (2011), p. 88); Example 11.3 made the same step with a ratio of six and got 36. The rest of the film retells this chapter, and a few of its lines need a qualifier. At 0:45 it has the water in the gaps pushing the grains apart; more exactly, it carries part of the load the grains would otherwise carry between them, so they press on one another less (Chapter 4). At 4:11 its pipe grows in a teaching model made for the film, which this book does not hold; Example 11.2 makes the same argument with a head budget you can check. At 5:39 it gives Teton's toll as eleven, the Bureau of Reclamation's count; the review panel associated fourteen deaths with the failure. At 5:41 it says that water carrying soil out has brought down more dams than floods pouring over the top. In the study it draws on, by Foster, Fell & Spannagle (2000), internal erosion caused 47% of the failures and overtopping 34%; count failed spillways and gates with the floods and they reach 48%, so the two come out about level. At 6:21 its filter at the toe raises our dam's margin against piping to about thirteen, while the leak rises about a tenth: in the lab the toe drain gives 12.8 and 2,647 L/day per metre against 2,366. At 6:57 it puts a filter's openings at about a fifth of its small grains, an older and rougher figure than the 0.11 measured since. The film stops here, before a closing question about tailings dams that Chapter 8 answered.

What to remember

  • Water rising through sand drags its grains upward. At the critical gradient, about 1, the drag matches the grains' weight under water, they stop pressing on one another, and the sand boils. Designers want an unfiltered exit at a third or a quarter of that.
  • Grains washed out at an exit leave a hollow that works back toward the reservoir, and every metre it gains shortens the path and steepens the push at its tip: a pipe feeds itself.
  • Water starts carrying soil out of a dam in four ways: a concentrated leak through a crack, backward erosion from an exit, suffusion of fine grains through coarse ones, and contact erosion along a boundary between soils.
  • A filter stops soil, not water. Its 15% size is at most four times the protected soil's 85% size, so the coarsest grains lodge at its face and hold back the rest, and at least about four times the soil's 15% size, so water leaves freely. Clay cores need the finer, tested rules of Sherard and Dunnigan.
  • Teton's leak met open rock and Bennett's met a transition. In both the water went on through; only at Bennett did the soil stay, and the grains of the core keep pressing on one another.

Questions

  1. 11.1

    Once a pipe has started under a dam, why does it speed up instead of slowing down?

    Hint

    The reservoir fixes the head. Over what length of sand does the water have to lose it once the pipe is open?

  2. 11.2

    A sand to be protected has D15 = 0.15 mm and D85 = 0.6 mm. Between what limits must its filter's D15 lie?

    Hint

    Use both rules of Eq. 11.2, with a factor of 4 in each.

  3. 11.3

    Which of the four ways of Fig. 11.3 began the boil in Fig. 11.2, the failure of Teton, and the first sinkhole at Bennett?

    Hint

    Ask where each began: at an exit in the ground, or in an opening inside the dam?

  4. 11.4

    Bennett and Teton both leaked through their cores into something coarser. Why did one survive?

    Hint

    Compare the gaps in what each core leaked into with the size of the core's grains.

  5. 11.5

    In Fig. 11.2, switch the toe to Filter + weighting berm and play the sequence. What does the filter stop, and what does it let through? What is the berm's weight for?

    Hint

    Think of the tube of Fig. 11.1: what would a weight on top of the sand do to the press between its grains?

Further reading

  • Stene, E.A. (1996). The Teton Basin Project. U.S. Bureau of Reclamation, Denver. https://www.usbr.gov/pn/snakeriver/dams/uppersnake/teton/projhistory.pdf gentle The Bureau of Reclamation's own history of the Teton project, from the first surveys through its building, the failure and what came after. Short, plain and free.
  • Federal Emergency Management Agency (2011). Filters for Embankment Dams: Best Practices for Design and Construction. Washington, DC, October 2011. https://fema.gov/sites/default/files/2020-08/filters_embankment_dams_update.pdf serious The standard practical guide in the United States. Its first chapter explains how filters work and how internal erosion starts; its fifth works the design rules step by step; an appendix tells the history, from Terzaghi to Sherard and Dunnigan. Free from FEMA.
  • Independent Panel to Review Cause of Teton Dam Failure (1976). Report to U.S. Department of the Interior and State of Idaho on Failure of Teton Dam. December 1976. https://damfailures.org/sites/default/files/wp-pdf/IR_Independent-Investigation_Teton.pdf primary source The independent panel's report. Its summary and conclusions take a few pages and are a model of plain engineering judgment; the second chapter rebuilds the morning of 5 June from the witnesses' own words.
Part IV

Will it stand?

By the end of Part IV

An engineer judges a slope by one number: the grip the soil can offer, divided by the grip it needs. Part IV shows how that number is worked out, by hand and by computer, and then hands you a dam to design. The last chapter tests the number against a dam that passed it.

  1. Chapter 12Will the slope stand?

    The factor of safety on a curved surface, worked by hand and by computer.

  2. Chapter 13Build your own dam

    A workbench, five missions, and Mount Polley's answer.

Chapter 12 · Part IV · Will it stand?

Will the slope stand?

This dam is standing, with water breaking out of its face. How close is it to sliding?

So far. Chapter 4 gave a block on a ramp a score, the factor of safety, and showed that water seeping out of a sand face roughly halves the steepest slope it can hold, measured as rise over run. Chapter 10 followed the water through our dam, and Chapter 11 what it does where it comes out. One question is left to ask of any dam with water inside it: will its slope stand, and by how much?

After this chapter you can

  • work out the factor of safety of a long, straight slope, dry and with water seeping out of it;
  • work one slice of Bishop's method by hand, say why the method has to go round in a loop, and add up a whole slip circle;
  • explain why emptying a reservoir quickly can bring down its upstream slope, and say what a factor of safety of 1.5 does and does not promise.

This chapter adds a slip circle through the downstream slope

3 figures · about 25 min to read · Film 12.1, 6 min 4 s

An earth dam rarely announces that it is in trouble. The first sign is often a strip of its downstream face that stays green through a dry summer, or a band of dark, soft ground a few metres up the slope where water seeps out into the light. Walk our dam, the 24 m embankment of the last three chapters, built all of one compacted soil on clay ground, and you would find that band 8 m up its downstream face, exactly where the seepage lab of Chapter 10 said the water would come out. Nothing is moving, and the crest road is level.

The engineer who checks the slope gets a single number back, a factor of safety of 1.12. Above 1 the slope stands, so this one stands, with 12% to spare. The owner has money for one fix. Make a choice before you read on:

A dam made all of compacted fill has a factor of safety of 1.12, with water breaking out 8 m up its downstream slope. Which change helps it most?

Whatever you chose, the choice rests on a number, and a number is only worth what you know about how it was made. To judge the fixes we need to make that number ourselves, starting where Chapter 4 left off: with a slope so long and even that one small piece of it stands for the whole.

12.1 The long slope, scored

Chapter 4 found the steepest face a sand can hold by treating one patch of the face as the block on its ramp. A designer doesn't build at the steepest face; she builds flatter and asks how much flatter. So ask the same patch a different question: how much grip does it have, against how much it needs? That ratio is the factor of safety, and for a long, even slope of grains with no glue between them, the slope Chapter 4 called an infinite slope, it takes one line.

On a dry slope at angle β, a slab of soil parallel to the face presses into the slope below it with its weight times cos β, and is pulled down it by its weight times sin β. Friction offers tan φ′ times the press. The weight cancels, as it did for the block:

FS = tan φ′ ÷ tan β(12.1)

φ′ the soil's friction angle · β the slope's angle; a slope of n:1 has tan β = 1/n, so FS = n · tan φ′. It holds dry, and it holds under still water, where buoyancy lightens the press and the pull alike.

The formula hides a pleasant fact: the factor of safety grows in proportion to how flat the slope is. Make a 2:1 slope into a 3:1 slope and its safety rises by half. But the water has its own say. Where water seeps out of the face, parallel to it, its drag adds to the pull and nothing to the press, and Example 4.5 showed what that costs: the grip falls in the ratio of the soil's weight in water, its buoyant unit weight γ′, to its soaked weight γsat:

FS = (γ′ ÷ γsat) · tan φ′ ÷ tan β(12.2)

γ′ = γsat − γw, the soaked weight less the water's, 9.81 kN/m³. For most soils γ′/γsat is close to one half.

Here are both rules drawn for two materials of the workbench in Chapter 13, the tailings sand and the rockfill, with two results from the full slope check marked beside them:

Fig. 12.1 · The long-slope rule, and the solver beside itChart · factor of safety against slope
0 0.5 1 1.5 2 2.5 1:1 1.5:1 2:1 2.5:1 3:1 3.5:1 4:1 flatter → across for every 1 up factor of safety below 1, it slides target 1.5 our dam’s 2:1 rockfill, dry tailings sand, dry tailings sand, water seeping out of the face 1.299 0.645 1.801 solver: 0.637 solver: 1.814

For a long slope of grains without cohesion the factor of safety rises in a straight line as the slope flattens, and water seeping out of the face halves it. Each line is Eq. 12.1 or Eq. 12.2 for one material: the top line for dry rockfill (φ′ = 42°), the middle line for dry tailings sand (33°), and the bottom line for the same sand with water seeping out of its face. The dashed level lines mark 1, where the slope slides, and 1.5, the usual target with a full reservoir; the dotted upright marks the 2:1 slopes of our dam. The two open rings are not from the formula: they are the workbench's full search of curved slip surfaces through a whole dam, one built all of tailings sand (left lower ring) and one built all of rockfill, checked right after its reservoir was emptied (upper ring). The number to remember: at 2:1 the tailings sand scores 1.30 dry and 0.645 with water seeping out.

  • Read off how flat a dry tailings-sand slope must be to reach 1.5, then how flat the seeping one must be.
  • Compare each open ring with the point on the line beside it.
Answers

Dry, about 2.3:1 (1.5 ÷ 0.6494 = 2.31); seeping, about 4.6:1, off the right of the chart. The rings sit almost on the lines: the solver's 0.637 against 0.645, and 1.814 against 1.801.

Notice that the seeping line is the dry line pressed down to half its height, everywhere: γ′/γsat does not depend on the slope. Notice too the two rings. The workbench knows nothing about long slopes. It searches thousands of curved surfaces through a whole section, with the water pressures its seepage solver computes, and for these two dams it lands within about 1% of a one-line formula. The example says why:

Example 12.1

A tailings-sand slope, dry and seeping

Given
The workbench's tailings sand: φ′ = 33°, no cohesion, soaked γsat = 19.5 kN/m³; rockfill φ′ = 42°; water 9.81 kN/m³. Both slopes at 2:1, so tan β = 0.5.
Find
The factor of safety of a long slope of tailings sand, dry and with water seeping out parallel to its face, and of a long dry slope of rockfill.
The idea
One slab parallel to the face stands for the whole slope (Eq. 12.1); seeping water adds a pull and no press (Eq. 12.2).
  1. Dry sand: tan 33° ÷ 0.5 = 0.6494 ÷ 0.5 = 1.299grip over pull for one slab; the weight cancels
  2. γ′ = 19.5 − 9.81 = 9.69 kN/m³the sand's weight in water
  3. 9.69 ÷ 19.5 = 0.4969the share of the pull that comes with a press
  4. Seeping sand: 1.299 × 0.4969 = 0.645the drag of the leaving water adds pull and no press
  5. Dry rockfill: tan 42° ÷ 0.5 = 0.9004 ÷ 0.5 = 1.801rougher, more angular grains, the same rule

Dry, a 2:1 slope of tailings sand scores 1.30; with water seeping out of it, 0.645, and it slides. Dry rockfill at 2:1 scores 1.80.

Check. The workbench's whole dam of tailings sand on clay scores 0.637, and its worst surface is a skin 2.0 m deep, the shallowest the check counts: nearly a slab parallel to the face, the very thing the formula assumes. Its rockfill dam, right after its reservoir has been emptied, scores 1.814 on a skin 2.0 m deep. The ratio of the two sand answers, 0.645 ÷ 1.299 = 0.497, is γ′/γsat and nothing else.

What it means. Water seeping out of a face halves its safety. A slope that grips well enough dry can slide with water coming out of it, without a grain changing; only the water moved.

Check yourself A long dry rockfill slope is built at 1.5:1, steeper than our dam's. What is its factor of safety?

1.5 × tan 42° = 1.5 × 0.9004 = 1.351: it stands, with a third to spare, as long as it stays dry. Rockfill dams are often built that steep.

12.2 Slopes slide on curves

One could assume that every slope fails the way the long-slope rule pictures it, a thin skin peeling off its face like snow off a roof. Clean sand and rockfill do fail that way, which is why the rule and the solver agreed. Our dam's fill does not. It has a little cohesion, c′ = 10 kPa, a grip that doesn't depend on how hard its grains are pressed, and a thin skin has too little weight for its friction to matter and plenty of cohesion for its size. A deep slip is the cheaper failure: it has more weight pulling, through soil that is wetter and so presses less.

The shape engineers give a deep slip comes from Sweden. In March 1916, a year after it was built, the Stigberg Quay in Gothenburg harbour slid, and a committee was set up to find out why.Petterson told the story himself in English, nearly forty years later (Petterson (1955)). Wolmar Fellenius, who chaired Sweden's state geotechnical commission, extended the method to soils with both friction and cohesion in the 1920s and defined the factor of safety for a slope as the ratio of the moments holding it to those turning it (Massarsch & Fellenius (2012)). The commission's final report of 1922 is the first publication known to use the word "geotechnical". The committee's analysis used a method worked out by Knut Petterson, an engineer with the harbour, and Sven Hultin, a professor at Chalmers: take the surface the ground slides on to be an arc of a circle. A circle has a convenient property: a mass sliding on one turns about the circle's centre as a solid piece, so its balance can be checked with one sum of turning effects. The Swedish engineers cut the sliding mass into upright slices and added them up. Most checks of an embankment slope still start there.

Look at one slice and you have met it before. It is Chapter 4's block on a ramp, except that its ramp is a short stretch of the circle, tilted at an angle α that changes from slice to slice: steep under the crest, nearly level near the bottom of the circle, tilted the other way where the circle climbs back out beyond the toe. Each slice's weight W pulls it down its stretch with W sin α; the soil under it grips back with cohesion over the length of the base, l, plus friction from the grains' press on it, N′. This is the method of slices. The slip surface's factor of safety is the grip of all the slices over the pull of all of them:

FS = Σ (c′·l + N′·tan φ′) ÷ Σ W·sin α(12.3)

Σ the sum over every slice · c′·l + N′·tan φ′ the grip on a slice's base (kN per metre of dam) · W·sin α its pull. The ratio comes from balancing turning effects about the circle's centre: each slice's weight acts at a lever arm of R sin α from it, each grip at R, the presses on the bases point at the centre and turn nothing, and the radius cancels.

The pull is easy: weights and angles, measured off the drawing. All the difficulty is in N′, how hard the grains on each base press, and that is where the water comes in.

12.3 One slice by hand

The water in the gaps pushes on the base of a slice with its pressure u over the base's length, u·l, and whatever part of the slice's load the water carries, the grains don't. The earliest method of slices took the press on the base to be the part of the weight at right angles to it, less the water: N′ = W cos α − u·l. That treats each slice as if its neighbours weren't there. But they are, and they push on it.

Alan Bishop's step, published in 1955, was to let them push, but only sideways, and to balance each slice up and down.Bishop, at Imperial College London, published the method in Bishop (1955). Taking the slices' pushes on each other as horizontal is why it is called his "simplified" method; his paper also gave a fuller version that keeps their upright parts, which matter little on a circle. E. Spencer's method of 1967 (Spencer (1967)) balances the forces and the turning effects together, taking the pushes as parallel, and works on surfaces of any shape. Sideways pushes have no upright part, so they drop out of that balance. What holds the slice up is then the water and the grains pressing on its base, both at right angles to it, and the grip along the base, which at the point of balance is the grip available divided by the factor of safety. Solve that balance for N′ and you get Bishop's method:

N′ = (W − u·b − c′·b·tan α ÷ FS) ÷ mα,  mα = cos α + sin α · tan φ′ ÷ FS(12.4)

b the slice's width (m) · u·b the upright part of the water's push on the base · c′·b·tan α ÷ FS the upright part of the cohesion's share of the grip · mα the factor that turns what is left into a press at right angles to the base.

Look where FS is: on the right-hand side. To know how hard each slice presses, you must already know the factor of safety, and to know the factor of safety you must know how hard every slice presses. The way out is to guess, work every slice with the guess, add them up with Eq. 12.3, and use the answer as the next guess. Here is the critical circle of our dam, the worst of the 2,000 or so the workbench tried, with one of its slices drawn large:

Fig. 12.2 · Our dam's worst slip circle, and one of its forty slicesA to scale, from the solver · B to scale, forces to scale with each other
reservoir, 20.4 m compacted fill clay ground water level inside the dam 40 slices slice 13 water breaks out, 8 m up the face centre of the circle R = 42.9 m α worst circle, FS = 1.116 A · The critical circle B · Slice 13, enlarged forces to scale with one another weight W 319.0 kN neighbours push sideways only grip 129.3 kN α pull, W sin α 168.8 kN water, u·l 122.3 kN grains, N′ 181.3 kN

The weakest surface through our dam is a circle that cuts deep, through the soaked fill and a little way into the clay beyond the toe, and slice 13 of its forty is a block on a ramp tilted at 31.9°. In A the dashed line is the water level inside the dam, from the reservoir down to where it breaks out of the downstream face (the heavy brick-red stretch); below it the fill is soaked. The brick-red arc is the slip circle the solver found, with its centre high above the slope; the thin uprights cut the soil above it into 40 slices of equal width. The angle α between the circle's radius and the vertical is the tilt of the slice's base. In B, slice 13 carries its weight straight down; on its base the water (u·l, blue) and the grains (N′, ochre) press at right angles, the grip (green) acts uphill along it, and the dashed arrow is the pull, the part of the weight along the base. The dashed side arrows are the neighbours' pushes. The number to remember: 1.116, the whole circle's grip over its pull.

  • In A, measure the circle's depth below the face at its deepest, with the slices' uprights as a ruler: each is 1.36 m wide.
  • In B, compare the grip with the pull. Is slice 13 holding itself up?
Answers

About 12 m. Slice 13 is not: its grip is only 0.77 of its pull (129.3 ÷ 168.8 = 0.77), and the slices lower down make up the difference.

Notice that the circle goes deep: about 12 m below the face, far below any skin, and out 6 m beyond the toe. Notice also that most of it runs below the dashed line, in soaked fill. Now take slice 13, the thirteenth from the top, and work it the way the solver does, with the factor of safety the loop ends on:

Example 12.2

Slice 13 of 40, by hand

Given
From the solver, per metre length of dam: width b = 1.359 m, weight W = 319.0 kN, water pressure on its base u = 76.40 kPa, base tilted at α = 31.94° (sin 0.5291, cos 0.8486, tan 0.6235). The fill: c′ = 10 kPa, φ′ = 32°, tan φ′ = 0.6249. The loop's answer, FS = 1.116.
Find
The grains' press on the base, N′; the grip on the base; the pull along it.
The idea
Balance the slice up and down (Eq. 12.4), then grip = cohesion along the base plus friction from the press (Eq. 12.3).
  1. u·b = 76.40 × 1.359 = 103.8 kNthe water's push up on the base
  2. W − u·b = 319.0 − 103.8 = 215.2 kNthe weight the water leaves for the grains
  3. c′·b·tan α ÷ FS = 10 × 1.359 × 0.6235 ÷ 1.116 = 7.59 kNthe cohesion along the tilted base also holds the slice up a little
  4. mα = 0.8486 + 0.5291 × 0.6249 ÷ 1.116 = 1.1449the factor that turns an upright load into a press at right angles to the base
  5. N′ = (215.2 − 7.59) ÷ 1.1449 = 207.6 ÷ 1.1449 = 181.3 kNthe grains' press on the base, Eq. 12.4
  6. l = 1.359 ÷ 0.8486 = 1.601 mthe tilted base is longer than the slice is wide
  7. Grip: 10 × 1.601 + 181.3 × 0.6249 = 129.3 kNcohesion over the base's length, plus friction from the press
  8. Pull: 319.0 × 0.5291 = 168.8 kNthe part of the weight along the base, W sin α

The grains press on slice 13's base with 181.3 kN, giving a grip of 129.3 kN against a pull of 168.8 kN.

Check. The water carries 103.8 of the slice's 319.0 kN, about a third, close to the share it carries on the whole circle (0.35). The solver, which carries every digit, gets 181.38 and 129.35 kN: the rounding on this page costs less than a tenth of a kilonewton.

What it means. On its own, slice 13 would slide: its grip is 0.77 of its pull. A slip circle stands or falls as a whole, and a third of this slice's weight is carried by water, not grains.

then press Test in the workbench

Do the same for all forty slices and add up. The grips come to 3,473.6 kN and the pulls to 3,111.8 kN, and 3,473.6 ÷ 3,111.8 = 1.1163, the readout's 1.12. The sum gives back the factor of safety the slices were worked with, which is how you know the loop has finished. Start it where the solver does, at 1.5, and the forty slices return 1.181; feed that back and they return 1.129, then 1.119, 1.117 and 1.116, and after that the third decimal stops moving.

The slices that save slice 13 are the ones near the bottom of the circle. Where the circle climbs back up toward the toe, a slice's base tilts the other way, and its weight pushes back against the slide instead of driving it; 11 of the forty are like that. Their grip counts in full, and their pull is negative: it takes away from the total. The older method that ignored the neighbours scores this same circle at 0.978: it says the dam should already have slid. It errs on the safe side, by a lot when the water pressure is high, and that is what Bishop's sideways pushes fixed.

Check yourself If you started Bishop's loop at a guess of 1.0 instead of 1.5, would it end somewhere else?

No. The guess only sets the starting point; the loop keeps going until the slices return the number they were worked with, and only one number does that, 1.116. From 1.0 it climbs to it instead of falling.

12.4 What the water costs

Slice 13 lost about a third of its weight to the water. So did the whole circle, and that is the number that explains the fixes. Bishop and Morgenstern, writing in 1960, put a slope's water into a single ratio, the pore-pressure ratio ru, the share of the soil's weight that the water pressure carries.In Bishop & Morgenstern (1960) ru is the pore pressure at a point divided by the weight of soil above it, and the paper's charts give a slope's factor of safety as a straight line falling with ru. The workbench reports one value for a whole surface, the slices' Σ u·b ÷ Σ W. The paper's second author, Norbert Morgenstern, is the engineer the Prologue met at the head of a review panel. Work it out for our circle and then take the water away:

Example 12.3

The same circle without its water, and the three fixes

Given
The critical circle of Fig. 12.2, FS = 1.116; its slices weigh 8,375.9 kN in all, and the water pushes up on their bases with Σ u·b = 2,955.1 kN. The dam is 1,296 cells of 1 m² in section. The workbench's results for each fix.
Find
How much of the weight the water carries, what the circle would score without it, and what each fix buys for what it costs.
The idea
A fix works by adding grip, adding weight where it resists, or taking the water away. Compare them on one scale.
  1. ru = 2,955.1 ÷ 8,375.9 = 0.3528the share of the weight the water carries
  2. With u = 0 on the same circle: FS = 1.741the same slices, weights and angles, the water pressure set to zero
  3. 1.741 ÷ 1.116 = 1.560the water costs the circle more than a third of its score
  4. Flatter, 2.5:1: 1,440 ÷ 1,296 = 1.111; FS = 1.27311.1% more fill
  5. Flatter, 3:1: 1,596 ÷ 1,296 = 1.231; FS = 1.41823.1% more fill, still short of 1.5
  6. Cutoff wall under the dam: FS = 1.142the leak falls from 44.4 to 40.4 L/day per metre
  7. Drain: 126 ÷ 1,296 = 0.097; FS = 1.628filter sand in place of 9.7% of the fill, no soil added

The water carries 0.35 of the circle's weight. Without it, the same circle would score 1.74. The drain reaches 1.63 without adding any soil; the flatter slopes reach 1.27 and 1.42 for 11% and 23% more fill; the cutoff barely moves it.

Check. Why does flattening help so little? The water still breaks out 8 m up the face at 3:1, as at 2:1, and the share of weight it carries on the critical circle grows, to 0.380 at 2.5:1 and 0.412 at 3:1: the extra fill sits in the wet part of the slope. The US Bureau of Reclamation's design manual says it plainly: in a dam of one soil, seepage will reach the downstream slope "regardless of its flatness".USBR (1987), p. 190, on the homogeneous section, which it says has been replaced in practice by one with "small amounts of carefully placed pervious materials" to control the seepage.

What it means. The fill was strong enough all along; its water was taking the grip. A fix that removes the water does more than one that adds soil, and the cutoff, which stops a little water going under the dam, leaves the water in the dam where it was.

then press Test in the workbench

Notice what the drain does to the leak: the drained dam lets through 109.7 L/day per metre, 2.47 times as much as without it. It is safer because it leaks more, through a path the engineer chose, the trade Example 10.4 found for piping. A slope feels the water's pressure; how much water passes never enters its sum.

12.5 Emptying the reservoir

So far the reservoir has been full and the danger on the downstream side, where the water comes out. The upstream face lives the other way round. A full reservoir is good for it: Example 2.5 showed that water standing on a sloping face presses it down harder than it pushes it across, and under still water the grains of the upstream shell lose their press and their pull in the same proportion. The danger comes when the reservoir is let down fast, to reach a damaged outlet, to take the load off a dam in trouble, or because the water is needed downstream.

Emptying a reservoir takes days or weeks. Draining the water out of compacted fill takes far longer: at 10⁻⁷ m/s, the fill passes water a hundred times more slowly than the tailings sand and ten thousand times more slowly than the filter sand of a drain, and Chapter 5 showed that the time a layer takes to drain grows with the square of its thickness. So the fill's water stays in while the reservoir's water goes. The water inside still has pressure, now with nothing outside to balance it; it seeps out through the face, and its drag pulls the face's grains downhill, exactly as on the beach face of Fig. 4.4. Here is the upstream half of our dam before and after:

Fig. 12.3 · The upstream slope, before and after a fast drawdownTo scale, from the solver · upstream half of our dam
reservoir, 20.4 m soaked fill A · Reservoir full reservoir gone still soaked worst circle, FS = 0.94 B · Emptied fast

Empty the reservoir fast and the fill keeps its water, so the upstream slope loses the reservoir's support while the water inside still presses. Both panels show the upstream half of our dam, all compacted fill on clay. In A the reservoir stands 20.4 m deep and presses on the face at right angles (the brick-red arrows, longer with depth); the fill below the dashed water level is soaked. In B the reservoir has gone but the fill is soaked to the same line, and its water seeps out through the face (the short blue arrows crossing it). The brick-red arc is the worst slip surface the workbench finds on this side once the reservoir is gone. The number to remember: 0.94, below 1: this slope would slide.

  • Find the deepest arrow in A. How hard does the reservoir press on the foot of the face? (Depth times 9.81.)
  • In B, find where the slip circle comes out. Is it inside the dam?
Answers

The reservoir presses on the foot of the face with 20.4 × 9.81 = 200.1 kPa, all of it gone in B. The circle comes out on the reservoir floor 1.8 m beyond the upstream toe, through the clay ground.

Notice that the water inside the dam is the same in both panels, below the same line. Its pressure has fallen only by the weight of the water taken off above it, while the reservoir that pressed on the face from outside has gone altogether. In Chapter 4's terms, the upstream slope went from a heap under still water to a heap with water seeping out of its face. Engineers call this rapid drawdown, and the workbench checks it the classic way: water pressure on each slice comes straight from the seepage solution; the free-draining materials (anything that passes water at 10⁻⁵ m/s or faster: rockfill, filter sand, tailings sand and the sand & gravel foundation) drain with the reservoir, while clay and fill keep their water pressure, less only the weight of the water taken off above them. The reservoir is taken all the way down at once, and the fill is given no time at all to drain: a deliberately pessimistic picture of a drawdown that in life takes days or weeks.

Example 12.4

Four dams, emptied fast

Given
Our dam's section with its 20.4 m reservoir emptied at once to the original ground, four ways of building it: all compacted fill on clay; the same with the downstream drain; a clay core between rockfill shells, on clay; all rockfill, on sand and gravel. The workbench's upstream check for each.
Find
Which upstream slopes stand, and why.
The idea
Only material that can drain as fast as the reservoir falls gets its grip back as the water goes. Compare each result with the long-slope rule.
  1. All fill: FS = 0.940, ru = 0.441the trapped water carries more of the weight than at full reservoir downstream (0.353)
  2. Same circle, water pressure set to zero: 1.705the soil itself is strong enough
  3. With the downstream drain: FS = 1.045a drain on the far side does little for this face
  4. Clay core, rockfill shells: FS = 1.094this core is wide, and the circle still cuts its clay
  5. All rockfill: FS = 1.814, against the long-slope 1.801the rockfill drains as the reservoir falls: a dry slope

The fill slope slides at 0.94; the drain and the core leave it near 1.0 to 1.1; all rockfill stands at 1.81, the score of a dry slope.

Check. The rockfill dam's worst surface is a skin 2.0 m deep, a long slope in all but name, and its 1.814 is within 1% of tan 42° × 2 = 1.801. The fill dam's surface is a circle 9.3 m deep: its cohesion pushes the slip deep, into the water.

What it means. A dam that must be emptied quickly needs a free-draining upstream shell, thick enough that the slip cannot run behind it through wet fill. Example 9.2 made the same point with a long slope; the workbench's "Empty it fast" mission in Chapter 13 asks you to design one.

then press Test in the workbench

Real reservoirs have done this. San Luis Dam, a large earth dam in California whose reservoir is filled and drawn down as water is stored and released, was found on 14 September 1981 to have slid along its upstream slope, after an unprecedented drawdown of its reservoir, by about 55 m in 120 days.Stark (1987), abstract: the slide's discovery on 14 September 1981, the drawdown of about 55 m (180 ft) in 120 days, and seven earlier drawdown cycles. US Bureau of Reclamation measurements at Alcova Dam, Wyoming, lowered 120 ft in 40 days, showed "appreciable pore water pressures" left in the embankment afterwards: USBR (1987), p. 242. It had been drawn down seven times before, some of them nearly as far. The investigation found clayey soil under the slope that had softened over fourteen years of the reservoir rising and falling, so the grip the slope had on the day was less than the grip it had been built with.

Check yourself A reservoir is lowered by the same 20 m over a year instead of a week. Is the upstream fill slope in the same danger?

Much less. The fill's water has time to leave as the level falls, so its pressure follows the reservoir down, and the slope passes through a series of nearly steady states. Rapid drawdown is a matter of time, like drained and undrained loading in Chapter 5.

12.6 What the check does, and what it leaves out

For circular slip surfaces the workbench uses Bishop's simplified method, the standard hand-and-computer method since 1955. It cuts the sliding mass into 40 vertical slices, works out how hard each slice presses on its base after taking away the water pressure there, and balances the turning effect of the whole mass about the circle's centre. It tries more than two thousand circles, from many entry and exit points and radii, then searches around the worst few, and keeps the lowest score.It ignores any slip shallower than 2 m. On a face of clean sand or rockfill the long-slope rule gives every depth the same score, so a search left to itself drifts to the thinnest skin it may, a few grains deep. Grains rolling off a face are called ravelling, and mending them is a maintenance job; the dam itself has not moved. On the weak-layer foundation it also tries sliding blocks with a flat base along the layer, using Spencer's method, which balances forces and turning effects together and so works on surfaces that are not circles.

Suction above the water level is ignored, as is usual practice, because it can vanish in a wet spell. Each set of conditions a slope is checked for is a load case, and the workbench checks two: the downstream slope with the reservoir full and the seepage steady, and the upstream slope after a rapid drawdown. It leaves out the others a designer must run: the end of construction, when the fill has only now been placed and the water squeezed into it by the weight of the layers above has not yet drained; an earthquake; a flood that raises the reservoir above its normal level. The workbench in Chapter 13 fixes the upstream slope at 2:1, which is steep for an earth dam; real ones are often flatter because of rapid drawdown.

12.7 What 1.5 means

Each load case comes with a target. The US Army Corps of Engineers, whose manual on slope stability is a common reference, asks a new earth or rockfill dam for 1.5 under long-term steady seepage, 1.3 at the end of construction, and 1.1 to 1.3 right after a rapid drawdown.USACE EM 1110-2-1902, p. 3-2, its table of minimum factors of safety for new earth and rock-fill dams; 1.1 for drawdown from the maximum surcharge pool, 1.3 from the maximum storage pool. The manual says on p. 3-1 that the values "continue to be based on experience". The workbench's missions use 1.5 and, for the drawdown, 1.2.

A factor of safety of 1.5 means the grip available on the worst surface is one and a half times the grip needed to hold it. Turned round, the soil could lose a third of its strength before that surface moved. Nobody derived 1.5 from physics, and it says nothing about the chance of a failure. It is a margin chosen by experience to cover what nobody can measure exactly: the strength of soil that was sampled in a few places and tested in a few small pieces, water pressures that piezometers read at a few points, and a model that is only a model. Designers accept less for the short-lived cases partly because they last days or weeks rather than the life of the dam, and partly because the way they are checked is already pessimistic, as the workbench's instant, total drawdown is.

The margin covers ordinary uncertainty. It cannot cover a wrong input. The pull in Eq. 12.3 is weights and angles, and those are hard to get badly wrong. The grip depends on the strength you feed it, c′ and φ′, and a factor of safety is only as good as that strength. Chapter 5 showed that strength depends on how fast the soil is loaded: given time, the water in a soil leaves as the load comes on and the grains take the press; loaded fast, the water cannot leave, and in a soil whose grains want to pack closer it takes part of the press itself. A calculation fed the slow strength scores a slow loading. It says nothing about a fast one.

The film Every dam leaks retells the opening question of this chapter, from the dam at 1.12 to the drain:

Film 12.1 · Every dam leaks6 min 4 s · captions

Before you watch

This chapter scored the drained dam at 1.63 and stopped there. With the drain in, the circle of Fig. 12.2 is no longer the worst one; the solver finds a different curve. Where would you expect the new worst curve to run, and how much of its weight will the water carry? The film shows it at 4:34.

This film was made before The sand that boils (Film 11.1), which calls it "the last episode". Its closing question is the one Chapter 11 answered.

In this film

  1. A dam at 1.12 the question
  2. Where the water comes out Fig. 12.2
  3. Bishop's forty slices Example 12.2
  4. Fix one: a fatter dam Example 12.3
  5. Fix two: a wall underneath Example 12.3
  6. Fix three: a strip of sand Fig. 13.1
  7. The new worst curve
  8. Controlling seepage §10.6
Read the transcript

This is a dam made of packed dirt, twenty-four metres tall. Water from the lake creeps through it, and comes out of its face. Engineers score a slope by the grip it has, divided by the grip it needs to stay put. At one, it slides. This dam scores one point one two, so it's barely standing. There are three obvious fixes: make it fatter, put a wall in the ground under it, or bury a strip of sand inside it. Which one saves it?

To judge a fix, you need to know why the dam is weak. Soil holds together because its grains press on each other, and friction needs that press. Water in the gaps between the grains pushes them apart, and takes away part of the press. You've seen it at the beach. A wall of damp sand stands up straight, until a wave soaks its base, and then it slumps. So where is the water in this dam?

Every dam leaks. Water finds its way through a dam, under it, through the rock around it, and along any pipe that passes through it. Here, the path that matters is straight through the dam. This is the course's seepage solver, working out where the water goes in a dam of one ordinary soil, standing on clay. A drop of water takes decades to cross it.

This line is the water level inside the dam. Engineers call it the phreatic surface. Where it meets the back slope, the water comes out, about eight metres up. The Bureau of Reclamation's manual says a dam of one soil will always do this, however flat its slope. About two thirds of its leak comes out of that face. So this dam leaks out of its own face. But why would that make it slide?

In nineteen fifty-five, the engineer Alan Bishop published a way to check. Draw a curve through the slope where it might slip. Then cut the soil above the curve into forty slices. Each slice's weight pulls it downhill, and the soil underneath grips back. That grip is friction, and it comes from how hard the grains press. Below the water level, water pressure pushes up on each slice from underneath, and the grip shrinks.

Add up the grip, divide by the pull, and that's the score for this curve. The solver tries more than two thousand curves, and keeps the worst. On the worst one, water carries about a third of the soil's weight. Take that water pressure away, and the same soil on the same curve would grip more than half again as hard. That's why this dam scores so low.

The first fix most people reach for is to make the dam bigger. Flatten the back slope, using about a tenth more dirt, and the score rises to one point two seven. But the water still breaks out at the same height. And all that extra dirt has to be dug, hauled and packed. Flatten it further, with a quarter more dirt, and it reaches one point four two. That's still short of one and a half, the score engineers usually want with the lake full. The manual says it plainly: the water comes out of the slope, however flat you make it.

So try stopping the water instead. Drive a wall down through the ground under the dam. The score barely moves, to one point one four. On this clay ground, only about a quarter of the water was going under the dam in the first place. The wall turns that water up into the dam, and the leak falls by just a tenth. The back slope still weeps. On sand and gravel, where much more of the water goes under, a wall like this can be the right fix. Here, it isn't.

Every dam leaks, so you can't stop the water. Instead, give it a way out. Bury a strip of sand inside the dam, standing just behind the crest, and join it to a layer of sand along the base. Sand lets water through about ten thousand times more easily than this soil, so the water takes it. It takes up a tenth of the dam, and it adds no soil: the sand simply replaces some of the fill. Engineers call this a drain. Now watch the water level inside the dam. It dives into the sand strip before it ever reaches the back slope. The back half of the dam is left with almost no water pressure in it. A drop now gets through in years, instead of decades.

The score jumps to one point six three. The new worst curve runs shallow, near the face, and water carries almost none of its weight. With no water in it at all, the same dam would score one point six seven. The soil was strong enough all along. The water in it was the trouble, and the drain takes that trouble away.

And here's the twist: the drained dam leaks about two and a half times more. But it leaks through a way out the engineer chose.

That should feel familiar. The back slope of a dirt dam is that sandcastle wall, with the tide inside it. The drain keeps the tide out of the wall, so the grains in the slope press with their full weight.

Draining is one of several ways engineers manage seepage. They cut it off, make its path longer, drain it, filter every exit, weigh down the ground where it comes out, and watch it all with instruments. The Bureau of Reclamation says dam design has moved from trying to stop seepage to controlling it. Today, a strip of sand like this is standard practice in dams built of one soil.

But all that water comes out somewhere. Where it leaves the ground, it pushes on the grains it passes. What if it pushes hard enough to carry them away?

After you watch: the answer

Shallower, and almost dry. The drained dam's worst surface runs from the crest to the toe at most 6.3 m deep, against 11.9 m for the circle of Fig. 12.2, and the water carries less than a thousandth of its slices' weight (ru = 0.0007, against 0.353). The drain empties the downstream half of water pressure, so every surface the solver tries runs through soil gripping with its full weight, and the best it can find scores 1.63. Two of the film's phrasings deserve a note. It says the water in the gaps "pushes them apart"; the careful version, from Chapter 4, is that the water carries part of the load the grains would otherwise carry between them, so they press on one another less. And its line that design has moved from stopping seepage to controlling it is the Bureau of Reclamation's own history: its manual records the dam of one soil giving way to one with carefully placed drains. You can paint the drain yourself in Fig. 13.1. The film ends on the question Chapter 11 answered: water leaving the ground can carry the grains with it, unless a filter holds them.

What to remember

  • A slope's factor of safety is its grip over its pull. For a long slope of clean grains it is n · tan φ′ for an n:1 slope, and water seeping out of the face multiplies it by γ′/γsat, about a half: the water adds pull and no press.
  • Real embankments slide on deep curves. Cut the mass into slices, each a block on a curved ramp, and add up their grips and pulls. Bishop's method balances each slice up and down, so the factor of safety sits on both sides of the sum and must be found by going round a loop.
  • On our dam's worst circle the water carries 0.35 of the slices' weight, and costs the slope a third of its score. A drain that takes the water away does more than a quarter more fill.
  • Rapid drawdown takes the reservoir's help away from the upstream face and leaves the fill's water in it. Only a free-draining shell gets its grip back as fast as the water falls.
  • 1.5 is a margin chosen by experience for what nobody can measure exactly. The number is only as good as the strength fed to it, and a soil's strength depends on how fast it is loaded.

Questions

  1. 12.1

    With the drain in, our dam leaks 2.47 times as much water as without it, and its factor of safety rises from 1.12 to 1.63. How can more water make it safer?

    Hint

    What does the slope check use: how much water passes, or its pressure on the slices' bases?

  2. 12.2

    A long slope of the tailings sand of Example 12.1 is built at 3:1, with water seeping out of its face parallel to it. Will it stand?

    Hint

    Use Eq. 12.2 with tan β = 1/3.

  3. 12.3

    The earliest method of slices ignored the neighbours' pushes, taking the press on a base as N′ = W cos α − u·l. Work slice 13 that way, with the numbers of Example 12.2. Does it give the slice more grip than Bishop's method, or less, and what does that do to the dam's score?

    Hint

    You need l = 1.601 m from the example; nothing in this formula depends on the factor of safety.

  4. 12.4

    Why do designers accept about 1.2 for an upstream slope right after a rapid drawdown, but want 1.5 for the downstream slope with the reservoir full?

    Hint

    How long does each condition last, and how was each one checked?

  5. 12.5

    Name one way a dam could pass its slope check at 1.5 and still fail.

    Hint

    A factor of safety is a calculation. What goes into it?

  6. 12.6

    In the workbench, start from the homogeneous dam, set the downstream slope to 3:1 and test it. What is its factor of safety, and where does the water come out? Then try the drain at 2:1.

    then press Test in the workbench

    Hint

    Read the downstream slope's factor of safety, and look for the seepage face on the downstream slope.

Further reading

  • U.S. Bureau of Reclamation (1987). Design of Small Dams, 3rd ed. Denver. https://www.usbr.gov/tsc/techreferences/mands/mands-pdfs/SmallDams.pdf gentle Its chapter on earthfill dams explains in plain words why the dam of one soil weeps from its face, why drains replaced it, and why a dam that may be emptied fast needs a flatter or free-draining upstream slope, with the pore pressures measured at Alcova Dam after a drawdown. Free from the Bureau of Reclamation.
  • Duncan, J.M., Wright, S.G., Brandon, T.L. (2014). Soil Strength and Slope Stability, 2nd ed. Hoboken: Wiley. serious The standard modern account of what this chapter did by hand: soil strength for stability work, the methods of slices from the ordinary method through Bishop's and Spencer's, and how to analyze rapid drawdown properly.
  • Bishop, A.W. (1955). The use of the slip circle in the stability analysis of slopes. Géotechnique 5(1): 7–17. doi:10.1680/geot.1955.5.1.7 primary source Eleven pages that set the method every dam engineer learned for the next half century. The derivation is the one in Eq. 12.4, with the side forces kept.

Chapter 13 · Part IV · Will it stand?

Build your own dam

Can you design a dam that passes every check, and what can the checks not see?

Effect of minor geologic details on the safety of dams

The title of a paper by Karl Terzaghi, 1929

So far. Part I was the ground: grains that grip only as hard as they press, and water in their gaps that takes part of that press away, in seconds in sand and over years in clay. Part II set the dam that holds by its weight beside the dam that holds by friction, and the tailings dam that is never finished. Part III gave the embankment its zones and followed the water through them and out at the toe. Chapter 12 turned all of it into one number for a slope. Now you design the dam.

After this chapter you can

  • design a section that passes the leak, piping and slope checks, on both of its faces;
  • weigh what each fix costs, in fill, clay or water, against the safety it buys;
  • say what the slope check leaves out, and why a dam can pass it and still fail.

This chapter adds the 1 m grid of the workbench over the section

4 figures · about 20 min to read · Film 13.1, 8 min 46 s

When the review panel opened Mount Polley's design files, it found that the side of the dam that later failed had been checked the way Chapter 12 checks a slope. Trial slip surfaces were drawn through it, the grip the ground could offer along each was set against the weight trying to slide it, and the weakest surface gave a factor of safety. During the raise of 2007 to 2011 that number was 1.7, where the designers required 1.3. The Chief Inspector of Mines found that, on the engineers' understanding of the ground, the calculated factor of safety of that embankment was 1.5 or more.Mount Polley Review Panel (2015), p. 64 (1.7 against 1.3); Chief Inspector of Mines (2015), p. 2 (1.5 or greater). By the rules of the trade the dam was safe, and on the night of 4 August 2014 it gave way (the Prologue).

Before finding out what that number missed, design a dam yourself and run the same kind of check on it. This chapter hands you a workbench wired to the tools of the last four chapters, and five missions, each a brief of the kind engineers are given. The last is Mount Polley, rebuilt, and your check will pass it too.

13.1 A workbench with three checks

The workbench is our dam, the 24 m embankment of Plate 0, drawn on a grid of 1 m cells. You paint the cells with five materials, choose the ground it sits on, and set the reservoir. Then the page tests your design with two calculations.

The first is the seepage solver of the lab in Fig. 10.3. It finds where the water inside your dam settles, how much leaks through, and how hard it pushes on the soil where it comes out.

The second asks whether the downstream slope would slide. It tries more than two thousand curved surfaces through the slope (1,960 on a grid, then a few hundred more homing in on the weakest), each a possible slip surface, and on a foundation with a weak layer it also tries flat-bottomed blocks that slide along the layer. On each it adds up, slice by slice as in Chapter 12, the grip the soil can offer (friction, plus the cohesion of a clayey fill) against the pull of the soil's weight trying to slide it downhill. Friction depends on how hard the grains press together, so wherever the seepage solution puts water pressure on a surface, the grip there drops.

Grip divided by pull is the factor of safety, the number of the block on the ramp in Chapter 4 and of the slices in Chapter 12. Below 1 the slope slides. With the reservoir full, dam engineers usually want 1.5 or more, to cover what they can't know about the soil.USACE EM 1110-2-1902, table 3-1, p. 3-2: 1.5 for long-term steady seepage in a new earth or rockfill dam; 1.1 to 1.3 after a rapid drawdown. A third verdict, the upstream slope after the reservoir is emptied fast, waits for the fourth mission. The page reports the three checks in the language of the chapters that built them:

Leaks

Seepage in litres per day for each metre of dam, the flow of Chapter 10. Some is normal. A lot means lost water, or polluted water for a tailings dam.

Pipes

Where water leaves unprotected soil, it pushes on the grains, as in Chapter 11. Rising out of level ground it lifts them; out of a slope it drags them along the face, and the same number is a guide to how hard. Safety against piping is the push that would float sand away divided by the push it gets. For water rising out of level ground the US Bureau of Reclamation wants at least 3 for an existing dam and 4 for a new one (about 2 to 2.5 where the soil is well measured); the missions here ask for 3 at every exit.USBR DS-13 ch. 8, section 8.2.2.2, p. 8-7. The number warns of sand boils; it can't see every kind of internal erosion.

Slides

The lowest factor of safety of any slip surface at least 2 m deep, as in Chapter 12. Shallower slips are surface damage that maintenance fixes.

Here is the workbench. It opens on a dam of compacted fill, untested; the tabs above the drawing hold the five missions, each with a brief, targets and hints:

Fig. 13.1 · Your dam, tested by the seepage solver and a slope stability check Drawn to scale · 24 m embankment · 1 m cells
Paint with

Brush
Finger
Undo
Start from

Drawing scrolls sideways → · set Finger to Scrolls to move it

Foundation
Cutoff
Downstream slope
20.4 m
Seepage–per metre of dam
Max exit gradient–unfiltered exits only
Safety vs piping–ic ≈ 1.0 / exit gradient
Slope factor of safety–downstream, full reservoir
Water in the slope–
Clay used–m³ per metre of dam
Show the check for

Paint a design, or start from one of the presets, then press Test my dam.

Water pressure in the soil (darker is higher) Water level inside the dam (phreatic surface) Critical slip surface and the soil that would slide Water pressure on the slip surface, as the height it would push water up a tube

Paint a dam and test it: the drawing shows where the water sits inside it, and which slip surface that water weakens most. The section is to scale on 1 m cells, with the reservoir on the left. After a test the blue tint is the water pressure in the soil, darker where it is higher; the dashed line is the water level inside the dam, the phreatic surface; thin lines with moving dots are the paths the water takes. A thick brick-red stroke on the downstream face marks the seepage face, where water breaks out, and a ringed dot the steepest unfiltered exit. The heavy line is the weakest slip surface, a circle or a block, with the soil above it tinted by its verdict: green stable, amber low margin, red sliding. The short bars standing on it are the water pressure there, drawn as the height it would push water up a tube. Show the check for switches to the upstream slope after the reservoir is emptied fast. A mission fixes some controls. The number to remember: the dam of fill scores 1.12; a drain inside its downstream half lifts it to 1.63.

  • Start from Homogeneous fill and test it. Then paint a strip of sand filter about 2 m wide standing immediately downstream of the crest, joined at its foot to a 2 m layer along the base out to the toe. Watch the dashed line drop and the bars on the slip surface shrink.
  • Start from Clay core + rockfill, set the foundation to Sand & gravel and test; then set the cutoff to Full. Read in the verdict where the water was going.
  • Start from Homogeneous fill and switch Show the check for to the upstream slope. Then start from Leaky, which is all rockfill.
Answers

A drain like the one in the preset Fill + drain takes the factor of safety from 1.12 to 1.63. The cored dam leaks 5,434 L/day per metre, 99.98% of it under the dam; the cutoff brings it to 2.18. Upstream, after the reservoir is emptied fast, the fill scores 0.94 and the rockfill 1.41.

Notice that the slip surface and the water belong together. Test the dam of fill and the blue bars stand tallest where the surface runs lowest under the dashed line; that is where the water carries most of the soil's weight, and where the grip is thinnest. Every mission below is, one way or another, about putting the water where it can do no harm.

13.2 Mission 1: dry the slope

The first brief is the dam that opened Chapter 12: compacted fill on clay, standing at 1.12 with water breaking out 8 m up its downstream face. There you chose a fix on paper. Here you paint it. Before you do, it is worth asking whether you could have known where the water would come out without any solver at all. Dam designers had a rule of thumb for it long before computers:USBR (1987), section 6.3(c), p. 190, after Justin, Hinds and Creager: in a completely homogeneous dam, seepage "will emerge on the downstream slope regardless of its flatness", to "roughly one-third the depth of the reservoir pool". in a dam of one material, water held long enough behind it will break out on the downstream slope at about a third of the reservoir's depth, however flat that slope is.

hout ≈ H / 3 (13.1)

hout height above the base at which water breaks out on the downstream face of a homogeneous dam (m) · H depth of the reservoir (m)

Example 13.1

A rule of thumb against the solver

Given
Our dam built of fill alone, with the reservoir 20.4 m deep. The solver's answer: water breaks out 8 m up the downstream slope.
Find
Where the one-third rule puts the breakout, and how well it agrees with the solver at three slopes.
The idea
A third of the depth (Eq. 13.1), whatever the slope.
  1. hout ≈ 20.4 ÷ 3 = 6.8 mthe rule
  2. Solver: 8 ÷ 20.4 = 0.392 of the depththe solver's breakout as a share of the reservoir
  3. At 2.5:1 and 3:1: 8 m and 8 ma flatter face moves the toe out but not the breakout up or down

The rule says 6.8 m and the solver 8 m, at every slope the workbench offers.

Check. The rule is rough in both directions. Lower the reservoir to 60%, 14.4 m: the rule gives 14.4 ÷ 3 = 4.8 m and the solver 3 m. The solver works in 1 m cells, so its breakout is good to about a metre.

What it means. Flattening a dam of one material keeps its face wet; the water comes out at the same height and has further to go to the toe. That is why the Bureau of Reclamation's manual says the fully homogeneous section "should seldom be used" (USBR (1987), p. 190), and why the first fix to try is a drain.

Mission 1 holds the slope at 2:1 and asks for a factor of safety of 1.5 with a piping safety of 3. The question that opened the last chapter already told you which fix wins, and the workbench agrees: paint the drain and the dam scores 1.63. But a designer has to pay for a fix, and the three fixes cost different things. Put them side by side as cubic metres changed against safety gained:

Fig. 13.2 · What each fix buys, per cubic metreFrom the workbench's solver · per metre length of dam
1.0 1.2 1.4 1.6 0 100 200 300 m³ of the section added or swapped, per metre of dam factor of safety, downstream target 1.5 as built, 1.12 2.5:1, +144 m³ fill: 1.27 3:1, +300 m³ fill: 1.42 drain, 126 m³ swapped: 1.63

A drain buys about four times as much safety per cubic metre as a flatter slope, and only the drain reaches the target. Across is how much of the section each fix changes, per metre length of dam: the extra fill a flatter downstream slope needs, or the fill the drain replaces with filter sand. Up is the downstream factor of safety the workbench computes. Each line runs from the dam as built to a fix, so its steepness is the safety each cubic metre buys; the dashed green line is the target. The number to remember: 0.41 of factor of safety per 100 m³ for the drain, against 0.11 for the first flattening.

  • Measure the steepness of the two fill lines in factor of safety per 100 m³. Does the second 156 m³ of fill buy as much as the first 144?
Answers

0.1085 per 100 m³ to 2.5:1 and 0.1007 to 3:1, averaged from the start. The second step alone adds 156 m³ and gains 0.1458: 0.1458 ÷ 156 × 100 = 0.0935, less than the first. Flattening gives diminishing returns.

Notice that the two fill lines nearly lie on one another: a flatter slope buys safety at roughly the same, low rate however far you take it, and stops short of 1.5 at 3:1. The drain's line is in a different class. Here is the arithmetic behind the drawing:

Example 13.2

The price of a point of safety

Given
The dam of fill at 2:1, 1,296 m³ per metre, factor of safety 1.1163. Flattened to 2.5:1 it holds 1,440 m³ and scores 1.2726; to 3:1, 1,596 m³ and 1.4184. The drain swaps 126 m³ of fill for filter sand and scores 1.6285.
Find
The safety each fix buys for every 100 m³ of the section it changes.
The idea
Divide what each fix gains by what it changes. Fill added and fill swapped both cost money; the drain's sand must also be graded to the fill, as Chapter 11 requires.
  1. 2.5:1: 1,440 − 1,296 = 144 m³; gain 1.2726 − 1.1163 = 0.1563the extra downstream wedge, and what it buys
  2. 0.1563 ÷ 144 × 100 = 0.1085 per 100 m³the rate for the first flattening
  3. 3:1: 300 m³; 0.3021 ÷ 300 × 100 = 0.1007gain 1.4184 − 1.1163 = 0.3021
  4. Drain: 0.5122 ÷ 126 × 100 = 0.4065gain 1.6285 − 1.1163 = 0.5122, for sand swapped in, not added
  5. 0.4065 ÷ 0.1085 = 3.7the drain's rate over the flatter slope's

Cubic metre for cubic metre the drain buys about 3.7 times as much safety as fill, and it alone passes 1.5.

Check. The flatter slope adds weight on the toe and lengthens the slip, but the water still breaks out 8 m up the face (Example 13.1) and still carries part of the weight on the slip surface. The drain removes that water instead, and on the dam as built the water carried 0.353 of the weight of the soil on the weakest slip.

What it means. No fix is free. The drain's cost is paid in water: it collects what reaches it, so the dam now leaks 2.47 times as much, 110 L/day per metre against 44.4, and leads it away where it can do no harm.

Check yourself You paint a block of filter sand inside the downstream half of the fill, but not joined to the toe. Will it help as a drain does?

Barely. Water that collects in a drain has to leave it; a patch with no way out fills up, and the water in it sits at the same pressure as the fill around it. The workbench treats such a patch as permeable soil, and drains only sand joined to the toe along the base.

13.3 Mission 2: hold the water

The second brief is the opposite problem. A dam of rockfill on sand and gravel stands well but leaks like a sieve: 32,490 L/day for each metre, 93% of it through the dam itself. The brief asks for 100 L/day or less without losing the slope or the piping safety, and with no more than 350 m³ of clay per metre. One could assume that a clay core, the watertight wall of Chapter 9, settles it. Here is what happens when you paint one:

Example 13.3

Where the water goes when you block one path

Given
All rockfill on sand and gravel: 32,490 L/day per metre. Clay passes water at 10⁻⁹ m/s, rockfill at 10⁻⁴ m/s. The preset clay core is 21.4 m wide at the base and 3 m at its top, 23 m up.
Find
The leak with the core, the leak with the core and a full cutoff, and the clay it takes.
The idea
Water takes every path open to it (Chapter 10). A core closes the path through the dam; a cutoff closes the path under it.
  1. Core alone: 5,434 ÷ 32,490 = 0.167the leak falls to a sixth
  2. Core and full cutoff: 2.18 ÷ 5,434 = 0.0004so 99.96% of the core's leak was going under it
  3. Mean core width: (21.4 + 3.0) ÷ 2 = 12.2 mthe core is a trapezoid
  4. Clay: 12.2 × 23 = 280.6 m³ per metremean width times height; the workbench counts whole cells, 286

Core and cutoff together leak 2.18 L/day per metre, with 286 m³ of clay, a factor of safety of 1.81, and every exit far from piping.

Check. The core passes water a hundred thousand times more slowly than the rockfill it replaces (10⁻⁹ against 10⁻⁴ m/s), yet the leak fell only sixfold. A mismatch that large means the water had another way round, and the workbench's verdict names it: of the cored dam's leak, 99.98% passes under the dam, through the gravel.

What it means. Before choosing a fix, find where the water goes. The cutoff is useless on its own (the water would still pour through the rockfill) and nearly everything once the core is in.

13.4 Mission 3: tailings on a budget

The third brief comes from a mine. It wants at least 70% of the dam built of cycloned tailings sand, the coarse part of its own waste (Chapter 8), because that is far cheaper than hauling fill from a borrow pit. Built of tailings sand alone at 2:1 the dam scores 0.64 and slides. The sand has no cohesion, and water seeping out of a slope of it takes about half its grip (Chapter 12). The answer is the first mission's drain with a flatter face: a chimney and blanket drain inside tailings sand at 2.5:1 scores 1.63 with 90% of the dam in sand. Its price is again paid in water, and here it is a large price: the drain collects 7,980 L/day per metre, polluted water that a real mine must catch below the dam and pump back.

13.5 Mission 4: empty it fast

Every brief so far tested the downstream slope with the reservoir full. The fourth tests the other face. A reservoir sometimes has to be emptied in a hurry, to reach a damaged outlet or to take the load off a dam in trouble, and Chapter 12 showed what that does: while the water stands against the upstream face it presses the slope in place, and when it goes, the water trapped in fill and clay stays, pushing out with nothing pushing back. This is rapid drawdown, and for such a short-lived event engineers accept a lower factor of safety, 1.1 to 1.3 depending on the case;USACE EM 1110-2-1902, table 3-1, p. 3-2: 1.1 from maximum surcharge pool, 1.3 from maximum storage pool. the mission asks for 1.2. Empty the dam of fill and its upstream slope scores 0.94.

The obvious remedy is a coat of rockfill on the face, since rockfill drains as fast as the reservoir falls. Paint a skin 5 m thick and test: the factor of safety goes down, to 0.89. The weakest slip runs deeper instead, behind the skin, through fill that still holds its water. A free-draining zone helps only if the slip cannot get round it. Rockfill over the whole upstream part of the dam, from the face in to 3 m short of the crest, scores 1.24; carried in to the crest's edge it scores 1.32; stopped 7 m short, 1.14, not enough. All rockfill drains best of all, 1.41, and leaks 29,621 L/day per metre, which is why the brief caps the leak at 1,000.

Now flip the view back to the downstream slope. The upstream shell has made it worse, 1.03 against 1.12: the reservoir's water now crosses the rockfill without losing head and reaches the fill with nearly all of it. So the shell solves one problem by handing the next zone a harder one. Combine it with the first mission's drain (start from Fill + drain, then paint the upstream shell) and the dam passes everything at once: downstream 1.63, upstream after a fast drawdown 1.30, a leak of 240 L/day per metre and a piping safety over 50. A free-draining shell upstream, a fill that holds the water in the middle, a drain that catches it downstream: you have rebuilt the zoned dam of Chapter 9 from its jobs.

Check yourself The downstream drain does almost nothing for the upstream slope after a drawdown (1.05 against 0.94). Why?

A drawdown's danger is the water trapped in the upstream fill once the reservoir has gone. A drain under the downstream half lowers the water there, far from the upstream slip, which still runs through fill that holds its water.

13.6 Mission 5: Mount Polley

The last brief is the dam of the Prologue, scaled to the workbench: a clay core between rockfill shells, on clay, with a weak layer in the ground below. In the real foundation the weak layer lay 8 to 10 m under a dam almost 40 m high, about a fifth of its height down;Mount Polley Review Panel (2015), p. 135 (the layer) and p. 75 (the height). the workbench's lies 4 to 6 m under a 24 m dam. The layer has no cohesion and a friction angle of 19°, against 28° for the clay around it. The brief asks for a downstream factor of safety of 1.5. The dam starts at 1.21. Decide first how you would spend your money:

In Chapter 12 a drain was the best of three fixes. This dam stands on a weak layer at 1.21. Which change lifts it most?

To see why one answer differs from the others, start with the shape of the slide the check finds. Choose tab 5 and test the dam as it stands.

Example 13.4

A block on a weak layer

Given
The Mount Polley mission: clay core and rockfill at 2:1, on clay (c′ = 10 kPa, φ′ = 28°) with a layer 4 to 6 m down (c′ = 0, φ′ = 19°).
Find
How much weaker the layer is, what shape of slide it allows, and what it takes to reach 1.5.
The idea
Friction is press times tan φ′ (Eq. 4.3), so a smaller angle on a flat, continuous layer offers a flat, weak path.
  1. tan 19° = 0.3443; tan 28° = 0.5317grip per kilopascal of press, layer and clay
  2. 0.3443 ÷ 0.5317 = 0.648the layer grips about 35% less for the same press
  3. Weakest circle 1.313; weakest block 1.209a block that drops at the crest, slides along the layer and rises beyond the toe; it governs
  4. Filters and drains throughout: 1.215the drains never reach the layer
  5. 2.5:1: 1.390; 3:1: 1.559weight on the layer beyond the old toe, and a longer path along it

At 3:1, for 23% more fill, the section scores 1.56 and passes.

Check. Take the layer away: the same 2:1 dam on plain clay scores 1.553. Flattening to 3:1 buys back about what the layer cost, and no more.

What it means. By the workbench's rules this is a dam you could build, and the real panel's verdict points the same way: had the downstream slope been flattened as the original design proposed, it found, the failure would have been avoided. Hold on to the passing number all the same, and ask which strength it used.

13.7 What the number can't see

Most people would stop here, with a passing number. The workbench gave the layer a drained strength, the grip σ′ tan φ′ of a soil whose water keeps pace with the load and leaves as the grains are squeezed. Mount Polley's designers did the same for the clays they knew of under the dam: silts and clays laid down on the bed of a lake fed by melting glaciers, which engineers call glaciolacustrine. After a field program in 2011, their checks gave these soils a peak friction angle of 28°, the same angle the workbench gives its clay foundation. The layer that failed, a higher one, was not in the checks of the side that failed at all.Mount Polley Review Panel (2015), p. 69 (the 28° peak angle after the 2011 field program) and p. 76 (only the lower of two such units was in the Perimeter Embankment's stability analysis). The workbench's 19° stands in for the layer's weakness. Neither is the strength the layer had on the night it failed.

The dam had pressed that clay harder than anything in its past. A clay in that state is loose for its load, like the loose sand of Chapter 5, and when it is sheared it tries to contract. Over the years of raising, the load came on slowly enough for its water to seep away, and the panel found the loading had been essentially drained. A slide is fast. In Chapter 3 water took 11 years to cross a single metre of clay; sheared in minutes, the clay's water could not leave; its pressure rose, the press between the grains fell, and the grip fell with it. That grip is the clay's undrained strength, and Eq. 5.4 gave its rough size for soft clays. The panel's laboratory tests on samples of the layer measured it directly. On average:Mount Polley Review Panel (2015), pp. 76 and 135 (normally consolidated, so contractive, under the dam), p. 101 (a loading path essentially drained), table 5.3.1, p. 49 (peak undrained strength ratio, average 0.24 over all tests), p. 98 (after strain weakening, perhaps 0.13).

su ≈ 0.24 · σ′v against σ′v tan 28° ≈ 0.53 · σ′v (13.2)

su undrained strength of the layer, sheared fast (kPa) · σ′v the vertical effective stress the dam had pressed it to (kPa), which on a level layer is the press across it · 0.24 the panel's average from its tests · tan 28° the drained grip the design gave such clays

Drawn against the press, the two strengths are two straight lines from zero, the strength envelope of Chapter 4 drawn twice:

Fig. 13.3 · One clay, two strengthsBoth axes in kPa, to the same scale
200 400 600 800 0 100 200 300 400 effective stress across the layer, σ′v (kPa) shear strength (kPa) at 600 kPa: slow, drained, tan 28°: 319 workbench stand-in, tan 19°: 207 fast, undrained, 0.24: 144 after it moved, perhaps 0.13: 78 grip that needs the water to leave

Sheared slowly the layer's clay grips with about half its press; sheared fast, with about a quarter, and less once it starts to move. Across is the effective stress pressing across a level layer, up the shear strength it can offer; both axes have the same scale. The ochre line is the slow, drained strength the design gave such clays, tan 28° times the press. The dashed line is the workbench's stand-in, tan 19°. The brick-red line is the fast, undrained strength the panel measured, 0.24 times the press, and the dotted one what the clay may have kept once the slide had moved and softened it. The shaded band is grip that existed only if the water could leave. The dots read each line at 600 kPa, one of the presses the panel tested at; the dam put up to about 800 kPa on parts of the layer (Chief Inspector of Mines (2015), p. 5). The number to remember: fast, the clay offers 0.45 of its slow grip, less than half.

  • Read the slow and the fast strengths at 600 kPa and divide the fast one by the slow one.
  • Where does the workbench's stand-in sit between them, and which way does that err?
Answers

144 ÷ 319.0 = 0.451. The stand-in, 207 kPa, is well above the fast strength: it errs on the generous side, so a check that passes with it says nothing about the fast failure.

Notice that the gap grows with the press. The higher the dam, the more strength the drained line promises and the more of it vanishes if the clay is sheared fast. The clay was also strain-softening: in the panel's tests its grip peaked and then fell as it kept moving, so its peak strength held only until it began to move. A slide that starts on such a clay weakens the ground it slides on, which is how a dam that had stood for years could go in a night.Why check drained strength at all? Because it is the right check for loads that last: over years a clay's water does leave, and a slope must stand on the strength it has then. The undrained check is a separate calculation, run with the clay's measured strength ratio and its stress history. The two answer different questions, and a design needs both wherever a clay can be sheared fast.

Your workbench passed the dam at 3:1, and with a flatter slope the real dam would probably have stood. But the number that approved it was computed for a strength the layer did not have when it mattered. The panel was blunt about where the fault lay: "the dominant contribution to the failure resides in the design." Its foundation investigations had not found a continuous layer near the breach, nor recognized that it could fail undrained.Mount Polley Review Panel (2015), p. iv and p. 105 (the quotation), p. iv (2H:1V would have avoided the failure).

Check yourself The real layer lay 8 to 10 m under a dam about 40 m high; the workbench's lies 4 to 6 m under 24 m. Is the workbench's at the same share of the height?

Nearly. 8 ÷ 40 = 0.20 and 10 ÷ 40 = 0.25; 4 ÷ 24 = 0.17 and 6 ÷ 24 = 0.25. Scaled to its height, the workbench's layer starts a little higher.

13.8 A layer has to be found first

A calculation can only weigh the ground it is told about. Finding out what that ground is, layer by layer and how strong each is, is the work of site investigation: mapping the geology, drilling boreholes and testing what comes up. A borehole samples a column of ground about a hand's width across. Under the side of Mount Polley that failed, the panel found the critical soil at depths of about 8 m, and only four locations where holes went deeper than 8 m and the soil was tested. None was where the breach occurred.Mount Polley Review Panel (2015), pp. 28 and 80. The Chief Inspector counted three deep sampled drillholes along the 2 km of that embankment: Chief Inspector of Mines (2015), p. 2. The layer was there, a few metres thick, and the drilling passed it by.The epigraph is the title of a paper Karl Terzaghi published in 1929, in a collection on geology and dams issued by the American Institute of Mining and Metallurgical Engineers (Terzaghi (1929)). Eighty-five years later, a layer a few metres thick under a 40 m dam was such a detail. After the failure, British Columbia's engineering association wrote guidelines for investigating dam foundations, in answer to one of the panel's recommendations.EGBC (2016), preface: written in response to the panel's Recommendation 6, and applying to all dams in the province.

Then there was the slope on top. The design had called for a downstream slope of 2H:1V, but for years the outer face had been left at the steepest angle tipped rockfill will stand, about 1.3H:1V. (You measured about 1.4 across for 1 up in Fig. P.1, from the crest's edge to the toe; the face itself stood at 1.3H:1V.) The panel put the two causes in one sentence: "If constructing unknowingly on the Upper GLU stratum, and not recognizing the potential undrained failure constituted loading the gun, building with a 1.3H:1V angle of repose slope over this stratum pulled the trigger." Here the Upper GLU is the panel's name for the layer. That night the crest dropped at least 3.3 m, more than the 2.3 m by which it stood above the pond. The pond spilled over the lowered crest and cut the breach.Mount Polley Review Panel (2015), p. 75 (an unbuttressed 1.3H:1V slope, almost 40 m high), p. 108 (the quotation), p. 137 (the crest's drop and the pond 2.3 m below the core). The Chief Inspector of Mines listed among the causes a "misplaced faith in the Factor of Safety".Chief Inspector of Mines (2015), p. 5.

Fig. 13.4 · Case file: Mount Polley, 2014Section to scale, from the Review Panel
moved about 11 m slip along the layer, analysed by the panel 2.3 m pond El. 966.8 m tailings core El. 969.1 m core of till rockfill face about 1.3H:1V upper till glaciolacustrine silt and clay lower tills original ground about 40 m 970 950 930 920 El. (m) 0 20 m
Dam
The Perimeter Embankment of the tailings storage facility at the Mount Polley copper and gold mine, about 65 km northeast of Williams Lake, British Columbia.
Built
From 1995–1996 (the starter dam) in stages, a few metres at a time; the ninth was being finished in 2014.
Height
Almost 40 m at the breach, with an unbuttressed downstream slope of 1.3H:1V.
Type
Zoned: a core of compacted glacial till, a filter and a transition each about 1 m wide, a rockfill shell.
What happened
On the night of 3–4 August 2014 the foundation failed in a layer of glaciolacustrine silt and clay about 8–10 m below the original ground, undrained. The embankment slid about 11 m downstream, its crest dropped at least 3.3 m, and the pond, 2.3 m below the top of the core, spilled over and cut the breach.
Toll
Nobody was killed. Between 21 and 25 million m³ of water and tailings were released into Polley Lake, Hazeltine Creek and Quesnel Lake.
Sources
Section traced from Mount Polley Review Panel (2015), p. 93, with the pond and core levels of p. 6; height, slope and the filter and transition widths p. 75; the layer p. 135; movement and uplift p. 21; the factor of safety at failure p. 92; crest drop p. 137. Location, the starter dam and stages, the till core and the volume: Chief Inspector of Mines (2015), pp. 1, 3 and 13.

The dam slid on a thin layer of lake-bed clay that its site investigation never found, under a downstream slope as steep as rockfill will stand. The section is the panel's own analysis at the breach, as the dam stood in August 2014: tailings and a thin pond on the left, the core and rockfill shell in the middle, and beneath the original ground the upper till, the glaciolacustrine layer and the lower tills. The dashed brick-red line is the slip surface the panel analysed: it drops through the core, runs flat along the layer under the whole downstream half of the dam and rises through the till beyond the toe, where the small caret marks the ground pushed up about 3 m. With the layer's measured undrained strength, the panel's calculations put this surface at a factor of safety of about 1.

What this chapter explains. A check of drained strength cannot see a clay that fails undrained, and no check can weigh a layer nobody sampled. A flatter slope would have held, but the number that said so would have been right for the wrong reason.

  • Use the scale bar to measure how far below the original ground the slip runs along the layer, and compare it with the dam's height.
Answers

About 8 m below the ground, under a dam about 40 m high: about a fifth of the height, 0.2, as the workbench's layer is near the top of its 4–6 m.

Notice how little of the drawing the layer takes up, and how much of the slide runs along it. Everything above the layer is the dam the checks were written for; the layer is the detail they were never shown.

13.9 What the check leaves out

The workbench is honest about its own blind spots, and so should you be. Material strengths and permeabilities are typical values, not design values. The sandbox does not check earthquakes, the liquefaction of loose saturated sand (Chapter 5), or whether each zone is a proper filter for the one beside it. In a real dam, clay painted straight against rockfill would wash into it, which is why real cores have filters.

Four more limits matter to anyone who would trust its numbers. It checks no clay for its undrained strength, as Mount Polley shows. It skips the end-of-construction case, when a clay core or foundation is still carrying the weight of fresh fill in its water, for which designers want 1.3.USACE EM 1110-2-1902, table 3-1, p. 3-2. Its exit gradients right at the toe depend on the 1 m grid, so the piping safety there is a guide. And it shows one cross-section: a real dam is a kilometre or more of sections, each on its own ground, and the weakest may be the one nobody drew.

The last film of the book tells Mount Polley's story in nine minutes, with the workbench in it:

Film 13.1 · The layer nobody found8 min 46 s · captions

Before you watch

The workbench stops at 1.5:1, steeper than our dam but not as steep as the 1.3:1 face at Mount Polley. At that slope, what would the check say for the Polley section with the weak layer, and without it? Make a guess; the film runs both at 4:22. Then, near 5:32, watch for the moment the clay's water can't get out of the way.

In this film

  1. One in the morning, August 2014 the Prologue
  2. It didn't burst
  3. The layer the drilling missed the boreholes
  4. The sandbox: a circle, then a block Example 13.4
  5. The trigger: a steep face Fig. 13.4
  6. Slow and fast: why it went in one night Fig. 13.3
  7. What would have held it your guess
  8. Seven dams, one idea
Read the transcript

This dam held a mine's waste: rock ground to powder, mixed with water. At about one in the morning, in August twenty fourteen, it let go. This was Mount Polley, in British Columbia. About twenty-five million cubic metres of water and tailings poured out, into a creek and the lakes below it. From space, the creek went from a thread to a wide grey scar. No one was killed. But a dam like this is built to stand for good. What gave way?

The answer most people reach for is the wall: it burst. An independent panel of experts looked hard at the wall. It found flaws: a core cut thinner than first planned, and samples of filter sand that were too coarse. But none of them caused the breach. It found no sign that the pond had poured over the top first, or that water had washed the dam out from inside. And what was left didn't look like a burst.

The crest had dropped more than three metres, along one clean break. The rockfill's layers were tilted, but still in order. The whole mass had moved eleven metres outward. And beyond the toe, the ground itself had been shoved up, about three metres. A wall that bursts doesn't lift the ground beyond it. Something underneath had slid. The crest dropped further than the water stood below it, so the pond spilled over, and cut the breach.

Eight to ten metres under the old ground surface, there was a thin layer of silt and clay. It was laid down in a lake at the end of the last ice age. The ground had been drilled, to see what was down there. But only four holes with tests went that deep, and none of them where the dam broke. The layer was missed where it mattered. Picture the castle at the beach, standing on firm sand. A hand's depth below it, there's a thin stripe of wet grey sand. Push a stick in here, and here, and here. Every time, you stop short of the stripe. So how much can a thin seam matter, under a dam forty metres tall?

The course has a sandbox for exactly this. It's a teaching model, not the panel's analysis. Here's the series' dam: a clay core, with rockfill on either side, on clay with a weak layer four metres down. That's the same share of its height as the real one. Engineers judge a slope by its factor of safety: the grip the soil has, divided by the grip it needs. Below one, it slides. First, the water.

The seepage solver says this dam barely leaks: about two buckets a day, for each metre of dam. And the sand at its toe is safe from boiling, with room to spare. Then the slope. The course's check tries more than a thousand circles through the dam and the ground. For each one, it weighs the grip along the circle against the weight trying to slide. The weakest circle gives one point three one. But a circle is only the shape the check expects.

A slide doesn't have to be round. The seam offers a flat path, so try a block instead. It drops at the crest, slides along the seam, and comes up beyond the toe. A second method, made for shapes like this, finds one point two one: weaker than any circle. And look at its shape: the crest drops, and the far end rises beyond the toe. That's the real slide. Now take the seam out of the ground. The same dam gives one point five five, and passes the course's target of one point five. That's the ground the design pictured. But one point two one is still above one. Something had to push it over.

That was the slope. The design called for two across for every one up. But for years, the outer face had been left at one point three across for every one up. That's rockfill tipped from trucks and left at the steepest it will stand. It's steeper than a cone of dry sand at the beach. And rock was still being placed there the week it failed.

The sandbox can't go that steep, but try its steepest slope. Without the seam, it gives one point two eight. With the seam, one point oh one. Each one alone leaves a margin. Together, they leave almost none. In the panel's words, building on the clay without knowing it was loading the gun. And the steep slope pulled the trigger.

But the seam had held for seventeen years. Standing that long didn't prove it safe. Each raise took the dam closer to the edge, and by that summer, the panel found, its factor of safety was about one. A few hours before it went, the night shift's log read: good slope, no visible cracks. So why did it go in one night, with no warning?

Back to the idea behind this whole series. Soil's grip comes from its grains pressing on each other. Water pressure in the gaps pushes them apart, and takes that press away. Stand still on the wet sand at the waterline, and it holds you. Wiggle your feet, and you sink. Same sand: what changed is how it was pushed.

Seventeen years of raising had pressed the clay harder than anything in its past. Clay in that state, when it's sheared, tries to pack tighter. Pushed slowly, its water has time to seep out, and the grains keep their press. Clay's gaps are tiny, so water creeps through it very slowly. Over seventeen years of raising, there was time.

In the minutes of a slide, there isn't. The water is trapped, its pressure rises, and the grip falls. The water pressure came from the shearing, not from the weight. The panel's tests found that, sheared fast, the clay had less than half its slow grip. And as it moved, it lost more. The design was checked against the slow grip. But a slide that starts is fast.

The dam didn't break. The ground under it lost its grip, the moment it began to move.

So what would have held it? Back in the sandbox, try the fixes from the episode on leaks. Drains saved that dam. Here, the factor barely moves: the seam lies sealed in clay, below the dam, where drains can't reach. A wall down through the ground stops nearly all the leak, but the block still falls short of the target. What works is a flatter slope.

Two and a half across is still short of the target. At three across for every one up, it passes. The extra weight beyond the toe holds the block back. The course calls a gentler slope, or a heavy berm at the toe, the usual fix for a weak foundation. At the real dam, the panel found that flattening the slope to two across, as first designed, would have avoided the failure.

So why did so much come out? At the breach, there was no beach of tailings between the pond and the dam. The water stood right against it. The panel called that high water the last link in the chain. With the water a metre lower and a wider beach, it wrote, that last link might have held until dawn. Instead, the water carried out about a third of the tailings. Then the panel set out what to aim for, from the first principles of soil mechanics.

Keep the water off the top. Keep the tailings drained. And pack them dense. That's this whole series, in one list. It also asked for a harder look at the ground before building, and for independent boards to review tailings dams. Its warning: if nothing changed, British Columbia could expect two failures every ten years. In twenty sixteen, the province rewrote its rules for them.

Seven dams, one idea. A beach, a concrete dam, a clay core, a leak, a boil, a beach of mine waste, and a seam nobody found. Every time, the soil held while its grains pressed on each other. And every time, water in the gaps could take that away. Dry sand pours. Damp sand stands. And soaked sand, pushed, turns to soup.

After you watch: the answer

Without the layer the section scores 1.28 at 1.5:1; with it, 1.01. Each alone leaves a margin, and together they leave almost none: the panel's loaded gun and its trigger, in two numbers. Then the film shows what the workbench cannot, the clay's water trapped by a fast slide, so that a drained check's margin was never really there. One line at 5:17 says the water pressure "pushes" the grains "apart"; read it as Chapter 4 does, as the water taking part of the grains' press away. The film's last minute is the whole book in one list: keep the water off the top, keep the tailings drained, and pack them dense.

What to remember

  • A design is a plan for the water: a core and a cutoff to hold it back, a drain to let it out before it reaches the downstream slope, a free-draining shell where the reservoir may fall away. Each keeps water from taking the grains' press where a slip would run.
  • Every fix costs something, in fill, clay or water. A drain buys safety cheapest because it removes water instead of adding weight, and pays for it in leak.
  • A factor of safety is only as good as the strength fed to it. A drained check assumes the water can leave; a clay sheared fast keeps its water, and Mount Polley's gave less than half its drained grip.
  • No calculation weighs ground nobody sampled. The layer had to be found before any strength could be measured.

Questions

  1. 13.1

    The workbench passes the Mount Polley section at 3:1, with a factor of safety of 1.56. Give two reasons the real dam could still have failed at that number.

    Hint

    Ask which strength the check used, and what the check was told about the ground.

  2. 13.2

    Meet Hold the water with the least clay you can. What sets the floor in the workbench, and what would set it in a real dam?

    Hint

    Clay passes water a hundred thousand times more slowly than rockfill, so in the model even a thin wall works if it has no gap. Then think about the gradient across a thin wall.

  3. 13.3

    The workbench lets you paint clay straight against rockfill. What would really happen, and which chapter said so?

    Hint

    Compare the size of the clay's particles with the gaps between rocks, and think about water crossing the boundary.

  4. 13.4

    In Empty it fast, start from Leaky (all rockfill) and test. Why does its upstream slope score so much better than the dam of fill, and why does it still not complete the mission?

    Hint

    The workbench lets anything that passes water at 10⁻⁵ m/s or faster drain as the reservoir falls.

  5. 13.5

    Raise the reservoir of the dam of fill to 95% of its height. Where does the one-third rule say water will break out on the downstream slope? Then test it.

    Hint

    95% of 24 m, then a third of that.

  6. 13.6

    You may drill four more holes at Mount Polley before the dam is built. Where do you put them, how deep do they go, and what do you do with what comes up?

    Hint

    Where would a weak layer hurt most? How deep can a slip under a 40 m dam reach? What test tells you a clay's fast strength?

Further reading

  • Independent Expert Engineering Investigation and Review Panel (2015). Report on Mount Polley Tailings Storage Facility Breach. Province of British Columbia, January 2015. https://www.mountpolleyreviewpanel.ca/sites/default/files/report/ReportonMountPolleyTailingsStorageFacilityBreach.pdf primary source Free and clearly written. The executive summary takes two pages; section 5.4 tells the dam's design history raise by raise, and sections 6.2 and 6.5 work the failure through with the undrained strength this chapter quotes.
  • U.S. Bureau of Reclamation (1987). Design of Small Dams, 3rd ed. Denver. https://www.usbr.gov/tsc/techreferences/mands/mands-pdfs/SmallDams.pdf gentle Free from the Bureau of Reclamation. Its sixth chapter, on earthfill dams, explains zoning, drains and the one-third rule with drawings of real sections.
  • Duncan, J.M., Wright, S.G., Brandon, T.L. (2014). Soil Strength and Slope Stability, 2nd ed. Hoboken: Wiley. serious The standard text on choosing strengths for slope checks: drained and undrained, peak and softened, and the loading cases of end of construction and rapid drawdown.

Epilogue

The dam after the drawings

What happens to a dam once the drawings are done?

So far. Every number in this book has been a calculation: a pressure, a leak, a grip, a factor of safety, each worked out for a dam whose soil you chose. Chapter 13 ended on a dam that passed its calculation and failed anyway. Real dams are also watched and governed by law, and some fail in ways no calculation in this book would catch.

The dam, piece by piece complete

2 figures · about 10 min to read

On Friday 11 June 2010 a hiker on Mount Kobau, high above the Okanagan valley south of Oliver, British Columbia, found a small lake full to the brim and running over a dirt road, fast enough to cut a channel across it. He reported it at the tourist booth in Osoyoos. The message passed from the booth to the police, and from the police to the Ministry of Forests, where it ended as a voice mail about a washout on a minor road. Nobody heard it until it was too late.Morhart (2010), p. 10, the province's review by its Deputy Solicitor General.

The road ran along the crest of a dam. Testalinden Dam was a bank of silty sand and glacial till about 2.5 m high, finished in 1937 to store irrigation water, built with horses and a horse-drawn scraper to no engineer's design. That spring a deep snowpack was melting in the year's first hot spell. On the afternoon of Sunday 13 June, the water flowing over the bank cut through it. The lake ran down a steep, narrow gully toward the valley floor 1,500 m below and arrived as a mudflow at least ten times the volume of the water that started it. It destroyed or badly damaged five homes and buried 200 m of Highway 97. Nobody was hurt.The dam, the snowmelt and the mudflow: ASDSO, Testalinden case study, quoting the field investigation in Tannant (2015), which found no rain recorded in the three days before (the province's review, written a month after, spoke of heavy rain). The date, the year built, the toll: Morhart (2010), pp. 7, 11 and 12.

None of it surprised the people whose job was to know. The province's files record water seeping near its outlet pipe as early as 1942. By 1960 a road had been built across it, and its spillway, the channel that lets a flood past (Chapter 6), had been filled with earth and a pipe laid through it. In 1977 an inspector wrote that "the dam is in deplorable condition and should, in my (the inspector) opinion, be breached or rebuilt." In 1988 another inspector added, "No attempt has been made to carry out our instructions." The review found warnings going back to the 1960s, and no sign that the repairs were made or that anyone held the owner to them.Morhart (2010), pp. 3 and 7–9; the quotations are from the inspections of 1977 and 1988.

Every fault on that list is in this book: water seeping along a pipe, a spillway too small for a flood, a freeboard of a foot or less. Nothing in the dam's calculations was wrong, because it never had any. What failed was the arrangement around it: inspections and letters, with nobody made to act on them. That arrangement is part of the dam. Here is the dam, and the fall below it, each drawn to its own scale:The foot of freeboard and the 12-inch spillway pipe: Morhart (2010), p. 8, the letter of 1978.

Fig. E.1 · Case file: Testalinden Dam, 2010A: heights to scale, widths schematic · B: heights to scale, distance not to scale
A · The dam high water, 1978 13 June 2010: water over the crest crest, a road 2.5 m 1 m the spillway pipe, 30 cm, same scale B · The fall the dam, El. 1,810 m 2.5 m high: too small to draw at this scale the mudflow 500 m about 1,500 m lower: the valley floor, orchards and Highway 97
Dam
Testalinden Dam, on the upper south fork of Testalinden Creek, Mount Kobau, above the Okanagan valley south of Oliver, British Columbia. Privately owned.
Built
Finished in 1937 to store irrigation water. A road was built across the dam by 1960, when the spillway was filled with earth and a pipe laid through it.
Height
About 2.5 m, at El. 1,810 m, about 1,500 m above the valley floor. Full to the crest it held about 29 acre-feet, 36,000 m³.
Type
Earthfill of silty sand and glacial till, built with horses and a horse-drawn scraper, with no engineering design.
What happened
On 11 June 2010 a hiker reported the lake running over the road on the crest; the message never reached anyone who knew the road was a dam. On 13 June, at about 2:15 pm, snowmelt flowing over the crest cut through the bank, and the release grew into a mudflow at least ten times its volume.
Toll
No one hurt. Five homes destroyed or badly damaged, 200 m of Highway 97 buried, about 25 homes evacuated; losses over C$20 million.
Sources
Morhart (2010), pp. 7–12 and 15; ASDSO, Testalinden case study (height, materials, elevations, volume, how it failed, losses), quoting Tannant (2015). The spillway pipe and the high water of 1978: Morhart (2010), p. 8.

A bank no taller than a room, perched 1,500 m above a valley, failed when the lake ran over its top, after decades of warnings on file. Panel A cuts across the dam with heights to scale and widths drawn for illustration, since its slopes were never published: the lake on the left, the bank with a road along its crest, the dashed high-water line an inspector recorded in 1978, and the brick-red arrow of water running over the crest on 13 June 2010. The circle below is the spillway pipe at the same scale. Panel B draws the hillside below with its heights to scale and its length shortened: at that scale the dam itself would be far thinner than the line of the ground. The numbers to remember: a dam 2.5 m high, and a fall of 1,500 m below it.

What this chapter explains. A dam's calculations hold it up only while someone keeps it the dam that was calculated. Testalinden had no calculations to keep, and it failed in the half of failures this book has hardly touched: water over the top.

  • Use panel A's scale bar to read how much freeboard the 1978 high water left. How many spillway pipes, stacked, would reach from the ground to the crest?
Answers

About 0.3 m, a foot. About eight pipes: 2.5 ÷ 0.3 = 8.3.

Notice how close the dashed line sits to the crest: in 1978 the lake already stood within a foot of the road. Notice too the spillway pipe, a circle you could cover with two hands. The lake was small, about 36,000 m³ full to the crest;An acre-foot is the water that would cover an acre, 4,047 m², one foot deep: 1,233.5 m³. Testalinden held about 29 of them (ASDSO, Testalinden case study). Mount Polley released some 600 times as much, at the low estimate. What made it dangerous was the 1,500 m below it.Mount Polley released 21 to 25 million m³: Chief Inspector of Mines (2015), p. 13.

E.1 The other half

Testalinden failed in the commonest way an embankment dam fails: water came over the top. Chapter 6 counted the record. Of the large embankment dams that failed in service up to 1986, about a third failed because water poured over the crest, and about half once the failed spillways and gates are counted in.Foster, Fell & Spannagle (2000), its table 3, failures in operation: overtopping 34.2%, with spillway and gate failures 48.4%. This book has spent its chapters on the other half, and the title page gives it one sentence:

Soil holds only as well as its grains press on one another, and the water between them can take that press away.

It carried us from the beach to the workbench, but it does not reach Testalinden. Water running over a crest does not need to take the grains' press away: it drags them off the surface, and each one it takes uncovers the next. So the sentence needs an honest second half:

And water that pours over the top can wash the whole heap away, however hard its grains press.

Other omissions deserve a line each. Earthquakes do more than liquefy loose sand: shaking can crack a core and settle a crest. The flood a spillway must pass comes from rain and snowmelt records that a changing climate makes harder to trust; one lesson drawn from Testalinden is to recheck the spillways of dams built before those records existed. Arch dams, gates and chutes have mechanics of their own.The lessons on floods and old spillways: ASDSO, Testalinden case study.

E.2 Watching the water you painted

For the half this book does cover, the calculation is where a dam's life starts. A real dam's fill is never quite what was sampled, the ground under it was seen only down a few boreholes, and drains silt up over the decades. So engineers build instruments into a dam, and each measures a quantity you have already computed. Here they are on our dam, the workbench's 24 m section on clay with the chimney and blanket drain that Chapter 12 used to dry it; the water level inside it and the slip circle are the solver's:

Fig. E.2 · What a dam's instruments readSection to scale · instruments schematic
1 2 3 4 5 6 reservoir level · Ch 2 piezometer A · Ch 3 survey markers · Ch 5 piezometer B · Ch 12 inclinometer · Ch 12 weir · Ch 10 reservoir clay foundation 10 m
  1. Reservoir level. The depth of water that loads everything else (Chapter 2). Every other reading is judged against it.
  2. Piezometer A, in the upstream half. The head at its tip, the level water would climb to in a straw (Chapter 3).
  3. Survey markers on the crest and slope. How far the surface settles and moves as water leaves the fill and the clay below (Chapter 5).
  4. Piezometer B, a little below the slip surface. The water pressure there, the u in the factor of safety (Chapter 12). Dry while the drain works.
  5. Inclinometer. Sideways movement at each depth, so where a slip is running and how fast (Chapter 12).
  6. Measuring weir. The leak q (Chapter 10), and whether it runs clear (Chapter 11).

Each instrument reads one quantity this book has computed; none reads the factor of safety. The section is the workbench's dam on its clay foundation, to scale, with a chimney drain close downstream of the crest joined to a blanket drain along the base. The dashed blue line is the water level inside the dam as the seepage solver finds it, falling from the reservoir into the chimney; the dashed arc is the critical slip circle the stability check finds. The numbered marks are the instruments, named in the key; in a standpipe the short blue tick is the level of the water inside it, and piezometer B has none. The number to remember: a rise of 1.57 m in the piezometers along our dam's slip costs it 0.177 of its factor of safety (Example E.1).

  • Piezometer A's water stands below the dashed water level above its tip. Why?
  • Use the scale bar on the inclinometer. How far below the slope's surface would a kink show if the slope began to slide along the arc?
Answers

Here the water flows down toward the drain, losing head as it goes, so the head at A's tip, 12.5 m above the base, is lower than the water level above it, 17.1 m: a piezometer reads the head at its tip, not the water level overhead (Chapter 3). The kink would show where the tube crosses the arc, about 5.1 m below the surface.

Notice that each instrument reads an ingredient of a calculation. A piezometer is the straw of Fig. 3.1 made permanent: it reads the water pressure u at one point, as the bars along the slip surface did in Fig. 13.1. The measuring weir reads the leak of Fig. 10.3. The inclinometer, a grooved tube down which a probe measures the tilt, shows where the ground is moving sideways, and at what depth. No instrument reads the factor of safety: it exists only in the calculation, and the readings are what engineers feed it.

So the question an engineer asks of a rising piezometer is how much safety it costs. For our dam, Chapter 12 has supplied what is needed: on its critical circle the factor of safety is 1.116, the water carries 0.353 of the slices' weight (the ratio engineers call ru), and the same circle without water stands at 1.741. In 1960 Alan Bishop and Norbert Morgenstern showed that for a given slope the factor of safety falls very nearly in a straight line as ru grows:The Morgenstern of that paper (Bishop & Morgenstern (1960)) is Norbert Morgenstern, who fifty-five years later chaired the panel that investigated Mount Polley (Mount Polley Review Panel (2015), p. i).

FS ≈ FS0 − n · ru(E.1)

FS0 the factor of safety with no water pressure · ru the share of the slices' weight the water carries, Σu·b ÷ ΣW · n the safety each unit of ru costs, set by the slope's shape and soilWhy nearly straight: in Bishop's sum the water enters only by taking u·b off each slice's weight before friction acts. If every slice lost the same share of its weight, the grip, and with it the factor of safety, would fall in a straight line as that share grew, except that the factor of safety also sits inside each slice's mα. That small feedback bends the line slightly.

Example E.1

What a rising piezometer costs (illustrative)

Given
Our dam on clay, as Chapter 12 left it: FS = 1.116 with ru = 0.353, and FS0 = 1.741 on the same circle with no water. The slices on that circle weigh, on average, 154.1 kN for each square metre of their bases. A metre of water in a piezometer is 9.81 kPa.
Find
The safety lost for each 0.1 added to ru, and the rise in the piezometers along the slip that adds it.
The idea
Draw the line of Eq. E.1 through the two points. Raising the water pressure on every slice by the same amount adds that amount, divided by the slices' weight per square metre, to ru.
  1. 1.741 − 1.116 = 0.625the safety the water takes, at ru = 0.353
  2. n = 0.625 ÷ 0.353 = 1.77the slope of the line: safety lost per unit of ru
  3. 1.77 × 0.1 = 0.177lost when ru rises by a tenth
  4. Δu = 0.1 × 154.1 = 15.41 kPathe extra water pressure that carries another tenth of the slices' weight
  5. 15.41 ÷ 9.81 = 1.57 mthe same rise as a height of water in a piezometer

A rise of about 1.6 m in the piezometers along the slip costs our dam about 0.18 of its factor of safety.

Check. Do it without the straight line. Rerun Bishop's sum on the same circle with every slice's water pressure 15.41 kPa higher: the solver's slices give 0.943, against 1.116 − 0.177 = 0.939 from the line. The two agree to the second decimal.

What it means. A dam like ours designed to 1.5 holds 0.5 above failure, and a season's rise of a metre and a half in its piezometers would spend more than a third of that. Our dam, at 1.116, would go below 1. So an engineer watches each piezometer's trend against the reservoir, and decides in advance the reading at which something must be done.

Then press Test my dam.

Check yourself Over a summer the reservoir behind our drained dam rises 3 m, and piezometer B stays dry. Good news or bad?

Good news. The drain is carrying the extra water away before it can build pressure in the downstream half. Water appearing in B, and climbing with the reservoir, would mean the drain was not keeping up.

E.3 Peck's bargain

Most people take a dam that has stood for decades as proof that it is safe. Testalinden stood for seventy-three years and Mount Polley for eighteen, and both failed. Instruments let an engineer do better than wait. Ralph Peck, Terzaghi's co-author, set out the approach in his Rankine Lecture of 1969: the observational method. You design for the ground you think most likely, and before you build you decide what you will measure, what readings would mean the ground is worse than you assumed, and what you will do then. It is a bargain with ignorance: you need not know everything about the ground at the start, provided it tells you in time and you have a response ready.Peck (1969); the method as the tailings standard defines it, citing Peck: GISTM (2020), glossary. Peck and Terzaghi wrote Terzaghi, Peck & Mesri (1996) together.

Both conditions failed at Mount Polley, though its designers had invoked the method from the start. Its few instruments stood beyond the toe, because the steep slope could not be reached, and too far apart to cover the place where it broke. More fundamentally, as the panel wrote, "piezometers cannot measure pore pressures induced by undrained shearing because the location of the failure surface on which to measure them cannot be known in advance." Had they warned, there was no way to build a buttress in time. With the undrained failure never recognized, the panel concluded, the method had been reduced to "mere trial and error".Mount Polley Review Panel (2015), p. 77 (the method, the instruments' reach, the quotation, no time to respond), p. 80 (no instruments near the breach) and p. 107 (trial and error).

The lesson is now in the international standard for tailings facilities, published in 2020. It asks for the observational method in full for failures that are not brittle, and adds: "Brittle failure modes are addressed by conservative design criteria." A brittle failure, like that of Mount Polley's clay sheared too fast for its water to get out (Chapter 5), gives little warning: the soil loses strength once it starts to move. Against it, the only defence is margin built in before anyone is watching.GISTM (2020), Requirement 7.2.

Check yourself Why would even a piezometer on the slip surface give little warning of an undrained failure?

The pressure that weakens the soil appears only as it shears. By the time the piezometer reads it, the slip is already under way, and a soil that loses strength as it moves leaves no time to respond.

E.4 The law around the dam

An instrument nobody reads, or a reading nobody acts on, does nothing, and Testalinden had warnings on file for decades. Closing that gap is what regulation is for. In British Columbia a water dam answers to the Dam Safety Regulation, and a dam on a mine to Part 10 of the Health, Safety and Reclamation Code for Mines. Both have been rewritten since these two failures.

In 2010 Testalinden was rated "Low", which meant an audit about once in ten years; within two weeks of the failure the province had assessed 628 dams. The regulation of 2016 sorts dams into the Canadian Dam Association's five consequence classes, from low to extreme, by what a failure would do downstream, to lives, property and the environment. The class, not the dam's height, sets the duties. A dam of high consequence or above is checked on site weekly unless its manual says otherwise, inspected formally at least once a year, run from a written manual with an emergency plan, and given a dam safety review by an engineer every seven or ten years. Under today's scale Testalinden would have been significant, not low.The rating and the 628 dams: Morhart (2010), p. 15. The class today, and the match with the Canadian Dam Association: ASDSO, Testalinden case study. The classes and duties: Dam Safety Regulation (2016), Schedules 1 and 2, ss. 8, 9 and 20.

Mount Polley rewrote the mining code. In July 2016 the province announced limits on the steepness of a tailings dam's downstream slope, a minimum factor of safety, flood and earthquake criteria, and an independent review board for every mine with a tailings facility. Every tailings facility and dam on a mine has an engineer of record, and if that engineer reports a safety problem the mine leaves unfixed, the manager must tell the chief inspector within 72 hours, or the engineer must. Some rules read like the panel's findings turned round. A downstream slope steeper than 2H:1V (two across for every one up) needs the engineer of record's justification and the regulator's approval; Mount Polley's stood at 1.3H:1V, and the panel found that a 2:1 slope would have avoided the failure. The factor of safety in operation must be at least 1.5. And the facility's manual must say what is to be done as readings worsen, where the piezometers of Mount Polley's failed embankment had no trigger levels at all.BC Ministry of Energy and Mines (2016); Mines Code, Part 10, ss. 10.4.1, 10.4.3, 10.4.4, 10.5.8, 10.5.9 and its table 10-5, and 10.6.6. The slope: Mount Polley Review Panel (2015), p. iii; no trigger levels, p. 79.

The same ideas now cross borders. The Global Industry Standard on Tailings Management of August 2020 asks every operator that adopts it for an engineer of record, an accountable executive answering to the chief executive and, where a failure would be very high or extreme in consequence, an independent tailings review board. None of it makes a dam safe by itself. It puts named people and outside reviewers where Testalinden had letters in a file.GISTM (2020), Requirements 8.4, 8.7 and 9.1.

Check yourself Two earth dams are each 10 m high. One holds a farm pond above an empty valley; the other holds the same pond above a town. Which needs the more frequent review?

The one above the town. The consequence class, which sets how often a dam is inspected and reviewed, follows what its failure would do downstream, not its size.

What to remember

  • Grains pressing on one another, and the water between them taking that press away, explain about half the ways embankment dams fail. Water pouring over the top explains most of the other half.
  • Instruments read the ingredients of the calculation: piezometers read u, weirs read q, inclinometers find the slip. Nothing reads the factor of safety itself.
  • For our dam each tenth added to ru, about a metre and a half of water in the piezometers along the slip, costs about 0.18 of the factor of safety.
  • The observational method works when a failure gives warning and there is time to act. An undrained, brittle failure gives neither, so it must be met with margin in the design.
  • Regulation is part of the dam: a class set by what lies downstream, a named engineer, reviews, and readings with actions attached to them.

Questions

  1. E.1

    Inspectors visited Testalinden in 1961, 1977, 1978, 1985, 1988 and 1992. List what they found. Which of those faults would a yearly walk along the crest still have shown in 2010, and what would it have taken to prevent the failure?

    Hint

    The opening pages and the case file name most of the faults. Then ask what the inspections lacked, not what they missed.

  2. E.2

    Name three things British Columbia's mining code requires today, and for each say which part of the Mount Polley failure it answers.

    Hint

    Chapter 13 found a layer nobody had looked for, a steep downstream face and a design checked only against slow loading. Add what the instruments lacked.

  3. E.3

    The drained version of our dam (Fill + drain in Fig. 13.1) stands at 1.628. Its engineer of record wants an alarm before it falls to 1.5. Using the straight line with our dam's n = 1.77, and the drained circle's average weight of 83.4 kN per square metre of base, how far may the piezometers along its slip rise?

    Hint

    Run Example E.1 backwards: margin, then ru, then kPa, then metres of water.

  4. E.4

    A wet spring fills our dam's reservoir from 85% to 95% of its height. Set the workbench to the homogeneous dam on clay at 95% and test it. How far up the downstream slope does the water break out, how much does the dam leak, and what is its factor of safety? Which instruments in Fig. E.2 would show the change?

    Hint

    At 85% the water breaks out 8 m up, the dam leaks 44.4 L/day per metre, and its factor of safety is 1.12.

Further reading

  • Morhart, D. (2010). Review of the Testalinden Dam Failure. Victoria: Province of British Columbia, Deputy Solicitor General, 13 July 2010. https://damfailures.org/sites/default/files/wp-pdf/testalinden_slide_failure_review.pdf gentle Eighteen pages: the dam's history from 1937 in the inspectors' own words, the warning that went to the wrong office, and twelve recommendations, almost all about people and paper rather than soil.
  • Independent Expert Engineering Investigation and Review Panel (2015). Report on Mount Polley Tailings Storage Facility Breach. Province of British Columbia, January 2015. https://www.mountpolleyreviewpanel.ca/sites/default/files/report/ReportonMountPolleyTailingsStorageFacilityBreach.pdf serious Section 5.5 and the pages before it: what the instruments at Mount Polley could and could not see, and why the panel judged that the observational method had failed there. Free from the panel's website.
  • Peck, R.B. (1969). Advantages and limitations of the observational method in applied soil mechanics. Géotechnique 19(2): 171–187. doi:10.1680/geot.1969.19.2.171 primary source Peck's Rankine Lecture on the observational method: what it needs to work and, as its title promises, its limits as well as its advantages. Free to read from the Institution of Civil Engineers.

Go back to the beach where this book began, on an evening when the tide is coming in. The sea covers the firm, dark middle of the beach, and the sand there, already full of water, hardly changes. Then it reaches the castle you built from damp sand, whose walls stood straight up, held by the water clinging where the grains touch (Chapter 1). The first wave fills every gap, the pull is gone, and the walls slump into a low heap. The grains are the same grains; only the water has changed.

Pick up a handful before the next wave comes in: grains pressing on one another, and the water between them. Every earth dam in this book was made of that, and so is the ground under every dam you will ever stand on.

Reference

Answers and hints

Chapter 1 · The beach

1.1 · back to the question

The amount of sand, which in a section is its area. The dry heap's triangle has area ½ × 58.27 × 19.653 = 572.6 cm², the same as the castle's 572.6 cm². The draining heap is lower again, 14.0 cm tall, and wider in proportion.

1.2 · back to the question

Packed densely, the heap holds. Packed loosely, it flows flat and the flag sinks; the readout says the grains packed tighter and the trapped water took their weight. Both heaps were saturated, so saturation alone does not make soup: the sand must also be loose. Chapter 5 explains why the dense heap holds, and with it the pale ring round your foot.

1.3 · back to the question

30 cm: 18 × 0.3 = 5.4 kN on each square metre. 30 m: 18 × 30 = 540 kN on each square metre, a hundred times more. The water rings grip the grains with the same small force in both, so the short wall can stand and the tall one cannot. A kilonewton on a square metre has a name, the kilopascal, which Chapter 2 uses throughout; Chapter 5 compares these numbers with the grip.

1.4 · back to the question

It depends on what you pour, which is the point: the angle is set by the shape and friction of the grains. For example, a heap 6 cm high on a base 18 cm across gives tan β = 12 ÷ 18 = 0.667 and β = 33.7°. Rounder, smoother grains give lower angles. Dry sands usually come out in the low-to-mid thirties of degrees; report what you measured.

1.5 · back to the question

1.5:1 is arctan 0.667 = 33.7°; 2:1 is 26.6°; 3:1 is arctan 0.333 = 18.4°. All three are flatter than 34°, so dry sand stands at each, though at 1.5:1 it has almost nothing to spare.

1.6 · back to the question

Only 3:1, at 18.4°, is flatter than 19.0°, and barely. 2.5:1 is 21.8°, too steep where water comes out. The other choice is to stop the water reaching the face, with a drain, and keep the steeper, cheaper slope. Real fill is not clean beach sand and designers want a margin as well; Chapter 12 and Chapter 13 do the full check.

Chapter 2 · Depth, not size

2.1 · back to the question

From 7.85 MN to 17.66 MN per metre, a factor of (60/40)² = 2.25. Half as deep again more than doubles the push.

2.2 · back to the question

24.0 MN per metre, as heavy as 2,450 tonnes: 1.96 times the push at 50 m, because 70² = 4,900 is nearly twice 50² = 2,500.

2.3 · back to the question

Three metres of water changes the pressure by 29.4 kPa. Thirty metres of sea-level air, at 1.225 kg/m³ (U.S. Standard Atmosphere (1976)), changes it by only 0.36 kPa, because water is about 816 times denser than air. The ears feel the difference.

2.4 · back to the question

16.7 m above the base, a third of the way up. Doubling the depth doubles that height and quadruples the push, so the moment grows 8 times, from 204 to 1,635 MN·m per metre.

2.5 · back to the question

Mostly a help. On a 2:1 face the water presses down 2 times as hard as it pushes across (Example 2.5), and pressing down holds the dam onto its foundation. That is exact for a watertight face; in an earth dam some of the water soaks into the slope (Chapter 10). The sideways part still has to be resisted, and the help disappears when the reservoir is let down quickly, the subject of Chapter 12.

2.6 · back to the question

At the foot of the face 9.81 × 175 = 1,717 kPa, or 1.72 MPa (a megapascal is 1,000 kPa). The push is 150 MN per metre, as heavy as about 15,300 tonnes. Its base is 0.72 times its height, against 0.8 in the example: the same family of shape, a little slimmer. Dimensions from BC Hydro (2013) and BC Hydro (2016).

Chapter 3 · Which way the water goes

3.1 · back to the question

The heads are 25 m at A and 22 m at B, so the water flows from A toward B. But B feels the higher pressure: its tip has 22 − 2 = 20 m of water above it, against 15 m at A, so 9.81 × 20 = 196.2 kPa at B and 9.81 × 15 = 147.2 kPa at A. The water flows from lower pressure to higher, because it flows toward lower head.

3.2 · back to the question

Every point in still water has the same total head. Going deeper trades elevation head for pressure head metre for metre, so a straw anywhere in the lake, or in the soaked bed under it, fills to the lake's surface (Fig. 3.1, panel A). With no difference in head there is no gradient, and Darcy's law gives no flow.

3.3 · back to the question

The share stays at 99.996%, because the resistances haven't changed. The head difference falls from 1.40 m to 0.20 m, so the flow falls by 1.40 ÷ 0.20 = 7 times, from 110 to about 16 litres a year through each square metre.

3.4 · back to the question

0.40 × 2 ÷ 10⁻⁷ = 8.0 × 10⁶ s, which is 93 days: twice the race's 46 days for one metre, because at the same gradient the drop moves at the same speed and has twice as far to go.

3.5 · back to the question

The gradient halves to 3, so the leak halves, to about 7.8 litres a day per metre of dam. The drop moves at half the speed over twice the distance, so it takes four times as long: 4 × 9.24 = 37 years.

Chapter 4 · Grip

4.1 · back to the question

FS = 0.6745 ÷ 0.5774 = 1.17. The water must carry 1 − 0.5774 ÷ 0.6745 = 0.144 of a press of 17.32 kN, and 0.144 × 17.32 = 2.49 kPa: only 14% of the press. In the figure it holds at 2.0 kPa and slides at 2.5. The steeper the ramp, the less margin it has to lose.

4.2 · back to the question

Less likely. The same 5 kPa over the same 1 m² takes 5 kN off each block's press, but the heavier block presses with twice as much, so it loses a smaller share of its grip. With no water the weight cancels out of the factor of safety; water pressure is what makes weight matter.

4.3 · back to the question

σ = 18 × 2 + 20 × 8 = 196 kPa; u = 9.81 × 8 = 78.48 kPa; σ′ = 117.52 kPa, which is 65% of the 180 kPa the grains press with when the water table is below the column.

4.4 · back to the question

No. The extra water adds 9.81 × 20 = 196.2 kPa to the total stress and the same to the pore pressure, so σ′ and the strength are unchanged, as the lake on the column showed. The danger from a deeper reservoir is elsewhere: more head drives more water under the dam, and where it comes up again on the downstream side the upward gradient takes away the press (Eq. 4.4). Chapter 11 follows it there.

4.5 · back to the question

The drag is 0.6 × 9.81 = 5.89 kN per cubic metre against a buoyant weight of 10.19, which leaves 4.30 kN, and 4.30 ÷ 10.19 = 0.422: about 42% of the press, and of the strength, is left. The gradient is under the critical 1.04, but more than half the grip has gone.

4.6 · back to the question

tan β = 0.5095 × 0.5774 = 0.2942, so β = 16.4°. Dry, the same sand would stand at 30°.

Chapter 5 · When the water can't get out

5.1 · back to the question

The firm sand is dense. Sheared by the foot it must swell, and the water it needs is drawn in from the surface, which sinks below the top grains and leaves them dry for a moment; meanwhile the stretched water pulls the grains together and the sand stiffens. The sand at the edge is loose. Wiggled, its grains try to drop into the gaps, the trapped water's pressure rises and takes their weight, they press less and grip less, and the foot sinks.

5.2 · back to the question

With the drain shut no water can get in, so the water already there is stretched, Δu falls below zero and σ′ = σ − Δu rises above the load. With the drain open water flows in as the sample swells, Δu stays near zero, and the grain stress stays near the load. The fall stops at −100 kPa*, the model's stand-in for cavitation: real water lowered to about zero absolute pressure, roughly −100 kPa gauge, breaks into bubbles of vapour and air, so trapped water can only pull so hard. After that, as the result line says, a little swelling gets through.

5.3 · back to the question

Shut, it falls in steps, a little with each shake, as the extra water pressure ratchets up toward the load, until after a few shakes the grains hardly press and the house tips. Open, the displaced water leaves through the drain, Δu stays small and the grains keep pressing; the loose sand still compacts, so the house still settles a little, but it does not sink and tip. Read the trend, not the model's numbers.

5.4 · back to the question

Sand: 3² ÷ 0.20387 = 44.1 s, three times the shaking, so it is undrained while the ground shakes. Clay: 5² ÷ 0.08807 = 283.9 days, and 283.9 ÷ 365.25 = 0.78 years: about as long as the raise itself, so the clay is partly drained as the dam grows, carrying a good part of each raise in its water for months. If it starts to slide, in minutes, it is undrained.

5.5 · back to the question

Its fill was hydraulic fill, sluiced into place with water, like tailings: loose and saturated. Fifteen seconds is shorter than the time water needs to leave even 2 m of sand (about 20 s in Example 5.2), so every shake was undrained, the water pressure in the loose fill ratcheted up, and the fill near the base of the upstream slope liquefied. Coquitlam Dam in British Columbia, a hydraulic fill dam of the same era, is in Chapter 8.

Chapter 6 · Two dams, two jobs

6.1 · back to the question

The power goes from 74.2 to 121.8 MW, a factor of 1.64, because H goes from 56 to 92 m. The load goes from 17.7 to 45.2 MN per metre, a factor of 2.56: the depth at the heel goes from 60 to 96 m, 1.6 times, and the load grows with its square, 1.6 × 1.6 = 2.56. The last metres of water earn the least and cost the most.

6.2 · back to the question

Neither. H is measured to the tailwater; the heel is 4 m lower, 88.8 m below the surface, and 9.81 × 88.8 = 871.1 kPa (Example 6.2).

6.3 · back to the question

Merriespruit began as overtopping: water flowed over the crest for one to two hours and washed away the outer wall, and only then did the loose tailings behind it liquefy and flow. Mount Polley was not a liquefaction of the tailings at all; what it was is the subject of Chapter 13. The other four were liquefaction of the tailings, set off by an earthquake at El Cobre, hastened by small ones at Fundão, and with nothing from outside at Stava and Brumadinho.

6.4 · back to the question

It keeps a small residual strength, the grip left between grains that still touch while the deposit flows. That remainder slows the flow and decides where it stops, so it sets how far the tailings run and what they reach.

6.5 · back to the question

Twice the flow, 1,039 m³/s, needs h3/2 = 2 × 5.196 = 10.392, so h = 10.3922/3 = 4.76 m: 1.59 times the design rise, not twice. The lake stands at 29.76 m, only 0.24 m below the crest: almost all of the 5 m between the normal level and the crest is gone.

Chapter 7 · A block on a cushion of water

7.1 · back to the question

Both carry part of the weight on water, so the block or the dam presses on what is under it with less, and the friction, which comes from that press, shrinks while the push or pull trying to move it stays the same: W − U in Eq. 7.1 is the "press left over" of Fig. 4.1.

7.2 · back to the question

hd = 5 + 45 ÷ 3 = 20 m. Heel to drains: 35 × 6 = 210 m²; drains to toe: 12.5 × 42 = 525 m². U = 9.81 × 735 = 7,210 kN. FS = 27,570 ÷ 12,140 = 2.27.

7.3 · back to the question

No: at 59 m it is still 1.57. It falls faster near the top because the push grows with the square of the depth while the uplift grows only in proportion to it; at 57 m one more metre of water adds about 559 kN of push, and with the drains working even less than the 235 kN of uplift it would add with them clogged.

7.4 · back to the question

At 0.66 B, inside the middle third but close to its end, so the heel is still pressed, but barely. With V = W − U = 19,712 kN, Eq. 7.3 gives 23 kPa at the heel, against an even press of 411 kPa. The factor of safety against sliding is 1.16. A crack opening at the heel would let in the full reservoir pressure and push the resultant further downstream.

7.5 · back to the question

At 50 per cent the head at the drains is 5 + 0.5 × 45.5 = 27.75 m and the factor of safety 1.61; at 25 per cent, 5 + 0.75 × 45.5 = 39.13 m and 1.45, below 1.5. The one-third rule amounts to removing 62 per cent here, above the Corps' usual range and close to the 67 per cent it allows only when tests justify it (USACE EM 1110-2-2200, para. 3-3.d(1)(c)). With the result this sensitive, a careful designer would want pressure gauges in the drain holes to show what the drains actually achieve, and would keep watching them.

7.6 · back to the question

It prevents a crack, once open, from being pushed wider by the water in it: if the masonry presses the crack shut harder than the water pushes it open, water is squeezed out instead of forced in. The middle-third rule only keeps the face from being pulled apart by the weight and the push; it allows the face to be pressed with almost nothing, and any crack then fills with the reservoir's full pressure and spreads. At Bouzey's break the Bureau of Reclamation's back-calculation found the upper wall failing even the middle-third rule, with its face in tension (Fig. 7.3).

Chapter 8 · Never finished

8.1 · back to the question

Upstream: 32 × 34 + 320 = 1,408 m³ at 34 m, and 2,496 ÷ 1,408 = 1.77. Centreline: 34² + 20 × 34 + 160 = 1,996 m³, and 6,144 ÷ 1,996 = 3.08. Upstream fill grows roughly in step with height, because each raise is the same small dyke; downstream fill grows nearly with its square, because each raise widens the whole base; centreline is in between, closer to downstream.

8.2 · back to the question

About 10% of the centreline dam's 2,976 m³ per metre is on the beach, against 44% of the upstream dam's 1,728. The centreline wedges are the pond-side edge of each raise; the crest and the whole outer shell, the part that holds the dam up, stand on compacted fill and the starter. If the wedge is sandy, drained and compacted, as it must be, it holds too. The upstream dam's crest has nothing under it but beach.

8.3 · back to the question

Ending the raising stopped adding load, but it did not make the tailings dense or dry. They stayed loose and saturated, the water level near the toe stayed high, and they were brittle, so a very small strain could set off a large loss of strength. Slow creep, and the loss of suction in the drier tailings above the water table after heavy rain, supplied that strain. With the water unable to leave in the time the slope took to move, the tailings were undrained, the water took the load, and the dam liquefied from within (Robertson et al. (2019)).

8.4 · back to the question

With the drain shut the water takes the load and the house sinks: that is an upstream raise standing on soaked beach. With the drain open the water leaves, the sand settles a little and keeps its grip: that is a raise built on drained fill, as downstream and centreline raises are (and they are dense as well, which helps more). The sandbox is a two-dimensional teaching model, so read its numbers as trends only.

8.5 · back to the question

20,000 ÷ 1.4 = 14,286 m³ a day; 14,286 × 365 ÷ 1,000,000 = 5.214 million m³ a year; 5.214 ÷ 1.5 = 3.48 m a year. That is above the 2 to 2½ m a year of SANCOLD (2020). To hold it to 2½ m a year: 5.214 ÷ 2.5 = 2.09 km².

8.6 · back to the question

Because it was built by hydraulic fill, which leaves soil loose and saturated, the same state as the beach under an upstream raise. Loose, saturated, contractive fill can liquefy when it is shaken or pushed, whatever the water behind it is for. The cure was the same too: a dense, drained, compacted dam against its toe that does not depend on the old fill.

Chapter 9 · Every zone has one job

9.1 · back to the question

Under still water, buoyancy lightens the grains' press and the pull down the slope by the same share, so the face is as safe as when dry: 1.98 for the rockfill face of Example 9.2. Rockfill drains as fast as the lake falls, so its water stays still. The danger is rapid drawdown of a shell that drains slowly: the water left inside seeps out through the face, and the same face falls to about 1.10.

9.2 · back to the question

The barrier is the concrete slab on the upstream face, sealed to the rock by the plinth, with a grout curtain drilled from the plinth into the rock below. Water through a crack in the slab enters rockfill that has no fine grains to lose and drains straight out at the base. A cushion of crushed rock and sand under the slab limits the leak, and a fine transition under the joint between slab and plinth catches silt washed into a gap there and seals it (the cards of Fig. 9.1).

9.3 · back to the question

The concrete-face rockfill dam: its slab goes on in the fourth of seven stages, once the rockfill is in place, and it stays on the face, where it can be seen and repaired whenever the reservoir is drawn down. An earth core rises with the shells, layer by layer, and is buried for good; its condition can be read only from instruments inside it and the water that seeps out of the dam.

9.4 · back to the question

Seepage crosses the core from the reservoir side to the downstream side, so the grains it could carry leave through the core's downstream face; the downstream filter keeps them in, and doubles as the chimney drain. The upstream filter matters when the flow turns round, as it can after a fast drawdown, when water in the core flows back toward the lowered reservoir.

9.5 · back to the question

γd = 21.0 ÷ 1.14 = 18.421 kN/m³, and 18.421 ÷ 18.9 = 0.975, a relative compaction of 97.5%. It fails 98%, narrowly; it would pass a 95% specification.

9.6 · back to the question

Because density alone does not decide how a core behaves when soaked. Bennett's panel found that core soil placed drier let water through 60 to 140 times more easily, once soaked, than the same soil placed wetter, and a dry, stiff core cracks more readily as the dam settles. Proctor showed the same thing in 1933: one soil packed to the same density at 6.2% and at 22% water was very hard in the first case and like soft putty in the second, and could become like the second once it soaked up water (Proctor (1933)). Too wet, on the other hand, and the soil cannot be packed dense at all, which is the wet side of Fig. 9.2.

Chapter 10 · Where the water goes

10.1 · back to the question

q = 2 × 10⁻⁵ × 6 × 4 ÷ 8 = 6 × 10⁻⁵ m³/s, which is 0.00006 × 86,400 = 5.184 m³ a day, about five cubic metres. Each drop loses 6 ÷ 8 = 0.75 m of head.

10.2 · back to the question

2,366 × 400 ÷ 86,400 = 10.95 L/s, or 0.01095 m³/s. Then h = (0.01095 ÷ 1.342)1/2.48 = 14.4 cm. Over a year that dam would leak about 350,000 m³, and a regular reading of a 14 cm head is how anyone would know.

10.3 · back to the question

The water goes under it. The cutoff reaches 7.5 m down, so the path grows from 102 m to about 102 + 2 × 7.5 = 117 m, and Darcy's law predicts about 13% less water; the lab gives 9%. The water still reaches the toe with most of its head, so the exit gradient there falls only a little. The Corps of Engineers' manual puts it bluntly: theory and model tests show a cutoff must reach at least 95% of the way through a uniform pervious foundation before seepage falls appreciably (USACE EM 1110-2-1901, para. 9-4).

10.4 · back to the question

Very nearly. The depth goes from 14.4 m to 22.8 m, a ratio of 1.58, and the leak from 1,652 to 2,657 L/day per metre, a ratio of 1.61: a little more, because a fuller reservoir also wets more of the fill. The safety against piping falls from 2.26 to 1.07, since the exit gradient grows with the head too. That is why engineers read piezometers and weirs against the reservoir level.

10.5 · back to the question

The chimney, and the extra 60 litres a day per metre are the price. A fill compacted layer by layer passes water more easily along its layers than across them, and a loose or badly bonded layer can carry water over a toe drain and out of the downstream slope. A chimney cuts across every layer, so whatever comes through the upstream half is caught. The Bureau of Reclamation calls an inclined filter drain with a horizontal blanket almost standard for this reason, citing loose lifts, poor bond between lifts and inadvertently pervious layers (USBR (1987), sec. 6.3(c)). The lab's uniform fill cannot show this, which is a reason to read the lab as a guide and not a guarantee.

Chapter 11 · When water carries soil

11.1 · back to the question

Water in the open pipe loses almost no head, so the whole head between the reservoir and the toe is spent on the sand still ahead of the pipe. As the pipe grows that stretch shortens and the gradient across it steepens: under our dam the average goes from 0.20 with no pipe to 0.39 with half the path eaten and 1.7 with 12 m left (Example 11.2). A steeper push carries grains away faster, which lengthens the pipe faster.

11.2 · back to the question

Retention: D15F ≤ 4 × 0.6 = 2.4 mm. Permeability: D15F ≥ 4 × 0.15 = 0.6 mm. So 0.6 mm ≤ D15F ≤ 2.4 mm; a filter near the middle of the band, about 1.2 mm, leaves room for the variation of a real pit.

11.3 · back to the question

The boil is backward erosion: it begins at the exit beyond the toe and works back. Teton and Bennett were both concentrated leaks: water running through openings inside the core, at Teton along the base of the key trench and into open joints, at Bennett most likely through cracks opened by hydraulic fracture. At Bennett some of the transition was also prone to suffusion, which the panel thought let a little core erode before the transition held.

11.4 · back to the question

Teton's silt leaked into open joints in volcanic rock, gaps far larger than its grains, so nothing stopped the eroded silt and the tunnel grew to a breach in hours (Fig. 11.6). Bennett's core leaked into its transition, a soil graded to act as a filter for the core: the eroded grains lodged in it and the erosion stopped (Fig. 11.7). Both leaked water; only Teton let the soil go.

11.5 · back to the question

It stops the sand and lets the water through: the sequence never gets past steady seepage. The berm's weight presses down on the ground at the exit, so the grains there press on one another harder, and the upward drag of the rising water has more press to overcome before the sand can heave. It is Terzaghi's first filter: sand with weight on top.

Chapter 12 · Will the slope stand?

12.1 · back to the question

The slope check never asks how much water passes; it asks how hard the water presses where the soil might slip. Without the drain the water rises through the downstream half and breaks out 8 m up the face, and on the worst circle it carries 0.35 of the slices' weight. The drain gives it an easy path down and out: it flows faster, 109.7 L/day per metre against 44.4, but it no longer presses in the downstream half, so the grains there press with their full weight. Example 10.4 found the same trade for piping.

12.2 · back to the question

No. FS = 0.6494 × 3 × 0.4969 = 0.968, a little below 1. With water seeping out of it, this sand stands only at about 3.1:1 or flatter, where FS = 1; to reach 1.5 it needs about 4.6:1, or a drain to keep the water from reaching the face.

12.3 · back to the question

W cos α = 319.0 × 0.8486 = 270.7 kN; u·l = 76.40 × 1.601 = 122.3 kN; N′ = 270.7 − 122.3 = 148.4 kN; grip = 10 × 1.601 + 148.4 × 0.6249 = 108.7 kN, against Bishop's 129.3. Less grip: the old method takes the water's push on the whole tilted base off the press, and lets no neighbour help. Over all forty slices it scores the circle 0.978, against Bishop's 1.116.

12.4 · back to the question

A full reservoir with steady seepage is the dam's everyday state for decades, so it gets the full margin. A rapid drawdown lasts days or weeks before the water drains out of the fill, it happens rarely, and the check for it already leans toward the worst: the reservoir falls all the way at once and the fill gets no time to drain. The manual's own range for it is 1.1 to 1.3 (USACE EM 1110-2-1902, in its table of minimum factors).

12.5 · back to the question

Any wrong input. The strength fed to the check may be the wrong one: a slow, drained strength for a soil that is loaded fast (Chapter 5). The water pressures may be higher than the model's, because a drain clogs or the soil is more layered than drawn. The load may be one the check left out, such as an earthquake or a flood over the crest. Or the slip may run where nobody looked. Chapter 13 follows one real case.

12.6 · back to the question

At 3:1 it scores 1.418 for 23.1% more fill, and the water still breaks out 8 m up the face. The drain at 2:1 reaches 1.628 with no fill added, and no water comes out of the face at all.

Chapter 13 · Build your own dam

13.1 · back to the question

First, the check uses a drained strength, and the real layer failed undrained: the panel measured an average undrained strength ratio of 0.24 against the drained tan 28° the designers gave such clays (Mount Polley Review Panel (2015), pp. 49 and 69) = 0.532, so the fast grip was 0.45 of the slow one, and it fell further once the clay moved (strain softening). Second, the check knows only the ground it is given; at Mount Polley the layer had not been found where the breach occurred, so no check, drained or undrained, included it. A third answer is also fair: the workbench's layer angle is a stand-in, and the real layer's properties had never been measured before the failure.

13.2 · back to the question

In the workbench the floor is continuity: a clay wall 2 m wide from the cutoff to above the water, 46 m³ per metre, leaks 9.8 L/day and passes every goal, with a factor of safety of 1.81; any gap, and the water finds it. In a real dam the floor is set by what the model leaves out. The whole head is lost across the wall, a gradient of about 20.4 ÷ 2 = 10.2, which would drive the core's fines into the rockfill without filters on both sides (Chapter 11); a thin core also cracks more easily as the shells settle round it. The panel noted that Mount Polley's core had been cut from 8 m to 5 m wide, making it more vulnerable to cracking (Mount Polley Review Panel (2015), p. 75).

13.3 · back to the question

Water leaving the clay would carry its fine particles into the large gaps of the rockfill, and the core would erode from its downstream side, internal erosion of the kind Chapter 11 describes. Real cores sit between filters and transitions graded so that each zone holds back the one beside it (Chapter 9). At Mount Polley the filter and transition were each about 1 m wide, and the panel found many samples of the filter too coarse; it did not find that piping or cracking caused the breach (Mount Polley Review Panel (2015), pp. ii and 75).

13.4 · back to the question

It scores 1.41 against 0.94. Rockfill drains as fast as the reservoir falls, so no trapped water pushes the slope out after the reservoir's support has gone, and its grains keep their press; fill and clay keep their water. But the dam leaks 29,621 L/day per metre against the mission's cap of 1,000. (On sand and gravel the same dam scores 1.81 upstream; the mission fixes the foundation as clay.)

13.5 · back to the question

The reservoir is 22.8 m deep, so the rule gives 22.8 ÷ 3 = 7.6 m. The workbench says 11 m: the rule is a rough guide, and here it errs low.

13.6 · back to the question

Put them under the planned dam where it will be highest and its slopes steepest, in a line across the footprint from the upstream toe to beyond the downstream toe, because that is where a slip would run along a layer; and space them by the geology, closer where the glacial deposits change from place to place. Take them deeper than any slip the dam could cause, through all the glacial soils to firm ground. Recover undisturbed samples of every fine-grained layer and test them for their stress history and their undrained strength, not only their friction angle. Then run the stability check with that undrained strength as well as the drained one. Four holes cannot cover kilometres of dam; their job is to tell you where the next forty should go.

Epilogue · The dam after the drawings

E.1 · back to the question

A spillway reduced to a 12-inch pipe under a road; water rising to within a foot of the crest, so almost no freeboard; seepage near the outlet pipe; an outlet valve nobody could reach, so the lake could not be let down. A walk along the crest would have shown nearly all of them, as the inspections did. What was missing was action: an owner obliged to repair the dam, lower the lake or breach the bank as the letter of 1978 asked, and a regulator able to make that happen; a rating that matched what lay below; and a sign on the dam, so the hiker's report would have said "dam" and not "road". The review's twelve recommendations included the signs and a stronger regulation (Morhart (2010), pp. 4–5 and 7–8).

E.2 · back to the question

Any three of these. A downstream slope steeper than 2H:1V needs the engineer of record's justification and approval before it is built: Mount Polley's face stood at 1.3H:1V, and the panel found 2:1 would have held. A minimum factor of safety of 1.5 in operation: the design target had been 1.3. An operating manual with trigger levels and the actions they call for: the failed embankment's piezometers had none. An engineer of record who must report unfixed safety problems, reaching the chief inspector within 72 hours. An independent review board, and an independent dam safety review every five years. The guidelines for investigating the ground under tailings dams, published alongside the code changes in 2016, answer the layer nobody found (BC Ministry of Energy and Mines (2016); Mines Code, Part 10; Mount Polley Review Panel (2015), pp. iii, 64 and 79).

E.3 · back to the question

The margin is 1.628 − 1.5 = 0.128. As ru: 0.128 ÷ 1.77 = 0.0723. As water pressure: 0.0723 × 83.4 = 6.03 kPa. As head: 6.03 ÷ 9.81 = 0.61 m. Bishop's full sum on the drained circle gives 0.64 m. The alarm sits about two-thirds of a metre above today's reading. The drained dam's critical circle is shallower than the homogeneous dam's, so the same metre of water carries a larger share of its slices' weight.

E.4 · back to the question

With 22.8 m of water the seepage breaks out 11 m up the slope instead of 8, the leak rises from 44.4 to 58.1 L/day per metre, and the factor of safety falls from 1.12 to 1.02. Piezometers in the downstream half would rise, the weir would read more, and anyone walking the slope would see the wet patch climb. Each change is read against the reservoir gauge, which is how an engineer tells a dam obeying Darcy's law from a dam whose drains are failing.

Reference

Glossary

Every dotted term on the page, in plain words. Hover or tap a term anywhere to see its definition.

Reference

Case files

  • Bouzey Dam · 1895 · Water in the joints of a wall too thin above its buttress; sliding on the foundation in 1884 · Chapter 7
  • Coquitlam Dam · 1913 · Loose hydraulic fill that could liquefy in a large earthquake; a new compacted earth-core rockfill dam built against its downstream toe (2007–2008) · Chapter 8
  • Teton Dam · 1976 · Internal erosion of a windblown-silt core into open rock joints, across the base of a key trench, on first filling · Chapter 11
  • W.A.C. Bennett Dam · 1996 · Internal erosion of the core around a survey benchmark casing; caught by the transition zone · Chapter 11
  • Mount Polley tailings dam · 2014 · See Chapter 13 · Chapter 13
  • Testalinden Dam · 2010 · Overtopping: snowmelt flowed over a crest with almost no freeboard and a spillway reduced to a small pipe, after decades of unheeded repair orders · the Epilogue

Reference

Bibliography

Books

  • Cedergren, H.R. (1989). Seepage, Drainage, and Flow Nets, 3rd ed. New York: Wiley. Cited in Chapter 3.
  • Darcy, H. (1856). Les fontaines publiques de la ville de Dijon. Paris: Victor Dalmont. Cited in Chapter 3.
  • Duncan, J.M., Wright, S.G., Brandon, T.L. (2014). Soil Strength and Slope Stability, 2nd ed. Hoboken: Wiley. Cited in Chapter 12 and Chapter 13.
  • Dupuit, J. (1863). Études théoriques et pratiques sur le mouvement des eaux dans les canaux découverts et à travers les terrains perméables, 2nd ed. Paris: Dunod. Cited in Chapter 10.
  • Fell, R., MacGregor, P., Stapledon, D., Bell, G., Foster, M. (2015). Geotechnical Engineering of Dams, 2nd ed. Leiden: CRC Press/Balkema.
  • Feynman, R.P., Leighton, R.B., Sands, M. (1964). The Feynman Lectures on Physics, Vol. II. Reading, Mass.: Addison-Wesley. https://www.feynmanlectures.caltech.edu/II_40.html Cited in Chapter 2.
  • Freeze, R.A., Cherry, J.A. (1979). Groundwater. Englewood Cliffs: Prentice-Hall. https://gw-project.org/books/groundwater/ Cited in Chapter 3.
  • Heyman, J. (1972). Coulomb's Memoir on Statics: An Essay in the History of Civil Engineering. Cambridge: Cambridge University Press. Cited in Chapter 1.
  • Holtz, R.D., Kovacs, W.D., Sheahan, T.C. (2011). An Introduction to Geotechnical Engineering, 2nd ed. Upper Saddle River: Pearson. Cited in Chapter 4.
  • Hvorslev, M.J. (1951). Time Lag and Soil Permeability in Ground-Water Observations. US Army Corps of Engineers Waterways Experiment Station, Bulletin 36. Cited in Chapter 3.
  • Minnaert, M.G.J. (1976). Stevin, Simon. In C.C. Gillispie (ed.), Dictionary of Scientific Biography, vol. 13. New York: Charles Scribner's Sons. The entry begins on p. 47. https://www.encyclopedia.com/people/science-and-technology/mathematics-biographies/simon-stevin Cited in Chapter 2.
  • Stevin, S. (1586). De Beghinselen des Waterwichts. Leiden: printed at the office of Christoffel Plantijn by Françoys van Raphelinghen. Cited in Chapter 2.
  • Terzaghi, K. (1925). Erdbaumechanik auf bodenphysikalischer Grundlage. Leipzig and Vienna: Franz Deuticke. Cited in Chapter 4.
  • Terzaghi, K., Peck, R.B., Mesri, G. (1996). Soil Mechanics in Engineering Practice, 3rd ed. New York: Wiley. Cited in Chapter 5 and the Epilogue.
  • Vick, S.G. (1990). Planning, Design, and Analysis of Tailings Dams. Vancouver: BiTech. Cited in the Prologue and Chapter 8.

Papers

  • Adamo, N., Al-Ansari, N., Sissakian, V., Laue, J., Knutsson, S. (2020). Dam safety problems related to seepage. Journal of Earth Sciences and Geotechnical Engineering 10(6): 191–239. https://www.diva-portal.org/smash/get/diva2:1444095/FULLTEXT01.pdf Cited in Chapter 6.
  • Bishop, A.W. (1959). The principle of effective stress. Teknisk Ukeblad 106(39): 859–863. Cited in Chapter 5.
  • Bishop, A.W. (1955). The use of the slip circle in the stability analysis of slopes. Géotechnique 5(1): 7–17. doi:10.1680/geot.1955.5.1.7 Cited in Chapter 12.
  • Bishop, A.W., Morgenstern, N. (1960). Stability coefficients for earth slopes. Géotechnique 10(4): 129–153. doi:10.1680/geot.1960.10.4.129 Cited in Chapter 12 and the Epilogue.
  • Bordes, J.-L. (2010). Les barrages en France du XVIIIe à la fin du XXe siècle : histoire, évolution technique et transmission du savoir. Pour mémoire 9 (hiver 2010): 70–121. https://www.barrages-cfbr.eu/IMG/pdf/barrages-jlb.pdf Cited in Chapter 7.
  • Casagrande, A. (1936). Characteristics of cohesionless soils affecting the stability of slopes and earth fills. Journal of the Boston Society of Civil Engineers 23(1): 13–32. Cited in Chapter 5.
  • Casagrande, A. (1937). Seepage through dams. Journal of the New England Water Works Association 51(2): 131–172. Cited in Chapter 10.
  • Cooke, J.B., Sherard, J.L. (1987). Concrete-face rockfill dam: II. Design. Journal of Geotechnical Engineering 113(10). Cited in Chapter 9.
  • Coulomb, C.-A. (1776). Essai sur une application des règles de maximis et minimis à quelques problèmes de statique, relatifs à l'architecture. Mémoires de mathématique et de physique, présentés à l'Académie Royale des Sciences par divers savans 7: 343–382 (read 1773). https://archive.org/details/mmoiresdemath07acad Cited in Chapter 1.
  • Cruz, P.T. (2008). A partial review of Barry Cooke and James Sherard 1987 papers on concrete face rockfill dams (CFRD). Soils and Rocks 31(2): 57–66. doi:10.28927/sr.312057 Cited in Chapter 9.
  • Dobry, R., Alvarez, L. (1967). Seismic failures of Chilean tailings dams. Journal of the Soil Mechanics and Foundations Division, ASCE 93(SM6): 237–260. doi:10.1061/JSFEAQ.0001054 Cited in Chapter 8.
  • Fannin, J. (2008). Karl Terzaghi: from theory to practice in geotechnical filter design. Journal of Geotechnical and Geoenvironmental Engineering 134(3): 267–276. doi:10.1061/(ASCE)1090-0241(2008)134:3(267) Cited in Chapter 11.
  • Foster, M., Fell, R., Spannagle, M. (2000). The statistics of embankment dam failures and accidents. Canadian Geotechnical Journal 37(5): 1000–1024. doi:10.1139/t00-030 Cited in Chapter 6, Chapter 11 and the Epilogue.
  • Fourie, A.B., Blight, G.E., Papageorgiou, G. (2001). Static liquefaction as a possible explanation for the Merriespruit tailings dam failure. Canadian Geotechnical Journal 38(4): 707–719. doi:10.1139/t00-112 Cited in Chapter 6.
  • Franks, D.M., Stringer, M., Torres-Cruz, L.A., Baker, E., Valenta, R., Thygesen, K., Matthews, A., Howchin, J., Barrie, S. (2021). Tailings facility disclosures reveal stability risks. Scientific Reports 11: 5353. doi:10.1038/s41598-021-84897-0 Cited in Chapter 6.
  • Horn, H.M., Deere, D.U. (1962). Frictional characteristics of minerals. Géotechnique 12(4): 319–335. Cited in Chapter 4.
  • Hornbaker, D.J., Albert, R., Albert, I., Barabási, A.-L., Schiffer, P. (1997). What keeps sandcastles standing? Nature 387: 765. doi:10.1038/42831 Cited in Chapter 1.
  • Houben, G., Batelaan, O., Birk, S., Kacimov, A. (2025). The force of Forchheimer: the contributions of Philipp Forchheimer to groundwater hydrology. Hydrological Sciences Journal 70(3): 436–455. doi:10.1080/02626667.2024.2438337 Cited in Chapter 10.
  • Jewell, R.J. (2012). Putting beach slope prediction into perspective. In R.J. Jewell, A.B. Fourie and A. Paterson (eds), Paste 2012. Perth: Australian Centre for Geomechanics, pp. 85–92. doi:10.36487/ACG_rep/1263_08_Jewell Cited in Chapter 8.
  • Lane, E.W. (1935). Security from under-seepage: masonry dams on earth foundations. Transactions of the American Society of Civil Engineers 100. Cited in Chapter 10.
  • Luino, F., De Graff, J.V. (2012). The Stava mudflow of 19 July 1985 (Northern Italy): a disaster that effective regulation might have prevented. Natural Hazards and Earth System Sciences 12: 1029–1044. doi:10.5194/nhess-12-1029-2012 Cited in Chapter 6 and Chapter 8.
  • Lévy, M. (1895). Quelques considérations sur la construction des grands barrages. Comptes rendus hebdomadaires des séances de l'Académie des sciences 121(6): 288–300 (séance du 5 août 1895). https://archive.org/details/mobot31753002064894 Cited in Chapter 7.
  • Majmudar, T.S., Behringer, R.P. (2005). Contact force measurements and stress-induced anisotropy in granular materials. Nature 435: 1079–1082. doi:10.1038/nature03805 Cited in Chapter 4.
  • Massarsch, K.R., Fellenius, B.H. (2012). Early Swedish contributions to geotechnical engineering. Full-Scale Testing and Foundation Design, ASCE Geotechnical Special Publication 227: 239–256 (GeoCongress 2012, Oakland). https://www.fellenius.net/papers/304%20Early%20Geocontributions.pdf Cited in Chapter 12.
  • Mesri, G. (1989). A reevaluation of su(mob) = 0.22σp using laboratory shear tests. Canadian Geotechnical Journal 26(1): 162–164. doi:10.1139/t89-017 Cited in Chapter 5.
  • Pakpour, M., Habibi, M., Møller, P., Bonn, D. (2012). How to construct the perfect sandcastle. Scientific Reports 2: 549. doi:10.1038/srep00549 Cited in Chapter 1.
  • Peck, R.B. (1969). Advantages and limitations of the observational method in applied soil mechanics. Géotechnique 19(2): 171–187. doi:10.1680/geot.1969.19.2.171 Cited in the Epilogue.
  • Pelletier, J.D., Sherman, D.J., Ellis, J.T., Farrell, E.J., Jackson, N.L., Li, B.Y., Nordstrom, K.F., Maia, L.P., Omidyeganeh, M. (2015). Dynamics of sediment storage and release on aeolian dune slip faces: A field study in Jericoacoara, Brazil. Journal of Geophysical Research: Earth Surface 120(9): 1911–1934. doi:10.1002/2015JF003636 Cited in Chapter 1.
  • Petterson, K.E. (1955). The early history of circular sliding surfaces. Géotechnique 5(4): 275–296. doi:10.1680/geot.1955.5.4.275 Cited in Chapter 12.
  • Proctor, R.R. (1933). Fundamental principles of soil compaction. Engineering News-Record 111(9), 31 August 1933, beginning p. 245. Cited in Chapter 9 and the Answers and hints.
  • Reynolds, O. (1885). On the dilatancy of media composed of rigid particles in contact. With experimental illustrations. Philosophical Magazine, Series 5, 20(127): 469–481. doi:10.1080/14786448508627791 Cited in Chapter 1 and Chapter 5.
  • Richefeu, V., El Youssoufi, M.S., Radjaï, F. (2006). Shear strength properties of wet granular materials. Physical Review E 73(5): 051304. doi:10.1103/PhysRevE.73.051304 Cited in Chapter 5.
  • Salençon, J. (2024). About the critical height of a vertical cut. Comptes Rendus. Mécanique 352: 269–275. doi:10.5802/crmeca.276 Cited in Chapter 5.
  • Seed, H.B., Lee, K.L., Idriss, I.M., Makdisi, F.I. (1975). The slides in the San Fernando dams during the earthquake of February 9, 1971. Journal of the Geotechnical Engineering Division, ASCE 101(GT7): 651–688. doi:10.1061/AJGEB6.0000178 Cited in Chapter 5.
  • Sherard, J.L., Dunnigan, L.P. (1989). Critical filters for impervious soils. Journal of Geotechnical Engineering 115(7): 927–947. doi:10.1061/(ASCE)0733-9410(1989)115:7(927) Cited in Chapter 11.
  • Sherard, J.L., Dunnigan, L.P., Decker, R.S., Steele, E.F. (1976). Pinhole test for identifying dispersive soils. Journal of the Geotechnical Engineering Division, ASCE 102(1): 69–85. doi:10.1061/AJGEB6.0000236 Cited in Chapter 11.
  • Simmons, C.T. (2008). Henry Darcy (1803–1858): immortalised by his scientific legacy. Hydrogeology Journal 16: 1023–1038. doi:10.1007/s10040-008-0304-3 Cited in Chapter 3.
  • Spencer, E. (1967). A method of analysis of the stability of embankments assuming parallel inter-slice forces. Géotechnique 17(1): 11–26. doi:10.1680/geot.1967.17.1.11 Cited in Chapter 12.
  • Stewart, R.A. (1998). WAC Bennett Dam: the sinkhole crisis. Australian National Committee on Large Dams (ANCOLD), paper. https://ancold.org.au/?p=25173 Cited in Chapter 11.
  • Tannant, D.D. (2015). Changes to dam safety management in British Columbia triggered by failure of a small earth dam. Georisk: Assessment and Management of Risk for Engineered Systems and Geohazards 9(3): 148–157. doi:10.1080/17499518.2015.1055278 Cited in the Epilogue.
  • Terzaghi, K. (1936). The shearing resistance of saturated soils and the angle between the planes of shear. Proceedings of the First International Conference on Soil Mechanics and Foundation Engineering, Cambridge, Massachusetts, vol. 1: 54–56. https://www.issmge.org/uploads/publications/1/44/1936_01_0017.pdf Cited in Chapter 4.
  • Terzaghi, K. (1923). Die Berechnung der Durchlässigkeitsziffer des Tones aus dem Verlauf der hydrodynamischen Spannungserscheinungen. Sitzungsberichte der Akademie der Wissenschaften in Wien, Mathematisch-naturwissenschaftliche Klasse, Abteilung IIa, 132: 125–138. Cited in Chapter 5.
  • Terzaghi, K. (1922). Der Grundbruch an Stauwerken und seine Verhütung. Die Wasserkraft 17(24): 445–449. Cited in Chapter 11.
  • Terzaghi, K. (1929). Effect of minor geologic details on the safety of dams. American Institute of Mining and Metallurgical Engineers, Technical Publication 215, February 1929; also in Geology and Engineering for Dams and Reservoirs (AIME, 1929). Cited in Chapter 13.
  • Turner, I. (1993). Water table outcropping on macro-tidal beaches: a simulation model. Marine Geology 115(3–4): 227–238. doi:10.1016/0025-3227(93)90052-W Cited in Chapter 3.
  • Yan, L., Sy, A., Thavaraj, T. (2008). Seismic deformation analyses of the new Coquitlam Dam. Geotechnical Earthquake Engineering and Soil Dynamics IV, ASCE Geotechnical Special Publication 181, pp. 1–10. doi:10.1061/40975(318)180 Cited in Chapter 8.

Reports and investigations

Standards and guidelines

British Columbia law and regulators

On the web

Appendix

How the models work

Reference

Errata and changes

  • 2026-10-24 · The permeability race, description and caption. Was: a wet front moves down each column; gravel is soaked through in seconds. Now: each column is already soaked, and the shading follows the first water poured on top, as if it were dyed. The race's times, gap fraction × length ÷ (k × gradient), are how long a drop takes to cross soil that is already full of water. Wetting dry soil from the top is a different process.
  • 2026-10-24 · The permeability race, side note. Was: time = gap fraction × length ÷ (k × push). Now: time = gap fraction × length ÷ (k × gradient). The note used a word, push, that the page never defined; the quantity is the hydraulic gradient.
  • 2026-10-24 · The opening of the chapter. Was: This step explains the three kinds of sand on the beach, and with them almost every way a dam can fail.. Now: What water does instead was worked out by Karl Terzaghi. By the end of this chapter it explains why water can make the castle stronger, why a heap under still water keeps its slope, the 19° face the beach chapter left unexplained, and the number every later chapter uses to say whether a slope will stand.. Overclaim: effective stress does not explain dams that fail by overtopping, which the book treats separately.
  • 2026-10-24 · The paragraph introducing effective stress. Was: When the gaps are full of water, the water pushes on every grain from all sides, like the cushion under the block.. Now: The water in the gaps meets almost all of the surface, and it presses on all of it with its own pore pressure; the water therefore carries u of each square metre's σ, and only the rest is left for the grains to carry.. Pore pressure does not push grains apart or outward; it carries part of the load the grains would otherwise carry between them.
  • 2026-10-24 · Where Part I comes back. Was: Soil is strong because its grains press on each other, and water pressure in the gaps pushes them apart.. Now: Soil holds as well as its grains press on one another, and water in the gaps, pushing or pulling, changes that press.. Water pressure acts equally on every grain round a gap; it takes part of the load the grains carry between them rather than pushing them apart.
  • 2026-10-24 · Grain sandbox, the house experiment. Was: With nothing pressing the grains together, friction between them vanishes. Now: Grains that hardly press can hardly grip. The grains still touch and still press a little; their grip falls with the press, toward zero, rather than vanishing outright.
  • 2026-10-24 · The water dam and tailings dam figure, caption. Was: the head H, the drop from the reservoir to the tailwater, drives the turbines. The load on the dam comes from the water's full depth at the heel, a few metres more than H. Now: the head H drives the turbines; the pressure on the upstream face comes from the full depth of water at the heel, a few metres more than H, as a worked example now shows. Readers who checked the heel readout with H got 0.83 MPa against the figure's 0.87; the example works both and shows the 4 m of tailwater that separates them.
  • 2026-10-24 · Why tailings dams fail differently, the list of failures. Was: a timeline of five failures under the liquefaction paragraph, with no raise method or mechanism for some. Now: a table of six failures giving each dam's ore, raise method, trigger, what happened, toll and volume, with sources. The list implied that every failure in it was a liquefaction of the tailings. Merriespruit began with water over the crest, and Mount Polley was not a liquefaction of the tailings.
  • 2026-10-24 · The paragraph on zones. Was: Read any section from the middle outward and you'll find the same order: tight in the centre, coarser and more free-draining toward the faces.. Now: The order (tight in the centre, coarser toward the faces) is stated for dams with a core; the chapter then shows the concrete-face rockfill, gravity and tailings dams breaking it.. Three of the four sections in the figure right below the sentence contradicted it.
  • 2026-10-24 · Concrete-face rockfill dam, face slab card. Was: Reinforced concrete, about 0.3 m plus 0.002–0.003 × depth … Leakage of hundreds of litres per second has been recorded at large CFRDs.. Now: Reinforced concrete, about 0.3 m thick near the crest and thicker with depth … Leaks of 1,000 to 3,000 litres a second have been measured at large CFRDs, and the rockfill passed them without harm.. The thickness rule could not be checked against a source; the measured leaks at Barra Grande and Campos Novos are reported by Cruz (2008).
  • 2026-10-24 · Film: A wall made of dirt, marks. Was: 2:59 All clay, and no drier, linked to the question about a metre of clay. Now: 2:59 The same dam, all of clay, linked to the question this chapter asks about it. The mark pointed to a question about travel time, and its label gave away the answer to the question placed before the film.
  • 2026-10-24 · Caption of the figure of core grains meeting rockfill and a filter. Was: The pores are too small, the particles jam at the interface, and only water passes.. Now: The filter's gaps are smaller than the core's coarser grains: those stop at the face, and the finer grains pile up behind them, so the water passes. The drawing stops every grain at the face by rule; in a real filter only grains bigger than its openings stop there, and the layer they build holds back the rest.. A filter's gaps are larger than the core's fine grains; it holds them because the coarse grains bridge its openings and build their own filter. The caption also disagreed with the drawing's own label.
  • 2026-10-24 · The filter rules. Was: A filter's pore throats are roughly a fifth of its D15.. Now: A graded filter acts like a sieve with openings about 0.11 times its D15 (FEMA 2011, pp. 275–276, after Sherard and colleagues' tests).. A fifth is an old, rough estimate; the measured effective opening of a graded filter is about a ninth of its D15.
  • 2026-10-24 · The Bennett Dam case. Was: Engineers lowered the reservoir and filled the voids by compaction grouting … R.A. Stewart's 1998 account of the crisis ranks it as the gravest dam safety concern in BC Hydro's history.. Now: Both sinkholes were repaired by compaction grouting (the lowering of the reservoir and the ranking are left out).. Neither the lowering of the reservoir nor the ranking could be found in the sources available; the panel's report and Stewart's abstract support the repair and its dates.
  • 2026-10-24 · The workbench's mission briefs. Was: The briefs of Dry the slope, Hold the water, Tailings on a budget and Mount Polley did not say which slope they check.. Now: Each brief says it checks the downstream slope only; Empty it fast checks the upstream one.. A reader who passed a mission and switched the view to the upstream slope saw a failing verdict with no explanation.
  • 2026-10-24 · Tailings on a budget, the completed mission. Was: The mission could be completed while the dam leaked about 8,000 L/day per metre, and the completion text did not mention it.. Now: The completion text names the leak as the price of the drain, and says what a mine does with that water.. For a tailings dam the leak is polluted water; a passed mission should not hide it.
  • 2026-10-24 · Mount Polley case. Was: A case card: 'released about 25 million m³ of tailings and water'.. Now: A case plate drawn to scale from the Review Panel's section: between 21 and 25 million m³, as the Prologue gives it.. The Chief Inspector of Mines reports the release as a range, 21 to 25 million m³; the book now gives one figure throughout.