- A dam is not a structure that holds water. It is a controlled failure that decides where the water goes when something goes wrong. Every design decision — spillway, freeboard, foundation treatment — is a decision about the failure path.
- Half of all dam failures occur in the first ten years. For arch and buttress dams, the figure is 100%. The risk is concentrated at first filling, not at old age.
- Since 2000, 70% of failures happened during floods. Overtopping accounts for more than 40% of failures historically; foundation failure accounts for 53% of concrete dam failures.
- Sedimentation is the slow failure. The world loses over 7% of freshwater storage capacity each decade, and half of all reservoirs may be functionally dead by 2060.
- Decommissioning is now a budget line. Removing 20% of US dams over 75 years is projected at $6.4B by 2025 and $22.4B by 2075.
Dam engineering is the discipline of deciding which failure you can afford. You cannot eliminate all failure modes — you can choose which one you design for, which one you monitor, and which one you accept. This guide teaches the basics through that lens: the four types and what each refuses to do, the failure statistics that should govern your design, the three loads that actually matter, and the slow failures — sedimentation and aging — that most introductions ignore.
On this page
- A dam is a controlled failure
- The four types, and what each one refuses to do
- What actually kills dams: the statistics
- The three loads that govern design
- The foundation is the dam
- The spillway: designing the flood you cannot stop
- The sedimentation clock
- Decommissioning: the decision nobody budgets for
- What you actually instrument
- The climate reclassification
- A practical dam engineering checklist
- Frequently asked questions
1. A Dam Is a Controlled Failure
The standard definition — a barrier that impounds water — is accurate and useless. It describes what a dam is, not what a dam engineer does.
What a dam engineer does is decide, explicitly or by omission, where the water will go when the dam does not hold. A spillway is a controlled release. A freeboard allowance is a controlled margin against overtopping. A grout curtain is a controlled seepage path. A downstream warning system is a controlled evacuation. Every one of those is a decision about failure — not about success.
The One-Sentence Reframe
A dam is a structure designed to fail in a direction you chose. The spillway is the failure you want. The breach is the failure you do not. Everything between them — freeboard, filters, grout curtains, monitoring — is the engineering.
2. The Four Types, and What Each One Refuses to Do
Most introductions describe dam types by how they carry load. That is correct and incomplete. A better way to choose a type is to ask what it cannot do — because the constraint usually decides the selection before the load path does.
| Type | Load path | Best foundation | Fatal weakness |
|---|---|---|---|
| Concrete gravity | Self-weight resisted by vertical reaction at the base | Sound rock; can be adapted to most foundations | Uplift and sliding; foundation failure accounts for 53% of concrete dam failures |
| Arch | Load carried in compression to the abutments | Competent rock abutments; narrow valley | Differential settlement; sensitive to faulty geology |
| Embankment (earth/rockfill) | Shear strength of the compacted fill | Almost any foundation, including alluvium | Overtopping and internal erosion; 46% of embankment failures are internal erosion |
| Buttress | Inclined slabs transfer load to the base | Sound rock; wide valleys | Construction cost and complexity; formwork-intensive |
| RCC gravity | Self-weight, placed by roller compaction | Sound rock | Lift joints and seepage control; needs continuous placement |
Failure percentages from ICOLD incident data. RCC is a construction method for gravity and arch-gravity dams, not a separate structural type.
Why Embankment Dams Are the Most Common Type
Embankment dams are the most numerous type worldwide because they can be built on foundations that would rule out a concrete dam, and because they use locally available fill. The trade-off is that they cannot be overtopped — a concrete gravity dam can survive water flowing over its crest in some conditions; an embankment dam usually cannot. That is why embankment dams depend so heavily on spillway capacity and freeboard.
3. What Actually Kills Dams: The Statistics
The failure record is the most useful design document in dam engineering, because it tells you which failure modes the profession keeps getting wrong. The numbers are consistent across ICOLD, national databases, and recent statistical analyses.
| Finding | Value | Implication for design |
|---|---|---|
| Failures in the first ten years of a dam’s life | 50%; 100% for arch and buttress dams | First filling and early operation are the highest-risk period, not old age |
| Failures during flood events since 2000 | 70% | Hydrological loading dominates modern failure statistics |
| Overtopping share of historical failures | More than 40% | Spillway capacity and freeboard are the first line of defence |
| Foundation share of concrete dam failures | 53% | The foundation is the dam; treat it as a structural element |
| Internal erosion share of embankment failures | 46% | Filters, zoning, and seepage control are not optional details |
| Piping failures during first filling | About 50%, with 64% within five years | Instrumentation must be active before impoundment begins |
| Large dam failures expected worldwide, 2023–2035 | 23 (95% CI 14–33) | Failure is not a historical curiosity; it is a forecast |
| Large dams with probability of failure above 1/10,000 | 4.4% | Most dams are safe; a small population carries most of the risk |
Sources: ICOLD incident database and Bulletin 99 update; Nature Water 2025 statistical analysis; Foster et al. internal erosion statistics. Percentages are not directly comparable across datasets.
The Infant Mortality Finding
Dam failures are concentrated at the start of life, not the end. The first ten years account for half of all failures, and for arch and buttress dams the figure is 100% in the ICOLD record. The mechanism is usually first filling: the reservoir loads the foundation and the structure for the first time, and any defect in the foundation treatment, the filters, or the spillway is revealed. This is why commissioning instrumentation, not just design instrumentation, matters.
The Forecast
The 2025 Nature Water analysis estimates 23 large dam failures worldwide between 2023 and 2035, with a 95% confidence interval of 14–33. It also finds that 4.4% of large dams have a probability of failure greater than 1 in 10,000, while 85% are at least twice as safe as that threshold. The risk is concentrated: a small population of dams carries most of the expected failures, and those dams are disproportionately old embankments between 15 and 70 m high, built in the second half of the last century.
4. The Three Loads That Govern Design
Textbook load lists run to a dozen items: self-weight, hydrostatic pressure, uplift, silt, wave, ice, temperature, earthquake, construction. Three of them govern the design of almost every dam.
4.1 Water: The Obvious Load
Hydrostatic pressure increases linearly with depth and acts normal to the upstream face. For a gravity dam, the overturning moment from the reservoir is resisted by the weight of the concrete. For an arch dam, the same pressure is carried in compression to the abutments. For an embankment, it drives seepage through the fill and the foundation.
4.2 Uplift: The Load That Is Easy to Forget
Water seeps under the dam and pushes up on the base. Uplift reduces the effective weight available to resist sliding and overturning. It is controlled by a grout curtain to limit seepage and a drainage system to relieve pressure. Uplift accounts for about 7% of concrete dam failures in the ICOLD record, but it is a contributing factor in far more — because it reduces the margin on the other failure modes.
The Uplift Arithmetic
A gravity dam is stable when the resultant of all forces passes within the middle third of the base under normal loading. Uplift reduces the effective weight, moves the resultant downstream, and can push it outside the middle third. Once that happens, the upstream face goes into tension — and concrete is weak in tension. The fix is not more concrete; it is drainage and a grout curtain that keep the uplift pressure low.
4.3 Time: The Load Nobody Puts in the Load Combination
Time is not a force, but it is the load that decides whether the dam is still there in 100 years. It acts through three mechanisms: sedimentation (which adds silt pressure upstream and reduces storage), material degradation (alkali-aggregate reaction, sulfate attack, freeze–thaw), and aging of the foundation (progressive erosion, dissolution, or weakening of the rock mass). Most design codes treat these as durability requirements rather than loads. In practice, they determine the asset’s life.
| Load | Nature | Primary control | Typical design factor |
|---|---|---|---|
| Hydrostatic | Static, predictable | Section geometry and weight | Factor of safety on sliding: ≥1.5 normal, ≥1.3 flood, ≥1.1 seismic |
| Uplift | Static, depends on seepage control | Grout curtain and drains | Reduced uplift assumed where drainage is verified |
| Silt | Increases over time | Sediment management, flushing, bypass | Often treated as an additional horizontal load |
| Seismic | Dynamic, probabilistic | Monolithic behaviour, contraction joints, freeboard | Design earthquake with return period typically 1,000–10,000 years |
| Time / degradation | Progressive | Material specification, monitoring, maintenance | Service-life modelling rather than a single factor |
Factors of safety and return periods vary by jurisdiction and dam type. Values shown are indicative of common international practice.
5. The Foundation Is the Dam
Foundation failure accounts for 53% of concrete dam failures. That single number should reorganise how a dam engineer thinks about the structure: the concrete is the easy part. The rock beneath it is where the uncertainty lives.
5.1 The Treatment Toolbox
| Method | What it does | Best suited to | Verification |
|---|---|---|---|
| Consolidation grouting | Fills voids and fractures in the upper rock mass | Fractured or weathered rock | Grout take records, check holes, water pressure tests |
| Curtain grouting | Creates a low-permeability barrier below the dam | Pervious rock, karst, fault zones | Water pressure tests, seepage monitoring |
| Concrete cut-off wall | Physical barrier through alluvium | Deep alluvial foundations | Continuity testing, excavation records |
| Drainage galleries and relief wells | Relieves uplift pressure downstream of the curtain | Gravity dams and abutments | Piezometer readings, seepage weirs |
| Slurry trench / deep soil mixing | Low-permeability barrier in soils | Soft alluvium, sand and gravel | Permeability testing, integrity testing |
The choice depends on the foundation geology, the depth of the permeable zone, and the consequence of seepage. Verification is not optional; an unverified curtain is an assumption.
Why Grout Curtains Are Installed Deeper Than You Think
Good practice is to extend the curtain at least 50 feet below and beyond the originally foreseen extent of the treatment zone. The reason is not conservatism for its own sake — it is that the most damaging foundation defects are the ones the investigation did not find. A curtain that stops at the expected boundary leaves the unanticipated feature untreated, and that feature becomes the seepage path that fails the dam.
6. The Spillway: Designing the Flood You Cannot Stop
The spillway is the most important structural element on most dams, because it is the one that converts an uncontrolled failure into a controlled release. Overtopping accounts for more than 40% of historical dam failures, and since 2000, 70% of failures have occurred during floods.
6.1 The Design Flood Is a Policy Decision
ICOLD recommends estimating a flood discharge with a 1,000-year return period for spillway design and a 10,000-year return period for the safety of the dam structure itself. Those are not natural constants. They are policy choices about how much protection society is willing to pay for, and they vary by jurisdiction.
| Element | Function | Design basis | Failure consequence |
|---|---|---|---|
| Service spillway | Passes normal floods | Often 1,000-year return period | Repairable damage |
| Auxiliary / emergency spillway | Passes extreme floods | Probable maximum flood or 10,000-year event | Controlled damage, no breach |
| Freeboard | Margin against overtopping from waves and gate failure | Wave run-up plus a minimum allowance | Loss of margin; potential overtopping |
| Energy dissipator | Removes kinetic energy before the river | Design flood plus a safety margin | Scour, erosion, damage to the toe |
| Gates and operating rules | Control release | Operating rules tested against flood forecasts | Gate failure is a recurring cause of overtopping |
Return periods and freeboard requirements vary by country and dam hazard classification. The table shows common international practice, not a universal code.
The Gate Failure Problem
Many overtopping failures are not caused by a spillway that is too small — they are caused by a spillway that could not be opened. Gate failure, power loss, or operating error during a flood turns a designed release into an uncontrolled one. That is why gated spillways need redundant power, manual operation, and operating rules that are tested before the flood arrives, not during it.
7. The Sedimentation Clock
Most dam engineering introductions treat sedimentation as an environmental footnote. It is not. It is the slow failure mode that determines whether the dam still performs its function in 50 years.
Dams trap sediment by design. The reservoir fills from the upstream end, storage capacity declines, and the dam’s ability to attenuate floods and supply water declines with it. The global numbers are stark.
| Metric | Value | Implication |
|---|---|---|
| Global freshwater storage lost to sedimentation each decade | More than 7% | Storage is a depleting asset, not a fixed one |
| Annual storage loss | More than 36 km³ — comparable to the Three Gorges Reservoir | The scale is comparable to building a major new reservoir every year just to stand still |
| Reservoirs “functionally dead” by 2060 | Over half globally | Functionally dead means more than 50% filled with sediment |
| Worst-affected countries by 2060 | 85% of Australian and 75% of Spanish reservoirs | Arid and semi-arid regions are disproportionately exposed |
| Namibia and Western Australia | Over 99% and nearly 96% of dams at risk | Some regions face near-total loss of reservoir function |
| Cost of dredging and bypass measures | Up to US$100 billion | Sedimentation is a capital budget problem, not just an environmental one |
| Sediment reduction from combined catchment BMPs | About 51% in a semi-arid case study | Catchment management is cheaper than dredging but slow to take effect |
Values from the 2026 Nature Sustainability global reservoir analysis and the Kiri Dam catchment study. Sedimentation rates vary widely with catchment condition, land use, and climate.
The Design Life Assumption
Many dams were designed with an assumed sediment yield and a design life of 50–100 years. In catchments where land use changed after construction — deforestation, agriculture, urbanisation — the actual sediment yield can be several times the design assumption. The reservoir then fills faster than planned, and the dam’s functional life is shorter than its structural life. That gap between structural life and functional life is the core of modern dam asset management.
8. Decommissioning: The Decision Nobody Budgets For
Every dam has an end-of-life decision: repair, upgrade, replace, or remove. Most introductions omit it. The economics are now large enough that it belongs in the basics.
| Scenario | Projected mean cost | Context |
|---|---|---|
| Remove 20% of US dams over 75 years old | $6.4B by 2025; $22.4B by 2075 | 3,935 dams by 2025; 15,165 by 2075 |
| Remove all US dams in poor condition | $10.6B | 7,475 dams; 50% prediction interval $2.2B–$29.2B |
| Remove poor-condition dams with high hazard rating | $6.5B | 3,621 dams |
| Remove dams that are poor, high-hazard, and over 50 years old | $5.6B | 3,023 dams |
| Standardised cost per unit dam height | US$0.14M per metre | Range by state: $0.05M/m in Georgia to $0.39M/m in Idaho |
US National Inventory of Dams analysis, 2026. Condition assessments are unavailable for 63% of NID dams, so these estimates are conservative.
The 63% Blind Spot
Condition assessments are unavailable for 63% of dams in the US National Inventory. That means the cost estimates above are based on a minority of the inventory, and the true liability is larger. The same pattern appears in many national dam inventories: the regulator knows a dam exists, but not what condition it is in. That is not an engineering problem — it is an information problem that precedes the engineering.
The decommissioning decision is not only about cost. Removal restores downstream sediment flow, reopens habitat, and eliminates a hazard. Retention preserves water supply, hydropower, and flood attenuation. The engineering question is which set of benefits the basin needs more, and the answer is different for a 5 m farm dam and a 100 m hydropower dam.
9. What You Actually Instrument
Monitoring is the feedback loop between the design assumption and the actual behaviour. A dam that is not instrumented is a dam whose design assumptions are never tested.
| Instrument | What it measures | Failure mode it detects | Typical location |
|---|---|---|---|
| Vibrating-wire piezometer | Pore water pressure | Uplift, internal erosion, seepage | Foundation, core, abutments |
| Inclinometer | Lateral displacement with depth | Slope instability, foundation movement | Embankment slopes, abutments |
| Extensometer | Displacement along a line | Crack opening, joint movement | Concrete dams, fault zones |
| Crack meter | Movement across a crack or joint | Structural distress | Concrete faces, galleries |
| Seepage weir / V-notch | Flow rate of seepage | Increasing seepage, turbidity | Drainage galleries, toe drains |
| Settlement points / survey | Surface settlement | Consolidation, foundation deformation | Crest, downstream slope |
| UAV photogrammetry / InSAR | Surface deformation | Broad-area movement, slope creep | Reservoir rim, abutments |
Instrumentation must be installed before first filling, not after. Piping failures cluster in the first five years, and the first filling is when the foundation is first loaded.
The Diamer Basha Framework
The Diamer Basha Dam — the world’s tallest RCC dam — uses an integrated real-time monitoring framework that combines UAV photogrammetry, GIS-based deformation mapping, and geotechnical instrumentation including piezometers, inclinometers, extensometers, and crack meters. The integration is the point: a piezometer reading alone tells you the pressure; combined with an inclinometer and a deformation map, it tells you whether the dam is behaving as designed.
The other half of monitoring is the seepage record. Increasing seepage flow, or seepage that becomes turbid, is one of the most reliable precursors of internal erosion. A V-notch weir in the drainage gallery is a cheap instrument that detects one of the most dangerous failure modes.
10. The Climate Reclassification
Dam design has always been based on a rainfall record. Climate change makes that record unreliable, and the effect on dam safety is not marginal.
The probable maximum precipitation (PMP) and probable maximum flood (PMF) are the design basis for spillways on high-hazard dams. Both are estimated from historical data and meteorological theory. As the atmosphere warms, it holds more moisture, and the extreme rainfall that drives PMP increases. A dam designed to a 1960s PMP may not pass a 2020s PMP.
| Climate effect | Mechanism | Dam safety consequence |
|---|---|---|
| Increased extreme rainfall intensity | Warmer atmosphere holds more moisture | Spillway capacity may be inadequate; freeboard may be insufficient |
| Increased sediment yield | More intense rainfall and changed land use | Faster storage loss; increased silt load |
| More frequent and severe floods | Changed flood frequency | Higher loading on spillways and downstream channels |
| Glacier retreat and GLOF risk | Glacial lake outburst floods | New upstream hazards that the original design did not consider |
| Drought and low reservoir levels | Reduced inflow | Operational stress, water supply conflict, reduced hydropower |
The non-stationarity of extreme rainfall is the central issue. Historical records are a lower bound, not a design basis.
Risk-Informed Decision Making
The modern response is not to rebuild every spillway to a new PMP. It is risk-informed decision making: estimate the probability and consequence of overtopping under a range of climate scenarios, compare that risk with the cost of spillway upgrade, and prioritise the dams where the risk reduction per dollar is highest. That is a different question from “does it pass the code?” — and it is the question most dam owners are now asking.
11. A Practical Dam Engineering Checklist
- Name the failure you are designing for. Every dam has a controlled failure path. State it, design it, and monitor it. The breach path is the one you are trying to make incredible.
- Select the type by constraint. Foundation, valley shape, and material availability usually decide the type before the load calculation does. A gravity dam on poor rock is a foundation failure waiting to happen.
- Treat the foundation as the structure. Foundation failure accounts for 53% of concrete dam failures. Grout curtains, cut-offs, and drainage are not ancillary works; they are the dam.
- Design the spillway for the flood you cannot stop. Overtopping is the most common historical failure mode, and 70% of failures since 2000 occurred during floods. Check the gates, the power supply, and the operating rules, not just the concrete.
- Instrument before first filling. Half of failures occur in the first ten years, and piping clusters in the first five. Commissioning instrumentation is a safety requirement, not a commissioning nicety.
- Model the sedimentation clock. A reservoir is a depleting asset. Estimate the sediment yield, plan the management method, and include the cost in the project economics.
- Budget for decommissioning. Every dam has an end-of-life decision. The cost is large enough to be a project economics line item, not an afterthought.
- Re-examine the design flood under climate scenarios. Historical PMP is a lower bound. Use risk-informed decision making to prioritise the dams where the risk reduction per dollar is highest.
12. Frequently Asked Questions
What is the most common cause of dam failure?
It depends on the dam type. For concrete dams, foundation failure accounts for 53% of failures. For embankment dams, internal erosion accounts for about 46%, with overtopping also a major cause. Historically, overtopping accounts for more than 40% of all dam failures, and since 2000, 70% of failures have occurred during flood events.
When in a dam’s life is it most likely to fail?
In the first ten years, which account for 50% of all failures. For arch and buttress dams, the ICOLD record shows 100% of failures in that period. The mechanism is usually first filling, when the reservoir loads the foundation and the structure for the first time. This is why instrumentation must be installed before impoundment, not after.
What is uplift, and why does it matter?
Uplift is water pressure acting upward on the base of the dam, caused by seepage under the structure. It reduces the effective weight available to resist sliding and overturning. It accounts for about 7% of concrete dam failures directly, but it is a contributing factor in many more because it reduces the margin on other failure modes. It is controlled by a grout curtain and a drainage system.
Why are embankment dams the most common type?
Because they can be built on foundations that would rule out a concrete dam, including alluvium and weak rock, and because they use locally available fill. The trade-off is that they cannot survive overtopping, which is why spillway capacity and freeboard are so critical for embankment dams.
What is the design flood for a dam?
ICOLD recommends a flood with a 1,000-year return period for spillway design and a 10,000-year return period for the safety of the dam structure. These are policy choices, not natural constants, and they vary by jurisdiction and dam hazard classification. The probable maximum flood (PMF) is used for high-hazard dams in many countries.
How long does a dam last?
The structural life is often 100 years or more, but the functional life is usually shorter because of sedimentation. The world is losing more than 7% of freshwater storage capacity each decade, and over half of reservoirs may be functionally dead by 2060. A dam can be structurally sound and functionally obsolete at the same time.
How much does it cost to remove a dam?
Removing 20% of US dams over 75 years is projected at $6.4 billion by 2025 and $22.4 billion by 2075. The mean cost for removing dams in poor condition is about $10.6 billion, and for poor-condition high-hazard dams about $6.5 billion. Standardised cost averages $0.14 million per metre of dam height, ranging from $0.05M/m in Georgia to $0.39M/m in Idaho.
How is climate change affecting dam safety?
Warmer air holds more moisture, so extreme rainfall intensity is increasing. A dam designed to a historical probable maximum precipitation may no longer pass a modern estimate. The consequence is that spillway capacity and freeboard may be inadequate. The modern response is risk-informed decision making: estimate the probability and consequence of overtopping under climate scenarios and prioritise upgrades by risk reduction per dollar.
Related reading on BuildersMeasures: Geotechnical Site Investigation: The Four Questions Every Ground Model Must Answer and Mass Concrete Placement: Thermal Control, Lifts & Real-World Failures.