- Resilience is not a bigger design storm. Infrastructure performs acceptably until a climate variable crosses a threshold, then fails nonlinearly. The engineering artefact is a threshold register, not a higher load factor.
- The economics are settled. A resilience premium of about 3% on new infrastructure generates roughly $4 of benefit per $1 invested; a decade of delay adds about $1 trillion in costs.
- Targeted interventions beat blanket overdesign. Measures costing 2.4–8% of asset value deliver the highest benefit-to-cost ratios — up to $8.60 of protected asset value per $1 in one sector study.
- Nature-based defences often out-earn grey. Mangrove restoration shows a global benefit–cost ratio of 3–6; in one coastal study, levees and home elevation returned below 1.
- The gap is implementation, not knowledge. Only 12–15% of surveyed infrastructure professionals report fully following resilience measures, and financing still favours post-disaster reconstruction over prevention.
Most climate resilience guidance tells engineers to design for a more severe future. That is necessary but insufficient, and it is the reason so much adaptation money is spent badly. This guide reframes the problem as threshold management: identify the climate variables that matter, find the value at which service degrades, define the trigger that starts the next intervention, and keep the option open rather than buying the maximum protection today. That framing is what adaptation pathways formalise, and it is what separates a resilience investment from an overdesign.
On this page
- The wrong mental model: resilience as a bigger load
- The threshold reframe
- The resilience dividend: 3%, 4:1, and the trillion-dollar delay
- The threshold register: four types of threshold
- Adaptation pathways: deciding now which trigger you will act on later
- Where the thresholds actually are, hazard by hazard
- Three ways resilience investment goes wrong
- Nature-based versus grey: what the benefit–cost data says
- The standards are catching up, unevenly
- The implementation gap: 12–15%
- A threshold audit checklist
- Frequently asked questions
1. The Wrong Mental Model: Resilience as a Bigger Load
The standard response to climate risk in design offices is to increase the design value. Raise the design rainfall intensity. Raise the design flood return period. Add freeboard. Increase the thermal movement allowance. Each of those is defensible in isolation, and together they produce a characteristic failure: the asset is designed for one future climate number, and the actual climate arrives as a sequence of surprises.
The problem is not that the numbers are wrong. The problem is that a single design value encodes a single assumption about the future, and it spends the entire adaptation budget at the moment of construction. If the future turns out milder than assumed, the money is wasted and the embodied carbon is emitted for nothing. If it turns out more severe, the asset has no capacity left to respond.
The Overdesign Trap
Attempting to prepare infrastructure for every plausible climate scenario “could result in overdesign, which would not only be expensive financially, but lead to unnecessary carbon emissions”. That is the assessment of the Institution of Civil Engineers in its introduction to the adaptation pathways approach now being formalised as PAS 3090. The point is not that overdesign is always wrong — it is that overdesign is the predictable outcome of treating an uncertain future as a single design case.
2. The Threshold Reframe
Infrastructure does not degrade smoothly as the climate shifts. It performs acceptably, then it does not. A drainage network designed for historic rainfall “may function adequately until extreme precipitation crosses a design threshold, after which failures become frequent.” A coastal ecosystem absorbs storm energy until degradation reduces its protective capacity. A health system copes with heat illness until simultaneous heat, wildfire smoke, and power outages overwhelm it.
In each case, the relationship between the climate variable and the service level is not a gentle slope. It is a plateau followed by a cliff. And the cliff is where the losses stop being linear.
The practical consequence is a change in what the engineer produces. A load-based design produces a number: the asset is designed for a 1-in-100-year event, or for a projected 2050 rainfall intensity. A threshold-based design produces a relationship: at what climate value does the asset stop meeting its service level, at what value does it fail, and what intervention is available at each point.
Why the Cliff Matters More Than the Slope
If performance declined linearly with the climate variable, you could buy resilience by buying a proportional margin. The cliff breaks that logic. Once the variable crosses T1, the marginal cost of maintaining service rises steeply, and once it crosses T2, the asset stops delivering the service altogether — regardless of how much margin was built in at the plateau. A margin sized on the slope will be exhausted before the cliff, and the cliff is where the losses are concentrated.
3. The Resilience Dividend: 3%, 4:1, and the Trillion-Dollar Delay
The economics of resilience are no longer contested. What is contested is why the investment is not happening. The benefit–cost evidence is consistent across independent bodies.
| Source | Finding | What it means |
|---|---|---|
| Global Commission on Adaptation | Benefit–cost ratios commonly 2:1 to 10:1; $1.8 trillion invested across five areas between 2020 and 2030 could generate $7.1 trillion in net benefits | The return is not marginal; it is multiples of the investment |
| World Bank, Lifelines analysis | Resilience in new infrastructure in low- and middle-income countries adds about 3% to construction cost and generates about $4 in benefits per $1; a decade of inaction adds roughly $1 trillion | The premium is small, the return is large, and delay has a price |
| IFC / AXA Climate / Scientific Climate Ratings (2026) | Investing less than 10% of asset value preserves multiples of it; in three Brazilian sectors, up to $8.60 of protected asset value per $1; highest ratios at 2.4–8% of asset value | Focused, lower-cost interventions outperform blanket hardening |
| Global Infrastructure Resilience Report 2025 | Disaster losses exceed $732 billion annually; indirect impacts average 7.4 times direct infrastructure damage | Measuring only the repair bill understates the loss by an order of magnitude |
| CDRI Africa working paper | Africa loses $12.7 billion annually; climate change is projected to increase losses by 27% | The exposure is growing faster than the asset base |
| McKinsey adaptation cost mapping | $190 billion spent annually today protects 1.2 billion people; $540 billion would protect all 4.1 billion exposed; the gap is $350 billion | The shortfall is a financing and prioritisation problem, not a technology problem |
Benefit–cost ratios are not directly comparable across studies: they use different discount rates, hazard sets, asset classes, and counterfactuals. The consistent finding is that the ratio exceeds 1 by a wide margin for targeted interventions.
The Most Important Number in This Article
The World Bank figure of a 3% construction premium for roughly 4:1 in benefits is the single most useful number in resilience engineering, because it is small enough to fit inside a normal project contingency. It means resilience is not a separate capital programme competing with the asset. It is a design decision available inside the existing budget — which is why the finding has been taken up by the OECD as a reason to mainstream resilience into ordinary infrastructure finance rather than treat it as a niche expenditure.
There is a second-order economic argument that is less often made. Early resilience protects the credit quality and investment capacity of the asset owner. Utilities and municipalities that suffer repeated losses face higher repair bills, weaker revenues, and greater borrowing needs. If resilience reduces outages, it improves the predictability of cash flows — which lowers the cost of capital for everything else the owner wants to build. The benefit therefore compounds beyond the asset itself.
4. The Threshold Register: Four Types of Threshold
If resilience is threshold management, the first deliverable is a register of thresholds. Most projects have none, or have one buried in a maintenance manual. A usable register distinguishes four types, because they are triggered by different variables and owned by different people.
| Threshold type | What it defines | Example variable | Decision it triggers |
|---|---|---|---|
| Performance threshold | The point at which the asset stops meeting its design service level | Rainfall intensity exceeding drainage capacity; temperature exceeding material tolerance | Operational change, restriction, or staged upgrade |
| Service threshold | The point at which users lose the service, not just some performance | Road closure depth; substation flooding level; water treatment turbidity limit | Emergency response, temporary works, rerouting |
| Financial threshold | The point at which expected annual damage or insurance cost makes the status quo uneconomic | Expected annual loss as a share of asset value | Capital investment case, insurance review, divestment |
| Ecological or social threshold | The point at which damage becomes irreversible or socially unacceptable | Habitat loss beyond recovery; displacement of a community | Redesign, relocation, or refusal of consent |
The four types are not hierarchical. A financial threshold may be crossed before a service threshold, which is why an asset can be economically obsolete before it is physically failing.
Two design rules follow from the register.
Rule one: each threshold needs a named owner and a defined trigger. A threshold without an owner is an observation. A threshold with an owner and a trigger is a decision. The difference is the whole point of the exercise.
Rule two: the intervention should be chosen at the threshold, not at the design stage. This is the logic of adaptation pathways, and it is the mechanism that avoids both overdesign and underdesign.
5. Adaptation Pathways: Deciding Now Which Trigger You Will Act On Later
Adaptation pathways are the formal method for managing thresholds under uncertainty. Rather than committing to one design case, a pathway approach identifies the trigger points at which climate conditions breach the thresholds where infrastructure services become ineffective or redundant, and sequences a range of measures over time in response.
The advantage is symmetrical. It mitigates the risk of over-investing in adaptation at any one time, and it mitigates the risk of under-investing. It keeps climate, technology, socioeconomic, and political factors in view, and it de-risks investment by making the next decision explicit rather than implicit.
5.1 What PAS 3090 Changes
The approach is being formalised. PAS 3090, expected in late 2026 and sponsored by the Institution of Civil Engineers with BSI, provides a framework for climate adaptation decision-making under uncertainty, guidance on embedding pathways into planning and governance, clarity on roles, responsibilities, metrics and data, methods for using climate scenarios and assessing impacts, approaches to address design-life constraints in codes, and tools for flexibility, thresholds, nature-based solutions, and business cases. It is intended to integrate with PAS 2080 on carbon management, because adaptation and decarbonisation decisions interact and should not be made in separate silos.
The Political Economy of Pathways
A pathway approach is politically harder than a single design case, because it requires a decision-maker to commit to a future action they may not be in office to take. That is precisely why the trigger must be written into an asset management plan, a regulatory instrument, or a concession agreement rather than left to a future committee. The engineering content of a pathway is straightforward; the governance content is the difficult part.
6. Where the Thresholds Actually Are, Hazard by Hazard
The threshold register only works if it names the right variable. The variables differ by hazard, and so do the asset types most exposed.
| Hazard | Threshold variable | Most exposed assets | What typically triggers first |
|---|---|---|---|
| Pluvial and fluvial flooding | Rainfall intensity and duration; river level | Drainage, roads, substations, treatment works | Drainage capacity exceeded; access routes cut |
| Coastal flooding | Sea level, surge height, wave energy | Ports, coastal roads, wastewater systems | Overtopping of defences; saltwater intrusion |
| Extreme heat | Peak temperature and duration | Rail, roads, power cables, bridges, water mains | Material tolerance; thermal movement; cable derating |
| Wildfire | Fire weather index; fuel continuity | Transmission lines, telecoms, water infrastructure | Asset ignition; smoke affecting air quality and solar output |
| Drought | Cumulative rainfall deficit; reservoir and aquifer level | Water supply, hydropower, navigation, cooling | Supply restrictions; reduced head; cooling constraint |
| Wind and ice | Gust speed; ice accretion | Bridges, overhead lines, towers | Design load exceedance; debris loading |
| Compound events | Joint occurrence of two or more hazards | All of the above | Loss of redundancy; simultaneous failure of response capacity |
Threshold variables are indicative. A site-specific register must be built from local climate projections and asset condition data, not from a generic table.
The compound event row is the one most often omitted and most often decisive. The threshold model assumes a single variable crossing a single line. In practice, the dangerous events are the ones where heat, smoke, and power outage coincide, or where a flood arrives while the drainage system is already at capacity. Compound events are where the nonlinear loss region is widest.
The Drainage Threshold Is Usually an Access Threshold
For a road or rail asset, the first service threshold is rarely structural. It is the depth at which the route becomes impassable, or the wind speed at which operations are suspended, or the temperature at which speed restrictions apply. Those are operational thresholds, and they are often crossed years before the asset suffers any physical damage. A resilience programme that only counts repair costs will miss them entirely, even though the indirect economic impact of a closed route averages several times the direct damage.
7. Three Ways Resilience Investment Goes Wrong
The 3% premium and the 4:1 return are averages. Individual investments can still destroy value, and they do so in three recognisable ways.
7.1 Maladaptation
Maladaptation is an intervention that reduces one risk while increasing another, or that creates a permanent demand for protective spending. Building a hospital, road, or housing district in the wrong location creates exactly that: a permanent obligation to defend an asset that should not have been there. The cost of integrating climate risk into location, design standards, and maintenance from the beginning is far lower than retrofitting once the asset exists — which is why the location decision is the highest-leverage resilience decision available.
7.2 Overdesign
Overdesign is the failure mode of the load-based mental model. It spends the whole adaptation budget at construction, emits the embodied carbon of the additional material, and leaves no capacity to respond if the climate arrives faster than the design case assumed. It also consumes the political and financial capital that would have funded the next intervention.
7.3 Lock-in
Lock-in is the temporal version of maladaptation. An asset designed around a single future climate, with no provision for staged upgrade, forecloses the options that would have been cheap to preserve. A bridge with no space for additional scour protection, a drainage system with no capacity for a downstream storage retrofit, a substation with no room to raise the equipment — each of these is an option that was free to preserve at design and expensive to recreate later.
The Design Question That Prevents All Three
For every major element, ask: if the climate variable crosses this threshold ten years earlier than expected, what is the cheapest action available, and have we preserved the ability to take it? That single question screens for maladaptation (is the action possible at all?), overdesign (are we buying capacity we may not need before the next decision point?), and lock-in (have we foreclosed the option?).
8. Nature-Based Versus Grey: What the Benefit–Cost Data Says
Nature-based solutions are often presented as a sustainability preference. The economics are stronger than that: in several assessments, green and hybrid measures outperform grey defences on benefit–cost ratio, not just on co-benefits.
| Measure | Benefit–cost ratio | Context |
|---|---|---|
| Mangrove restoration with foreshore dikes | 3–6 globally under future conditions; NPV $44–125 billion | Would reduce expected annual damage by about $800 million and affected population by 140,000 annually; viable in 85–105 of 208 subnational regions assessed |
| Nature-based coastal measures | 3.7–4.9 | Highest BCR in a coastal adaptation valuation study |
| Levees and home elevation | Below 1 | Same study — the grey options did not recover their cost |
| Green versus grey infrastructure | Green more than double grey | Stated-preference valuation and cost–benefit analysis of coastal adaptation |
| Integrated grey and green (Shanghai) | Positive BCR | Combining storage tanks with green measures was both effective and economically viable |
Benefit–cost ratios depend heavily on the discount rate, the value assigned to non-market benefits, and the counterfactual. The pattern — nature-based and hybrid measures frequently outperform single-purpose grey defences — is robust across studies, but the specific ratio is not transferable.
Where Nature-Based Solutions Do Not Win
Mangroves do not offer the same level of protection as a seawall, and the cost of a seawall in a low-income rural location “may far exceed the magnitude of avoided damages” — which is exactly the situation in which mangroves are more cost-effective. The engineering judgement is not “green is better”; it is which measure delivers the required protection at the lowest lifecycle cost for this shoreline, this community, and this hazard profile. In dense urban frontages with high asset value behind the defence, grey may still be correct. In sparsely developed low-lying coasts, it usually is not.
Mangrove restoration also has a distributional property that matters for public investment. Because lower-income communities are disproportionately exposed to coastal flooding and disproportionately located in areas suitable for mangrove restoration, the benefits accrue mainly to poorer populations. That is unusual among infrastructure investments, and it strengthens the case where equity is part of the appraisal framework.
9. The Standards Are Catching Up, Unevenly
A threshold-based approach needs standards that support it. The current landscape is transitional: the principles are established, the sector-specific design codes are lagging, and the newest instruments are still arriving.
| Instrument | Status | What it provides |
|---|---|---|
| ISO 14090:2019 | Published | Principles, requirements and guidelines for adaptation to climate change; integrating adaptation into business as usual; identifying impacts and uncertainties |
| ISO 4931-1:2024 | Published | Principles, framework and guidance for resilience design adaptive to climate change in buildings and civil engineering works |
| PAS 3090 (BSI, sponsored by ICE) | Expected late 2026 | Adaptation pathways for infrastructure systems: trigger points, scenario use, design-life constraints in codes, governance, metrics |
| PAS 2080 | Published | Carbon management in infrastructure; intended to integrate with PAS 3090 |
| CEN-CENELEC Guide 32, Edition 2 (2026) | Published | How to address climate change adaptation within European standards |
| CEN/TR 18365:2026 | Published October 2026 | Guidelines on using climate projection data in infrastructure standards |
| CSA S6:25 (Canadian Highway Bridge Design Code) | Published 2025 | First Canadian bridge code to explicitly integrate predictive climate modelling; provisions for hydrology, hydraulics, scour, wind, ice, and a “uniform risk” approach |
| ASCE 7-22 | Published | Hazard-specific design loads including flood, tsunami, snow, rain, atmospheric ice, and wind; a load-based rather than threshold-based framework |
| WFEO Model Code of Practice and Interpretive Guide | March 2026 | 15 principles of climate change adaptation and resilience at a practical level, for engineering organisations worldwide |
The standards landscape is moving faster than most design offices. A project specified today may need to comply with instruments published after its design freeze.
9.1 The Non-Stationarity Problem in Codes
Design codes are built on the assumption that the climate record is statistically stationary — that the past is a reliable guide to the distribution of future extremes. That assumption has failed. Canada has warmed by about 1.7 °C since 1950, roughly double the global average, and by more than 3.5 °C in northern regions. The research underpinning CSA S6:25 confirmed that the stationary assumption is no longer valid and recommended a “uniform risk” approach, so that bridges in all provinces are designed for comparable levels of safety and reliability despite regional differences in climate change.
That is the mechanism by which a code adapts: not by adding a blanket climate factor, but by replacing a stationary statistical basis with a non-stationary one and calibrating reliability across regions.
The Hybrid Standard
A 2026 UNESCAP comparative analysis of port design standards across 17 countries concluded by recommending a hybrid approach combining international base standards with localised adaptation. That is likely to be the pattern across sectors: a common framework for method and reliability, with locally derived climate data and thresholds. A single global design value is neither scientifically defensible nor politically achievable.
10. The Implementation Gap: 12–15%
If the economics are so favourable, why is the investment not happening? The Global Infrastructure Resilience Report 2025 survey — the most extensive global assessment of infrastructure resilience, drawing on more than 3,000 experts from over 100 countries — gives a clear answer.
| Finding | Implication |
|---|---|
| Only 12–15% of respondents report fully following resilience measures | Awareness has risen; implementation has not followed at the same rate |
| Regulatory compliance remains insufficient | Frameworks are improving, but enforcement is weak in many regions |
| Risk assessments are inconsistent and data systems fragmented | The threshold register cannot be built without consistent asset and hazard data |
| Capacity deficits, particularly in public institutions, constrain planning, maintenance, and regulation | The bottleneck is often institutional rather than technical |
| Financing favours post-disaster reconstruction over prevention | The incentive structure pays for recovery, not for avoided loss |
Survey findings from the CDRI Global Infrastructure Resilience Report 2025 working paper on infrastructure professionals. Regional capability varies: East Asia and the Pacific show stronger institutional capacity.
The Financing Bias Is the Root Cause
Post-disaster reconstruction is visible, politically rewarded, and easy to finance. Prevention is invisible, deferred, and competes with other capital priorities. That asymmetry explains why a measure with a 4:1 return remains unfunded while reconstruction after a failure is funded immediately. The fix is not better engineering economics — the economics are already compelling. The fix is financing structures that pay for prevention: contingency funds, resilience bonds, layered financing combining bonds, insurance, and contingency reserves, and longer financing tenors, which the IFC finds improve bankability more effectively than interest rate reductions alone.
There is a geographic dimension too. The $350 billion annual resilience gap is not evenly distributed: about $200 billion of it falls on low-income places, roughly 2.7 times the figure for middle- and high-income places. Closing rural gaps requires 1.0 to 1.5 percentage points more of GDP than closing urban gaps in middle- and low-income countries. The largest single gap is heat protection for low-income places, at about $125 billion, followed by rainfall flooding in low- and middle-income regions at about $40 billion and wildfire protection at about $50 billion.
11. A Threshold Audit Checklist
- Replace the single design case with a threshold register. For each critical asset, record the climate variable, the value at which performance degrades, the value at which service is lost, and the owner of each threshold.
- Separate performance, service, financial, and ecological thresholds. They are crossed at different times and trigger different decisions. A financial threshold crossed before a physical one means the asset is economically obsolete while still standing.
- Name the trigger and the action before the project is approved. A threshold with an owner and a pre-agreed action is a decision. A threshold without them is an observation.
- Preserve the option rather than buying the maximum. Space for additional scour protection, capacity for a downstream retrofit, room to raise equipment — these are cheap at design and expensive to recreate. Option preservation is the cheapest form of resilience.
- Use the 3% premium as a budget test, not a target. If a resilience package costs far more than a few percent of construction cost, check whether it is buying capacity that a staged pathway would deliver later at lower cost.
- Screen for maladaptation at the location decision. The highest-leverage resilience decision is where the asset goes. A site that requires permanent protective spending is a liability created at planning, not at design.
- Appraise nature-based and hybrid options on lifecycle cost, not preference. Mangroves and hybrid defences frequently outperform grey on benefit–cost ratio, especially where asset values behind the defence are low. Grey remains correct on dense, high-value frontages.
- Build the threshold register into the asset management plan, the regulatory instrument, or the concession agreement. A pathway that depends on a future committee's goodwill will not survive a change of government.
- Check the standard set at design freeze and at each gate. The standards landscape is moving faster than project cycles. A design that complied in 2024 may not comply with an instrument published in 2026.
- Count indirect losses. Indirect economic impacts average 7.4 times direct infrastructure damage. A business case that counts only the repair bill understates the benefit of resilience by roughly an order of magnitude.
12. Frequently Asked Questions
What does climate-resilient infrastructure actually mean?
It means infrastructure that continues to deliver its service as the climate shifts, or that fails in a planned and recoverable way when a threshold is crossed. The practical definition is not “designed for a more severe climate” but “designed with a known set of thresholds, triggers, and staged interventions” so that service is maintained or restored without rebuilding the asset.
Why is resilience described as a threshold problem rather than a load problem?
Because infrastructure does not degrade in proportion to the climate variable. It performs acceptably until a variable crosses a service threshold, after which failures become frequent and losses become nonlinear. A load-based approach buys a margin against a single design case; a threshold-based approach identifies where the cliff is, what triggers the next intervention, and how to keep the option open. The second is cheaper and more robust under uncertainty.
How much does climate resilience cost, and what does it return?
World Bank analysis found that making new infrastructure more resilient in low- and middle-income countries adds about 3% to construction cost and generates roughly $4 in benefits per $1 invested, with a decade of delay adding about $1 trillion in costs. The Global Commission on Adaptation found benefit–cost ratios commonly between 2:1 and 10:1, and the IFC found up to $8.60 of protected asset value per $1 in a three-sector Brazilian study.
What are adaptation pathways?
Adaptation pathways are a planning method that identifies the trigger points at which climate conditions breach thresholds where infrastructure services become ineffective, and sequences a range of measures over time in response. They avoid both over-investing and under-investing in adaptation at any one moment, and they make the next decision explicit rather than implicit. PAS 3090, expected in late 2026, formalises the approach for infrastructure systems.
Are nature-based solutions cheaper than engineered defences?
Frequently, on a benefit–cost basis, but not universally. Mangrove restoration shows a global benefit–cost ratio of 3–6 under future conditions, and one coastal valuation study found nature-based measures at 3.7–4.9 while levees and home elevation returned below 1. But mangroves do not provide the same level of protection as a seawall, and grey defences remain appropriate on dense, high-value frontages. The correct test is lifecycle cost for the required protection at that site.
What is maladaptation?
An intervention that reduces one risk while increasing another, or that creates a permanent demand for protective spending. Building in the wrong location is the classic example: it locks in an obligation to defend an asset that should not have been there. Maladaptation is cheapest to avoid at the location decision, before any design work begins.
Which standards govern climate-resilient infrastructure?
The core international framework is ISO 14090:2019 for adaptation principles and ISO 4931-1:2024 for resilience design in buildings and civil engineering works. PAS 3090, expected late 2026, will cover adaptation pathways for infrastructure systems and is intended to integrate with PAS 2080 on carbon. Sector codes are moving at different speeds: CSA S6:25 is the first Canadian bridge code to integrate predictive climate modelling explicitly, while ASCE 7-22 remains a load-based framework.
Why is the implementation gap so large if the returns are so good?
Because financing favours post-disaster reconstruction over prevention. Reconstruction is visible, politically rewarded, and easy to finance; prevention is invisible, deferred, and competes with other capital priorities. The Global Infrastructure Resilience Report 2025 survey of more than 3,000 experts in over 100 countries found that only 12–15% report fully following resilience measures, with inconsistent risk assessments, fragmented data systems, and public-sector capacity deficits compounding the problem.
How should an engineer start on an existing asset?
Build a threshold register for the assets that matter most, starting with the variables that drive service loss rather than physical damage. For most networks, the first thresholds crossed are operational — road closure depth, wind speed for suspension, temperature for speed restrictions — and they are crossed years before any structural damage occurs. Those thresholds are cheap to identify and they usually reveal that the resilience case is stronger than the repair-cost case suggests.
Related reading on BuildersMeasures: Dam Engineering Basics: A Dam Is a Controlled Failure, Water Conservation Engineering: The Three-Bucket Audit, and Mega-Project Risk Management: The Risk Is in the Estimate, Not the Execution.