A reservoir is still the only way to hold a year's worth of water for a whole region at a cost of pennies per cubic metre, and the only water-supply asset that also absorbs a flood and runs a power grid. It is also the option that hands the most water back to the sky, loses capacity every year to silt, and is emptiest precisely when the drought it was built for arrives. That is the whole trade-off in two sentences — and in 2026 it is being settled differently in different places, because the alternatives that did not exist at scale in 1970 (reverse osmosis, potable reuse, managed aquifer recharge, serious leak repair) now compete on cost in the rich, water-short world, and still do not in most of the water-poor one. This is the honest ledger: what reservoirs beat, what beats them, and the five questions that decide which one a given region should build next.
Key takeaways
- Reservoirs win on scale and on carryover. Nothing else banks water across years at a whole-basin scale. About 30–40% of the world's irrigated land depends on dams (World Commission on Dams, 2000), and hydropower supplies roughly 14% of global electricity from about 1,400 GW of installed capacity — with pumped storage still making up over 90% of the world's grid-scale energy storage.
- Reservoirs lose on depreciation and on evaporation. Large dams across 150 countries have already lost an estimated 13–19% of their original storage to sediment and are headed for about 26% by 2050 (Perera et al., UNU-INWEH, 2023). In the Colorado basin, roughly 1.9 million acre-feet — some 13% of the entire river — evaporates every year, most of it off Lake Mead and Lake Powell.
- Alternatives are rainfall-independent; reservoirs are not. Seawater desalination runs at roughly 3–4 kWh/m³ and advanced water recycling at 1–1.5 kWh/m³, against about 0.1–0.5 kWh/m³ for a gravity-fed reservoir. You are buying drought-proofing with electricity.
- The cheapest new water is usually the water you stop losing. Global non-revenue water is estimated at about 126 billion m³ a year, worth roughly $39 billion (Liemberger & Wyatt, 2019), and Cape Town cut municipal demand by roughly half during its Day Zero scare without building anything.
- The verdict is regional, not universal. In basins that are already fully dammed, a new reservoir is usually the worst remaining option. In much of South Asia and sub-Saharan Africa, where per-capita storage is a fraction of the rich world's, it is still the right one — and whether a country actually needs another is a question only open data can answer, which is what the Reservoir Transparency Index scores.
1. What a reservoir still does better than anything else
Four things, and no competitor does all four.
It stores across years, not days. This is the property that gets forgotten. Desalination and recycling produce a flow — a steady number of cubic metres per day. A reservoir holds a stock, which is what carries a region through a bad year and into a worse one. Lake Mead and Lake Powell together were built to bank several years of the Colorado River's flow for some 40 million people; no plant on any coast can substitute for that function, only for part of the flow it delivers.
The marginal water is nearly free. Once the concrete is poured, water in a reservoir arrives by gravity. Treatment and distribution typically cost on the order of 0.1–0.5 kWh per cubic metre — an order of magnitude less than seawater desalination. In an energy-constrained or high-tariff country, that gap is the whole argument.
It absorbs floods. A treatment plant has no flood-control function whatsoever. A reservoir with empty space above the conservation pool converts a destructive peak into a managed release. As rainfall intensifies — warmer air holds about 7% more moisture per degree of warming — this half of the job is getting more valuable, not less, even as the water-supply half gets harder.
It runs on and stabilises the grid. Hydropower is around 14% of world electricity generation, and pumped-storage hydro remains more than 90% of installed grid-scale storage capacity worldwide — the flywheel that lets solar and wind expand. Countries like Norway, Colombia and New Zealand do not report their reservoirs in cubic metres at all; they report them in gigawatt-hours, because the reservoir is the battery.
2. The bill the twentieth century didn't price
None of the following was in the cost-benefit sheet when most of the world's 58,000-plus large dams were approved.
- Sediment. A reservoir is a depreciating asset. Every river carries silt; a dam stops it. Large dams have already lost an estimated 13–19% of original storage and are on track for ~26% by 2050, taking global capacity from roughly 6,316 to 4,665 billion cubic metres (Perera et al., 2023). Taiwan is the vivid case: Zengwen, the island's largest reservoir, has lost close to 40% of its design capacity to silt. You cannot expand a reservoir back; you can always add another treatment train to a plant.
- Evaporation. The desert reservoir gives water to the sky. Around 1.9 million acre-feet a year — about 13% of the Colorado River — evaporates, with Lake Mead alone giving up an estimated 600,000–860,000 acre-feet. Storing water in a hot open basin means paying a permanent tax on the balance.
- Greenhouse gas. Decomposing vegetation under reservoir water releases methane. The world's reservoirs emit roughly 1 gigaton of CO₂-equivalent a year, about 1.3% of human-caused emissions (Deemer et al., BioScience, 2016). A hydropower reservoir is low-carbon; it is not zero-carbon, and tropical reservoirs are the worst offenders.
- People and rivers. The World Commission on Dams estimated 40–80 million people were physically displaced by dams during the twentieth century. Add fragmented fish migration, sediment-starved deltas and drowned valleys — the reasons the United States removed 108 dams in 2024, tied for the most on record.
- Dams overrun their budgets, systematically. An Oxford study of 245 large dams (Ansar et al., Energy Policy, 2014) found an average real cost overrun of 96%, with schedules running about 44% long. Eight in ten came in over budget. This is not incidental — it is the base rate you should assume for a new project.
- It fails in exactly the scenario it was sold for. A reservoir is emptiest when the drought is worst. In late 2024 Lake Kariba, the world's largest man-made reservoir by volume, fell to around 1% of usable storage, forcing long daily blackouts across Zambia and Zimbabwe. Rainfall-dependent supply is rainfall-dependent in the bad year too.
3. The alternatives, and what each is actually good at
Seawater desalination — the drought-proof option you pay for in electricity. Some 16,000 operational plants produce roughly 95 million m³ a day globally (Jones et al., 2019). Modern reverse osmosis runs at about 3–4 kWh/m³ against a thermodynamic floor near 1, and prices have collapsed at the top end: Israel's Sorek B was contracted at roughly $0.40 per cubic metre, among the lowest ever signed. But that is a best case built on cheap power, a warm sea and enormous scale. California's Carlsbad plant cost about $1 billion for 50 million gallons a day, and its water is several times the price of the region's imported surface water. The other cost is brine: Jones et al. put global brine output at 141.5 million m³ a day — an estimate other researchers consider high, but nobody argues the discharge is negligible. Best when: you are coastal, rich, and need a supply that does not care whether it rains.
Water recycling — the supply that scales with population, not weather. Advanced purification runs around 1–1.5 kWh/m³, well under half of seawater desalination, because you are cleaning water that is already fresh. Israel reuses close to 90% of its municipal wastewater, Singapore's NEWater meets up to 40% of national demand, Orange County purifies 130 million gallons a day, and Windhoek has drunk recycled water since 1968. California's rules for direct potable reuse took effect on 1 October 2024. We argue the fuller case in water reuse and fewer new reservoirs. Best when: you have a large city already producing wastewater — which is to say, whenever you have a large city.
Groundwater — the world's biggest reservoir, and the easiest to overdraw. Aquifers supply about half the world's drinking water and around a quarter of irrigation withdrawals (UNESCO, 2022), with no evaporation, no valley flooded and no dam to build. The catch is that they are invisible, which makes them easy to spend. GRACE satellite data found 21 of the world's 37 largest aquifers past their sustainability tipping points (Richey et al., 2015), and parts of California's San Joaquin Valley have sunk roughly 8.5 metres since the 1920s from pumping — permanently destroying the storage space itself. Best when: it is genuinely being recharged and someone is measuring it. Otherwise it is a loan, not a supply.
Managed aquifer recharge — the reservoir with a roof on it. Deliberately banking surplus wet-year water underground gets you most of a reservoir's carryover function with none of the evaporation and no drowned valley. Arizona's Water Banking Authority has stored on the order of 4 million acre-feet of Colorado River water underground for exactly this purpose. Best when: you have a suitable aquifer and occasional surplus water to put in it — which is a real constraint, not a detail.
Long-distance transfer — moving the problem, at a price. China's South–North Water Diversion has cost well over $60 billion and, by official figures, has delivered more than 70 billion cubic metres since 2014, supplying most of Beijing's tap water. It works. It also imports the donor basin's politics and its drought years, and the pumping bill is permanent. Best when: the donor basin has a genuine, durable surplus and the political settlement holds for decades.
Leak repair and demand management — the cheapest water there is. Roughly 126 billion m³ a year is lost as non-revenue water globally, worth about $39 billion (Liemberger & Wyatt, 2019); England and Wales alone leak around 3 billion litres a day. And demand is more elastic than utilities assume: Cape Town cut consumption by roughly half — from about 1.2 billion litres a day to near 500 million — and that, more than any new infrastructure, is why it never reached Day Zero. Best when: always, first, before anything else on this list. Its limit is that it is finite — you can only save what you are currently wasting.
4. Side by side
| Supply option | Typical energy (kWh/m³) | Rain-dependent? | Typical lead time | Binding constraint |
|---|---|---|---|---|
| New surface reservoir | ~0.1–0.5 | Entirely | 10–20 years | Good sites are gone; silt, evaporation, politics |
| Groundwater pumping | ~0.4–0.6 (deeper costs more) | Slowly | Months–2 years | Depletion, subsidence, contamination |
| Managed aquifer recharge | ~0.3–0.7 | Yes — needs a surplus to bank | 2–5 years | Needs both a suitable aquifer and wet-year water |
| Water recycling (advanced) | ~1–1.5 | No | 5–8 years | Capital, brine, public acceptance |
| Seawater desalination (RO) | ~3–4 | No | 3–6 years | Energy, brine disposal, coastal siting |
| Long-distance transfer | ~1–3 | Yes, at the source | 10–20+ years | Cost and the donor basin's politics |
| Leak repair / demand management | ~0 (it saves energy) | No | 1–3 years | Finite — capped by current waste |
Energy figures are indicative ranges from published reviews and utility reporting; real values are site-specific, and lead times vary enormously with permitting regime. Use the table to rank options, not to price a project.
5. What actually changed by 2026
Five things have moved since the last generation of these decisions was made.
- Reverse osmosis got cheap and solar got cheaper. The energy penalty that ruled desalination out of most budgets is now roughly a quarter of what it was in the 1990s, and it can increasingly be paid with daytime solar.
- Potable reuse became legal and normal. California's direct-potable-reuse regulations took effect in October 2024; the question moved from is it safe to how fast can we build it.
- The sediment arithmetic got public. A 26%-by-2050 storage loss is no longer a footnote — it means a share of the world's existing reservoir capacity has to be replaced just to stand still.
- The climate signal cuts both ways. More intense rain raises the value of flood storage; longer dry spells and hotter surfaces lower the reliable yield of the same reservoir. A basin under a strong El Niño can experience both within a single year.
- The rich world stopped building and the rest did not. The United States removed 108 dams in 2024 while India and China kept building storage — rationally, because per-capita storage in South Asia is a fraction of North America's and the monsoon delivers most of the year's water in a few weeks.
6. Five questions that decide it
Before defending or attacking "reservoirs" in the abstract, answer these for the specific basin:
- Is the basin already fully developed? If the good sites are taken, the marginal dam is small, expensive and ecologically costly. If storage per capita is low and rainfall is concentrated in a short season, it is not.
- Do you need a stock or a flow? Multi-year carryover means storage or aquifer recharge. Filling a persistent daily deficit means reuse or desalination.
- What does a kilowatt-hour cost you? At high power prices, a 3–4 kWh/m³ option is not a plan. At low or surplus-renewable prices, it is the safest supply you can buy.
- How much are you currently losing? If non-revenue water is 25%, no new-supply project should be approved before the leak programme.
- Can you see the numbers? If nobody outside the ministry knows how full existing storage runs, "we need another dam" is an assertion, not a finding. That is the argument behind why reservoir data must be public.
The bottom line
Reservoirs are not obsolete and they are not the default any more. They remain unmatched at holding a large stock of water cheaply, absorbing floods and balancing a grid, and they remain the correct answer wherever storage per capita is low and rain falls in a narrow season. They are also a depreciating, evaporating, rainfall-dependent asset whose ecological and displacement costs are now measured rather than assumed, and whose budgets overrun by an average of 96%. In a mature, fully dammed basin, the ranking in 2026 usually runs: fix the leaks, recycle the wastewater, bank surplus underground, desalinate if you are coastal and can afford the power — and build the reservoir only if you genuinely need a stock that none of those can hold.
Which case applies is an empirical question, and you can check it on the data: how full storage actually runs, year after year, across the countries we track — from a comfortable Spain to a structurally short United States Southwest.
FAQ
Are reservoirs obsolete in 2026? No. They still supply the water for 30–40% of the world's irrigated land, generate about 14% of global electricity, and provide flood control and multi-year carryover storage that no treatment plant can replicate. What has ended is their status as the automatic answer: in basins that are already fully dammed, recycling, leak reduction, aquifer recharge and desalination now usually beat a new dam on cost, speed and ecological impact.
Is desalination cheaper than building a new reservoir? Per cubic metre of water delivered, almost never — a gravity-fed reservoir runs at roughly 0.1–0.5 kWh/m³ against 3–4 kWh/m³ for seawater reverse osmosis. Desalination wins on different terms: it is drought-proof, it can be built in three to six years instead of ten to twenty, and it needs no new valley. Where power is cheap and the sea is close, that trade is worth it; where power is expensive, it is not.
What is the single biggest drawback of a reservoir? That it depends on the same rainfall it is meant to insure against, so it is emptiest exactly when it is needed most — Lake Kariba at about 1% of usable storage in late 2024 is the extreme case. The slower-burning one is sedimentation: large dams have already lost an estimated 13–19% of their original capacity to silt, heading for roughly 26% by 2050, and that loss is effectively irreversible.
Which option is most drought-resistant? Water recycling and seawater desalination, because both scale with population rather than with rainfall — a city produces wastewater and the sea stays full in a dry year. Managed aquifer recharge is the drought-resistant way to keep the carryover function of a reservoir, since banked groundwater does not evaporate.
Why do countries keep building large dams if the alternatives are better? Because "better" is regional. In South Asia and much of Africa, storage per capita is a fraction of the rich world's and most of the year's rain arrives in a few weeks, so a reservoir is the only technology that can bridge from wet season to dry at that scale — and it delivers electricity and flood control at the same time. The reuse-over-reservoirs shift is a conclusion about mature, fully developed basins, not a universal rule.
How can I check whether my region actually needs more storage? Look at the long record, not the headline: how full existing reservoirs run in a normal year, how deep they draw down in a bad one, and whether the trend is falling. We publish that for every country we track — see the United States, India or Spain — and we score countries on whether they publish enough for anyone to check in the Reservoir Transparency Index.
Sources: irrigated land and displacement — World Commission on Dams (2000). Sedimentation — Perera et al., "Present and Future Losses of Storage in Large Reservoirs Due to Sedimentation," Sustainability (UNU-INWEH, 2023). Reservoir greenhouse gases — Deemer et al., BioScience (2016). Cost and schedule overruns — Ansar, Flyvbjerg, Budzier & Lunn, Energy Policy (2014). Colorado River evaporation — US Bureau of Reclamation-linked research. Dam removals — American Rivers (2024). Desalination capacity and brine — Jones et al., Science of the Total Environment (2019). Groundwater — UNESCO World Water Development Report (2022); Richey et al., Water Resources Research (2015). Non-revenue water — Liemberger & Wyatt, Hydrogeology Journal (2019). Hydropower capacity and pumped storage share — International Hydropower Association and IEA. Reuse figures — Israel Water Authority; Singapore PUB; Orange County Water District; City of Windhoek; California State Water Resources Control Board. Energy-intensity ranges — published reviews of water-supply energy use and utility reporting. Live reservoir levels are our own, from the sources documented in how we get our data.