Reservoir operating dilemma

Why can a reservoir not be ready for a drought and a flood at the same time?

A reservoir is sold as insurance against both of a river's moods. It holds water back when there is too much, and hands it out when there is too little. Two problems, one structure.

But ask what the operator actually controls, and the promise starts to look strained. There is only one lever: how much water is in the pool right now. Drought readiness means that number is high. Flood readiness means that number is low. What is the operator supposed to do with a single dial that two people want turned in opposite directions?

Notice first that the conflict is not about money, engineering, or politics. It is arithmetic. A reservoir stores exactly one thing, and the thing that protects you from drought — water present — occupies precisely the space that protects you from flood, which is water absent. The same cubic metre cannot be both a reserve and a void. Any scheme that appears to escape this is either enlarging the reservoir or moving the conflict somewhere else.

So how do real dams manage at all? Mostly by separating the two risks in time. In many climates the floods and the droughts do not arrive in the same season, so the operator can follow a rule curve: a target storage level that varies through the year, drawn down before the storm season to open up flood space, then allowed to refill afterwards to carry supply through the dry months. The dial is turned back and forth on a schedule.

Which tells us something about when the dilemma actually bites. It bites hardest where the two seasons overlap or where the calendar is unreliable — and it bites hardest of all in the year when the schedule is wrong. Consider an operator sitting on a full pool in early spring, with a rule curve that says empty and a forecast that says this may be the last rain for a year. Emptying it wastes the drought reserve. Keeping it forfeits the flood space. Whatever is chosen, one of the two failures becomes possible, and the choice must be made before anyone knows which.

Now push on the two losses, because they are not symmetric. A drought loss arrives slowly, with rationing available as a partial remedy and months of warning. A flood loss arrives in hours, is often catastrophic, and is legally attributable to the person holding the lever. Which mistake would you rather be found to have made? Operators lean toward keeping the space, and in many jurisdictions they are not free to lean the other way — in the United States, flood operations at Corps of Engineers dams follow a written water control manual the operator may not simply deviate from because a forecast looks encouraging.

But that last clause is the interesting one, because it points at what the tradeoff is really made of. Ask why the space has to be kept empty. Because you do not know whether the storm is coming. If you knew for certain that the next ten days were dry, the flood space would be doing nothing at all and could safely be full. The tradeoff is not a law of hydrology. It is a consequence of uncertainty, and uncertainty can be bought down.

That is the premise of forecast-informed reservoir operations, tested over several years at Lake Mendocino in California, where skilful forecasts of atmospheric rivers were used to decide when the flood space could be allowed to hold water instead. After a multi-year viability assessment the approach was carried into the dam's updated operating rules. It does not repeal the conflict — it changes the exchange rate between the two risks.

And there is a second currency besides forecast skill: how fast water can be let out. A dam with large low-level outlets, and a channel downstream that can carry the flow, can hold water late and still evacuate the space when a storm appears. A dam that can only spill over the top, into a valley that floods at modest flows, must make the space in advance. So the real limit is not just what you know but how quickly you can act on it — and forecast lead time, typically about a week for these storms, has to exceed the time it takes to drain.

Some refinements that keep the picture honest.

First, this is a genuine tradeoff, not a design flaw waiting to be engineered away. The reason it feels like a failure is that the reservoir is usually sold as two separate services when it is physically one.

Second, most large dams are asked to do more than two things: hydropower wants a high head and steady flow, navigation wants a minimum depth, recreation wants a full pool through the summer, and downstream ecology may want a deliberate high-flow release at a particular time of year. Each additional purpose narrows the operating window further. Two-way tension is the simplest version of the problem, not the usual one.

Third, "make the reservoir bigger" is a real answer but a weak one — yield rises less than proportionally with volume, sediment consumes capacity, a larger surface evaporates more, and the sites are largely taken. Buying information is usually cheaper than buying concrete.

Fourth, the tradeoff shifts as the climate does. Where more precipitation falls as rain rather than snow, the snowpack that acted as a free upstream reservoir delivers less, so the same dam faces sharper inflow peaks and a longer dry season — the dial pulled harder in both directions at once.

• Rule curves, and how a changing climate invalidates the historical inflow records they were fitted to.
• Sedimentation, and why a reservoir's ability to serve either purpose quietly declines over decades.

Every term the collection defines is gathered in the glossary.