Between Christmas and January this year, a parade of nine atmospheric rivers — vast streams of water vapor flowing east from the tropical Pacific — pummeled California. The trillions of gallons of rain poured on the state caused widespread flooding. While the rain topped up some drought-depleted reservoirs and aquifers and filled out snowpack in the Sierra Nevada, much of the water quickly ran off into the sea, flowing off asphalt and farms or released from reservoirs to prevent further flooding.
For Patrick Sing, a water manager with the U.S. Army Corps of Engineers, the deluge was an opportunity to try something that would be dangerous anywhere else in the country.
Sing sits at the controls of Lake Mendocino, a reservoir on the Russian River near Ukiah, in northern California. Like reservoirs across the state, Lake Mendocino has seen years of extreme swings between wet and dry — it almost emptied after a dry stretch in 2021 and was near empty when the first atmospheric river let loose on December 26. Sing watched as the reservoir filled up, first topping off the zone reserved for water supply, then quickly rising into the flood zone.
In February, a month after the rains stopped falling, the reservoir was still in the flood zone, holding 11,000 acre-feet more than the usual limit, according to Sing. Normally, this water would have been released to prevent a flood in the event another storm came along. If more water came too quickly, there might not be time to safely release the extra water from the dam, risking an uncontrolled release or overtopping. But there was a dry forecast, so Sing retained the extra water. It could be a while before so much of it came again.
For a thirsty West, forecasts could reduce flood risk in wet years and increase water supply in dry years.
Lake Mendocino is the first and, as yet, only reservoir in the country authorized to use weather forecasts in making decisions about when to keep and when to release water. The idea is that when a forecast is dry, an operator can safely keep more water in a reservoir. When the forecast is wet, they can cut into stores normally reserved for water supply to make space for the coming rain. “It’s all about the timing, and forecasts give you time,” says Cary Talbot, a researcher at the Army Engineer Research and Development Center.
For thirsty farms and cities in the West, the approach could reduce flood risk during wet years and increase water supply for dry years. Lake Mendocino increased the water it stored by nearly 20 percent in its first two years. A retrospective study of the reservoir’s operations between 1985 and 2010 found water storage would have been 33 percent higher on average at the end of the flood season if forecasts had been in use.
Researchers working on the approach in the U.S. say they aren’t aware of any similar projects in other countries, but studies suggest that integrating forecasts has the potential to improve reservoir operations anywhere weather predictions are sufficiently reliable. The approach could also help aging dams respond to more variable precipitation seen with climate change. “We’re poising ourselves to be much more adaptive going forward,” says Talbot.
In addition to their work at Lake Mendocino, the Army Corps of Engineers, along with researchers at the Scripps Institution of Oceanography and other partners, are studying the approach at four other reservoirs within their jurisdiction in California and Washington, including Lake Oroville, which is held back by the country’s tallest dam. These projects will support a process starting this year to consider using weather forecasts at more than 700 other dams the Army Corps operates across the United States, a few of which are among the nation’s largest.
Those dams represent a fraction of the more than 90,000 dams in the U.S., and the approach won’t work at all of them. Precipitation forecasts may not be reliable enough in the Midwest or Gulf Coast to make decisions about water use. And many reservoirs aren’t built to control the release of water or may not have the resources to continuously monitor the weather and respond. But hundreds of reservoirs could potentially hold more water and reduce flood risk by watching and responding to the weather, say Talbot. “This is the next generation of water management.”
The change in thinking is largely enabled by improvements in weather forecasting. “Decades ago, when the dams were built, forecasts were lousy,” says Marty Ralph, a climate scientist at Scripps, where he directs the Center for Western Weather and Water Extremes. Precipitation forecasts in particular were notoriously tricky to get right. Without reliable forecasts, dams operated by the Corps were required by law to be managed only using data on streamflow, snowpack, and actual precipitation — water on the ground. The caution was justified, says Talbot. “Dams are dangerous. They have the potential to kill people and cause tremendous damage.”
In 2019 and 2020, Lake Mendocino stored 19 percent more water than it would have without using forecasts.
Since the 1960s, however, weather forecasting has become far more reliable, driven by advances in numerical modeling, satellite data, radar, and computing power. By the late 2000s, “it was clear to me there was real potential for forecasts to be useful in reservoir operations,” says Ralph. He saw the potential on the West Coast in particular, where atmospheric rivers meeting the mountains predictably drive precipitation. “Atmospheric rivers are the types of storms that matter for floods and for annual water supply,” he says.
Ralph proposed to collaborate with the Army Corps of Engineers on what he termed “forecast informed reservoir operations,” or “FIRO.” The project would involve research to improve forecasts of atmospheric rivers, and then use those improved forecasts to change how California reservoirs are operated. He says the Corps was interested in the idea, but it was slow to catch on. Then 2012 saw the start of one of California’s worst droughts on record. The drought, which lasted through 2017, drew attention to the state’s depleted water supply as well as the rigid water control manuals still used to make decisions at many reservoirs.
At Lake Mendocino, a wet 2012 winter saw three atmospheric rivers fill the reservoir above flood levels. As required by the reservoir’s control manual, operators released the additional water in case another storm came through, but no more big rains came that year or the next. “The reservoir never rebounded,” says Ralph. The Corps was “pretty well criticized” by politicians for relying on out-of-date manuals, says Sean Smith, the principal hydrologic engineer at the Corps. “Reservoir operations should be based on modern science and weather forecasts, not antiquated rulebooks,” U.S. Rep. Jared Huffman (D-CA) said at the time.
In 2016, Congress changed the law to give the Army Corps permission to consider weather forecasts at the reservoirs it manages. But before forecasts could safely be used, individual dams would have to be studied to make sure the approach was safe and that weather forecasts were reliable enough in that area.
Lake Mendocino was the first to be studied. Detailed analyses of the weather forecasts and hydrology there found that a reliable forecast five days out would give operators enough time ahead of a heavy storm to release any additional water and for that water to make it past flood-prone areas downstream. In 2019 and 2020, Lake Mendocino’s first two years operating with forecasts, the reservoir stored 19 percent more water than it would have without using forecasts, enough to supply around 30,000 households for a year.
The successful test at Lake Mendocino led to projects at four other reservoirs in California and Washington, which will test the use of forecasts in areas more affected by snowmelt, with more complex systems of reservoirs, or with greater flood risk downstream.
Prado Dam on the Santa Ana River just outside Orange County will serve to test the approach in a more urban context with high flood risks. An initial assessment showed the approach is viable at the dam and would increase water supply by 7,000 acre-feet a year without increasing flood risk. The extra water would go to recharge the groundwater that provides much of the county’s water supply. County water officials say they plan to start using forecasts at the reservoir by October this year.
The use of forecasts could help the U.S. cope with drought without building new water infrastructure.
Tests at Lake Oroville and New Bullards Bar Reservoir on the Yuba and Feather River watersheds in California’s Central Valley will gauge how forecasting can be used where snowpack and runoff is a major variable. They are also much larger than the other tests: Lake Oroville can store more than 3.5 million acre-feet of water, while New Bullard’s Bar can store just under a million acre-feet. For comparison, Lake Mendocino has a capacity of around 100,000 acre-feet. An initial assessment found using forecasts in conjunction with other changes to the dams’ spillways would reduce flood risk without impacting water supply.
The fourth test at Howard A. Hanson Dam on the Green River in Washington will test the approach in a very wet system. Forecasting could reduce flood risk and also help spawning salmon by allowing for greater control of water conditions, says Ralph.
Those tests will inform a screening process starting this year to determine where else forecasts could improve operations. An initial study shows most of the more than 80 dams in the Army Corp’s South Pacific Division — which includes California and most of the U.S. Southwest — could be viable, according to Smith. Elsewhere, the calculus is more complicated. “You have to take each project in each basin one by one,” Smith says.
New England, also affected by atmospheric rivers and with mountainous landscapes, has quite reliable forecasts, says Ralph, but precipitation is more difficult to predict in the middle of the country, where thunderstorms predominate and there aren’t mountains to shape weather patterns. Along coasts, tropical storms and hurricanes are also hard to forecast with enough skill to integrate into reservoir operations.
As the use of forecasts expands, “we have to be very cautious,” says Smith. Rigorous testing is needed to make sure forecasts are reliable and that dams have the resources to watch and respond to changing weather. Decisions have to be made daily and sometimes hourly.
But with sufficiently accurate forecasts and rigorous tests, dam managers should see no increased risk of floods or water shortages, says Michelle Ho, a researcher at the University of Melbourne who has studied U.S. water infrastructure. And the added flexibility could help the U.S. cope with worsening drought and more intense rain without needing to build expensive new infrastructure.
One thing is certain, Talbot says: As the climate change changes, forecasts will increasingly figure into dam management. “This represents a new paradigm.”