By Nicole Choi, AGU
In December 2012, heavy downpours from an atmospheric river—a long atmospheric pathway that transports large amounts of water vapor—deluged parts of Northern California, including around Lake Mendocino near Ukiah. Water levels behind the lake’s Coyote Valley Dam rose enough to trigger a required release downstream to make space in the reservoir for future rainfall.
But storms later in the 2012–2013 wet season never arrived to replenish the lake. Indeed, the state was entering a multiyear drought that resulted in low lake levels and cast the released water as a valuable resource lost.
At the time, the policy of the U.S. Army Corps of Engineers (USACE), which manages flood operations for Coyote Valley Dam and more than 700 others nationwide, was to drain all water from Lake Mendocino’s flood pool as soon as downstream conditions allowed for a safe release, regardless of forecasted conditions, wrote Patrick Sing of USACE’s San Francisco District in an email to Eos. Sing began managing the water in Lake Mendocino’s flood control space in 2015 alongside Sonoma Water, which manages releases from the reservoir’s water supply space.
In response to the Lake Mendocino incident, USACE wanted to learn how reservoir water managers might use forecasting proactively to better decide when and at what rate to release water. Together with several partner agencies and institutions, USACE launched the Forecast-Informed Reservoir Operations (FIRO) program.
A successful pilot study of FIRO at Lake Mendocino from 2014 to 2020 demonstrated that FIRO increased water storage by nearly 20% compared with standard, non-FIRO operations. Since then, additional testing has been underway at more dams in California, Oregon, and Washington. And now new research is further advancing efforts to improve the efficiency of U.S. reservoir operations.
Screening for Skill
To continue scaling FIRO for use across the country, USACE developed a screening process to determine at which dams FIRO might be applied most effectively. A key part of this process is evaluating the accuracy and reliability of forecasting, known as forecast skill, for the area around a given dam.
Water managers can use both weather and hydrologic forecasts to help plan reservoir operations. The latter predict how precipitation will interact with a landscape and the proportion of falling precipitation that will flow into a body of water (called the runoff ratio). Many variables shape precipitation-landscape interactions, such as topography, soil moisture, and snowpack levels, so hydrologic forecasts are highly localized.
Different regions where USACE manages reservoirs can thus seem like islands, “with each region using different strategies,” wrote Eric Shearer, a research hydraulic engineer at USACE’s Engineer Research and Development Center in Vicksburg, Miss., in an email to Eos. These forecasts are so localized that USACE can’t readily evaluate them with a screening process designed for general use, he explained.
Not evaluating hydrologic forecasting skill during screening complicates planning for FIRO’s nationwide rollout. A dam might appear to be a good candidate for applying FIRO on the basis of the skill of weather forecasts in the area, but a watershed’s unique hydrologic response could influence this potential.
A Better Sense of Watershed Responses
That’s why Shearer and Christine Albano, an ecohydrologist at the Desert Research Institute in Reno, Nev., analyzed the hydrologic responses of 42 reservoirs situated in dry, wet, and snowy watersheds across California, Colorado, Nevada, and Utah. They identified atmospheric rivers of the same strength that affected these watersheds over the past 40 years and quantified runoff ratios at different reservoirs for each storm using daily historical hydrologic and stream gauge data and modeled precipitation, soil moisture, and snowfall data. They calculated the variability in those ratios for each storm event at each reservoir and then compared those variabilities among the different reservoirs in the watershed.
A wide range of runoff ratios means a lot of uncertainty in a watershed’s hydrologic response exists, whereas less variability means more certainty. The researchers observed, for example, tighter spreads of runoff ratios for small Southern California basins at lower elevations, such as near Prado Dam, than for higher basins in the Sierra Nevada, Albano wrote in an email to Eos.
But tight spreads were also observed in the highest elevations, as at Pine Flat Dam, “where precipitation typically comes in the form of snow rather than rain,” Albano wrote. It’s the variability in precipitation and soil moisture conditions that could indicate variability in hydrologic response, she explained. “Tighter spreads in runoff ratios occur when both rain and soil moisture conditions are less variable and this could indicate higher hydrologic response predictability.” With climate change potentially bringing more rain than snow, predictability could change, which is a consideration for FIRO implementation.
The researchers will present their work on 13 December at AGU’s Annual Meeting 2023 in San Francisco.
Knowing which regions have higher uncertainty can inform decisions about where to install observational networks that will help improve watershed models, whereas FIRO can be initiated in regions of greater certainty. Understanding where flow magnitudes are high also helps USACE to identify dams that might need more flood management. According to Shearer, USACE plans to screen 419 of its dams across the United States within the next 5 years, and the more knowledgeable it is about the watersheds surrounding these dams, the better it can streamline the introduction of FIRO.
Shearer and Albano also studied lag times, or how long it took for precipitation from an atmospheric river to make its way into a reservoir. That information is important because it affects water managers’ operational planning. If a storm is forecasted 2 days out but the expected lag time is 2 days, then a water manager has 2 extra days to make a decision about a reservoir release, for example. Greater lag times increase the suitability of a dam for FIRO.
Imperfect, but Invaluable, Forecasts
Looking at historic events to see how different basins responded provides “great information for water managers and for helping inform how they might use FIRO,” wrote Savannah Lacy, chief of the Water Management Operations Unit in USACE’s Jacksonville District in Florida, in an email to Eos. “Understanding the hydrologic response of an event is key to helping manage our reservoirs in the most beneficial and efficient way possible.”
No matter how skillful, a forecast is not a crystal ball, however—it won’t always be accurate or provide enough lead time to guide safe water releases. Indeed, forecasts help manage many water releases, but not all, Lacy wrote, even in areas like South Florida, where forecasting has been used in water management for years. But forgoing forecasts is not a smart approach, she added. “It just doesn’t make sense to ignore the information that you have available.”
