In 2011, heavy snows in the Rocky Mountains filled the Colorado River, lifting reservoirs—and spirits—in the drought-stricken U.S. Southwest. The following year, however, water levels dropped to nearly their lowest in a century, imperiling a supply vital to millions of people and dozens of ecosystems. Now, scientists say they may have come up with a potential early warning system for the Colorado’s water levels—by watching temperature patterns in the Atlantic and Pacific oceans, thousands of kilometers away.
If the approach works, it could be a first step toward developing longer term river forecasts, the scientists say. “If we know now that next year we will have a severe drought, water managers can work proactively,” says Yoshi Chikamoto, a climate dynamics expert at Utah State University, Logan, who helped lead the research.
Scientists have long recognized links between ocean temperatures and continental weather patterns. Most famously, the central Pacific’s El Niño—a periodic warming of ocean waters—has been tied to drought in Africa, torrential rains on the Pacific coast of North America, and wildfires in South America. And as climate models have grown more precise and computers more powerful, scientists have searched for evidence of other ways that ocean changes ripple through the atmosphere.
Chikamoto turned his attention to the Colorado River in 2017. Utah and several nearby Rocky Mountain states sit in the upper reaches of the river system, which helps slake the thirst of Las Vegas, Phoenix, and Los Angeles. The river also churns out enough hydroelectric power to light more than 750,000 homes, and its waters turn California’s Imperial Valley from a desert into an agricultural cornucopia. But government forecasters typically predict its bounty only 7 months in advance, based in part on snow levels, temperature trends, and historical records.
Chikamoto and colleagues wondered whether they could make longer term forecasts by examining distant ocean conditions. They took an existing computer model of the interplay between global oceans and the atmosphere and added to it historical ocean data on temperature and salinity. They also added measurements of soil moisture—a common drought indicator—for land drained by the Colorado River and its tributaries.
The scientists found that in the seven most extreme drought years over the past 6 decades, including 2012, the downturn nearly always happened on the heels of a multiyear pattern of global ocean temperatures, they report this month in Communications Earth & Environment. Those patterns started with unusual warm spells in the tropical Atlantic 3 to 4 years before the drought, and continued with warming in the northern Pacific and cooling in the central Pacific 1 to 2 years before the drought.
To test the predictive power of this approach, the scientists used their model to create a series of 10-year forecasts starting in 1960. They compared these forecasts with those of two other models: one that was missing the extra ocean data, and one that used the prior year’s drought conditions to predict the following year. They also compared the forecasts with actual conditions. They found that their ocean-driven forecasts were as much as 40% more accurate than the other models, looking 2 years into the future.
But this new approach isn’t a perfect crystal ball, Chikamoto cautions. For forecasts beyond 2 years, results were no more accurate than a coin toss. Additionally, the ocean temperature patterns could explain only 38% of the fluctuations in Colorado River water, the researchers found. And the forecasts require so much computing time and power that it’s not practical for most water managers and government weather forecasters. But the researchers did develop a simpler statistical forecasting model that worked for 1-year forecasts, Chikamoto says. “This is just a first step.”
Other researchers warn of the danger of confusing correlations in ocean temperatures and drought with causation. Shang-Ping Xie, a climate dynamics expert at the Scripps Institution of Oceanography who studies how oceans influence climate, says the new work doesn’t explain how temperature changes in the tropical Atlantic might be connected to shifts in Pacific Ocean temperatures or precipitation in the Colorado River region. That makes him question how accurate any forecast might be. “I think until the physical causalities are being revealed and pinned down, statistical models are always suspect.”