by Robin Meadows
When it comes to water, winter is a time of promise and peril in California. Our fate is uncertain―and can swing wildly―from year to year. Will mountain snowpacks be plentiful enough to get us through the dry season? Will they melt so fast in the spring that we’re down to a trickle by summer? Will too many atmospheric river storms in a row cause devastating floods like those we suffered last year?
To help us prepare for what is to come, researchers are developing new models that zoom in on the Sierra Nevada snowpack and on individual atmospheric river events. Snowmelt from the Sierra Nevada provides, on average, about one-third of California’s annual water supply, and atmospheric rivers provide about half of the rain and snow statewide.
Until recently, historical records helped state water managers and planners know what to expect for a given year. But this approach is no longer reliable as the world warms.
“The past isn’t as informative about the future as it used to be,” says Michael Anderson, the state climatologist with the California Department of Water Resources.
Conventional global climate models, which cover the entire Earth, can’t tell us what we need to know at a local level either. “They’re too coarse,” says Paul Ullrich, a climate modeler who leads the Climate and Global Change Group at the University of California, Davis. “The whole Sierra Nevada range is just a couple of data points.”
Regional climate models are higher resolution and include details of terrain such as mountains and coastal areas, where the weather can be astonishingly different over distances of just 30 miles. But, because these models are limited to relatively small areas, they don’t account for large-scale weather systems such as the jet stream and extratropical cyclones, which can span continents.
BETTER MODELS OF THE SIERRA NEVADA SNOWPACK
So Ullrich and colleagues developed a model of the Sierra Nevada snowpack that combines strengths of both global and regional climate simulations. This approach, called a regionally refined Earth system model, simulates how large-scale weather systems interact with local terrain. Researchers start with a global model, which has a resolution of about 200 miles, and zoom on a small patch of the planet, in this case California.
The resulting regionally refined model has a resolution of less than 10 miles, giving the Sierra Nevada hundreds of data points. The model accounts for factors that determine how much snow falls and sticks on the mountains as well as how much melts. For example, trees catch snow in their branches, affecting how much reaches the ground; air temperatures affect how long snow sits there; and dust makes snow less reflective, making it absorb more solar radiation and melt faster.
To validate the model, the researchers compared its predictions with data from sources including automated weather stations. These sensor-equipped stations, called the Snow Telemetry Network (SNOTELs), were installed beginning in the 1960s in western mountains. SNOTELs collect information including air temperature, precipitation, and snow depth and water content―which varies considerably depending on how many air pockets the snowpack contains―and report these data several times a day or even hourly.
“This is really an exciting time,” Ullrich says. “Models of snowpack in the Sierra Nevada continue to get better―and they’re already pretty good.”
His wish list for making these models even more accurate includes adding data from spots outside the range of the mountain weather stations. “SNOTELs are in places people can reach,” he explains. “They’re not in the bottoms of steep valleys or on top of peaks.” Complete information is already available for snow depth, thanks to recently-developed aerial lidar surveys that scan mountain ranges in their entirety rather than just fixed points.
Ullrich would also like to run the models at even finer resolutions, down to one-third of a mile. “There’s so much variability in the Sierra Nevada,” he says. “They’re very steep, and the shaded sides keep snow longer.” Here the limitation is the cost of computing power: doubling a model’s resolution makes it eight times more expensive to run.
RECREATING INDIVIDUAL ATMOSPHERIC RIVER EVENTS
Now the researchers are turning their attention to how individual atmospheric river events affect the snowpack. “A big question is the impact of rain on snow―warm storms can melt snow and increase the flood risk,” Ullrich says. Recreating disastrous floods can help water planners mitigate future extremes.
The researchers started with the 1997 New Year’s flood, the second worst on record in California. The stage was set by storms in November and December, which built up the Sierra Nevada snowpack. Then, between Christmas to January 2, a trio of atmospheric rivers dropped 30 inches of warm rain at elevations as high as 9,000 feet, rapidly melting enormous quantities of snow. The ground was already as wet as could be from rainfall and water just ran across the land, overwhelming rivers.
“I was just in Yosemite and the whole valley flooded in 1997―you can still see the seven-and-a-half-foot tall water line,” says Alan Rhoades, a hydroclimate scientist who helped develop the regionally refined Earth system model as a graduate student with Ullrich and is now a research scientist at Lawrence Berkeley National Laboratory.
Martinez, where Rhoades lives, flooded to depths of four feet. Statewide, more than 300 square miles flooded, impacting more than 23,000 homes and 2,000 businesses, and causing more than one billion dollars in economic damages. Hardest hit were Yuba City and Marysville due to levee breaks on the Feather River.
Rhoades and colleagues recreated the 1997 flood with a regionally refined climate model that focused on California. The resolution was about two miles, fine enough to capture how the storms interacted with mountain terrain, and the model’s estimates of precipitation, snowmelt and flooding matched those measured during the actual event. A video of the recreated event shows a timelapse of the storms sweeping over the land and leaving a greatly diminished snowpack.
“The model did a remarkably good job at predicting precipitation and snowmelt,” says Rhoades, who was lead author on a 2023 study that validated the model. “We were really happy.”
Now the researchers are using the model to estimate how flooding from a 1997-like flood event might change in the future under varying degrees of warming. “You get to play around with a lot of things in a virtual world―you can prod it and see what happens,” Rhoades says.
The next thing he wants to play around with is how spring heatwaves affect snowpacks, which can sublimate straight from ice crystals to water vapor. “Anecdotal evidence suggests Mount Rainer lost about 30% of its snowpack in the 2021 Pacific Northwest heatwave, and we don’t know where it went,” Rhoades says. “It needs a rigorous analysis because it has such important implications for water managers.” Seattle temperatures soared as high as 108⁰F during this extreme event, which lasted from June 26 and July 2, breaking records.
UNPACKING EXTREME EVENTS ONE STORM AT A TIME
State climatologist Anderson welcomes the contributions of researchers like Ullrich and Rhoades who are pushing the boundaries of climate models. Recreating individual extreme events is a “key piece to navigating a warmer world,” he says. “There’s a lot to do and we get by with a little help from our friends.”
He hopes researchers recreate other past impactful events, including California’s destructive spate of storms in 2023. Early that winter, nine back-to-back atmospheric rivers hit in a three-week span, causing floods, power outages, and mudslides. At least 21 people died and economic losses were estimated to exceed $3 billion.
Understanding extreme events will help planners prepare effective emergency responses for the immediate future, and prepare for likely climate change impacts by the end of the century. “Recreating an event with a model lets you unpack it and say ‘it unfolded this way,’” Anderson says. “Each storm is its own beast and instead of just talking about 30-year averages, we can get into the nuts and bolts―when they say the devil’s in the details, they’re absolutely right.”
