Recent flood forecasts missed the mark in Sonoma County. Here’s why

The wavy, colored lines rose with each big storm and fell slightly with each break in the rain.

Each time, the “forecast” and “guidance” lines in the river charts approached the somewhat arbitrary, broken red line describing Russian River’s 32-foot flood stage in Guerneville.

During the seemingly endless parade of atmospheric rivers earlier this month, the colored, sinusoidal graphs — produced by the California Nevada River Forecast Center — became a go-to resource for predicting flooding along the river, advising first responders, business owners, residents and government officials.

But each time, predictions were off, in some cases, by several feet. Sometimes, predictions changed dramatically throughout the same day.

With the memory of the devastating 2019 flood still fresh, few in Sonoma County were willing to ignore flood predictions earlier this month. Four years ago, the river jumped roughly 10 feet in only 24 hours to crest at 45.38 feet on Feb. 27.

So why were predictions this year off by such a large margin?

Dynamic model

Alan Haynes, the hydrologist in charge of the California Nevada Forecasting Center, which produces river level forecasts for the Russian River, said that unlike regular storms, atmospheric rivers carry a tremendous amount of moisture. That leads to a huge margin of error.

If the storm’s path deviates from predictions, the forecasts will be way off.

“A lot of people describe it almost like a fire hose that you're trying to hold and this jet of moisture just sort of keeps moving around,” Haynes said. “The model is very dynamic. Models are trying to lock it in, but there's uncertainty in the atmospheric modeling.”

Haynes pointed to three separate forecasts made on Jan. 11:

• At 8:22 a.m. the forecast had the river rising above the 32-foot flood stage on Jan. 12 and Jan. 15.
• Five hours later, at 1:20 p.m., the forecasts for those two days were lowered by 2 feet, closer to the actual river levels for Jan. 15.
• Then, seven hours later, at 8:10 p.m., the forecast for Jan. 12 was spot on but the forecast for Jan. 15 once again rose above flood stage, 2 feet higher than what actually resulted that day.

The modeling on Jan. 14 was the most accurate.

Haynes said that to predict flood crests at Guerneville, his team relies on precipitation forecasts, which are largely based on atmospheric model predictions. The Russian River has a number of watersheds before it gets to Guerneville.

Rainfall and runoff in the upper part of the Russian River watershed can take up to two-three days to make it to Guerneville, while runoff in the middle and lower part of the watershed takes less time. He said to accurately predict a flood crest at Guerneville, hydrologists need an accurate prediction of rain several days ahead.

“Rainfall forecasts several days out can have significant errors,” Haynes said in an email.

A tricky proposition

Haynes described atmospheric rivers as weather systems that transport a tremendous amount of moisture that can result in a high amount of runoff and flooding wherever they strike. But forecasting their impact several days out can be tricky.

The margin of error for an atmospheric river can be much larger than a normal storm, he added. For example, if the forecast two to three days out calls for one inch of rain in the watershed — what one might expect from a typical storm — and the storm misses the watershed, it produces a relatively small error in the forecast river stage at Guerneville.

However, if the forecast for an atmospheric river calls for five inches of rain two to three days out and the storm misses the watershed, “it will produce a much, much larger error in the forecast river stage at Guerneville,” he said.

Haynes said more research on atmospheric rivers is needed. “We can greatly improve our forecasts for flood crests at places like Guerneville several days out if we can better predict where these atmospheric rivers will land,” he said.

Josh Leone, a local tech entrepreneur who lives in the west county and has begun tracking river level forecasts, said that while flood predictions aren’t perfect, they can’t be ignored.

“You definitely can't ignore it, and you have to prepare for the worst, you have to prepare for that forecast,” he said. “And you have to be prepared that there will be some frustration with this data until, except for maybe 24-to-30 hours out.”

Toward the end of the series of storms earlier this month, Leone wrote a program that would automatically download river forecasts every six hours. He plotted them, starting Jan. 11 and running through Jan. 15, at 3 p.m. each day.

The most accurate at that hour was Jan. 11 and the least accurate was Jan. 13.

“If you talk to any old-timer, they’ll tell you don't pay attention to that (forecast) stuff. What's gonna happen is gonna happen,” Leone said.

But Leone said the forecasting was helpful for those in the lowest lying areas of the west county, including many trailer home residents who were moved to higher elevations.

“It was good we moved them out because the water was there,” Leone said. “Even at 29, 30 or 31 feet, those lowest, lowest lying areas that had trailers had water.”

From 28 to 32 feet, the lower Russian River sees minor flooding. Then between 32 to 35, the region gets what’s considered “nuisance flooding,” where it starts to more deeply affect communities, he added.

Major flooding occurs at levels above 35 feet.

“Rio Nido goes underwater, major establishments start to get impacted,” Leone said.

While Sonoma County escaped that scenario, the atmospheric rivers pounded communities to the south, including Santa Cruz, Monterey and San Luis Obispo counties. “All it takes is a little jet stream north and it’s would have been us,” he said. “I pay attention to (predictions) and I advise the community, through social media pages, to pay attention.”

Crying wolf

During the storms, Leone set up an automatic text alert system that notified Supervisor Lynda Hopkins and others every time the forecast center’s charts got updated.

“Sometimes there are substantial adjustments that can bring either relief or concern to an exhausted community,” said Hopkins, whose district takes in west county. “My personal strategy is to prepare for the worst and hope for the best.”

Hopkins said the “whiplash” surrounding forecasts earlier this month was better than being caught off-guard, as was the case in 2019.

“In 2019, they didn't realize that the atmospheric river would put a bull's-eye on Sonoma County and stall out over us,” she said. “That year, the river forecast jumped up almost 10 feet overnight, and we were all scrambling to adjust from what was going to be maybe a 34, 35-foot flood to a 45-foot flood.”

Hopkins said she’s much rather see “conservative or aggressive” forecasts that describe worst case scenarios than predictions that “give us a sense of security and leave us scrambling to evacuate.”

You can reach Staff Writer Martin Espinoza at 707-521-5213 or martin.espinoza@pressdemocrat.com. On Twitter @pressreno.