MPR Photo/Conrad Wilson
By Dan Gunderson
This year’s Red River spring flood in Fargo-Moorhead is the latest on record, and perhaps one of the most over predicted floods.
Earlier this spring, the National Weather Service offered a probabilistic flood outlook – an early estimate of flooding potential – and predicted the Red River in Fargo-Moorhead would reach a crest of 38 to 42 feet. The upper range of that outlook would have been a record-breaking flood, and the possibility kicked flood preparations in Fargo into high gear. Volunteers filled more than 1 million sandbags and contractors constructed several miles of emergency levees.
Cost for that work will top $2 million.
But after several revisions lowering the crest levels each time, the Red River in Fargo is predicted to crest today at just over 33 feet. That’s more than six feet below the record level and more than four feet below the low end of the weather service outlook.
The sandbag dikes will all stay dry this year, Fargo Senior Engineer April Walker said.
“It can be frustrating but at the same time it’s way better to be prepared than to have the crest rise and not be prepared for that,” she said.
Walker said the city must prepare for the river levels that the National Weather Service predicts.
What happened with the predictions?
The conditions made for an ideal spring melt; gradual with temperatures above freezing during the day and below freezing at night. The Red River Valley also avoided several storm systems that could have added more water to the region.
But those factors were part of the initial outlook from the National Weather Service and should have resulted in a river level near the lower end of the prediction: 38 feet.
Steve Buan, a hydrologist for the weather service, said this spring flood also exposed a weakness in the flood prediction model. It didn’t adequately account for the drought that dried out the landscape last summer.
The late spring melt meant the soil thawed and absorbed much more water than forecasters expected.
“And also I think some of the factors of the dry fall with the depressional storage, the marshes and sloughs that typically hold water, that probably wasn’t taken as much into account numerically within the model as what’s actually occurring out on the landscape,” Buan said.
The same weakness in the flood predication model was evident in 2009 when the Red River reached a record level of 40.8 feet.
That was near the top of the range predicted by the National Weather Service, with about a five percent chance of happening, and the high water caught Fargo- Moorhead by surprise, prompting an intense effort to keep the cities from flooding.
In the spring of 2009, conditions were the exact opposite of 2013. Sloughs and wetlands were all full to the top after a wet summer the year before. The saturated soil then froze and a rapid snow melt in March meant very little water was stored on the landscape or in the soil. Most of the snowmelt flowed into rivers, leading to a record flood.
The National Weather Service flood prediction model can accurately adjust for soil moisture and wetland storage on a small scale, down to about four square miles, according to Buan.
But he said attempts to scale that prediction power up to hundreds or thousands of square miles for an entire watershed or the Red River Valley basin does not work.
There’s just too much variation across the landscape. And the land is constantly being manipulated, Buan said. Some fields have drain tile installed , new ditches dug or old ones cleaned. All of those activities change the dynamics of snowmelt runoff.
This was a tough year to be a river forecaster, but Buan said criticism for missing the forecast goes with the territory. He hopes this flood will raise some questions that improve future forecasting.
“Usually things like this spur research,” he said. “Somebody will take this on and want to explore exactly what happened and that will feed back into the predictive process.”
The information from the 2013 Red River flood will also influence future predictions. It will take about two years to review all of the data and check it for accuracy. Then the latest flood on record will be one of the 60 scenarios the computer model can consider for future forecasting.