Flat and fickle: Red River puts forecasters to testChris Laveau is on the front lines – make that the shorelines – of flood prediction science.
By: Patrick Springer, INFORUM
Chris Laveau is on the front lines – make that the shorelines – of flood prediction science.
Laveau, a hydrologist with the U.S. Geological Survey, maintains gauges that monitor the levels of the Red River and other streams.
That means he and his colleagues were extremely busy during this spring of record flooding, working to ensure the flow of information to river forecasters was accurate and current.
A case in point came a day or two before the Red reached its record crest on March 28 of 40.82 feet.
Laveau had checked gauge readings for the Red upstream at Abercrombie and at the confluence with the Wild Rice River – two key locations in predicting when the Red’s crest will reach Fargo.
After calling in the numbers to forecast hydrologists at the National Weather Service, Laveau heard on the radio that the predicted peak had been raised by a foot.
“It really brought home to me how important it is to provide good numbers, good information,” he said. “A lot of people are affected by the information we put out.”
River gauge readings, routinely delivered by satellite link, are just part of the information stream used by forecasters to predict flood crests.
The trickier parts of flood forecasting come in estimating water flowing over land – and when the spring thaw will release stored water in the snowpack and ground that eventually reaches the river.
As Red River Valley residents know, the river is perilous to predict when it spills over its shallow banks onto the surrounding tabletop landscape.
The river’s extremely gradual slope, its tendency to encounter ice jams as it flows north, and spring fluctuations between freezing and thawing temperatures all combine to make it precarious to predict the Red.
“Unfortunately, the Red River is one of the most difficult systems because of how flat it is,” says Gregg Wiche, director of the U.S. Geological Survey’s North Dakota Water Science Center in Bismarck, which provides river gauge readings and other data to forecasters with the National Weather Service.
Of all the variables that plague efforts to pinpoint the river’s crest, it’s the alternating freezing and thawing that most bedevils forecasters.
“There are no tools for that,” says Steve Buan of the weather service’s North Central River Forecast Center in Chanhassen, Minn. “It’s that freeze-thaw cycle that really makes a difference.”
The Red’s second spring crest in Fargo, at 34 feet on April 14, was far below the 41 feet forecasters predicted with 75 percent probability, with a 25 percent chance it could reach 42.8 feet.
Quite simply, the weather got colder than forecasters expected and stayed cold longer than they expected. The cold temperatures slowed the thaw and the flow of melted snow to the river, taming the second crest.
Forecasters had to estimate how much water remained on the landscape after the spring thaw – a prediction derailed by a two-week cold snap that kept water on fields and delayed runoff flowing into the river.
The freakish two-week cold snap – and lack of precipitation – spared many in the valley from disaster, but left forecasters scratching their heads.
“The model was still thinking the ground was frozen,” Buan says. That’s because it was influenced by a much more rapid thaw that exacerbated the 1997 flood.
“These are all things that go into research,” he says.
Pat Neuman, a retired hydrologist at the prediction center in Chanhassen, says there is a public perception that the too-high forecasts were based on the assumption heavy rains would fall.
Actually, he says, that’s not the case. “It’s not built into the models,” Neuman says. But he isn’t a fan of the “probabilistic forecasts,” with a range of possibilities, each associated with a chance of occurrence.
“They’re so hard for the people to understand,” Neuman says. “They can be used to explain away just about anything.”
He would prefer a single forecast level, plus or minus a foot or two.
Over time, as more gauges are posted along the main river and its tributaries, forecasters have a better information stream to refine their predictions, derived from 103 sub-basins covering 353 square miles in the valley.
“That’s going on all the time,” Buan says. In fact, adding stream gauges is one of the best ways to improve river forecasting, he says.
Daryl Ritchison, a WDAY meteorologist, credits the weather service for issuing a very accurate prediction for the Red River’s first crest, which surged to a record.
The final prediction called for a crest of 41 or 42 feet, with a 43-foot peak possible. The March 28 record crest, 40.82 feet, was “phenomenal,” in Ritchison’s view.
But he thought the weather service erred in issuing a forecast with a high probability that the second crest would top 40 feet, an outcome he thought was unlikely to occur.
Ritchison says it would be helpful if hydrologists somehow could issue forecasts that specify a predicted river level predicated on certain conditions, such as rainfall or temperatures.
A range of flood predictions, each associated with a different weather event that might – or might not – occur would be a useful guide for the public and decision-makers, Ritchison says.
“I think we will get some good changes out of this,” he says, with the errant second-crest prediction spurring forecasters to improve their techniques.
In fact, Buan says, upgraded forecasting software is in the works.
The new software will give forecasters more flood scenarios to consult when issuing their river forecasts.
“This is a big, revolutionary changeover,” Buan says. “We’ll have more options of different systems.”
Since the 1997 Red River flood, forecasting tools have seen some refinements, including better ways of mapping hydraulic water flows that can show property owners whether they will be flooded.
And experience adds up. Every year provides more information to give the model a better historical perspective to tap in making forecasts.
“We constantly change the internal workings of the model,” Buan says. “That’s a continuing process.”
The forecasting model has data going back to 1949, and evaluates 57 scenarios, allowing each subsequent flood forecast to be better informed.
The rise of ever-more sophisticated software models is really just a much more powerful version of the old slide-rule calculations hydrologists used when river forecasting was in its infancy during the 1930s and ’40s, Buan says.
But, as this spring’s second-crest predictions show, forecasters still can be stumped, even when armed with the best gadgetry and software models science can offer.
Readers can reach Forum reporter Patrick Springer at (701) 241-5522