You know the drill. The storm brews and threatens to flood. But where? That would sure be useful to know.
Researchers at UL Lafayette are developing a way to answer that question — at least within a range of probabilities. A beta version of a river stage forecasting model is live now and tracking Tropical Depression Nicholas. It predicts how high the water will rise at points along the river over the next several days, based on rainfall predictions provided by NOAA’s Global Forecast System.
Forecasting floods, the researchers hope, will help Lafayette more effectively respond and plan for events, given enough lead time. Data isn’t quite precise enough to predict whether individual homes or businesses will flood, but the UL model can help localize potential flood risks to a narrower scope than what’s generally offered by national weather predictions. When a storm like Nicholas approaches, the hope is, local governments can plot out which roads and bridges they’ll likely need to close, or more precisely develop response plans for affected areas.
“This is the first step in a complicated system,” says Emad Habib, a civil engineer and the director of the Louisiana Watershed Flood Center at UL. The goal is to refine the model and create forecasts at a “scale that impacts people.”
The idea is simple, but the science behind it is complex. Rain forecasts from an impending storm are plugged into a finely tuned model of the river, its tributaries and dynamic, which in turn projects how the river will behave.
Right now, forecasters can’t do much more than alert the public that flooding, in a general sense, is likely. Digging deeper requires a refined and local hydrological model, which hasn’t yet been developed.
While meteorological forecasting has matured, Habib says, forecasting hydrological events — floods — hasn’t kept up. But the science is within reach to produce a useful tool locally.
The National Weather Service provides a similar forecast at the Surrey Street bridge, but that’s it for pretty much the entire river in Lafayette Parish. And, as we know, the river’s elevations vary greatly from place to place. UL’s model has those elevations accounted for.
Predicting a flood requires essentially three ingredients: how much rain, where the rain falls and when. Forecasting the volume of rain is pretty reliable, Habib says, but the last two ingredients are still relatively elusive. The science is good enough to manage the variability, however.
“I can tell you over the next few days over that area there will be so much rain,” Habib says. “But if you want to forecast rainfall over a watershed, you’re talking about forecasting the right place and the right timing, not just over days but hour to hour, because that’s how the [weather] system responds.”
The scale of the weather system matters here. Modeling needs lead time to work. So predictions are best served for larger rain events, like hurricanes or tropical storms. A flash flooding event like May’s surprise attack would be too fast for the model to be of use. But that doesn’t mean storms smaller than named weather events are invisible. Consider the August 2016 storms that dumped rain over several days. UL’s model could project the flooding that would follow from a similar event and still help locals respond. Climate change is making those kinds of events much more likely, unfortunately, meaning the research here has plenty of opportunity to prove useful.
One of the researchers’ goals is to eventually run forecasts on any tributary of the Vermilion. That would mean forecasting flood risks along major coulees, the source of most of Lafayette’s flooding.
“This is tricky business. A hurricane is a large-scale phenomenon. The response time and the change time of the hurricane has its own scale,” Habib says. “Compare that to Coulee Ile de Cannes and Coulee Mine, the time scale is much smaller and the response time is much faster.”
Next steps are designed to make the forecasting more useful to the public and decision makers. First, the model will communicate probability — i.e. the Vermilion River at Rotary Point is 50% likely to rise between 2 and 3 feet — not unlike how meteorologists describe hurricane paths. Uncertainty is baked in, so assessing it is useful.
Second, the model will extrapolate the river stage data to map potential flooding. Instead of interpreting a line on a hydrograph, there’d be the ability to gauge with some confidence what areas would flood and to what degree and visualize it.
Habib emphasizes the model is still developing. Its predictions are only as good as the data fed into it. While the next phase, including a more user-friendly interface, is within reach, there will always be room to improve reliability.
“I don’t want to trivialize it,” Habib says of the business of flood forecasting. “Every storm we learn something new.”
Got some feedback on the forecast? Send them to [email protected]