A team of MU researchers is working on a new method of short-term weather forecasting that will enable meteorologists and hydrologists to better predict when and where heavy amounts of rainfall will occur.
The new model, known as “nowcasting,” could allow forecasters to issue flash flood warnings earlier and more accurately. Such floods are among the deadliest force of nature, killing an average of 140 people every year in this country, according to the National Weather Service.
The research is a collaboration among MU statistics professors Chris Wikle and Athanasios Micheas and professor of atmospheric science Neil Fox.
The researchers’ model uses history as well as current weather conditions to generate a range of outcomes for storms. The model assigns each possible outcome a statistical probability and an estimated margin of error.
Current nowcasting models, by contrast, only generate one possible scenario that does not measure the uncertainty of the forecast itself.
Unlike forecast models that analyze a week’s worth of weather data for the entire country, the nowcasting model being developed at MU restricts analysis to a smaller area dependent on the kind of radar used.
Nowcast models predict storm locations and intensities up to two hours in advance by using radar images to generate a new forecast every 10 minutes. Daily forecast models update every three to six hours, predicting general weather conditions three to seven days in advance.
Using such frequent data in the nowcast system allows the MU researchers to measure more precisely the amount of rainfall in an area as small as a single neighborhood at a specific time, Wikle said.
“A forecaster can put data into a model, and it will spit out a value, but the model does not tell the forecaster the margin of error involved when the value was determined,” he said. “By having this new information available, forecasters can know exactly what they are dealing with.”
Such a large amount of data takes a considerable amount of time to be processed. Creating a model that will account for every necessary variable and still find a solution quickly is difficult.
“Right now we are just starting to work on the first versions of the model,” Micheas said. “It is running a bit long, but we are learning what we can change to make it faster. Computers are also continually getting faster, which speeds up computational time.”
The researchers are in the first year of the four-year study, which is funded by a $750,000 grant from the National Science Foundation.
Right now, the ability to accurately issue flash flood warnings in enough time to warn the public is limited, meteorologists say.
“Ideally, we would like to issue flood warnings 15 to 30 minutes in advance. That is not always possible,” said Jim Kramper, the severe-storm-warning-coordination meteorologist for the St. Louis bureau of the National Weather Service.
“Flash flood warnings are issued when there are either reports of flooding or when the radar indicates that enough rain has fallen in an area that would cause creeks to leave their beds,” Kramper said.
Flood warnings are issued on a county-by-county basis although some more detailed information may be included if floods are affecting one specific area.