Floods have just ravaged Western Europe, fires ravage the west coast and forecasters predict a busy hurricane season in the Atlantic – extreme weather is on the rise and it won’t go away anytime soon. At Stanford, researchers are applying machine learning to understand them. Extreme weather events, especially extreme rainfall and flooding in the Midwest.
“We know the floods have gotten worse,” said lead author Frances Davenport, a PhD student at the Stanford School of Earth, Energy and Environmental Sciences, in an interview with Stanford’s Rob Jordan. “Our goal was to understand why extreme precipitation increases, which in turn could lead to better predictions of future floods.”
Floods in the Midwest are responsible for flooding the majority of the United States, even in the Midwest. Years ago, researchers trained a machine learning algorithm that uses network data to identify the circulation patterns that come with extreme rainfall. Using precipitation data from the last 40 years, the researchers trained a machine learning algorithm oriented around grid data to recognize the circulation patterns that accompany extreme precipitation.
The trained algorithm turned out to be very accurate. “The algorithm we used correctly identified over 90 percent of the extreme days of precipitation, which is higher than the performance of the traditional statistical methods we tested,” said Davenport. In addition, the researchers were able to use the algorithm to better understand the causal factors at the meteorological and climatological level: air pressure patterns that are steadily increasing from year to year, high humidity that flows out of the Gulf of Mexico, and more.
Next, the researchers want to broaden their focus and investigate a greater spatial-temporal variety of extreme weather conditions. “While we initially focused on the Midwest, our approach can be applied and used to other regions to better understand changes in extreme events”, Davenport “This will help society better prepare for the effects of climate change.” The researchers also want to forecast how extreme weather events will develop in the future.
“This new approach to using machine learning techniques opens up new avenues in our understanding of the underlying causes of the shift from extremes,” said co-author Noah Diffenbaugh, professor at Stanford College of Earth, Energy and Environmental Sciences. “That could enable communities and decision makers to better prepare for large impact events that are so extreme that they are beyond our historical experience.”
Source link
