Scientists from MIT and the Pacific Northwest National Laboratory have developed a way to increase the accuracy of large-scale climate models, allowing for more precise predictions of extreme weather incidents in specific locations. Their process involves using machine learning in tandem with existing climate models to make the models’ predictions closer to real-world observations. This method can be applied to all global climate models and could help to predict the frequency and location of extreme weather events in the coming years. It could be an invaluable tool for preparing for the impacts of climate change.
The current climate models can predict average temperatures, humidity, and precipitation around the world, but their coarse resolution does not capture finer details like clouds or storms, which can have significant impacts on local weather patterns. The researchers have now developed a method to correct the model’s predictions by using a machine-learning algorithm. This algorithm is trained using past data to identify and correct discrepancies between the model’s output and real-world conditions.
In a test run with the Energy Exascale Earth System Model from the U.S. Department of Energy, their method was able to produce climate patterns that more closely matched real-world observations from the past 36 years. The corrected model also accurately predicted the frequency of extreme storms at specific locations worldwide.
Themistoklis Sapsis, the William I. Koch Professor and director of the Center for Ocean Engineering at MIT, believes their approach offers a different path to improving climate models. Instead of tweaking the complex equations that underpin the models, they’re correcting the output, giving more accurate forecasts of extreme weather events, which will be crucial for preparing for climate change impacts on everything from biodiversity to food security to the economy.
The team’s results have been published in the Journal of Advances in Modeling Earth Systems. The work was partially funded by the U.S. Defense Advanced Research Projects Agency. The researchers’ ongoing work will focus on analyzing future climate scenarios.
