A climate modeling downscaling method developed by MIT scientists leverages a machine-learning technique called adversarial learning to produce simulations faster and at finer resolutions, making them relevant to use on local levels for assessing risks of extreme weather.