Most climate models in use today are based upon large-scale, well-understood physical relationships that drive global temperature and precipitation trends.
But the effects of complicated interactions that occur on smaller scales, which may still be significant, are harder to capture in these models.
That is why Dr. Laure Zanna of New York University and her colleagues are employing machine learning techniques, which can "learn" the effects of these interactions without explicitly solving the physics, to improve climate modeling. Climate Now spoke with Dr. Zanna to learn more.
Improving Climate Models with Machine Learning with Laure Zanna