Sara Shamekh and collaborators present a new machine learning framework that combines probabilistic modeling with symbolic equation discovery to uncover how different types of tropical rainfall—shallow convective, deep convective, and stratiform—depend on large-scale atmospheric conditions. Using satellite observations and reanalysis data, the study derives compact, physically interpretable equations that capture key environmental controls on rain area, providing new insights into tropical convection and paving the way for more realistic, stochastic precipitation parameterizations in climate models.

Read the paper here