Training machine learning models on climate model output yields skillful interpretable seasonal precipitation forecasts

Seasonal forecasting skill in machine learning methods that are trained on large climate model ensembles can compete with, or out-compete, existing dynamical models, while retaining physical interpretability.