Building Custom Batched Ensemble Weather Forecasting with NVIDIA Earth2Studio

This tutorial demonstrates how to build an ensemble weather forecasting system using NVIDIA Earth2Studio, which generates multiple weather predictions simultaneously to capture uncertainty. The work involves loading a forecasting model, retrieving atmospheric data, creating custom diagnostics that convert wind measurements into wind-power capacity factors, and applying physically appropriate noise to different atmospheric variables while keeping one unperturbed control forecast. The ensemble forecasts are written to a coordinate-aware data storage system and evaluated against reference analyses using metrics like latitude-weighted RMSE and ensemble spread-skill ratios, with results visualized through spatial maps and forecast skill curves.