UT-GraphCast Hindcast Dataset: A Global AI Forecast Archive from UT Austin for Weather and Climate Applications

TL;DR

The UT GraphCast Hindcast Dataset offers a comprehensive 45-year archive of global weather forecasts generated by a physics-informed graph neural network, enabling detailed climate and weather analysis.

Contribution

This work introduces a large-scale, high-resolution global hindcast dataset produced by a novel AI model, enhancing weather and climate research capabilities.

Findings

Provides daily 15-day forecasts from 1979 to 2024

Uses a physics-informed graph neural network trained on ECMWF ERA5 data

Delivers forecasts in under one minute on modern hardware

Abstract

The UT GraphCast Hindcast Dataset from 1979 to 2024 is a comprehensive global weather forecast archive generated using the Google DeepMind GraphCast Operational model. Developed by researchers at The University of Texas at Austin under the WCRP umbrella, this dataset provides daily 15 day deterministic forecasts at 00UTC on an approximately 25 km global grid for a 45 year period. GraphCast is a physics informed graph neural network that was trained on ECMWF ERA5 reanalysis. It predicts more than a dozen key atmospheric and surface variables on 37 vertical levels, delivering a full medium range forecast in under one minute on modern hardware.