Neural general circulation models optimized to predict satellite-based precipitation observations
This addresses the critical problem of unreliable precipitation simulation for climate scientists and forecasters, representing a substantial but incremental advance in climate modeling methodology.
The researchers tackled the problem of inaccurate precipitation simulation in climate models by developing a hybrid model trained directly on satellite-based precipitation observations, which demonstrated significant improvements over existing models including reduced biases, better representation of extremes, and outperforming ECMWF ensemble forecasts.
Climate models struggle to accurately simulate precipitation, particularly extremes and the diurnal cycle. Here, we present a hybrid model that is trained directly on satellite-based precipitation observations. Our model runs at 2.8$^\circ$ resolution and is built on the differentiable NeuralGCM framework. The model demonstrates significant improvements over existing general circulation models, the ERA5 reanalysis, and a global cloud-resolving model in simulating precipitation. Our approach yields reduced biases, a more realistic precipitation distribution, improved representation of extremes, and a more accurate diurnal cycle. Furthermore, it outperforms the mid-range precipitation forecast of the ECMWF ensemble. This advance paves the way for more reliable simulations of current climate and demonstrates how training on observations can be used to directly improve GCMs.