Increasing the accuracy and resolution of precipitation forecasts using deep generative models

TL;DR

This paper introduces CorrectorGAN, a deep generative model that enhances global precipitation forecasts by correcting biases and increasing resolution, achieving high accuracy and rapid predictions suitable for practical use.

Contribution

The paper presents a novel conditional GAN approach for simultaneous bias correction and super-resolution of precipitation forecasts, outperforming existing methods and enabling fast, high-resolution predictions.

Findings

CorrectorGAN outperforms interpolation and CNN-based methods.

The model approaches the accuracy of regional high-resolution models.

Predictions are generated in seconds on a single machine.

Abstract

Accurately forecasting extreme rainfall is notoriously difficult, but is also ever more crucial for society as climate change increases the frequency of such extremes. Global numerical weather prediction models often fail to capture extremes, and are produced at too low a resolution to be actionable, while regional, high-resolution models are hugely expensive both in computation and labour. In this paper we explore the use of deep generative models to simultaneously correct and downscale (super-resolve) global ensemble forecasts over the Continental US. Specifically, using fine-grained radar observations as our ground truth, we train a conditional Generative Adversarial Network -- coined CorrectorGAN -- via a custom training procedure and augmented loss function, to produce ensembles of high-resolution, bias-corrected forecasts based on coarse, global precipitation forecasts in addition to other relevant meteorological fields. Our model outperforms an interpolation baseline, as well as super-resolution-only and CNN-based univariate methods, and approaches the performance of an operational regional high-resolution model across an array of established probabilistic metrics. Crucially, CorrectorGAN, once trained, produces predictions in seconds on a single machine. These results raise exciting questions about the necessity of regional models, and whether data-driven downscaling and correction methods can be transferred to data-poor regions that so far have had no access to high-resolution forecasts.