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

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 Networkcoined CorrectorGAN-via a custom training procedure and augmented loss function, to produce ensembles of high-resolution, biascorrected 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 highresolution model across an array of established probabilistic metrics. Crucially, Cor-rectorGAN, once trained, produces predictions in seconds on a single machine. These results raise exciting questions about the necessity of regional models, and whether datadriven downscaling and correction methods can be transferred to data-poor regions that

spateGAN: Spatio‐Temporal Downscaling of Rainfall Fields Using a cGAN Approach

“…The CRPS showed an improvement with a value of 0.012 compared to the MAE of spateGAN det and spate-GAN prob 01 . Compared to other studies (L. Harris et al, 2022;Price & Rasp, 2022), the score of the cGAN model does not drop below the respective MAE of the CNN. This might be related to the fact that both models apply an MAE loss function during training and the model selection is not considering pixel accuracy.…”

“…Therefore, requirements for the downscaling model would include an additional bias correction step. The potential for bias correction and spatial downscaling of weather forecast data using generative networks has been demonstrated in L. Harris et al (2022) and Price and Rasp (2022) and resulted in a performance reduction compared to downscaling coarsened observations. A similar result should be expected for spatio-temporal downscaling.…”

“…Leinonen et al (2021) demonstrated the performance and capability of cGANs within a spatial super-resolution approach by downscaling coarsened precipitation data from a resolution of 16-1 km. The same idea has also been applied to downscaling global precipitation forecasts (L. Harris et al, 2022;Price & Rasp, 2022). Furthermore, cGANs outperformed traditional precipitation nowcasting algorithms (Ravuri et al, 2021).…”

WeatherBench Probability: A benchmark dataset for probabilistic medium-range weather forecasting along with deep learning baseline models

“…Generative adversarial networks (GANs) [Goodfellow et al, 2014] are one way of producing realistic weather output. Recently this has been used for precipitation nowcasting [Ravuri et al, 2021] and downscaling [Price andRasp, 2022, Harris et al, 2022]. While GANs are, in principal, a well-suited method to produce coherent output, in practice they can be hard to train and empirically don't always produce the correct distribution.…”

CLGAN: A GAN-based video prediction model for precipitation nowcasting

“…Since generator and discriminator are trained adversarially, the generator is encouraged to create predictions that share the same statistical properties as the ground truth data. This is considered to be useful for generating realistic precipitation forecasts which should exhibit the high spatial variability of the observed data (Ravuri et al, 2021;Price and Rasp, 2022;Harris et al, 2022).…”