A conditional Generative Adversarial network model for the Weather4Cast 2024 Challenge
Atharva Deshpande, Kaushik Gopalan, Jeet Shah, Hrishikesh Simu
TL;DR
The paper tackles rainfall nowcasting by mapping SEVIRI radiance data to cumulative rainfall using OPERA ground truth. It introduces a two-stage pipeline: dense optical-flow extrapolates future radiance frames for $16$ steps over a $4$-hour horizon, followed by a conditional GAN that translates these radiances to rainfall maps aggregated over the horizon. Preprocessing includes averaging four infrared SEVIRI channels and masking cloudy regions with an Otsu threshold, enabling a compact Pix2Pix-like network with a U-Net backbone and a PatchGAN discriminator operating at $256\times256$ resolution. The approach achieves a CRPS of $7.34$ on Weather4Cast 2024, outperforming the baseline ($10.84$), demonstrating promise for data-driven nowcasting while highlighting the need for temporal modeling and better handling of peak intensities.
Abstract
This study explores the application of deep learning for rainfall prediction, leveraging the Spinning Enhanced Visible and Infrared Imager (SEVIRI) High rate information transmission (HRIT) data as input and the Operational Program on the Exchange of weather RAdar information (OPERA) ground-radar reflectivity data as ground truth. We use the mean of 4 InfraRed frequency channels as the input. The radiance images are forecasted up to 4 hours into the future using a dense optical flow algorithm. A conditional generative adversarial network (GAN) model is employed to transform the predicted radiance images into rainfall images which are aggregated over the 4 hour forecast period to generate cumulative rainfall values. This model scored a value of approximately 7.5 as the Continuous Ranked Probability Score (CRPS) in the Weather4Cast 2024 competition and placed 1st on the core challenge leaderboard.