A Probabilistic U-Net Approach to Downscaling Climate Simulations
By: Maryam Alipourhajiagha , Pierre-Louis Lemaire , Youssef Diouane and more
Potential Business Impact:
Makes climate predictions more detailed and accurate.
Climate models are limited by heavy computational costs, often producing outputs at coarse spatial resolutions, while many climate change impact studies require finer scales. Statistical downscaling bridges this gap, and we adapt the probabilistic U-Net for this task, combining a deterministic U-Net backbone with a variational latent space to capture aleatoric uncertainty. We evaluate four training objectives, afCRPS and WMSE-MS-SSIM with three settings for downscaling precipitation and temperature from $16\times$ coarser resolution. Our main finding is that WMSE-MS-SSIM performs well for extremes under certain settings, whereas afCRPS better captures spatial variability across scales.
Similar Papers
Deep Learning-Driven Downscaling for Climate Risk Assessment of Projected Temperature Extremes in the Nordic Region
Geophysics
Predicts exact future temperatures for planning.
China Regional 3km Downscaling Based on Residual Corrective Diffusion Model
Machine Learning (CS)
Makes weather forecasts more detailed and accurate.
On Global Applicability and Location Transferability of Generative Deep Learning Models for Precipitation Downscaling
Machine Learning (CS)
Makes weather forecasts work everywhere.