U-Net and GAN models improve rainfall detail over West Africa
Statistical Downscaling of Daily Precipitation from ERA5 Atmospheric Fields to IMERG over West Africa
Machine Learning
Summary
Weather forecasts over West Africa struggle to capture the small storms that cause most rain because the data is too coarse. The authors used advanced computer models called U-Net and GAN to transform rough weather data into much finer rain maps for daily rainfall. They found that standard models miss a lot of rain variability nearby, so they added a special GAN model to do better. They also discovered that moisture around 500 hPa altitude is the most important factor for predicting rain in this region.
What this means in practice
- •For weather modelers: Create more detailed daily rainfall maps for West Africa from coarse reanalysis data to improve local weather predictions.
- •For hydrological planners: Use high-resolution precipitation data to support water management and flood risk assessment in West African monsoon regions.
Authors
David Sagna, Redouane Lguensat, Samo Diatta
Abstract
Global reanalyses are too coarse to resolve the convective systems that produce most West African monsoon rainfall. In this work, we use U-Net based architectures to downscale daily ERA5 fields (0.25°) to IMERG V07 satellite precipitation (0.1°) over West Africa for the June to October seasons of 2000 to 2024, using geopotential, temperature, humidity and wind at three pressure levels together with elevation and land fraction. Adding ERA5 precipitation as input slightly improves performance for this particular challenging region. We find that all the deterministic networks investigated for this work are too dry and lose almost all observed variance below a few hundred kilometers, which motivated our use of a GAN-based alternative. We also investigated permutation importance for the U-Net that ranks 500 hPa specific humidity first, consistent with the sensitivity of deep convection to mid-tropospheric moisture.