A Novel Reference-Based and Gradient-Guided Deep Learning Model for Daily Precipitation Downscaling

The spatial resolution of precipitation predicted by general circulation models is too coarse to meet current research and operational needs. Downscaling one way provide finer data at local scales. single-image super-resolution method in the computer vision field has made great strides lately been applied various fields. In this article, we propose a novel reference-based gradient-guided deep learning model (RBGGM) downscale daily considering discontinuity ill-posed nature downscaling. Global Precipitation Measurement Mission (GPM) data, variables ERA5 re-analysis topographic are selected perform downscaling, residual dense attention block constructed extract features them. By exploring discontinuous feature precipitation, introduce gradient reconstruct distribution. We also high-resolution monthly as reference resolve Extensive experimental results on benchmark sets demonstrate that our proposed performs better than other baseline methods. Furthermore, construct downscaling set based GPM data.