PrecipCorrDiff: A Deep Learning Framework for Generating 0.0625$^\circ$ High-Resolution IMD Gridded Precipitation Data

Abstract

High-resolution precipitation datasets are essential for weather forecasting, hydrological modeling, and climate risk assessment, yet observation-based gridded products often lack the spatial detail required for regional applications. Here we present PrecipCorrDiff, a diffusion-based downscaling framework built by modifying the original corrector diffusion based archetecture that incorporates a customized loss function and a multi-scale residual generation strategy to improve the precipitation especially the representation of extremes. PrecipCorrDiff generates 0.0625° resolution India Meteorological Department (IMD) precipitation fields from 0.25° gridded data while preserving precipitation statistics and reconstructing fine-scale spatial variability. Validation against independent IMD station observations, comparing the proposed framework with the SRGAN-based model and the original CorrDiff model, indicates that the proposed framework substantially improves the representation of spatial precipitation patterns, rainfall intensity distributions, and extreme rainfall events, while eliminating the systematic negative bias present in the previously developed SRGAN-based model. Across four climatologically distinct homogeneous Indian monsoon regions, PrecipCorrDiff reproduces observed precipitation characteristics with near-zero spatial bias, demonstrating the potential of diffusion-based downscaling for generating physically consistent, high-resolution precipitation datasets for hydrological, meteorological, and climate applications.