0.0625° Resolution IMD Gridded Precipitation Data generated by Diffusion based DL framework.

Description: This dataset contains high-resolution (0.0625°) daily precipitation fields over India generated using PrecipCorrDiff, a diffusion-based downscaling framework built on the Corrector Diffusion (CorrDiff) architecture. The dataset was produced by downscaling India Meteorological Department (IMD) gridded precipitation data from 0.25° to 0.0625° resolution for the test period 2014–2018. PrecipCorrDiff incorporates a customized loss function and a multi-scale residual generation strategy designed to improve the representation of fine-scale spatial variability and extreme rainfall events. The model was validated against independent IMD station observations and shown to substantially improve spatial precipitation patterns, rainfall intensity distributions, and extreme event representation relative to a previously developed SRGAN-based downscaling model, while eliminating its systematic negative bias. This release provides the downscaled precipitation output for the independent test period (2014–2018), enabling evaluation, benchmarking, and downstream applications in hydrological modeling, weather forecasting, and climate risk assessment over India. Spatial coverage: India (land domain)Spatial resolution: 0.0625° × 0.0625°Temporal coverage: 2014–2018 (daily)Temporal resolution: DailyFormat: NetCDF (.nc)Source input data: IMD gridded daily precipitation (0.25°)Variables: precipitation (daily rainfall, mm/day) Methodology: Precipitation fields were generated using a two-stage diffusion-based downscaling model (PrecipCorrDiff), trained on historical IMD gridded precipitation data. Full methodological details are provided in the accompanying manuscript. Related publication: [Kumar et al. 2026 "PrecipCorrDiff: A Deep Learning Framework for Generating 0.0625° High-Resolution IMD Gridded Precipitation Data"] How to cite: [10.5281/zenodo.21819376] Contact: Bipin Kumar, Indian Institute of Tropical Meteorology (IITM), Pune