PUnet-CDR: A Global High-Resolution Precipitation Climate Data Record for Hydroclimate and Drought Applications

Reliable long-term precipitation records are essential for hydrologic forecasting, climate analysis, and drought monitoring, yet existing satellite-based products face trade-offs among resolution, temporal frequency, and historical coverage. High-resolution datasets such as IMERG, CMORPH, and PERSIANN provide detailed precipitation estimates but are limited to recent decades, while long-term climate products such as GPCP and CMAP span multiple decades at coarse resolution. These constraints limit the characterization of sub-daily variability, extremes, and long-term hydroclimatic trends, particularly in data-sparse regions.

We present the PERSIANN-UNet Climate Data Record (PUnet-CDR), a global deep learning–based system that reconstructs high-resolution precipitation Climate Data Records (CDRs) from 1980 to 2025. Built upon the PUnet algorithm, PUnet-CDR integrates geostationary infrared (IR) satellite observations with monthly precipitation climatology to produce 3-hourly global precipitation estimates at 0.04° (~4 km) resolution. The system leverages GridSat-B1 (1980–February 2000) and CPC-4km (March 2000–2025) IR datasets standardized to a common grid.

Long-term consistency is achieved using a monthly GPCP-based bias correction, in which coarse-scale correction factors are transferred to high-resolution outputs. In addition, GPCP-corrected NASA MERRA-2 precipitation is used to fill gaps in the IR record, yielding a spatially and temporally complete precipitation CDR. Unlike regional mosaicking approaches, PUnet-CDR employs a globally trained framework, eliminating boundary artifacts and enabling consistent representation of large-scale precipitation patterns.

A key application of PUnet-CDR is global drought monitoring and prediction. The dataset supports multi-timescale drought indicators and machine-learning models for drought onset and severity, demonstrated through UCI’s global drought monitoring platform. PUnet-CDR thus provides a scalable, high-resolution foundation for hydroclimate research and operational decision support at global scale.