Skip to main navigation Skip to search Skip to main content

High-resolution ensemble precipitation and temperature datasets for CONUS based on a probabilistic geospatial estimation approach

  • Wuhan University
  • Colorado School of Mines
  • University of Oklahoma
  • United States Army
  • U.S. Department of the Interior

Research output: Contribution to journal › Article › peer-review

Abstract

Precipitation and temperature datasets are essential for diverse applications, yet existing high-resolution products often lack robust uncertainty quantification and transparent, reproducible station processing. This study addresses these gaps by developing a new daily 0.02° ensemble of surface precipitation and temperature for the Contiguous United States (CONUS) spanning 1950–2023 with 20 members, built on a framework combining station record reconstruction and spatial probabilistic estimation. We first construct a serially complete station archive from high-density networks (GHCN-D and MADIS) using a comprehensive set of gap-filling and reconstruction techniques—quantile mapping, interpolation, machine learning, and multi-source merging—yielding 25,887 precipitation and 20,998 temperature stations. Ensemble fields for precipitation, mean temperature (Tmean), and daily temperature range (Trange) are then generated using a probabilistic geospatial estimation framework that characterizes real-world uncertainty via cross-validation. Results show the reconstructed station records are highly accurate (median modified Kling–Gupta efficiency, KGE'': 0.89 for precipitation, 0.99 for Tmean, 0.93 for Trange). Spatial estimates validated by leave-one-out cross-validation also perform robustly (median KGE'': 0.74 for precipitation, 0.95 for Tmean, 0.71 for Trange), with larger challenges in complex terrain. Comparison with the widely used climate datasets (PRISM and Livneh) indicates that our new dataset reliably captures the climatological distributions and trend patterns of precipitation, Tmean, and Trange across CONUS. By contrast, the Livneh dataset exhibits substantially higher Trange and unrealistic spatial trends characterized by interpolation hotspots. Overall, this open-access ensemble—together with its station archive and gridded deterministic and probabilistic components—provides a critical resource for high-resolution hydrologic and climate research across CONUS.

Original languageEnglish
Article number134761
JournalJournal of Hydrology
Volume666
DOIs
StatePublished - Feb 2026

Fingerprint

Dive into the research topics of 'High-resolution ensemble precipitation and temperature datasets for CONUS based on a probabilistic geospatial estimation approach'. Together they form a unique fingerprint.

Cite this