Stakeholders such as water managers need high-resolution precipitation information on various timescales. However, accurate and reliable precipitation simulation is challenging for even fine-resolution climate models due to the highly variable nature of precipitation on both the spatial and temporal scales, inherent model biases, and imperfect parameterizations. Therefore, the quality of high-resolution simulated precipitation must be evaluated for the saliency and useability of the product. Our study evaluates 3 and 12h simulated mean and extreme precipitation across the contiguous United States (CONUS) in a 12-km Weather Research and Forecasting model version 4.2.1 (WRF v 4.2.1)-based dynamical downscaling of the fifth-generation ECMWF atmospheric reanalysis (ERA5) against the National Centers for Environmental Prediction/Environmental Modeling Center (NCEP/EMC) 4 km Stage IV and Oregon State University Parameter-Elevation Regressions on Independent Slopes Model (PRISM) data on 3 and 24 h timescales, respectively. The study shows that the WRF improves on many aspects of simulated precipitation, such as the magnitude and timing of diurnal and annual cycles of precipitation and probability density functions of 3 and 24-h annual maximum precipitation. However, WRF shows seasonally- and regionally-dependent precipitation biases, such as the slow propagation of mesoscale convective systems typically found in similar WRF simulations. The study shows that while a 12-km WRF simulation better represents precipitation than the native ERA5, further improvements are likely from downscaling to convection-permitting scales.

Attached figure: magnitude of the monthly average precipitation peak (MMPP) estimated over 2001–2020 (2003–2019 for Stage IV). The left column shows the magnitude of the peak in each dataset and uses the color scale along the bottom edge of the figure. The right column shows biases in the magnitude and uses the color scale along the right edge of the figure. Units: mm/day.