Winter precipitation characteristics in western US related to atmospheric river landfalls: observations and model evaluations

Abstract

Winter precipitation (PR) characteristics in western United States (WUS) related to atmospheric river (AR) landfalls are examined using the observation-based PRISM data. The observed AR-related precipitation characteristics are in turn used to evaluate model precipitation data from the NASA MERRA2 reanalysis and from seven dynamical downscaling simulations driven by the MERRA2. Multiple metrics including mean bias, Taylor diagram, and two skill scores are used to measure model performance for three climatological sub-regions in WUS, Pacific Northwest (PNW), Pacific Southwest (PSW) and Great Basin (GB). All model data well represent the winter-mean PR with spatial pattern correlations of 0.8 or higher with PRISM for the three sub-regions. Higher spatial resolutions and/or the use of spectral nudging generally yield higher skill scores in simulating the geographical distribution of PR for the entire winter. The PRISM data shows that the AR-related fraction of winter PR and associated daily PR PDFs in each region vary strongly for landfall locations; AR landfalls in the northern WUS coast (NC) affect mostly PNW while those in the southern WUS coast (SC) affect both PSW and GB. NC (SC) landfalls increase the frequency of heavy PR in PNW (PSW and GB) but reduce it in PSW (PNW). All model data reasonably represent these observed variations in the AR-related winter PR fractions and the daily PR PDFs according to AR landfall locations. However, unlike for the entire winter period, no systematic effects of resolution and/or spectral nudging are identified in these AR-related PR characteristics. Dynamical downscaling in this study generally yield positive added values to the MERRA2 PR in the AR-related PR fraction for most sub-regions and landfall locations, most noticeably for PSW by NU-WRF. The downscaling also generate positive added value in p95 for PNW, but negative values for PSW and GB due to overestimation of heavy precipitation events.