Attributing Causes to Biases in Shortwave Radiation from NOAA’s 3-km High Resolution Rapid Refresh (HRRR) Model Using NOAA’s High Quality Surface Radiation Measurement Network

This study evaluates causes of biases in surface shortwave radiation from NOAA’s 3-km HRRR model using observations from NOAA SURFRAD sites and 3 sites across the Columbia River Basin during the Wind Forecasting Improvement Project (WFIP-2). Biases in clear-sky and all-sky surface shortwave radiation are presented diurnally and seasonally and with respect to changing surface characteristics and for defined cloud-types. A cloud-type product has been developed at NOAA-ESRL Global Monitoring Division’s SURFRAD observatory network using surface-based radiation measurements, RadFlux products, ceilometer profiles, and sky imagers. The cloud-type product is developed using a random forest machine learning classification model that is trained on an enhanced cloud type product derived from vertically-pointing radar and lidar profiles from the DOE-ARM Southern Great Plains (SGP) site. Statistical correlations are explored between surface shortwave and longwave radiation with cloud-type, surface albedo, and 2-m temperature across these sites. Changes in these correlations will be investigated diurnally and seasonally, and how well these correlations are captured by HRRR simulations. During this period, the HRRR was targeted for specific improvements in sub-grid scale clouds, scale-aware aspects of turbulence parameterizations (PBL + shallow cumulus scheme), and land-surface physics.