Evaluating the Impact of Radar Data Availability on the Success of MRMS and HRRR Winter-time Precipitation in the Colorado Rockies

Accurate wintertime precipitation forecasts and estimates enhance public safety and are critical for water supply forecasts. Complex terrain, such as the Rocky Mountains, is known to make forecasts and estimates of precipitation challenging for the High Resolution Rapid Refresh (HRRR) model and Multi-Radar Multi-Sensor (MRMS) product. Since both HRRR and MRMS make use of radar data, we hypothesize that the quality of gridded precipitation products is related to the availability of radar observations. To assess this, cool-season precipitation from the HRRR 1-hour forecast, the HRRR 6-hour forecast, the MRMS pass 2 product, Snow Telemetry (SNOTEL) precipitation, and lidar-derived Snow Water Equivalent (SWE) data were used. The difference in seasonal accumulation from HRRR and MRMS was compared to radar quality index (RQI) values within five Colorado watersheds. These basins were selected for their windward/leeward position, average RQI, and the availability of lidar-derived SWE data. Since not all cool-season precipitation falls as snow, and lidar-derived SWE measurements observe the state of the snowpack which may melt before it is observed, it is expected that the HRRR and MRMS precipitation will be greater than what is observed by the lidar-derived SWE. Where adequate radar data existed (RQI >= 0.6), the HRRR and MRMS precipitation totals were greater than the lidar-derived SWE. Where inadequate radar data existed (RQI <= 0.3), the HRRR and MRMS precipitation was less than the lidar-derived SWE. This means that HRRR and MRMS were both underestimating wintertime precipitation. Where RQI was between 0.3 and 0.6, HRRR and MRMS precipitation were substantially lower than the lidar-derived SWE. Precipitation data from SNOTEL was used to verify the patterns seen with lidar derived SWE in relation to changing HRRR forecast lead-time. These results suggest that patterns of seasonal precipitation bias exist within HRRR forecasts and the MRMS pass 2 products in relation to RQI.