Evaluation of GPM Ku-band Radar Reflectivity Observation and E3SM Simulation Using NEXRAD Network over the Continental US in warm seasons
The Ku-band radar onboard of Global Precipitation Measurement (GPM) core observatory features frequent revisit time and high vertical resolution. Its nearly global coverage and large swath width can help reveal the vertical and horizontal storm structure in midlatitudes, which were not observed by the radar on the Tropical Rainfall Measuring Mission (TRMM) satellite. We have applied the storm classification algorithm of Houze et al. (2015) to the GPM radar data. The classifications of Wide Convective Core (WCC) and Deep Wide Convective Cores (DWC) on are thought to be strongly related to Mesoscale Convective Systems (MCSs). However, neither the GPM radar reflectivity field nor this classification scheme have been previously evaluated using a ground-based radar network. We compare Cartesian gridded 3D NEXRAD observations obtained during the warm seasons of April-September 2014-2016 to GPM reflectivity and find that: (1) the 3D radar reflectivity values from GPM agree with those from NEXRAD observations (correlation greater than 0.55 and averaged ratio of 1.01); and (2) the GPM defined WCC and DWC can capture more than 70% of MCS events based on tracking features in the NEXRAD radar data over the conterminous U.S. Through this evaluation, GPM demonstrates a strong capability in MCS detection. We apply the detection algorithm to other areas of the globe without dense ground-based radar network and provide a global observational MCS database. In addition, we use the NEXRAD and GPM data to test the ability of the E3SM global climate model to see precipitation features identified in the GPM and NEXRAD data by averaging the finer scale observations (horizontal resolution of 5 km and 2 km respectively for the GPM and NEXRAD) to the model output grid (1°). Overall, the E3SM simulated radar reflectivity field demonstrates moderate overestimation of reflectivity below the 4 km altitude and severe underestimation in the upper level compared to the two observations. The reasons for such model behaviors will be presented.