Using ground-based radar to investigate false detections in satellite precipitation
Dr. Tejas Gokhale, Department of Computer Science & Electrical Engineering.
Joey Mule, Department of Computer Science & Electrical Engineering.
Determining the causes of false detections from satellite-derived precipitation is of utmost importance due to the development and dependence of precipitation estimates in global water cycle modeling, weather forecasting, flood modeling, landslide hazard assessment, drought monitoring, watershed modeling, and climate modeling. Previous studies suggest that 30-50% of false detections from satellite-derived precipitation are due to virga since they are derived from cloud temperatures, emission-based retrievals, and/or scattering-based retrievals within the atmosphere and not at the surface. While virga occurs across the world, it might appear differently, depending on location and/or detection method. In arid climates, virga is caused by evaporation of precipitation falling from clouds high in the atmosphere. Whereas, in cold climates, virga is caused by sublimation of frozen precipitation falling from clouds. Often, this results in satellite sensors falsely detecting light precipitation even though the precipitation never reaches the surface.
The purpose of our study was to determine if false detections from the Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement (IMERG) (GPM) mission were due to virga. This was done using precipitation estimates from IMERG, surface precipitation observations from rain gauges, and vertical radar profiles (VRPs) from ground-based radars. We experimented with reflectivity data from VRPs from K-band, Ka-band, and S-band radars because they provide varying information about precipitation and cloud structure. The VRPs were from radars located at the U.S. Department of Energy’s Atmospheric Radiation Measurement (ARM) sites, the Next Generation Weather Radar (NEXRAD) sites, and the Global Precipitation Measurement (GPM) mission validation sites. Each radar had limitations in data availability and usability, which offered a chance to develop code for other users for easier processing of VRPs.
We analyzed VRPs to determine the causes of the false detections and to begin developing a virga events data set. Developing a virga events data set is necessary to train models to classify virga events using VRPs in future research. We also extracted information about cloud characteristics for each case to better understand if more general patterns exist for false detections. Our results were compared to precipitation estimates from IMERG to determine the proportion of virga versus other causes of false detection.