Understanding Overland Satellite-Based Precipitation Errors in IMERG Products as a Function of Input Sources

Abstract The multisatellite precipitation product from the U.S. Global Precipitation Measurement (GPM) mission Science Team, the Integrated Multi-satellitE Retrievals for GPM (IMERG), is a widely used GPM product. It merges precipitation retrievals from passive microwave (PMW) and infrared (IR) sensors on board a collection of satellites operated by several agencies around the globe. IMERG retrospectively provides over two decades’ global precipitation data at fine temporal–spatial resolutions. This overland study evaluates the satellite-based precipitation statistics in the latest versions of the IMERG (V06B and V07B) products over the conterminous United States. High-resolution, ground-based radar precipitation estimates from a quality-controlled version of the Multi-Radar Multi-Sensor system—developed under the GPM Ground Validation effort—are used as the reference product. The relative performance of the two IMERG versions is evaluated using volumetric and categorical statistical metrics. The precipitation retrieval errors are further separated into three individual components: hit bias, missed-precipitation bias, and false-precipitation bias, and all are traced back to the sensor-specific sources, as well as different levels of the IR input usage. The evaluation highlights the clear improvement in the IMERG V07B precipitation product for all seasons, especially for winter, with reduced systematic bias and uncertainty and increased precipitation detectability in comparison with V06B. The changes in V07B include the use of input data processed with improved algorithms for both PMW and IR retrievals, the addition of PMW retrievals over frozen surfaces, and upgrades in the intercalibration that address sources of bias in V06B. Significance Statement This study conducts the first extensive evaluation of IMERG V07B against V06B at their native resolutions using ground-based radar data over the conterminous United States (CONUS) and traces errors in the IMERG products back to specific sensors and the use of infrared data. The results indicate that the changes in V07B clearly improved the performance of the IMERG estimates, with reduced systematic bias and uncertainty and improved precipitation detection.