Validating a Satellite-Only Ensemble Precipitation Dataset (STREAM-Sat) Against Multiple Radar Networks

Uncertainty characterization within satellite-based precipitation products (SPP) that properly account for the time/space error correlation structure across spatiotemporal scales has been described as a grand challenge facing the satellite precipitation community. Solving this challenge is important if satellite precipitation is to reach its full potential in a wide range of applications (e.g., flood forecasting, drought monitoring, water resources management, climate change impact assessment). Ensemble methods, in which precipitation uncertainty is reflected via multiple realizations of precipitation fields, have been widely used in other contexts such as numerical weather prediction and climate modeling, but rarely in satellite contexts. To fill in this significant gap, Peng et al. (2023) generated a novel near-real time global satellite-only ensemble precipitation dataset (STREAM-Sat).

In this study, probabilistic validation of STREAM-Sat ensemble quantitative precipitation estimation (QPE) will be conducted using three high quality ground-based weather radar networks over the U.S., Europe, and Australia as references. These networks are the U.S.-based Multi-Radar Multi-Sensor (MRMS) radar-based QPE, the European RADar CLIMatology (EURADCLIM), and the Australian Unified Radar Archive (AURA). This study promotes the use of gauge-corrected ground-based radar estimates of precipitation as references in various regions of the world to analyze the performance of satellite-only ensemble precipitation dataset over diverse climatologies. The GPM-era IMERG V07B is used to create STREAM-Sat. Probabilistic validation is conducted for the 2015-2020 time period. IMERG, STREAM-Sat and all ground-based reference data are resampled to multiple resolutions to determine how well STREAM-Sat is able to model the structure of rainfall error across arbitrary space/time averaging intervals. We present STREAM-Sat’s performance in terms of (i) uncertainty characterization and representation and (ii) probabilistic validation while conducting a comparative analysis of regional differences.

This study demonstrates the importance of uncertainty estimates for satellite precipitation and the ability of the ensemble-based STREAM-Sat in addressing this gap, thereby realizing the full potential of satellite products for various applications.