STAR Satellite Rainfall Estimates - Current Work

SCaMPR Modifications

The Enterprise Rain Rate (ERR) / Self-Calibrating Multivariate Precipitation Retrieval (SCaMPR; Kuligowski 2002; Kuligowski, 2020) has run in real-time for GOES-East at NESDIS / STAR since November 2004, and has run globally since 2021. The quality of the estimates has improved considerably during that time. A significant driver has been improvements in available infrared imaging capability (e.g., the transition from the GOES I-M 5-channel / 4-km IR Imager to the GOES-R series 16-channel, 2-km IR Advanced Baseline Imager). Additional algorithm improvements have included the addition of an adjustment for subcloud humidity, adjustment of the size of the calibration regions, and separation into different cloud regimes based on brightness temperature differences.

Efforts to validate the algorithm and address known shortcomings are ongoing; however, it is also recognized that the skill of the algorithm is limited by both the calibration data set used (passive microwave (PMW) estimates of rain rate) and the linear numerical methods (i.e., discriminant analysis and regression). To address these issues, collaboration with the University of Oklahoma (Upadhyaya et al. 2022) has yielded a version of the algorithm that is calibrated against gauge-adjusted Multi-Radar Multi-Sensor (MRMS) data using Convolutional Neural Networks (CNN). Significant improvement has been demonstrated over the CONUS, but it is not clear how well (or not) these relationships will generalize outside the CONUS. Additional efforts are being made to evaluate whether calibrating against Global Precipitation Mission (GPM) Dual-frequency Precipitation Radar (DPR) would provide improved results outside the CONUS.