CISESS Internship Project in Collaboration with University of California, Irvine
Vesta Gorogh & CHRS
In collaboration with the Center for Hydrometeorology and Remote Sensing (CHRS), University of California–Irvine (a minority-serving institution–MSI), a PhD candidate, Vesta Gorooh, has completed CISESS’s 12-week internship program. To explore capabilities of fused satellite products for retrieval of precipitation rate, she investigated common and complementary information content of passive microwave and visible/infrared (VIS/IR) observations of precipitation processes. Under the supervision of CISESS Scientist Veljko Petkovic, she developed a Machine Learning (ML)-based model to optimize inputs from Low-Earth Orbit (LEO) satellites Passive Microwave (PMW) sensors, Geostationary Orbit (GEO) satellite Advanced Baseline Imager (ABI), and the Global Forecast System (GFS) for estimating instantaneous rainfall rates over the Eastern CONUS. The retrieved rain rates assessed against the current operational satellite and ground products showed improvements across all standard validation metrics. Using only raw information on the brightness temperatures and radiances from the GEO and LEO sensors, the new U-net model is capable of capturing features of precipitating systems at high accuracy and with improved spatial sampling (example in Figure 1).
Figure 1. July 2017 rainfall event. (left) U-net model retrieved rates; (middle) Ground MRMS reference; (right) NASA GMI GPROF precipitation product.
Gorogh presented these results to the lead scientists at NOAA, NASA and CHRS last week. Work on this promising model will continue through the continued collaboration between the UMD and UCI centers.