Rainfall Estimation From TEMPEST-D CubeSat Observations: A Machine-Learning Approach

In this study, a machine-learning model was used to produce surface rainfall estimates from Temporal Experiment for Storms and Tropical Systems – Demonstration (TEMPEST-D) microwave radiance observations CubeSat. The is based on an artificial neural network (ANN). space-borne TEMPEST-D sensor performed brightness temperature (TB) at five frequencies (i.e., 87, 164, 174, 178, 181 GHz) during its nearly three-year mission. TBs were as inputs, the multiradar/multisensor system (MRMS) radar-only quantitative precipitation estimation product ground truth train ANN model. A total of 19 storms identified that simultaneously observed by weather radar over contiguous United States. training dataset 14 storm cases. other cases, consisting three continental two land-falling hurricanes, independent testing. spatial alignment algorithm developed align with measurement storm. This study showed captured features well current-generation satellite sensors, such global mission imager. results demonstrated estimated matches MRMS in terms intensity, area, pattern. average structural similarity index measure score test cases 0.78.