A Neural Network Approach To Retrieve Atmospheric Temperatures From Observations Of Constellation Of Smallsats.

TROPICS (Time-Resolved Observations of Precipitation structure and storm Intensity with a Constellation of Smallsats) is a NASA mission featuring six 3U CubeSat satellites. These CubeSats provide rapid microwave measurements across the tropics, allowing detailed observation of tropospheric thermodynamics and precipitation structures in storm systems at both mesoscale and synoptic scales throughout their lifecycle. In this study, we present a neural network approach to retrieve atmospheric temperature and moisture profiles from the TROPICS 12 channel microwave brightness temperature observations and other auxiliary information. The model was trained using atmospheric temperature and moisture profiles from the ERA5 reanalysis, among which 80% of data were used for training and 20% for test. Quantitative evaluation with independent data from a different period will be presented, demonstrating the promising performance of the machine-learning based retrieval approach.