Precipitation retrieval in tropical cyclones by means of TROPICS constellation and neural networks

This study focuses on precipitation retrievals over oceans in tropical cyclones by means of neural networks using data from the new NASA TROPICS (Time-Resolved Observations of Precipitation structure and storm Intensity with a Constellation of Smallsats) mission. Accurate monitoring of tropical cyclones is a major concern for both the scientific community and emergency management services due to the severe damage they cause, especially when they fall on population centers and surrounding areas. The TROPICS constellation consists of four CubeSats carrying on-board passive microwave radiometers providing high revisit time measurements over the Tropics with the aim to study in detail the structure and evolution of TCs during their lifecycle. A NN architecture for precipitation retrieval is developed and trained with data from the GPM (Global Precipitation Measurement) constellation providing reference value of precipitation and it is tested on an independent dataset constituted by TROPICS observations. An automatic spatial-temporal collocation procedure between TROPICS brightness temperatures and IMERG (Integrated Multi-satellitE Retrievals for GPM) data is performed in order to set up the training dataset. In this study the tropical storm Ida is considered as test case, and the preliminary results obtained are promising showing a R² between modeled (NN precipitation outputs) and reference target (IMERG precipitation product) above 0.8.