Using AI-based Synthetic Passive Microwave Data to Increase Understanding of Tropical Cyclone Structure and Evolution

Satellite observations are a key tool for observing tropical cyclone formation, evolution, structure, and behavior; and calculating tropical cyclone position and intensity. While geostationary satellites provide visible and infrared observations of tropical cyclones with broad spatial coverage and high temporal frequency, these instruments are largely limited to observing features at cloud top. Passive microwave imagery, on the other hand, can see through clouds to provide information about tropical cyclone structure and evolution at various vertical levels; yet passive microwave instruments provide at most a few observations per day due to their low-Earth orbits. In recent years, a team at CIRA and NESDIS has used machine learning models to construct a synthetic passive microwave (SPM) dataset that is trained on the full suite of visible and infrared geostationary satellite imagery to produce simulated 89 GHz and 37 GHz passive microwave imagery at high temporal frequency. The thousands of simulated microwave images in this dataset are used to explore tropical cyclone structure and evolution at temporal resolution as high as every 20 minutes. Through principal component analysis of this dataset, we will discuss how the synthetic passive microwave images may provide additional information about tropical cyclone structure and evolution beyond what is available in infrared imagery. We further discuss the various ways this machine learning-generated dataset can be used to better understand physical processes of tropical cyclone structure, variability, and evolution around the globe.

Disclaimer: The scientific results and conclusions, as well as any views or opinions expressed herein, are those of the author(s) and do not necessarily reflect those of NOAA or the Department of Commerce.