Tropical Cyclone Intensity Estimation via Passive Microwave Data Features

Satellite passive microwave imager data provides a unique opportunity to extract and compute structural and ancillary characteristics of a tropical cyclone (TC) at the time of a given image. Extracting these characteristic features from historical data and applying a machine learning tool to the resultant data set can produce an empirically-based algorithm to estimate intensity for current TC's. Previous research by NRL in this area, focusing on a relatively small, but varied, set of Atlantic basin TC's with associated reconnaissance data, provided positive results on the potential of such an approach (RMSE = 11.9 kt/9.8 mb). To continue to build upon this research effort, an expanded database of microwave-based structural features is being developed with the statistical relationship and significance of those features as it relates to current intensity estimation examined. While the analysis methodology will be similar to the previous research, updates in the dataset, software, and analysis techniques suggest an improved, robust algorithm development. Leveraging previous research into statistical methods and previous metrics related to intensity estimation via satellite images, an assortment of methodologies from different fields including statistics, image processing, and morphometrics are gathered for their potential in characterizing microwave imagery for TC intensity. Exploratory work has begun on characterizing TC structural asymmetries that the previous work has shown is important to estimating intensity. These new features include the wavenumber structure of the eye and convective location relative to shear, as well as storm tilt. Initial results with an analysis of the structural features used and intensity estimation performance will be presented.