Cyclone Type Classification Using JPSS Data and Machine Learning

The transition between cyclone types has associated changes in storm structure, and understanding and predicting tropical to extratropical (ET) transition is crucial for improving tropical cyclone (TC) forecasts and mitigating their impacts. Thus the National Hurricane Center (NHC), Central Pacific Hurricane Center (CPHC), and Joint Typhoon Warning Center (JTWC) official forecasts include the storm type (e.g., tropical, subtropical, extratropical). The impact of ET transition varies, but generally includes a weakening of the storm, increased rainfall and wind spread over a larger area, and associated changes in weather patterns. For example, Hurricane Sandy (2012) was transitioning to an ET cyclone as it approached the U.S. east coast with an expanding wind field that resulted in a very large storm surge.

The majority of global TCs occur in remote areas, making satellite data the primary source for TC analysis. The microwave sounders and imagers that are only available on low-earth-orbiting (LEO) satellites can see through clouds and provide TC intensity and structure information not available from Visible and IR imagery. Currently, there are three JPSS satellites carrying the Advanced Technology Microwave Sounder (ATMS), and several NOAA and MetOP satellites carrying the Advanced Microwave Sounding Unit (AMSU) instruments. Thus, despite much higher latencies of polar-orbiting data (typically 1 - 2 hours) as compared to geostationary satellite data (10-15 minutes for the current geostationary satellites), there is an increasing amount of information about TC structure that microwave sounders provide in near real-time. Further, instruments similar to ATMS are now becoming available on smallsats (e.g. TROPICS) that have the potential to further improve temporal and spatial coverage.

The JPSS Extratropical Transition (JET) is a newly developed JPSS-TC product that uses as input ATMS-MiRS (Microwave Integrated Retrieval System) temperature and moisture retrievals, two-dimensional wind and geopotential height fields from the CIRA Hurricane Intensity and Structure Algorithm (HISA), as well as VIIRS imagery and ancillary data for storm type diagnostics. Preliminary results indicate that JPSS data offers information independent of the Global Forecast System (GFS) model, suggesting potential improvements in storm diagnostics with JPSS input. JET features image displays of JPSS data and derived quantities, such as temperature anomalies and temperature advection, contributing to qualitative storm type diagnostics and situational awareness. JET also includes JETClass - a new storm type classification machine learning model that builds on the operational ETClass method, which uses a linear discrimination method with predictors from infrared imagery and GFS, and is part of the NHC’s Statistical Hurricane Intensity Prediction Scheme (SHIPS). In this presentation, we will discuss the use of JPSS data to improve operational storm type classification, including examples of JET image displays, ETClass verification, new JPSS predictors, and preliminary JETClass results. We will also discuss the possible path for transitioning JET to NHC, CPHC, and JTWC operations.

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.