2A.5 An AI-Based System for Objective Tropical Cyclone Intensity Estimation

Monday, 7 January 2019: 11:30 AM
North 124B (Phoenix Convention Center - West and North Buildings)
Jeffrey Miller, University of Alabama in Huntsville, Huntsville, AL; and M. Maskey, R. Ramachandran, I. Gurung, B. Freitag, D. Bollinger, R. Mestre, D. Silva, A. L. Molthan, C. R. Hain, and D. J. Cecil

The ability to accurately estimate a tropical cyclone’s intensity is essential for disaster preparedness and response. Direct measurements of a tropical cyclone’s intensity are sparse making satellite-based techniques the primary approach to estimating tropical cyclone intensity. The National Hurricane Center estimates a 10-20% uncertainty in post-analysis intensity estimates when only satellite data is available. Current estimation techniques utilized by operational forecasters, such as the Dvorak technique, suffer from human subjectivity sometimes resulting in completely different estimations from experts looking at the same image. This study presents the development of a deep learning model for objective tropical cyclone intensity estimation in real time using satellite imagery. Topics to be discussed include methodology and process of developing the model, along with an analysis of the model's performance compared to that of the state of the art, the development of an interactive portal for real-time and historical evaluation, and lessons learned throughout development.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner