Start
September 18, 2020 - 3:00 pm
End
September 18, 2020 - 4:00 pm
Categories
Convective Meteorology (Mesoscale Dynamics)Developing an Automated Tornado Warning Guidance System
Dylan Steinkruger
Penn State University
Friday, September 18th
3:00 pm
Join Google Meet:
https://meet.google.com/ksh-txvg-kni
Artificial intelligence (AI) continues to be leveraged in the detection and prediction of severe weather hazards. In nowcasting of severe weather, rapid data processing is necessary to generate informed products that are useful to end users. This characteristic is particularly important in the forecasting of tornadoes, where radar observations and model output are rapidly updating and forecasters must make rapid decisions to provide useful warning lead times. AI provides an opportunity to quickly process data from multiple sources and produce consistent guidance and interpretation. Here, we explore the utility of using AI to produce automated tornado warning guidance.
The first half of the presentation will focus on the development of the AI system in an ensemble of idealized numerical simulations. The highly idealized setup allowed for the basic framework of the AI system to be developed without issues related to data quality (i.e., a setup where the observations are essentially perfect). Capabilities of the AI system include issuing warnings at high temporal and spatial resolutions and tunable warning thresholds to optimize a user’s preferred performance metrics (e.g., detections, lead time, etc.). The second half of the presentation will explore more recent work that has applied the AI framework to real-life data. Preliminary results from this work will be presented and the feasibility of using AI to provide tornado warning guidance in real-time will be discussed.