369242 Inferring Severe Convective Wind Gust Probabilities in Florida from NEXRAD Storm Structure Data

Wednesday, 15 January 2020
Hall B1 (Boston Convention and Exhibition Center)
Madeline R. Frank, Climate Forecast Applications Network, Atlanta, GA; and J. Miller and V. Toma

To support the real-time understanding of downed power lines in Florida caused by severe convective wind gusts, we investigate the relationship between severe convective wind gusts (> 50 knots) and NEXRAD-derived spatial and temporal storm structure data. Additional variables available from the Level III NEXRAD datasets are selected by implementing Regression- and tree-based methodologies. For a 5 x 5 km grid, available NEXRAD Level III data is integrated with surface wind gust data (SPC, METARS, SPECI, etc). Regression- and tree-based machine learning algorithms are developed to predict the probability of severe convective wind gust events for a given storm cell. These models are then verified using various evaluation metrics, such as receiver-operating-characteristic (ROC) curve, maximum critical success index (CSI), and Brier skill score (BSS).
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner