Monday, 13 January 2020
Hall B1 (Boston Convention and Exhibition Center)
Mesoscale snowbands have been the subject of extensive research over the past few decades, owing in large part to their ability to impart significant societal impacts. We previously developed an automated snowband detection algorithm and applied it to evaluate object-oriented band predictability in the operational High-Resolution Rapid-Refresh (HRRR) model. Results demonstrated poor predictability, with low probabilities of detection despite high false alarm rates. Here, we extend this previous effort into high-resolution ensemble prediction, with an emphasis on maximizing the extraction of actionable hazard information from operational ensemble modeling systems through novel probabilistic visualization strategies.
Three snowband cases which were poorly or not predicted by the deterministic HRRR were simulated with a nine-member initial condition ensemble as proof-of-concept of the utility of probabilistic mesoscale snowband forecasting. Forecast member simulated reflectivities were compared to observational reflectivities to calculate a verification metric based on similarities in centroid location, aspect ratio, orientation angle, and area. These individual member metrics were then aggregated to evaluate the overall forecast skill of the ensemble. Finally, using a variety of ensemble visualization strategies, such as heat maps, paintball plots, and attribute blocks, we demonstrate how forecast value can be gleaned for low-probability, high-impact events.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner