Lake Effect Snow Warning Polygon Experiment: Verification

Currently, National Weather Service Offices, including WFO Buffalo, issue Lake Effect Snow Warnings in a zone-based format. However, unlike traditional winter storm systems, lake effect snow typically occurs on a localized scale. Starting in the 2015-2016 winter season, WFO Buffalo gained permission to experiment with creating polygon-based warnings for lake effect snow within the operational zone-based warning framework for events off of Lake Ontario. The office expanded the polygon-based warning experiment to events originating off both Lakes Erie and Ontario in winter 2016-2017, and continued the experiment into winter of 2017-2018. Polygonbased warnings are familiar to other severe weather warning operations within the National Weather Service, and offer potentially attractive benefits to lake effect snow warning applications, such as reduced false alarm area and more specific hazard timing information over zone-based warnings. With several seasons of events containing parallel zone-based and polygon-based Lake Effect Snow warnings, this study examined the verification of both warning methods to determine if the polygon-based warnings are more skillful than zone-based warnings. Verification for the LES polygons is more complex than standard zone-based warnings, as the polygons allow for temporal variation in the location of the high-impact lake effect snow, and can also be updated more frequently to reflect the most current forecast information available. A spatial verification scheme is presented here that allows for equitable scoring of zone-based and polygon-based warnings using common statistics, such as probability of detection (POD), false alarm rate (FAR), and critical success index (CSI). In addition, given the unique nature of the LES polygons compared to most other NWS warning products, statistics like spatial and population savings, and the reduction of unnecessary warning time were also examined. Spatial verification showed a statistically significant increase in skill for polygon-based warnings over legacy zone-based warnings. Population statistics showed that the number of people that were correctly removed from the zone-based warning was much greater than the number of people that were incorrectly removed from the zone-based warning. Finally, polygon-based warnings were found to reduce unnecessary warning time by approximately half over zone-based warnings. The verification statistics presented show that there is sufficient forecast skill and significant value in switching to a polygon-based warning scheme from the current zone-based warning scheme.