Evaluation of Near Real-Time Preliminary Tornado Damage Paths
AuthorKarstens, Christopher D.
Shourd, Kacie N.
Smith, Travis M.
Erickson, Somer A.
AltmetricsView Usage Statistics
The ability to preliminarily diagnose areas damaged by a tornado is examined using both a manual and an automated approach. The manual method consists of using Weather Surveillance Radar-1988 Doppler base data to track radar-indicated centroids of low-level rotation over the entirety of a tornado event. The automated method utilizes 0.2- and 3.6-km AGL azimuthal shear from the Multi-Radar Multi-Sensor system to link together a series of strong azimuthal shear clusters and create a proposed damage path polygon. The quality of each method's diagnosis is evaluated using traditional verification metrics derived from an object-based geospatial verification technique. These metrics indicate that the preliminary damage paths from the manual method better detect damaged areas with less false area denoted, compared to preliminary damage paths from the automated method. However, the preliminary damage paths from all methods fall short of detecting the entirety of damaged areas, and they also denote large areas in the damage path vicinity that were not affected. Potential avenues for future research and applications are described.