Analysis of Uncertainty in Location and Motion Information in Tornado Warnings

Advances in geospatial and atmospheric sciences allow forecasts tailored to specific people, locations, and situations to facilitate timely and effective protective behaviors. Such decision-support systems are helpful for users who require more than usual lead time to prepare and implement a protective plan. This research is part of an effort to develop a user-centric tornado geospatial awareness and decision-support system to assist with timing of evacuation, sheltering, and other response behaviors. We analyzed data from the Time...Motion...Location section (TML) provided in every tornado warning; these data have not previously been evaluated to quantify uncertainty and inform potential use in spatial decision-support systems.

Using data from 2008 to 2020, we compared the latitude/longitude locations provided in warnings to the corresponding tornado paths that occurred in or near each tornado warning polygon. Results showed the tornado latitude/longitude locations were within 5 miles of reported tornado paths about 50% of the time, and within 2 miles in about 25% of cases. We also paired warning and watch data using space-time matching to quantify uncertainty of the estimates of translational speed using the forecast average storm motion vector from tornado watches in comparison to those in the TML data. We found the translational speed estimates from warnings to be slightly slower, on average, than the forecasted movement from tornado watches, and within 5 knots in about 50% of cases. Results provide basic uncertainty estimates to inform development of decision heuristics for protective actions relative to location and movement information provided in warnings.