56 A Bayesian Hierarchical Modeling Framework for Correcting Reporting Bias in the U.S. Tornado Database

Monday, 22 October 2018
Stowe & Atrium rooms (Stoweflake Mountain Resort )
Corey K. Potvin, OU/CIMMS and NOAA/OAR/NSSL, Norman, OK; and C. Broyles, P. S. Skinner, and H. E. Brooks

Handout (1.2 MB)

The Storm Prediction Center (SPC) tornado database is indispensable for assessing United States tornado risk and investigating tornado-climate connections. Maximizing the value of this database, however, requires accounting for systemically lower reported tornado counts in rural areas due to lack of observers. This study uses Bayesian hierarchical modeling to estimate tornado reporting rates and expected tornado counts over the central U.S. during 1975-2016. Our method addresses a serious solution non-uniqueness issue that may have affected previous studies. The adopted model explains 49 %, 62 %, and 82 % of the variance in out-of-sample reported counts at scales of 50 km, 100 km, and 200 km, respectively.

Population density explains much more of the variance in reported tornado counts than other examined geographical covariates, including distance from nearest city, terrain ruggedness index, and road density. The model estimates that only 42 % of tornadoes within the analysis domain were reported. The estimated tornado reporting rate decreases sharply away from population centers; for example, while > 90 % of tornadoes that occur within 5 km of a city with population >100,000 are reported, this rate decreases to ~ 70 % at distances of 20-25 km. The method is directly extendable to other events subject to under-reporting (e.g., severe hail and wind), and could be used to improve climate studies, tornado and other hazard models for forecasters, planners, and insurance/reinsurance companies, and development and verification of storm-scale prediction systems.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner