Statistical Models for Tornado Climatology: Long and Short-Term Views

Area of Science:

• Atmospheric Science and Meteorology
• Statistical Modeling
• Geospatial Analysis

Background:

• Accurate estimation of regional tornado risk is crucial for public safety and disaster preparedness.
• Previous risk assessments may not fully account for historical data biases or dynamic climate influences.

Purpose of the Study:

• To estimate long-term and short-term regional tornado risk using statistical models and historical event data.
• To identify major U.S. tornado corridors and assess the impact of climate variability on tornado activity.
• To provide updated tornado risk insights for government agencies, insurers, and the public.

Main Methods:

• A spatial statistical model was applied to county-aggregated tornado counts, adjusting for observational changes over time.
• A space-time statistical model was used with raster-aggregated tornado counts, incorporating climate factors.
• Model validation involved predicting tornado rates by excluding the target year from the training data.

Main Results:

• Identified U.S. tornado corridors with expected annual rates exceeding two tornadoes per 10,000 sq km, with some areas in Texas and Kansas exceeding four.
• Analysis revealed a southward and eastward shift in high-risk areas when weaker tornadoes (EF0) were excluded.
• Short-term risk assessment showed shifts in tornado activity away from the Ohio Valley during El Niño and the Southeast during positive North Atlantic Oscillation conditions.

Conclusions:

• The developed statistical models offer state-of-the-art, dynamic views of regional tornado risk.
• Findings highlight the importance of considering both long-term trends and climate-driven short-term variations in tornado risk assessment.
• The results are valuable for informing policy, insurance decisions, and public awareness regarding tornado hazards.