A View from David Talbot
Playing the Odds on Tornado Warnings
Pinpoint predictions are a long way off, but taking daily odds into account might help make the public more alert.
The devastation in Moore, Oklahoma, shows the limits of sensing, modeling, and warning technologies. While some technologies promise somewhat more accurate hurricane tracks and thus sharper evacuation orders (see “A Model for Hurricane Evacuation”), tornado warnings are another story altogether (see “The Limits of Tornado Predictions”).
Twisters can form, become more dangerous, and change direction in a matter of tens of seconds, wiping out one neighborhood but leaving another a quarter-mile away unscathed. And the conditions that form them often don’t exist for very long. Thus, the warnings necessarily cover larger geographic areas and time periods. Forecasters are wary of issuing too many warnings, lest they be seen as crying wolf and lose effectiveness.
But if micro-scale warnings are a long way off, a macro view might help. If you can’t get a street-level five-minute warning before the twister comes, at least you can know which days to be especially alert. This visualization put together by the National Oceanic and Atmospheric Administration shows, day by day and week by week, where the harsh weather is more likely to hit. And it shows just how bad late May can be in Oklahoma.
Be the leader your company needs. Implement ethical AI.
Join us at EmTech Digital 2019.