Predicting the expected number of U.S. lightning fatalities for a year or for a date within that year

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 5 January 2015
William P. Roeder, Private Meteorologist, Rockledge, FL
Manuscript (204.8 kB)

Handout (273.9 kB)

An update is presented for the statistical model that predicts the number of lightning fatalities in the U.S. for any near future year and the number expected by any day within that year. The previous statistical model was a best-fit curve of the declining rate of lightning deaths in the U.S. since 1940s. The model also included a percentile regression for the amount of lightning fatalities by day within the year based on data from 2006-2011. The updates add the lightning fatalities from 2012-2014 in both parameters, which is especially important for the fatalities-by-day-of-year prediction since that was 50% increase in sample size. In addition, a new logistic regression was performed on the various percentile regressions for a more robust overall result. Finally, 95% and 99% confidence intervals were added to the previous 90% confidence interval to allow additional levels of statistical significance in the hypothesis tests. The new statistical model predicts 22.5 lightning fatalities in the U.S. for 2015 with a 95% confidence interval of 19.6 to 60.9 fatalities. This median is significantly less than the running 30-year (1984-2013) running mean lightning of 51 fatalities. This shows the main problem of using a running 30-year mean for a phenomena with a significant trend during that period such as lightning—the running mean is not representative for the current year. While the prediction is more consistent with the running 10-year (2004-2013) running mean of 33 fatalities, the curve-fitting approach is still preferred since the 10-year running mean is very sensitive to any extreme events during that period. The new logistical regression indicates that the median of the lightning fatality year in the U.S. is 14 July. These results can be applied in lightning safety. The distribution of lightning fatalities during a year can be used to schedule lightning safety education more effectively, e.g. provide just-in-time training before a surge in lightning fatalities is expected. If the present lightning fatality total is above expectations, that can be used to motivate the public during education. Likewise, if lightning fatalities are far below expectations that can also be used to motivate the public by suggesting that lightning safety education is effective.