Statistical modeling of tornado intensity distributions

We address the issue to determine an appropriate general functional shape of observed tornado intensity distributions. Recently, it was suggested that in the limit of long and large tornado records, exponential distributions over all positive Fujita or TORRO scale classes would result. Yet, our analysis shows that even for large databases observations contradict the validity of exponential distributions for weak (F0) and violent (F5) tornadoes. We show that observed tornado intensities can be much better described by Weibull distributions, for which an exponential remains a special case. Weibull fits in either v or F scale reproduce the observations significantly better than exponentials. In addition, we suggest to apply the original definition of negative intensity scales down to F-2 and T-4 (corresponding to v=0 m s−1) at least for climatological analyses. Weibull distributions allow for an improved risk assessment of violent tornadoes up to F6, and better estimates of total tornado occurrence, degree of underreporting and existence of subcritical tornadic circulations below damaging intensity. Therefore, our results are relevant for climatologists and risk assessment managers alike.