1DLR Institute for Atmospheric Physics, Oberpfaffenhofen - Germany
2DWD German Weather Service, Offenbach - Germany
3NSSL National Severe Storms Laboratory, Norman OK - U.S.A.
The increased number of recorded tornadoes in Europe has yielded very similar tornado intensity distributions in most countries. These lin-log distributions for strong and violent tornadoes (F2 or higher) agree well with US tornado intensity distributions from the 1950s. Evidence is there that tornado wind speeds follow a universal probability distribution worldwide.
High observation efficiencies for tornadoes in the US during the 1990s revealed that most likely the complete range of tornado intensities is lin-log distributed, closely resembling the well-known Gutenberg-Richter intensity distribution of earthquakes.
However, other candidates for distribution functions of extreme events are, for instance the Gumbel distribution or simply the high-windspeed tail of Weibull distributions frequently used for "normal" wind speeds at a given site.
To determine the appropriate shape from the central portion of the observed distribution (which has a much better data base compared to the distribution's tails) would be extremely important for various reasons:
• the risk of strong and violent tornadoes (the right tail of the distribution) could be better estimated,
• the percentage of "missing", i.e. not observed weak tornadoes (the left tail) could be better estimated,
• as a consequence, the total number of tornadoes compared to the number of actually observed tornadoes could be derived.
Even if it is too early to determine the functional type of the tornado intensity distribution, based on the data available today, our analysis will allow for a discussion of the consequences of different distribution shapes for violent tornado risk and total number of tornadoes.