Discussion of particular topics

Determination of tornado or downburst intensity is a difficult
task, as it is mostly done using the damage occurring with a storm.
Only in very rare cases are there reliable measurements of windspeed.
A grading of intensity is done either using the Fujita-scale (F-scale)
or the TORRO-scale (T-scale), or using both classifications.
The scales reach from F-2 to F6, or T-4 to T13, respectively. But
only the grades F0 to F5 or T0 to T11 are really being applied.
The tornadoes with *negative* scale values are so weak (and
short-lived?!) that they do not cause any damage. For the existence
of F6 tornadoes (T12, T13), there is currently only little evidence.
And besides, there is an estimate of maximum tornado windspeeds
close to the F5 to F6 threshold, which comes from energy budget
calculations.

The above table (click to enlarge) gives an overview of the F- and
T-scale, the related wind speeds, and typical loss ratios S for
light (S-) and strong (S+) buildings in Central Europe. **The quantity
"loss ratio" is often applied in the insurance industry and denotes
the ratio of property damage to reinstatement value in percent.**
These values adapted for Central Europe were determined in cooperation
with
Munich Reinsurance
(cf.
Dotzek et al.,
2000,
in ).

The terms to coarsely classify tornado intensity in the table are also important: Weak (F0, F1), strong (F2, F3), and violent (F4, F5). Tornadoes with an intensity of F2 or greater are called significant, while tornadoes with negative F- or T-scale are named subcritical.

In order to determine which typical property-, building-, and vegetation damage occurs with the different classes of the F- or T-scale, one needs a verbal description of that damage. A simple adoption of the description valid in the USA is not very helpful in Central Europe with its much more solidly built homes compared to the USA. Information on F-scale damage analysis in the USA is extensively outlined in the current national weather service guidelines (NOAA-NWS, 2003)

Here we give a verbal desription of the TORRO- and Fujita-scales adapted for Central Europe (in ) as an HTM or PDF file. With this description, intensity of tornadoes and downbursts can be inferred. Also the loss ratios S are given for light (S-) and strong (S+) buildings (cf. Dotzek et al., 2000, in ).

An English-language description of the T-scale is available at TORRO. It is very similar to the verbal description of the F-scale, as given e.g. by Fujita (1981). Such descriptions of the Fujita-scale, more likely only valid for the USA can also be found at the NSSL and many other websites in the USA.

The TORRO and Fujita-scales displayed in the above table and verbally
described in a form adapted to Central Europe, already contained
typical values for the loss ratios S
(Dotzek et al.,
2000, in ).
This readily enables us to include building structure
representative for Central Europe when determining tornado or downburst
intensity. The following graph gives an descriptive overview of these
loss ratios S for **Central Europe**:

It is obvious that from T5/F2 up to T8/F4, the strongest increase in
damage occurs, but that on the other hand even an F5 tornado will not
destroy a well-constructed strong building in Europe completely (100%).
- **there is no better shelter from tornadoes or downbursts than a
well-constructed and well-built house!**

The following table shows another recent concept to consider the stability of building structures within the Fujita scale (Fujita, 1992): The socalled f-scale matrix.

Even though Fujita's f-scale concept appears promising, this decision matrix has not yet gained any widespread acceptance.

Forest damage was recently investigated by Hubrig (1999, 2002, in ). As the stability of e.g. trees is certainly much more uniform worldwide as that of buildings, a scientific overview of vegetation damage analysis is very important and desirable. In Europe since the 19th century, there has been a tradition to put emphasis on the assessment of forest damage occurring with winter cyclones or severe local storms like tornadoes and downbursts. A substantially expanded treatment of these matters has recently been published (Hubrig, 2004, in ).

Principal questions regarding building structure and vegetation when investigation tornado or downburst intensity are also addressed in the recent US national weather service guidelines (NOAA-NWS, 2003). The typical framehouse construction in the USA, however, is irrelevant in Europe.

The tornado intensity distribution p(F) or p(T) provides information on the percentage of weak (F0, F1), strong (F2, F3), or violent (F4, F5) tornadoes. Therefore, it is an extremely important quantity for risk assessment and climatological analysis alike! Using this distribution allows to determine the expected total number of tornadoes per year. Also, the risk or recurrence interval of tornadoes of a given intensity can be estimated.

**Only since the year 2001 do we have such an intensity distribution
for tornadoes in Germany from an analysis of the TorDACH data.**

Presently, 724 tornadoes (i.e. about 70.5% of all cases) have an F- or
T-rating: The histogram using the Fujita scale (red) or the
twice-as-fine TORRO scale (blue) reproduces a Weibull probability
density for tornadoes of F1 intensity or more. Only the weak F0
tornadoes make an exception here. For F larger than 1, the mean ratio
p(Fn+1)/p(Fn) is about 0.30. Like for the Mid-Western USA, this slope
implies a supercell-dominated intensity distribution.
The T scale ratings display a somewhat less homogeneous slope for
T-values larger than T2. Due to the doubled number of intensity
bins compared to the Fujita scale, *about* twice the number
of T-rated tornadoes would be necessary for a distribution just
as smooth as for the present F scale function.

Note that the Weibull distribution with F scale is found for American
tornadoes as well! So the common myth "Tornadoes in Europe are always
weak, US tornadoes are always strong or violent" is completely wrong.
The only difference relevant here is total number of events per year:
If about 0.2 % of all tornadoes are of F5 intensity, and the USA
experience about 1200 tornadoes a year, then *on average* the USA
have about two F5 events each year. In Germany, on the other hand,
where only about 20 to 30 tornadoes a year are
observed, it can take several centuries to get barely one F5 event.
However, F3 and F4 tornadoes should be expected every three years or
every few decades, respectively. It is just these long recurrence
intervals for violent tornadoes (F4, F5) what misleads some people
(even professionals in meteorology) to the false conclusion that
tornadoes in Europe are "always weak".

What's more, the US tornado intensity distribution in the 1950s looked very much like the one shown above in the range F1-F5, but with a low number of reported F0 tornadoes. Today, after 50 years of paying more attention to weak tornadoes, their number in the US distribution has substantially increased. Yet, over the same time period, no increase in number of strong or violent tornadoes was observed. It is therefore probable to expect that the same would be true in Europe - by now we still miss many of the very weak events. Either they are just overlooked or misclassified as non-tornadic storm damage. Therefore, we can expect future changes of the German tornado intensity distribution, which are outlined in the following figure:

Nevertheless, recent research work (Dotzek et al., 2005) suggests that an exact linear-logartihmic distribution cannot be the asymptotic case, for physical reasons. A certain curvature to the right (upward convex) of the histogram must remain in a lin-log frame of reference. Therefore, first of all a strong increase in reports of weak and subcritical tornadoes is to be expected, whereas the two F5 cases from Germany will most probably remain unchanged for a longer time.

Dotzek et al. (2003) investigated if there is a worldwide universal shape of tornado intensity distributions, and how this can adequately be modelled, e.g. in order to assess the risk of strong or violent tornadoes, to determine the total number of tornadoes per year, or the typical recurrence interval of tornadoes of a given F- or T-intensity.

Best-suited for this purpose, according to Dotzek et al. (2003), are the socalled Weibull functions, which are also applied to model ordinary wind speed distributions (also tornado path lengths and widths are Weibull-distributed in the USA).

After a fit to the German intensity data already shown, one obtains the smooth, right-curved, grey and black curves depicted in the figure above. The only difference between the two curves is that one fit only starts at F0 intensity (at which the first light damage occurs), while the other one sets in already at F-2, i.e. at v=0 m/s velocity. The latter therefore allows for an estimate of the subcritical tornadoes with negative F-scale, which are currently rarely observed.

In the range from F0 to F5, both fits are statistically about equivalent. But looking at the steep slope of the fit curves for the violent tornadoes, it becomes obvious how unlikely the (hypothetical) F6 tornadoes are in Germany. For the USA tornado data from the 1990s, however, a similar Weibull fit suggests that F6 tornadoes could be roughly a 10-year event in the USA ...