Extreme Wave Height Validation

The use of the NMIMET 3-parameter Weibull wave height distribution for the estimation of extreme events has been discussed in several publications [24] [26] [27], and comparisons given with estimates made by other means.

It is not of course possible to validate such estimates in absolute terms because one would need extremely long records of measurement (say 500 years to validate a 50 year wave estimate!). Apart from the impossibility of this, it would bring into question additional issues of long term variation in climate.

All such estimates must therefore be based on some sort of fitting of a probability distribution shape to the source wave height observations, followed by an extrapolation of this shape into the low probability 'tail' in order to arrive at the extreme event of interest. There are very many different ways of doing this. One can select different distribution shapes to fit, and one can select different ways of performing the fit. One can decide to use all the data in the population, or one can just fit the higher wave height data.

The NMIMET method is to use all the wave height data to fit a 3-parameter Weibull distribution using a moment matching technique.

To 'validate' the NMIMET method therefore we can only compare the estimates of extreme wave height predicted by it with those predicted by other fitting techniques based on other data sources for the same part of the ocean.

Figure 26 - Comparisons of H50 with Design Wave Height.

Figure 26 compares Hmax50 (the most probable maximum wave height in a 50 year period) derived from the NMIMET analysis with levels from chart contours of extreme height based largely on instrumental data, published by the UK Department of Energy [48]. The hatching defines ranges of contour level within the respective NMIMET catchment areas and it can be seen that the NMIMET estimates are generally inside or within 1 or 2 metres of these ranges. (Estimates based on raw visual observations are also shown, but these are mostly too high by widely varying amounts.)

(It should be noted that this comparison has been made using the NMIMET program, and using visual datasets specially selected to be close to the particular locations. If you attempt to repeat this validation using Global Wave Statistics Online the result will be somewhat different. You would have to use data from sea areas 9, 10, 11 and 17. And each of these would cover a much larger sea area than represented in the validation.)