In this discussion contribution, Holmes states that in an analysis of Australian data (by the GPD/GEV method of Conditional Mean Exceedance) ''the range of c (shape parameter) was À0.25 to +0.31, with an average value of around À0.1. The latter value of shape factor gives reasonable predicted physical bounds to the gust speeds, and has been adopted in the new Australia/New Zealand Standard on Wind Actions (AS/NZS1170.2:2002).''
The writer is very surprised by this statement. Given the reported spread of values, it seems to him highly dubious to introduce an upper bound. Holmes quotes the reference by Hosking et al. [1], in which the authors, particularly in these circumstances, recommend that a significance test with the null hypothesis c ¼ 0 should be applied, before the values of c are believed. Were any significance tests applied to these Australian results?
The debate about Australian data in JWEIA goes back to the analysis of results for thunderstorm downbursts at Moree, NSW, by Holmes and Moriarty [2]. They performed a GPD(CME)/GEV analysis and obtained a shape parameter of c ¼ À0:161: In a discussion contribution, Cook and Harris [3] showed that it was equally plausible to fit the Moree data with a Weibull parent with index 1.47 and a rate parameter of 47.7/year. Subsequently Cook, Harris and Whiting [4] were able to analyse annotated records from Onslow and Brisbane, and in both cases establish a reasonably accurate count of the total number of events. Again the parent distributions of the thunderstorm data were good fits to Weibull parents with index and rate parameters of (1.44, 11.9) for Onslow and (1.29, 66.2) for Brisbane. The