Staticulation on rainfall from the BOM
Dr David Jones, head of the Bureau of Meteorology's (BOM) National Climate Center, if accurately reported, relies on staticulation rather than evidence in remarks about drought and climate change:
"While climate scientists warn a single event such as Victoria's deadly bushfires cannot be blamed on climate change alone, Dr Jones said the conditions were 'totally typical of climate change at the most pessimistic end of the models'" Extract from The Age
It's worrying to hear insinuations like this from senior government scientists on the eve of the planned bushfire Royal Commission. Dr Jones seems to be misusing the BOM's own statistical reports to suggest an alarming conclusion that is not supported by the evidence.
The outlook for 'exceptionally low rainfall'—that is, drought in its usual meaning—in the BOM's own modeling for Southern Australia has a 'central tendency' of no change or slightly fewer droughts (see Section 5.2.2 quoted here). The modeled results show increased drought ('exceptionally low rainfall') only in the 'tail' of the distribution: below the 10th percentile of results.
What that means, exactly, is that increasing drought is completely untypical of any of the model results for Southern Australia because it occurs in fewer than one-tenth of all modeled outcomes. Statistics gives no warrant to characterize the 'typical' results of the tail. They're in the tail of the distribution precisely because they're a-typical.
Posted on 02/17 at 01:45 PM.
Your comments
IanC
David (not Eric) Jones was one of the 11 authors of the CSIRO/BoM Drought Exceptional Circumstances Report (DECR). So far as can be seen from the public record, he and the BoM played no part in the subsequent developments. There is no evidence that the CSIRO checked with BoM before initially denying David Stockwell access to the data supporting the modeling in the DECR, or before CSIRO subsequently reversed that stance. We don’t know whether David Jones and the other BoM authors were consulted about the simplification (dumbing down?) of the draft DECR in order to meet DAFF’s requirement for it to be accessible to a non-professional readership.
The use of the DECR in policymaking in Australia without due diligence has now become one of the examples cited in the study “Check the Numbers: The Case for Due Diligence in Policy Formation” by Bruce McCullough (Professor of Decision Sciences at Drexel University in Philadelphia) and Ross McKitrick (Associate Professor of Economics at the University of Guelph), which has just been published by Canada’s Fraser Institute (available at http://www.fraserinstitute.org/researchandpublications/publications/6511.aspx : the discussion of ‘Droughts in Australia’ is at pps. 25-26).
The review of drought assistance in Australia is proceeding in the light of the assessment in this and two other reports to Government. It is to be hoped that the critiques published on David Stockwell’s blog will now receive some further international attention from experts at Climate Audit.
Damn… Thank you for correcting that slip Ian, and my apologies to Dr Eric Jones (a former colleague) whose name somehow crept into my head while I was writing this.
And thanks for the link to the Fraser Institute document. It looks like an intriguing summary of a problem that I see happening quite a deal in Australian public health policy, too. The ‘noble-cause’ corruption problem that Aynsley Kellow has written about seems to me to be spreading to the management of ‘risk factors’, helped in part by the obscurity of our very paltry national health data. Not only is the national data (based on ‘self-reported’ health) of low quality, but some of the privately collected data (such as the 2001 and 2005 surveys by the modestly named “International Diabetes Institute”) are being held closely and selectively published.
The IDI’s data, for example, is being used to promote the idea that there is an ‘epidemic’ of obesity and type II diabetes in Australia. But they have not published basic statistics from their surveys, so we don’t know important facts about the distribution (of Body Mass Indices, diabetes) in the population such as the central tendency (they provide a statistic representing the change in the population mean but only age-stratified data for the rest) or, perhaps more important, changes in the variance of the distribution over time (which may be responsible for increased proportions in the upper tail of the distribution).