Thursday, September 01, 2016

Homogenised temperatures, and planning for bushfires

How hot was it at Rutherglen on Black Friday?  Black Friday was that tragic day — 13 January 1939 — when a firestorm swept across Victoria destroying whole towns.

Just to the north, at Rutherglen, did maximum temperatures reach 47.6°C on that fateful day — or only 42.2°C?  Experts will explain that when considering the likely impact of future bushfires, it is important to consider the historical temperature record — but which record?  The Australian Bureau of Meteorology keeps two records:  there is the official ACORN-SAT record used to report on climate change each year and easily accessible at their website through an interactive map;  and then there are the raw values in a back cupboard — metaphorically speaking.

The difference between the official-adjusted maximum temperature for Rutherglen on 13th January 1939 versus the actual measured value is rather large — more than 5°C.  Historical temperature data is used to model and forecast the likely impact of future bushfires, with Fire Danger Indices sensitive to small changes in temperature.  But which values should be inputted when modelling bushfire behaviour:  the raw data, or the improved-adjusted ACORN-SAT values?

According to the Bureau adjustments to the historical temperature record do "improve it" and are based on a, "Percentile-matching (PM) algorithm, which applies differing adjustments to daily data depending on their position in the frequency distribution.  This method is intended to produce data sets that are homogeneous for higher-order statistical properties, such as variance and the frequency of extremes, as well as for mean values."

Australian climate scientists generally prefer to work with the adjusted values as passed along by the ACORN-SAT unit in the Bureau, and increasingly refer to these adjusted values as "observations" (i.e. Sophie Lewis and David Karoly, Bulletin of American Meteorological Society, Volume 95).  But this is a travesty:  the adjusted values may be quite different from the actual recorded values — the real observations.  It is the real values that should be used for contingency planning in the real world — including for bushfire preparedness.

Of the temperature series from the 112 weather stations used to calculate year-on-year temperature increases in Australia, 109 have been adjusted by the Bureau's ACORN-SAT team — many by rather a large amount.  In the case of Rutherglen, the adjustments are dramatic, not only to the daily temperature maxima, but also the minima.  These adjustments are then incorporated into international datasets, in particular HadCRUT that informs IPCC deliberations.

This was pointed this out to News Limited journalist Graham Lloyd over two years ago:  that while the raw mean-annual minimum temperatures for Rutherglen showed a slight cooling trend of 0.35°C per century, after the ACORN-SAT adjustments, there is statistically significant warming of 1.73°C per century.  This warming has essentially been achieved by the ACORN-SAT team of just 2.5 persons, by progressively dropping down the measured raw values from January 1974 back to the beginning of the record, which is November 1912.

The Bureau attempted to fob Lloyd off explaining that the adjustments had been made at Rutherglen because the weather station had been moved:  as though a move between two paddocks in a relatively flat terrain should change the temperature trend so significantly! Regardless, this advice directly contradicts what is written in black-and-white in the Bureau's official weather-station catalogue:  the catalogue states "There have been no documented site moves" at Rutherglen.  When Lloyd queried this, rather than admit that they were-in-error, the Bureau advised Lloyd they would be assigning someone to find supporting evidence.  But, they would need more time, because the evidence was surely in an archival box:  except no-one knew which one.

When Lloyd reported in the Australian newspaper that the Bureau was now in search of evidence for a site move Ibid — information which, if found, would contradict the official catalogue — I thought there might be some public outcry.  But I was told by those less surprised by the apparent apathy that this is simply another example of the bureaucracy taking a revisionist approach to history.

I've since discovered that a "revisionist history" is not the same as "history written by the victors".  What I am referring to as regards the Bureau, is the rewriting of even benign events — and things as apparently politically irrelevant as an historical temperature series.

After much rummaging the Bureau came to the following conclusion regarding sites moves at Rutherglen:  "No document has been located which states explicitly that the observation site moved.  However, there are a number of documents from 1958 or earlier which make references to the site which are not consistent with it being in its current location, indicating that the site moved on one or more occasions at some point between 1958 and 1975.  There are also additional documents which indicate a strong likelihood of a move or other changes."

But if all the adjustments to the values as actually measured at Rutherglen correspond with breakpoints which were caused by site moves (as claimed by the Bureau), there should be evidence for a total of six "site moves"!  Because there are a total of six adjustments listed in the two advisories that have been published by the Bureau:  the first in August 2014, and the second in September 2014.

To be clear, when I run the unadjusted raw temperature data as measured at Rutherglen (the observational data) through standard QA software — as detailed in this submission to the Australian Auditor General — I can't find any breakpoints, let alone six.  But there should be six, if the Bureau's explanations are to be consistent with their improved ACORN-SAT remodelled temperature data for Rutherglen as available at their website via an interactive map.

Most peculiarly, according to the Bureau's own policy, in accordance with world's best practice as recommended by the World Meteorological Organisation, any significant site relocation should result in a new site number.  Yet Rutherglen has always been referred to by a single station number:  082039.  This is in contrast, for example, to the nearby location of Deniliquin that was number 074128 when the weather station was at the post office, and number 074258 when it was moved to the airport.

Putting all of this aside, is it reasonable to assume that moving a weather station between paddocks will create a difference of 5.4°C for 13 January 1939 between the measured and the adjusted values?  No, it is not.

Back in 2014, then Prime Minister Tony Abbott was right to suggest an inquiry into the Bureau of Meteorology and this revisionist approach to the historical temperature record.  Greg Hunt, Environment Minister at the time, apparently "killed" the idea during these discussions in cabinet claiming confidence in the Bureau by the Australian public was paramount — especially so we (the public) would heed bushfire warning.  But how reliable are these warning — and exactly which historical record are they based on?


Postscript

More information on temperature trends at Rutherglen is detailed in Temperature change at Rutherglen in south-east Australia, New Climate (2016).

The difference of 5.4°C between the adjusted and raw values on 13 January 1939 can be verified by scrutinising the ACORN-SAT versus CDO/raw daily data for Rutherglen available at the Bureau of Meteorology website online.  Specifically the ACORN-SAT TMax for Rutherglen versus the raw TMax for Rutherglen — scroll to 13 January 1939.

No comments: