Author: Bill Johnston
“… weather stations were not set-up to measure trend and change, but to monitor local weather …”
Note [1], [2], etc refer to references at the bottom of this page.
As a retired scientist and former weather observer with a keen interest in data, I have been researching Australian weather station datasets for almost two decades. I am disillusioned by the quality of the Australian Bureau of Meteorology (BoM) network; suspicious of data they rely on to monitor Australia’s climate (the Australian Climate Observations Reference Network – Surface Air Temperature: ACORN-SAT), and mightily concerned about constant adjustments to datasets that universally seem to result in on-going warming.
The main problem for ACORN-SAT and its precursor high-quality datasets [1, 2], is that weather stations were not set-up to measure trend and change, but to monitor local weather, compare climates, and provide day-to-day information relevant to commerce.
While agriculture was an early beneficiary – crops to sow and limits to farming for instance, weather maps and short-term predictions were important to shipping and trade, and later for aircraft flying major air routes. Using data telegraphed by post and telegraph offices, Australia’s first weather map was published by NSW Astronomer Henry Chamberlain Russell in The Sydney Morning Herald on 5th February 1877.
The bolt-on experiment that became ACORN-SAT commenced around the time the IPCC was preparing the 1990 First Assessment Report. Chapter 8 of the Scientific Assessment lamented the lack of terrestrial temperature data and recommended “setting up a climate change detection panel to coordinate model experiments and data collection efforts” [3]. Consequently, Australian scientists got busy stitching data together and applying adjustments to iron-out resulting kinks – they homogenised the data.
Homogenisation refers to the process of adjusting for non-climate effects, while resulting homogenised data are assumed to solely reflect the long-term climate.
All Australian weather station sites have moved and changed in ways that impact data. Consequently, trends in raw data are as fraught as poorly adjusted data. Thus, the overarching question for any long-term temperature record is whether methods used to adjust for non-climate impacts are appropriate and unbiased. Appropriate requires that adjustments directly align with significant changepoints in data, while unbiased requires they are directly proportional to the effect.
The 112 ACORN-SAT sites used to monitor Australia’s climate.
Hide Full Screen
The 112 ACORN-SAT sites used to monitor Australia’s climate.
So, if independently of rainfall, mean maximum temperature (Tmax) at Victoria River Downs, Northern Territory (BoM ID 14825) stepped-up 1.12oC in 2013 (t(56)=6.28, p=<0.001), an unbiased adjustment would result in a sign-reversed proportionate change of -1.12oC [4]. While missing observations was a problem (particularly in 1973, which was ignored in subsequent analysis), and data before 1993 lacked precision, Victoria River Downs is one of the 112 ACORN-SAT sites used to monitor warming in Australia.
The overall Tmax raw-data trend of 0.126oC/decade, which was weakly significant (p = 0.039, R2adj = 0.057), was spuriously related to the 2013 step-change, highlighting that unless underlying inhomogeneities are accounted for, naïve approaches to determining trend can be highly misleading.
The ACORN-SAT Catalogue [5] mentions that while the site moved 250 m northwest in August 1987 and an automatic weather station (AWS) was installed on 9 May 1997, surroundings were “watered up until around 2007” – perhaps sporadic watering continued until 2013, or a 60-litre screen was installed or something else. The step-change was detected as highly significant by two entirely different statistical tests and verified using categorical multiple regression. Additional tests confirmed that data consisted of two non-trending segments disrupted by the step-change, with no residual trend attributable to CO2, coal mining, electricity generation or anything else.
Meanwhile, ACORN-SAT V.1 homogenisation (AcV.1 – to 2017) adjusted for a change in 1976, which was not detectable or verifiable, and an alleged move that made no difference to the data in 1987, leaving the 2013 step-change intact. AcV2.0 (to 2018), ignored the change in 1976, adjusted for the move in 1987, and in addition, for no stated reason made an adjustment in 1996 while also leaving the 2013 step-change intact. AcV2.3 (to 2021) made adjustments in 1968 (screen), 1974 (statistical), and 2007 (vegetation), but still no adjustment in 2013. Similarly, for AcV2.4 to 2.5 (to 2024), which made adjustments only in 1987 (move) and 1996.
While not resulting in significant trends, ACORN-SAT adjustments were inconsistent in their timing, and don’t align with the step-change in 2013. Changes in data they aimed to correct could not be detected, while confirmatory and post hoc analysis (Section 4 in [4]) found changepoints adjusted by ACORN-SAT were inappropriate for the data.
Victoria River Downs is by no means a special case. Inappropriate and disproportionate adjustments are a common feature of the thus-far 19 ACORN-SAT sites investigated and reported-on at www.bomwatch.com.au, and sites reported earlier at https://joannenova.com.au (Bourke, Port Hedland, Canberra and Sydney Observatory). An additional 66 weather stations in total, that have been used to homogenise ACORN-SAT sites have also been investigated using the same robust protocols as applied to data for Victoria River Downs.
Prior to the development of BomWatch protocols [6, 7], statistically sound methods for assessing the quality of data for individual weather stations and those that had been homogenised, were seriously lacking, which led to conflicting outcomes [2]. Further, station metadata – data about the data used to identify changepoints could not be considered reliable.
For instance, aerial photographs and a file held by the National Archives of Australia (NAA) show coordinates of the original met-enclosure at Cairns airport (BoM ID 31011), which was adjacent to the 1939 Aeradio office [8], were incorrectly specified in site-summary metadata. Metadata also ignored that the site moved to a 30m-by-30m mound near the centre of the airport in 1965, and due to building a new taxiway, to another mound before September 1983. Another site apparently also established around that time to the northwest, near the location of the current automatic weather station.
Ignoring previous moves, the ACORN-SAT catalogue simply states “the site moved 1.5 km northwest (to the other side of the runway)” in December 1992, which is both wrong and incomplete.
The Garbutt Instrument file at NAA show BoM commenced negotiations with the Royal Australian Air Force (RAAF) to move the site at Townsville airport in 1965, and that observations commenced at the new mounded site in 1970 [9]. Aerial photographs and RAAF airport plans showed the site had moved at least three times while on the eastern side of the runway, while subsequent to the 1970 move to the western side, it likely moved twice more before finally relocating to the current site in December 1994. Despite station files being in their possession and other material, including aerial photographs available in the public domain, the ACORN-SAT catalogue claims: “There are no documented moves until one of 200 m northeast on 8 December 1994, at which time an automatic weather station was installed”, which their own records would confirm is untrue.
A more subtle example of data corruption is that replacement of 230-litre Stevenson screens with 60-litre screens at individual sites is routinely ignored by site-summary metadata, including for the 66 comparator stations mentioned previously, each of which was studied in-depth. While it appears that temperature extremes have increased across the network over recent decades, much of the effect is due to the staged rollout of smaller screens, which respond more rapidly to changes in outside temperature, not to CO2 or a change in the climate.
Considering that datasets may be affected by multiple issues, development of a robust protocol required an objective, broad-brush approach.
Multiple attributes / year are derived from daily station datasets using the statistical program R, which is much faster than using a spreadsheet program such as Excel. Absolute extremes, data counts, means and medians, standard deviations, observations greater than 95th and less than 5th day-of-year percentiles, their Hi/Lo and log10 ratio, and indices of precision based on earlier work by Chris Gillham (www.waclimate.net). Frequency analysis of daily temperature and rainfall observations within classes, may provide additional insights.
As a reference frame, the First Law of thermodynamics predicts a linear relationship between mean maximum temperature (Tmax) and annual rainfall such that the drier it is the warmer it gets. Significance and goodness-of-fit – R2adj (adjusted for the number of terms and datapoints [10]) provide objective, comparable measures of data-fitness. Should relationships be not significant, or R2adj be less than 0.5 (<50% of variance in Tmax explained), something is wrong – data are no good (random to rainfall), or more typically relationships are contaminated by site moves or changes. For more detail, and what happens next see [11].
Commencing with Neville Nicholls and Simon Torok in 1990 with their HQ datasets [1], and most recently with Blair Trewin’s ACORN-SAT [12], aforementioned and recent www.bomwatch.au studies, and others that to date have not been reported, show unequivocally that BoM’s homogenisation methods create warming trends that are unrelated to the climate.
This is achieved by combinations of:
• Ignoring changes that happened; or alternatively, adjusting for site changes that made no difference to the data. Arbitrary application of changepoints and the lack of replicable, transparent and objective methods is the antithesis of the scientific method.
• Reliance on poor quality metadata, particularly with regard original locations of Stevenson screens, and when and where they moved to, without independent corroboration using aerial photographs.
• Adjusting for the presumed effect on daily maxima and minima of time of observation changes from 3 am to 9 am at airports and lighthouses; rounding differences due to metrication; and the rollout of 60-litre Stevenson screens, that are disproportionate to their effect at individual sites.
• Selected “neighbouring” sites, some of which are >1000 km away, whose first-differenced monthly values are highly correlated with those of the site to be homogenised likely embed parallel faults. Consequently, their use as reference series may reinforce rather than unbiasedly adjust faults in ACORN-SAT.
As the rollout of 60-litre screens is almost complete, and Google Earth Pro shows that transpiring vegetation at many unmanned sites has already been sprayed-out, graded around or scalped, data for individual weather stations are such that there is limited scope for instrument or methodology adjustments to continue to cool the past and warm current data in order to maintain the claimed trend. Thus the time is near when further adjustments would create grossly implausible data, which will be when the whole edifice must collapse. It is therefore surely time that in order to save reputations, and cease further damaging the science, the Bureau of Meteorology must abandon the ACORN-SAT project altogether.
Dr Bill Johnston
www.bomwatch.com.au
31 January 2025
References
[1]. Torok, S.J. and Nicholls, N. (1996). A historical annual temperature dataset for Australia. Aust. Met. Mag., 45, 251-260.
[2]. Della-Marta, P., Collins, D. and Braganza, K. (2004). Updating Australia’s high quality annual temperature dataset. Aust. Met. Mag. 53, 15-19[2].
[3]. https://www.ipcc.ch/site/assets/uploads/2018/03/ipcc_far_wg_I_full_report.pdf
[4]. https://www.bomwatch.com.au/wp-content/uploads/2024/02/VictoriaRiverDowns-16-Feb-2024-1.pdf
[5]. http://www.bom.gov.au/climate/data/acorn-sat/stations/#/14825
[6]. https://www.bomwatch.com.au/wp-content/uploads/2020/08/Are-AWS-any-good_Part1_FINAL-22August_prt.pdf
[7]. https://www.bomwatch.com.au/wp-content/uploads/2020/01/Methods-CaseStudy_-GladstoneRadar.pdf
[8]. https://www.bomwatch.com.au/climate-of-gbr-cairns/
[9]. https://www.bomwatch.com.au/climate-of-gbr-townsville/
[10]. https://stats.stackexchange.com/questions/48703/what-is-the-adjusted-r-squared-formula-in-lm-in-r-and-how-should-it-be-interpret
[11. https://www.bomwatch.com.au/wp-content/uploads/2021/02/BOM-Charleville-Paper-FINAL.pdf
[12]. Trewin, Blair (2012). Techniques involved in developing the Australian Climate Observations Reference Network – Surface Air Temperature (ACORN-SAT) dataset. (Technical Report: https://cawcr.gov.au/technical-reports/CTR_049.pdf )
Disclaimer
Unethical or poor-qualityscientific practices including the manipulation of data to support political narratives undermines trust in science. While we are not accusing the persons mentioned of unethical conduct, we are gravely concerned about their approach to data processing, use of poor data or their portrayal of data in their cited and referenceable publications as representing facts that are unsubstantiated, statistically questionable or not true. The debate is therefore a scientific one, not a personal one.
Biography
Dr. Bill Johnston is a former senior research scientist with the NSW Department of Natural Resources (abolished in April 2007); which in previous iterations included the Soil Conservation Service of NSW. With colleagues he undertook weather observations for about a decade from August 1971.
Bill’s main fields of interest have been agronomy, soil science, hydrology (catchment processes) and descriptive climatology and he has maintained a keen interest in the history of weather stations and climate data.
Bill gained a Bachelor of Science in Agriculture from the University of New England in 1970, Master of Science from Macquarie University in 1985 and Doctor of Philosophy from the University of Western Sydney in 2002 and he is a member of the Australian Meteorological and Oceanographic Society (AMOS).
Bill receives no grants or financial support or incentives from any source.
Related
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
