ACORN-SAT routinely adjusts measured historical temperatures to correct for various equipment changes, site, and vegetation changes etc. Some of these changes are legitimate and some are not. A reduction in historical temperatures enhances any century scale warming. In the case of Darwin, historical temperatures have been lowered significantly causing Darwin’s warming rate over the last century to be nearly one full degree higher per century than it would have been if raw temperatures as measured were used.
I am interested here in two of the adjustments made by the Australian Bureau of Meteorology (BOM) to raw data for inclusion in the ACORN-SAT 2.2 Darwin maximum temperature reconstruction.
- The cooling of recorded and unadjusted historical temperatures prior to 1/1/37 due to the station change from the Darwin Post Office (DPO) (#014016) to Darwin Airport (DAP) (#014015) in 1942. This adjustment also included an estimate for any temperature change across the in-continuity from 1/1/37 to 1942. All temperatures prior to 1/1/37 were reduced by 1.02C based on only one reference station, Wyndham Port (WP) (#001005).
- The BOM reduced all maximum temperatures prior to 1/1/80 (including further reducing the maximum temperatures prior to 1/1/37) for statistical reasons based on the reference stations, Cape Don (#14008), Jabiru Airport (#14198), Oenpelli (#14042) and Katherine Council (#14902). All maximum temperatures from 1/1/80 back to the beginning of the reconstruction in 1910 were reduced by 0.67C.
For reference I have copied Jennifer Marohasy’s graph depicting Darwin’s ever changing temperature history, Figure 1, from her site,
I need to acknowledge Jennifer’s work in compiling this important data source.
Figure 1 – All changes over time to the Darwin maximum temperature series. Each subsequent change adds warming to the record. ADAM is the raw unadjusted data while the 4 ACORN-SAT datasets have all been heavily adjusted.
I’ve also copied below for your reference, a page from the Australian Bureau of Meteorology (BOM) website, Figure 2, that outlines all the changes to Darwin maximum temperatures including the two I am interested in.
Adjustments for ACORN-SAT version 2.2 for station number 14015
|Date applied||Temperature variable||Cause||Impact (°C)||Comparative stations|
|1/08/1995||Maximum||Screen||Annual +0.14||14272 14090 14142 14938 14198 14153 14042|
|1/08/1995||Minimum||Screen||Annual +0.03||14090 14272 14932|
|14213 14090 14932 14400|
|1/01/1980||Maximum||Statistical||Annual −0.67||14198 14042 14902 14008|
|1/02/1941||Minimum||Move (n)||Annual −0.83||14905 14626 1005|
|1/01/1937||Maximum||Site (vegetation)||Annual −1.02||1005|
Table 1 – All Acorn-Sat changes to Darwin temperatures and reference stations used for those changes.
1. THE ADJUSTMENT TO MAXIMUM DARWIN TEMPERATURES PRIOR TO 1/1/37
As mentioned, ACORN-SAT reduced all the raw, measured maximums prior to 1/1/37 for the Darwin record by 1.02C based on the reference station, Wyndham Port (WP). Some reduction is justified as there was a major cyclone in 1937 that altered vegetation near the station and a site move in 1942. The original site was also bombed in 1942.
The point I wish to make here is that using Wyndham Port as a reference station exaggerated the warming across the in-continuity in the homogenised Darwin record. The Wyndham Port, Darwin Post Office and Darwin Airport graphs of maximum temperature are copied below.
Figure 2 – Darwin Airport raw maximum temperatures. 1942 to 2018.
Figure 3 – Wyndham Port raw maximum temperatures.
Figure 4 – Darwin Post Office raw maximum temperatures to 1941.
The first thing to notice is that Wyndham port maximums warm by about 2.0C from 1925 through to 1955 while Darwin Post Office maximums are level from 1925 to 1936. Darwin Airport maximums are also level from 1942 to 1945 and only rise by about 0.4C from 1942 to 1955. This should immediately make us suspicious about using Wyndham Port as the only comparison station to adjust the Darwin temperatures for the station move, cyclone and Post Office bombing between 1/1/37 and 1942.
Acorn-Sat has justified the use of Wyndham Port maximums in this case by referring to their homogenisation program as described at http://www.bom.gov.au/climate/change/acorn-sat/documents/BRR-032.pdf and referred to here as Trewin, 2018.
By using this homogenization algorithm, they have transferred the anomalous warming in the Wyndham Port maximum record to the Darwin record.
I will show here why Wyndham Port (WP) should not have been used as a comparison station for Darwin.
- Trewin 2018, first chooses either the daily method or monthly method of comparing a reference station to the target station. The daily method is more complex and time consuming and is used when there is sufficient daily data. The monthly method is used when there is not enough daily data. I have chosen to use the monthly method here as it saves a lot of time and has no significant effect on the result. Trewin 2018, support this approach. See the quote below.
“This comparison showed that there was no evidence of any systematic differences between the daily and monthly methods. The mean adjustment for the daily method was –0.067 °C and for the monthly method –0.069 °C, whilst the mean absolute size of adjustments was 0.465 °C and 0.478 °C respectively”
- They then calculate the Pearson Correlation Co-efficient (PCC) for each of the 12 calendar months between the reference station and the target station over a period of 3 to 5 years (depending on data availability) either side of an in-continuity. 5 years, of course, being preferred. In this case Wyndham Port (WP) (The reference station) is compared to Darwin Post Office (DPO) for the 5 years, 1932 to 1936, and in a second calculation Wyndham Port is compared to Darwin Airport (DAP) for the 4 years, 1942 to 1945. These results are shown in Table, 2. Below.
Table 2 – The PCC (Pearson Correlation Co-efficient) for the three stations (WP v. DPO & DAP) over the years indicated is calculated for each individual month. These values are shown in the table along with their mean and median. The red values are the months with poor correlation
3. The next step is to decide how to treat these monthly PCCs. Trewin 2018, describes their method below.
“…the correlation index (for station comparison purposes) was determined as the median of those 12 monthly correlation values,…
In general, only stations with correlations of 0.6 or above were used as reference stations, but if fewer than three such reference stations were available, this limit was relaxed to 0.5. If there were still fewer than three available reference stations, the reference stations were weighted by inverse distance from the candidate station, rather than as per the procedure outlined above. (There was one case, at Darwin in 1937, where only one reference station was available; this adjustment was well supported by metadata.)”
Darwin is in a unique position in that it is the only station in Australia to have only one comparison station. The Wyndham Port inhomogeneity was, consequently, not picked up and was allowed to contaminate the Darwin record. The red PCCs in Table 2. are the poorly correlated months and it is in these months that the possibly spurious warming at Wyndham Port was transferred to the Darwin record.
Once a reference station has been cleared as suitable, its mean monthly temperatures either side of an in-continuity can be used to bridge that gap in the target station. The mean monthly temperatures for each of the 3 to 5 years either side of the in-continuity are then compared between the two stations and the target station is adjusted accordingly. In the case of the Darwin maximums for the in-continuity from 1936 to 1942, Wyndham Port was considered a suitable comparison station as the median of its monthly PCCs with Darwin Post Office and Darwin Airport were above 0.6 as shown in Table 2. Unfortunately, this method effectively hid any effect the poorly correlated months (red) would have had on the mean. The problem comes from using the mean to adjust the stations while using a different measure, the median, to decide if a comparison station is suitable.
On the evidence from the maximum data, Wyndham Port (WP) should not be used as a reference station for either Darwin Post Office (DPO) or Darwin Airport (DAP) as it clearly has strong warming from 1925 to 1955 that is not apparent in either Darwin record.
Another reason why Wyndham Port should not be considered as a reference station for Darwin is that its minimum temperatures from 1942 to 1950 are strongly negatively correlated with the Darwin Airport as is clearly seen in the two minimum graphs below. Darwin Airport minimums rise strongly from 1942 to 1950 while Wyndham Port minimums fall by nearly 2.0C during this same period. The temperature drivers in the two places appear to be unrelated for this period.
Figure 5 – Darwin Airport raw minimum temperatures.
Figure 6 – Wyndham Port raw minimum temperatures. Note the dramatic fall in temperature after 1942 compared to the rise in temperature at Darwin Airport for the same period.
Amazingly, Wyndham Port (#1005) was eligible to act as a reference station for Darwin Airport minimums for the in-continuity in 1941 (See Table 1) despite the strong negative correlation between the two after 1942. This almost certainly happened because they used monthly median PCCs to qualify the reference station. The fact that the mean temperatures are then used to bridge the in-continuity introduces a spurious warming or cooling to the target station.
Wyndham Port should have been disqualified as a reference station for the in-continuity in maximum temperatures in the Darwin record from 1937 to 1942. It was not disqualified because the homogenisation algorithm used is faulty. Acorn-Sat qualifies a reference station if the median monthly correlation co-efficient either side of an in-continuity is greater than 0.6 (or occasionally 0.5). The fact that they use the median and not the mean correlation co-efficient allows poorly correlated months to have limited or no effect on whether the reference station qualifies or not. In this way poorly correlated reference stations can qualify as a reference station, despite having large, incongruous temperature variation hidden in the months that are poorly correlated.
What a mouthful. Basically, Wyndham Port was used as a reference station for Darwin despite it having a large and obvious increase in temperature from 1925 to 1955 that does not appear in either Darwin record. This has occurred because the homogenisation algorithm used by Acorn-Sat is faulty as discussed above.
It is quite possible that this situation is common in the Acorn-Sat homogenised series in Australia and other similarly homogenised series around the world.
2. THE ADJUSTMENT TO MAXIMUM DARWIN TEMPERATURES PRIOR TO 1/1/80
Acorn-Sat reduced all Darwin maximum temperatures prior to 1/1/80 (including further reducing the maximum temperatures prior to 1/1/37) for statistical reasons based on the comparison stations, Cape Don (#14008), Jabiru Airport (#14198), Oenpelli (#14042) and Katherine Council (#14902) (see Table 1). All maximum temperatures from 1/1/80 back to the beginning of the reconstruction in 1910 were reduced by 0.67C.
This reduction occurred between Version 1 and Version 2 of the Acorn-Sat series (Figure 1) and was attributed to “statistical” changes (Table 1) by the BOM. Trewin, 2018, discusses all methodological changes from version 1 to version 2 and defines statistical changes as either, “removal of rounding bias”, “detection of date shifts”, or “Adjustment for negative diurnal ranges”.
None of these statistical changes apply to Darwin in any significant way.
- Rounding biases only apply when there is an even number of reference stations for an in-continuity. There was no in-continuity at Darwin in 1980. Also, the average adjustment is only 0.025C for any individual adjustment which is not significant in this case.
- Date shift changes are insignificant and only last a few months at most.
- Adjustments for negative diurnal ranges effect individual days and in Darwin’s case must be less than 0.3% of days if any (Trewin, 2018)
I am at a loss to know what possible statistical changes could reduce every maximum temperature from 1/1/80 back to the start of the record in 1910 by 0.67C. This is significant and not explained as far as I can see. What possible reason could Acorn-Sat have for changing a perfectly good data set by such a large amount many years after the original data was recorded.
The only useful thing I can see to do is graph the 4 Acorn-Sat reference stations with the Darwin Airport raw and adjusted data to see if the change makes any sense. This is done in Figure, 7 below.
All stations in this graph are equalised in 1980 for comparison purposes. You can see that the Darwin raw temperatures are a better match to the 4 reference stations than is the Darwin adjusted series indicating that the adjusted temperatures were less accurate than the raw Darwin series. You will notice that the Darwin adjusted series is significantly cooler than the other 5 series and that the raw Darwin series tracks the 4 reference stations well over 30 years.
Figure 7 – Darwin raw and Darwin adjusted are compared with the 4 reference stations used by Acorn-Sat to lower Darwin temperatures artificially in 1980.
There appears to be no valid reason to lower Darwin maximums by 0.67C for all years prior to 1980 all the way back to the start of the record in 1910. The reason given is “statistical” and is based on the 4 reference stations mentioned above. (Table1) & (Figure 7).
There is no statistical reason given that could explain this in the Trewin 2018 peer reviewed paper quoted by Acorn-Sat as their reference. The result of this adjustment is a significant increase in Darwin’s century scale warming trend. Unless someone can come up with a valid explanation, I would say that this adjustment was the result of either some sort of error or alternately was the result of politically motivated support for the global warming narrative.
Finally, just out of interest, that great and revealing email from climate-gate.
Figure 8 – Climate gate email from 2009.