DARWIN TEMPERATURES: UNSCRAMBLING THE ACORN SAT 2 SERIES
By Bob Irvine
There have been a number of adjustments made to the Darwin temperature record in recent years that have significantly increased its trend from 1910 to 2016. This trend increase is largely due to the lowering of historical temperatures.
The BOM has achieved this significant lowering of historical Darwin temperatures by, in my opinion, using badly correlated comparison stations and doubtful statistical methods.
Grounds for being suspicious of the Australian BOMs methods are outlined in the discussion below.
https://wattsupwiththat.com/2019/04/02/darwin-temperatures-what-is-going-on/
The Darwin station history and adjustment justifications can be accessed at the link below.
http://www.bom.gov.au/climate/data/acorn-sat/stations/#/14015
Raw Darwin temperatures have been changed 6 times before inclusion in the Acorn Sat 2 series. This post will not deal with the three most recent changes in 1990 and 1995. The earlier three adjustments were on 1/1/37, 1/2/41 and 1/1/80. Each of these adjustments lowered all previous readings back to the beginning of the record by the amount specified. In total the Darwin mean temperature prior to 1937 was lowered by approximately 1.26C by these three adjustments.
Each of these three adjustments will be dealt with separately starting with the earliest in 1937.
1/01/1937 MAXIMUM TEMPERATURE ADJUSTMENT
All maximum recordings before 1/1/37 at Darwin P.O. (014016) were lowered by 1.02C on this date due to changes in the local vegetation. These changes were made by the BOM based on maximum temperatures taken from the comparison station Wyndham Port (001005).
For the 10 years from 1927 to 1936 the correlation between Darwin PO and Wyndham Port is 0.39 which disqualifies Wyndham Port as a comparison station. Wyndham Port also has very poor or no correlation with Darwin Airport from 1942 to 1945 (correlation 0.03). The BOM should not have used Wyndham Port as a comparison station.
The problem then becomes, how do we change the Darwin PO maximum record to more accurately reflect the changes in vegetation after 1937 and the station move in 1941. The best and most obvious available data set to do this is the Darwin PO minimum record. Vegetation does not generally affect minimum temperatures as there is no shading problem at night.
The BOM have averaged the 1937 to 1941 maximums to get a figure of 31.54C. This they have subtracted from the 1927 to 1936 average of 32.56C to justify an adjustment of 1.02C. This process is not legitimate for three reasons. Firstly, the record from 1937 to 1941 is compromised by the vegetation issue and should be discarded. Secondly, they have used the non- correlated Wyndham Port series as justification. And thirdly, 1939, 1940 and 1941 were colder than average and have biased the 5-year 1937 to 1941 average down. This is confirmed by the nearby Port Keats Police Station (014905) maximum record. Port Keats is the only series nominated as a comparison series by the BOM that covers the period 1939 to 1041.
All Australian raw station records can be accessed by station number at the following link;
http://www.bom.gov.au/climate/data/
A more accurate way of quantifying the adjustment is to use the difference between the Darwin PO maximum temperature average from 1927 to 1936 and the average from 1942 to 1952 in the Darwin Airport record. This method, incidentally, has the advantage of including any changes to maximum temperatures that are needed to allow for the 1941 station move to Darwin Airport.
The adjustment then becomes the Darwin PO maximum average from 1927 to 1936 of 32.56C minus the Darwin Airport maximum average of 31.69.
By these methods, the adjustment should have been 0.87C instead of the BOMs figure of 1.02C.
1/02/1941 MINIMUM TEMPERATURE ADJUSTMENT
All minimum temperatures prior to 1/2/41 were lowered by the BOM by 0.83C at Darwin PO to allow for the station move to Darwin Airport. The BOM used three comparison stations to justify this adjustment.
These stations were Port Keats (014905), Daly Waters (014626) and Wyndham Port (001005).
Fig. 4, shows Darwin PO minimums. The other four relevant minimum temperature graphs are printed below.
Daly Waters has no record for the first four months of 1939 which affect correlation as this can change with the seasons. This leaves only one year of overlap with Darwin PO (1940). It also has a correlation of only 0.49 with Darwin Airport from 1942 to 1945. For these reasons the Daly Waters data has also been discarded.
Port Keats minimums, on the other hand, have a correlation of 0.62 with Darwin Airport from 1942 to 1945. Port Keats also has two years of overlap with Darwin PO that has a good correlation.
If Port Keats is used as a bridging station, then the adjustment between Darwin PO and Darwin Airport is 0.67C. This is arrived at by using the common 1939 and 1940 data with Darwin PO and the common data between 1942 and 1952 with Darwin Airport.
Darwin PO minimums from 1931 to 1940 average 23.7. Darwin Airport minimums from 1942 to 1952 average 22.94. This method gives an adjustment for minimum temperatures of 0.76 as an allowance for the station move. This is in reasonable agreement with the adjustment derived using the Port Keats data.
Taking these two methods together we arrive at a reasonable adjustment to Darwin minimum temperatures of about 0.7C. This should give a reasonable approximation of the change in minimum temperatures caused by the move from Darwin PO to Darwin Airport.
1/01/1980 MAXIMUM TEMPERATURE ADJUSTMENT
All maximum temperatures prior to 1/1/80 were lowered by 0.67C by the BOM. Their stated reason for doing this was “statistical”. They quoted four comparison stations as justification for this change.
They were Jabiru Airport (014198), Oenpelli (014042), Katherine Council (014902) and Cape Don (014008). These maximum temperature records are printed below and should be compared to the raw Darwin Airport maximums in Fig. 4.
I see no reason why all maximum temperature readings going all the way back to the beginning of the 20th century for Darwin should be lowered by 0.67C based on these comparison stations.
Unless someone can think of a reason for this adjustment, I think it is reasonable to simply ignore this 0.67C. The direct measurements taken at Darwin Airport are obviously more accurate than this type of speculation.
CONCLUSION
If the methods outlined in this post were used, then these three adjustments would lower the Darwin mean temperatures prior to 1/1/37 by about 0.79C. This is consistent with a total increase in Darwin temperature of approximately 0.85C for the period 1910 to 2016 when combined with the other three more recent BOM changes.
If the BOMs methods are used, then these three adjustments would lower Darwin mean temperatures by 1.26C. This increase is an outlier when compared to other global datasets and has been arrived at largely by using badly correlated comparison stations as well as the doubtful statistical adjustment mentioned above.
A perceptive article,Bob.
The familiar graph showing correlation coefficients versus separation distance for pairs of stations is at the core of the nonsense. Two straight lines can give 100% correlation as everyone knows, so the coefficient includes effects unrelated to climate. I have been looking at a geostatistical determination of the separation at which one station fails to forecast for another. It is not fully definitive but it urges caution for stations more than 300 km apart. There is also the matter of different climate regimes at distant stations. I was involved with early days at the mentioned Jabiru site 256 km by road from Darwin. The wet season storms had different paatterns at Jabiru vs Darwin, if that means anything about correlation coefficients.
The question is why the BOM has not looked at better criteria than correlation coefficients. Geoff
Good comment Geoff. This is typical of bureaucrats sitting in offices playing with statistics without out any local knowledge or ground reality. They seem to think they can study climate ( which is the use they are trying make for these poor quality data ) by ignoring climate and climate differences and just looking at it a s a bucket of numbers like stock exchange data.
Both Wyndham and Daly are well away from the coast so any expectation that they will reflect either the max or min of coastal Darwin seems pretty far fetched. BOM have consistently been anything but transparent about their data manipulations probably , like Phil Jones at UAE, they fear anyone “finding
something wrong” with their work.
That is a pretty silly statement. Taking a broken snippet from the beginning of a record. The correlation argument, if it has any value at all, is intended to be applied to the full common period of the data, not just any arbitrary snippet of arbitrary length.
Vegetation is considerably cooler than bare earth. Changes in vegetation are not confined to whether there is a tree in front of the thermometer. Why are you talking about “shade” ?
To be honest, at this stage I stopped reading. I think BOM have been rigging the record for the last few decades but to challenge that requires objective, informed analysis. I lost hope of this article providing it.
Greg
Your first objection is not correct. Here is the quote from Trewin 18 that describes the BOM method. The 4 years from 1942 to 1945 is valid.
“The first reference period was normally the five calendar years before the breakpoint and the second reference period the five calendar years after the breakpoint (for example, for a breakpoint in April1986, the first reference period was normally 1981 to 1985, and the second reference period 1987 to1991). A shorter reference period (with a minimum of three years) was used where there was another breakpoint within what would normally be the reference period.”
Your second objection is simply nit picking. The Wyndham Port data was compromised and should not have been used. It was 600km away and didn’t correlate with the Darwin data. The BOM needs to explain why they used it. The Darwin PO Max. data for 1937 – 1941 was obviously compromised by the vegetation issue. The BOM needs to explain why they used it. There were no other comparison stations for this period so the only other data available to correct the faulty Darwin PO Max. temp was the Min. temp. Unless you can think of a better way, this the best way to bridge the Max. temp. gap between 1936 and 1942. The vegetation issue appears to have made no difference to the Min. temps. although it is impossible to be sure.
Thanks for the reply Bob.
I was questioning the period over which the correlation was calculated. You are talking about the overlap. What does their method say about calculating the correlation. I don’t see any sense or validity in calculating correlation of 3 or 4 data points, it meaningless.
That is pretty much what I said just above. Comparing in-land to coastal will obviously be incompatible.
The thing is, if there is not a satisfactory way you don’t do it! There are two records for Darwin from different sites. If you can not find a credible method to combine them you remain with two separate records from two different sites.
The only value in combining is to have a long term “trend”. If grafting the two together is based on reading tea-leaves or other ad hoc tricks, the long term change is complete fiction, not “data”. You can not make data out of wombat shit, no matter how hard you want some.
Greg,
Different seasons affect correlation differently, particularly in the tropics. The yearly temperatures are not recorded by the BOM unless they have all 12 months recorded. The 4 data points you refer to are really 48 data points that cover all seasons equally. This is why the BOM considers a minimum of three years to be valid.
Australia’s BoM – the world’s worst at Kriging.
And don’t you just love arbitrary corrections to raw data.
Unless someone can think of a reason for this adjustment,
Yep,
IPCC AR6 WG1 chapters are in progress. i.e. the need to further adjust data to keep up with alarmist hypothesis/model outputs.
Nothing else in the climate hustle matters if that fails.
These adjustments must be all pre-1979 because it looks good in the post 1979 period when compared with satellite data (following up on the inspirational work by Dr. Roy Spencer)
https://tambonthongchai.com/2019/04/07/bomuah/
I think the inspiration belongs to John Christy, Spencer’s boss.
“Who controls the past, controls the future: who controls the present controls the past.” (George Orwell,1984)
So very true!!!!!!!
Table 1 should have an easy-to-see statement of date range that the adjustments apply to.
The changes are quite large. I suspect the “statistical” means to be in keeping with the theory. Manipulation of data to try to validate a theory is scientific FRAUD. Any science that has a foundation using manipulated and thus fraudulent data is flawed and of little or no value whatsoever. All researchers that have used the manipulated data need to completely review their work and redact those tainted papers.
The theory is only believed due to intimidation-if you do not come to the conclusions that those that funded the study want, you won’t get any more funding. There is no money in the truth.
Donald
“Each of these adjustments lowered all previous readings back to the beginning of the record by the amount specified.”
The problem lies in using adjustments to try and pretend the 1910 station is identical to the 2016 station. This is a nonsense as you can never hope to do this correctly, there are too many unknowns.
As such, you cannot draw a meaningful trend line through the data points, unless you includes error bars.
Which you can calculate by way of the central limit theorem by sampling the raw data.
Once you adjust the data however I’m afraid what you have is a pigs breakfast. Nothing anyone else would want to consume. Except maybe a climate scientist. They will believe all sorts of hogwash.
This. I can’t see how you can make these adjustments and pretend the error bars (whether with the original or the adjusted data) aren’t very large.
Darwin is not the only station to show this arbitrary tinkering. Alice Springs also in the Territory is another location with dodgy adjustments. There are many many others. See https://kenskingdom.wordpress.com/ and http://www.waclimate.net/acorn2/index.html.
Joel O’Bryan above is right- it’s all about AR6.
FWIW, the HadCRUT3 record, UK Met, shows no increasing trend since 1890, in fact decreasing recently:
https://datamarket.com/data/set/1loo/average-monthly-temperatures-across-the-world#!ds=1loo!1n6s=fu&display=line
Johanus, thanks for the signpost to the datamarket.com site. There is a notification on the site which reads “After April 15th, DataMarket.com will no longer be available.” Does anyone know why that is? Is it significant/sinister that such an easy source of a mass of data is closing down?
A copy of that website should be made for posterity.
“DataMarket.com will no longer be available.”
Details here:
https://datamarket.com
I wonder why they appear to have stopped recording in the Western Sahara .Actually i don’t , look at the drop at the end of the series https://datamarket.com/data/set/1loo/average-monthly-temperatures-across-the-world#!ds=1loo!1n6s=2u6&display=line .
I looked at Berkely Earth, their Darwin PO Station (Berkeley ID #152446). I believe it is a composite between Darwin PO and Darwin Airport with a discontinuity coinciding to the move from the PO to the Airport at ~1940. However, they annotate 3 station moves (red diamond symbol) on their graphs, none of which correspond to the 1940 discontinuity (marked as an empirical break).
http://berkeleyearth.lbl.gov/stations/152446
Judith Curry has a post about Berkeley Earth’s homogenization methodology here:
https://judithcurry.com/2015/02/09/berkeley-earth-raw-versus-adjusted-temperature-data/
Do not ignore the comment section, it is interesting.
Cheers.
Steve et al go to great lengths to demonstrate that “The impact of adjustments on the global record are scientifically inconsequential.” He states a bit later ” On balance the effect of adjustments is inconsequential.”
And again, later, “Since the true temperature distribution is unknown, determining the right amount of homogenization to best capture the local details is challenging, and an active area of research. However, as noted above, it makes very little difference to the global averages.”
Finally, “Globally however, the effect of adjustments is minor. It’s minor because on average the biases that require adjustments mostly cancel each other out.”
If the adjustments are “scientifically inconsequential,” “make[s] very little difference to the global averages”, and “mostly cancel each other out,” then the question becomes:
Why do them at all, then? What’s the point? Prettier graphs? Have to look like you’re doing SOMETHING other than just copying data around?
” …the Berkeley approach has a smoother field.”
Oh, well, then.
In the comments on Berkeley Earth’s homogenization methodology, there’s a comparison between geographically representative, nearly intact, century-long USA urban and non-urban stations that totally contradicts Mosher’s claim of “inconsequential differences.”
See: http://s1188.photobucket.com/user/skygram/media/Publication1.jpg.html?o=0.
Bob, there has been some warming in Darwin it seems (Great town, I was in Air Traffic there a few years back). Assuming some of adjustments may have been reasonable what is your best estimate of the actual increase from 1910 to 2016?
Loydo,
I haven’t checked the BOM adjustments in 1990 and 1995. Assuming they are justified, my best estimate would be that Darwin mean temperature has warmed by close 0.9C since 1910, and a little less since 1881.
This increase is an outlier when compared to other global datasets and has been arrived at largely by using badly correlated comparison stations as well as the doubtful statistical adjustment mentioned above.
But what really matters this is approach give them the results they ‘needed’
And that , not scientific validity is the objective.
Climate science is full of this type of approach , start with the results you want , model you way to them and facts be dammed . This is the normal not unusal way of working .
“Vegetation does not generally affect minimum temperatures as there is no shading problem at night”
I guess that frost is not too common in most of Australia, but elsewhere your car will have less frost on it on a winter morning if it is parked near a tree or bush, which radiate enough infrared energy to make a substantial difference to nearby temperatures.
Which actually is an interesting comment, we have seen a warming in Tmin at night in much of the data sets (assuming this is real) at the same time in both Europe and North America, we have also seen a relatively strong resurgence of forestation in Europe and North America over a similar time period could there be a connection?
My question is why do the adjustments always cool the past?
It’s a rhetorical question, btw. I suspect I know why they always cool the past. That’s where the money is.
Bob Irvine has focused on Darwin. Jennifer Marohasy has focused on several sites.
The challenge is to find a site which has not been bastardized in Australia by the low life fraudsters from the BoM.
The surface thermometer record made up of all the individual thermometer readings was installed locally as a means of telling locals and the national authorities what the local temperature was. If it was affected by UHI then so be it, that is what the locals would feel.
To aggregate that data and then pretent that it is in any way fit for the purpose of ascertaining a meaningful ‘global’ temperature, i.e. including all those vast areas unaffected by UHI effects, and its ‘trend’ over time is utterly ludicrous.
As for attributing any ‘trend’ to increase of CO2 in the amosphere well just ask yourself what has risen more, atmospheric CO2 concentration or the amount of bitumen, concrete and steel, particularly in the vicinity of said surface thermometers. The latter has increased by orders of magnitude more than CO2.
Why does the BOM want to start at 1910. What about starting before 1898, the records exist, so use them. That way we get the 1898 to 1902 Drought. But that will show how hot things were back then, make the temperature today seen a bit cool.
MJE VK5ELL
All adjustments are unjustified. If the data is known to be ‘bad’. then remove it from the series. The systematic adjustment most certainly has an impact on the analysis, else there is no reason to do it. Any claim that it is done to create a better record is a lie.
Time to start filing lawsuits for fraud against these a$$#at$
Looking for a moment at the data from the comparison stations. How many show warming in the minimum or maximum temperatures? Quite independent of Darwin does the data from these stations support a thesis of warming post 1960 or does it just show random or cyclic variations?
A very good test of a thesis is to compare the TOTALLY raw data with the adjusted data. If the thesis is real then the trend should show up in the raw data as well as the adjusted data. If it doesn’t, then the claimed trend is entirely a result of the adjustments. That may not mean the thesis is wrong but it is certainly VERY good grounds for scepticism. Also, when there are several adjustments one would expect some would go one way, some the other way. When all adjustments go one way, the way that better supports the thesis, its again grounds to seriously question objectivity and the thesis itself. If on top of that there is massive resistance against such a comparison the suspicion rises to near certainty. If there is nothing to hide don’t just allow the comparison, actively support it.
A very simple test, if an aircraft or a bridge had been designed and built according to the approach used by the global warming advocates would you be willing to be the first passenger? Then consider that if an aircraft crashes at most say 400 people lose their lives, if CAGW is wrong thousands of people will die from causes such as energy poverty, malnutrition in 3rd world countries and millions more in 1st world countries will have their lives majorly negatively impacted. Why does the theory of CAGW require less scrutiny than the design of an aircraft or bridge.
As a salutary lesson of exactly this point, read about the history of the R100 and R101 airships. No co-incidence that the R101 was a government project shrouded in secrecy that ultimately crashed and burnt on its maiden flight whereas the R100 was a private project (vickers) carried out with full audit and scrutiny that performed admirably but was ultimately destroyed politically by the R101 disaster.
Michael Hammer
“A very good test of a thesis is to compare the TOTALLY raw data with the adjusted data. If the thesis is real then the trend should show up in the raw data as well as the adjusted data.”
I made such comparisons more than once. You can, for example, compare raw (*) data from the GHCN daily data set and what GISS computes out of raw data in its GISS-land (!!) data set:
https://drive.google.com/file/d/1da0mAoC5vIHVGLUy4J9TCDmewcSKSKN4/view
The raw data trend for
– 1880-2018: 0.08 °C / decade. That for GISS: 0.10 °C / decade.
– 1979-2018: 0.21 °C / decade. That for GISS: 0.22 °C / decade.
GHCN daily is really a raw stuff. No adjustments, only error flags.
I got a jolt yesterday while looking at the GHCN-M unadjusted file from NOAA. I thought I’d look at just the US stations to see what the US anomaly looked like, so I fired up the database and started querying. First, I built a 1981-2010 baseline for Jan-Dec, and only used stations that had at least 345 of the full 360 records that make up a 30-year, 12-month record set. Then I whittled those down by deleting stations that didn’t have a full 12 months in the baseline table. That left me with 2442 stations, plenty good coverage for the US. I wrote the station ID, the baseline average, the number of records used in the calculation, and the standard deviation and standard error for each record into a table I named “US_BASELINE_1981_2010”.
Then for each station I took its monthly record, if valid, from Jan 1900 to Mar 2019 and subtracted from it the average temp for that month and that station out of the baseline table. I wrote each station ID, the NOAA average for the month, the baseline average for the month, into another table I created and named “US_BASE_1981_2010_ANOMALIES.”
Finally, I averaged up each calendar month from Jan 1900 to Mar 2019, and printed those values out, along with the count of how many stations were used in each, and loaded it all into an Excel spreadsheet and created a chart of the data.
When I saw the chart, my first reaction was, “Aw, crap, what did I do wrong?”, because the plot line bounced between ±3°C, and of course it was compacted to fit on the monitor, so it looked like a fuzzy caterpillar. It even had the sinuous nature, with a hump around the 1930s, and a dip around the 1970s. The real “Aw crap” moment was when added the trendline, and it was firmly along the 0° line on the chart with a barely discernible slope.
In other words, it was like there was zero warming in the US on average since 1900, though the individual stations did bounce up and down a lot. I ran the procedure again and got the exact same numbers, so I have to go back tonight and really look at it to see where I went wrong. Can’t believe those numbers.
Hi James
In Australia about 15% of series involved station moves from, usually, the post office in town out to the local airport. This artificially cooled the record and legitimately must be allowed for by cooling the older record.
A similar thing may have happened in the US.
Your unadjusted data may not allow for this, which could possibly explain your level series.
Just a thought.
Bob Irvine
Thanks for this convincing guest post which gives a good explanation for the strange gap between
and
https://drive.google.com/file/d/1F8O0kfafHjuqfMxVtX5d7CEio8CgSkDU/view
It would be quite interesting to compare, where possible, the linear estimates for raw data with those for the ACORN V2 variant.
Darwin is tropical and effects of global warming are reduced in the tropics compared to higher latitudes. People sitting in air conditioned offices then try and manipulate the tropical data to try and get it to conform to higher latitude trends, without knowing anything about what they are doing.
As a psychometric statistician, I see the issue somewhat differently. If the BOM is arguing that the originally reported data is incorrect, what they are in effect confirming is that the error of measurement uncertainty is substantially higher than has been acknowledged. Clearly error bands have to include the originally recorded data, which itself had error of at least +/-.5 degrees F. And despite claims re the laboratory-measured accuracy of modern recording devices, it is unlikely that the accuracy of human read instruments has improved much beyond this. So these need to be ADDED to the measurement uncertainty introduced by these alterations (because clearly it is possible for the true score to be somewhere between the old and new values). Psychometrics has always had to grapple with measurement uncertainty and its effect on measurement reliability, but these issues seem to be wished away by many climate scientists, many of whom confuse measurement uncertainty ( which cannot be reduced by taking more measures and averaging) and prediction uncertainty ( eg as indicated by 1-Rsquared in a regression equation). And given the magnitude of changes reported here, any increase in temperature since 1910 is largely within the (increased by these changes) measurement error window.