From Jennifer Marohasy’s Blog.
Jennifer Marohasy
Tuesday, 18th April @ 9.30pm US Pacific Time
Wednesday, 19th April @ 2.30pm Australian Eastern Standard Time
Listen on the internet:
The Australian Bureau of Meteorology has replaced most of its mercury thermometers with platinum resistance probes in automatic weather stations.
My assessment of nearly twenty years of parallel data from Mildura, and just three years of parallel data from Brisbane airport, indicates that the probes can record both hotter and colder than the mercury, but are not equivalent.
Statisticians are notorious for disagreeing about the most appropriate statistical test to apply, but nevertheless should at the very least provide the name of the test undertaken and the level of significance, as a lawyer would detail the case law referenced.
The Bureau claims an assessment of the full 2019 to 2022 period for Brisbane Airport finds no statistically significant difference between the probe and mercury thermometer. But without providing any details.
More generally, their climate scientist, Blair Trewin, claims the parallel data –- temperature series recorded using a mercury thermometer and a platinum resistance probe at the same sites and at the same time –- are not significantly different across their network of some 700 automatic weather stations though as far as I can tell there is only parallel data for 38 sites.
My analysis of the three years of Brisbane Airport parallel data — only recently made available following years of wrangling over an FOI request with the Bureau — shows that 41% of the time the probe is recording hotter than the mercury, and 26% of the time cooler. The difference is statistically significant (paired t Test, n = 1094, p < 0.05).
The differences are not randomly distributed, and there is a distinct discontinuity after December 2019.
I initially thought that this step-change from an average monthly difference of minus 0.28 C in December 2019 to plus 0.11 in January 2020 (a difference of nearly 0.4C) represented recalibration of the probe.
The Bureau has denied this, explaining there was a fault in the automatic weather station that was immediately fixed and operating within specifications from January 2020 onwards. After January 2020, the probe can record up to 0.7C warmer.
I have been asking for all the parallel data to be made public, beginning with requests for data from Wilsons Promontory lighthouse back in 2015.
In 2011, an expert panel commented that throughout the last 100 years, Bureau ‘guidance’ has allowed for a tolerance of ±0.5 °C for field checks of either in-glass (mercury) or resistance thermometers (probes), and for this reason the Panel did not rate the observing practices of the Australian panel amongst international best practices.
It is time for another expert assessment, and for all the parallel data held by the Bureau to be made public. There should be 15 years in total of parallel data for Brisbane airport and a similar amount for another 37 of the Bureau’s 700 official weather stations.
I will be talking about these issues this afternoon on TNT radio with the legendary Chris Smith.
Tuesday, 18th April @ 9.30pm US Pacific Time
Wednesday, 19th April @ 2.30pm Australian Eastern Standard Time
Listen on the internet:
So far, the Bureau has claimed a variety of reasons for not making the parallel data public. In the case of the Brisbane airport data, it initially claimed that manually scanning the handwritten A8 report would be too onerous. When the FOI request by John Abbot ended up with the Information Commissioner the Bureau somewhat bizarrely claimed that the parallel data for Brisbane airport did not exist.
While providing me with nearly 10,000 scanned A8 reports for Mildura back in 2017 after the intervention of then Environment Minister Josh Frydenberg, the Bureau continues to withhold some key pages pertaining to September 2012.
I can only estimate likely differences between the probe and a mercury for the month of September at Mildura until this parallel data is made public. In the meantime, the Bureau stands by its claim that the temperature of 37.7C recorded by the probe at Mildura on 22rd September 2017 is a new hottest record back to 1889, even though temperatures were recorded with different equipment back then, with a mercury thermometer.
I told you. You cannot compare the two methods. With mercury they usually took the tmin and tmax and figured the tmean is tmin + tmax/ 2. You need people for that job.
What if the weather changes during the day?That is really important. And this is apart from a growing error in the mercury thermometer due to a reaction with glass.
With the recorders and computers there is continous measuring and once calibrated there is almost no error in the measurement.
Also you don’t need people. Computers can pump out the relevant data anytime you need it.
I am finding that the difference in measurement could be 0.3 or 0.4 K.
Personally I stopped looking at data before 1975.
“With the recorders and computers there is continous measuring and once calibrated there is almost no error in the measurement.”
Except for the bias, drift, different size Stevenson screens. 😉
Even with the same size screen there can be differences in airflow, screen conditions, and microclimate over short periods of time, e.g. grass being mowed, nearby soybean/corn fields being harvested, etc.
There are a lot of things that can contribute to measurement uncertainty. Random errors may be reduced but not systematic ones. Of course in climate science, systematic bias is always assumed to be zero!
You’re right, Tim. If the PRT is in a naturally ventilated Stevenson screen, systematic error is inevitable.
The error pretty much varies on all time scales. The only recourse is a field calibration and apply the calibration uncertainty to every measurement.
PRTs have the additional problem of self-heating error, if the wiring is not ideal. In the same design of naturally ventilated screen, the LiG thermometer is comparable to a PRT in accuracy.
“With the recorders and computers there is continous measuring and once calibrated there is almost no error in the measurement.”
The computers and recorders help assure that there is little error in the recording process, particularly when compared with human observations and logging. There are still ample reasons for errors in the electronic data itself.
Dear Henry,
You can (and must) compare two instruments when one supersedes the other. In Australia average temperature is by definition (Tmax + Tmin)/2. ’cause that is what it is.
Irrespective of what happens between dawn and dusk, Tmax is still Tmax, and between dusk and dawn, Tmin is still Tmin. We also measure temperature in degC, because that is the standard unit, not in K, which is a more abstract unit. Although a barometric adjustment is usually applied, zero degC is also the melting point of ice and 100 degC is the boiling point of water, so that linear scale can be calibrated.
The point is that within their respective error envelopes, two instruments operating in the same space (a Stevenson screen), should measure the same temperature (i.e., T +/- instrument uncertainty values). While they may not measure exactly the same value, the difference between them is NOT significant, in other words immaterial.
Jennifer is trying to turn immaterial differences into material ones. The paired t-test is particularly sensitive to autocorrelation and in the case of Brisbane data, because early data were cooler, differences carry-through implying later differences were warmer. However it is a product of the test not the data.
In order to rescue her statistical credibility she should put the data in the public domain.
All the best,
Bill Johnston
Bill
Thanks. I would like to know. Are you saying that your station is also seeing that it is getting cooler in Canberra every night?
Dear Henry,
I have not looked at Canberra data for several years, so I don’t know is the short answer.
All the best,
Bill Johnston
Tune into me, look at me, its me, all about me.
You have potentially bought Graham Lloyd and The Australian into disrepute.
It is in the interests of everyone involved including the IPA, that you describe the tests you used and put Mildura and Brisbane data in the public domain. Otherwise you have no message worth listening to. It is that simple.
The use of paired t-tests on time-correlated differences has always been and still is invalid.
Yours sincerely,
Dr Bill Johnston
http://www.bomwatch.com.au
I stopped looking at data from BOM when I found that at some places data had been fiddled with, compared to the original data from my own (trusted) site.
Anyway, for those interested here, looking at it from the mid seventies, I find virtually no warming in the SH from that time, measured on the ground.
https://breadonthewater.co.za/2021/11/25/an-inconvenient-truth/
The problem comes in when you add data from 50 and 60’s. All of a sudden, then there is a ‘clear’ warming trend. How can that be?
Cooling the past.
Yeah. Maybe. I found data had also been fiddled in Gibraltar. I chucked them out because the MET data were completely different compared to three surrounding stations. MET, BOM and KNMI are not trusted by me. It seems they like fiddling with their data. BEST came here in RSA and glued stuff on stations with older data.
But remember that my stations were random from all over. When I looked at earlier data (then 1974) I always came out lower in the past. So, going by my observations I think the difference between the two methods is ca. -0.3K
“The use of paired t-tests on time-correlated differences has always been and still is invalid.”
The tests may well be invalid. However, with enough data, the difference will become statistically significant. The test is utterly pointless.
A test rejects or not a null hypothesis. The hypothesis here is that the instruments are the same (zero difference). But we know very well that they are not the same, and do not need a statistical test to tell us so. The question is not whether the difference is non-zero but whether they are too large. The test cannot tell you that.
Dear Nick,
The test is whether daily data for the two instruments is the same, not whether the instruments themselves are different!
Looking at Marohasy’s headline graph
?w=1024&ssl=1) there are as many differences LESS than -0.5 DegC (N =4) as the number of differences she is sweaty about that are more than 0.5 DegC (N = 4). My eyeball assessment is there is zero likelihood that data measured by the instruments is significantly different.
As a scientist she should justify her conclusions. However, a woman on a mission is to be feared not revered. This whole thing is out of control and it up to her to justify her case.
All she has to do to is explain her test and place the data in the public domain.
All the best,
Bill Johnston
Bill,
“All she has to do to is explain her test and place the data in the public domain.”
More importantly, explain what is the useful conclusion from the test. Other than to say “like all those other scientists, we have statistical significance!”
A useful exercise too would be to state the null hypothesis she claims her test rejects.
But yes, after all this kerfuffle about the BoM being allegedly stingy with data, the least she could do is to post the numbers on which the graph is based.
Tonight I received an email newsletter from Chetna Mahadik at the IPA entitled IPA this week.
The leading story was “This week, I want to share an IPA achievement that brings home how our research can sometimes take a long time to come to fruition – in this case, more than a decade.
I am talking about IPA Senior Fellow Dr Jennifer Marohasy’s decade-long investigation into discrepancies in temperature data of the Bureau of Meteorology (BOM).
The discrepancies arose when the BOM switched its temperature reading instruments from mercury thermometers to platinum resistance probes across its various weather stations in Australia“.
“This week, The Australian published the results of her research ….”
For all the reasons I’ve previously discussed and defended, the whole story including in The Australian, her piece published at https://joannenova.com.au/, and here at WUWT is bullshit.
She has not not defended her use of paired t-tests and no interests are served by IPA-Fellows who are also porkie-tellers.
As an IPA member interesed in scientific integrity it is surely time to have a serious conversation with the IPA board.
Cheers,
Dr Bill Johnston
http://www.bomwatch.com.au
“As an IPA member interested in scientific integrity”
I am emphatically not an IPA member, and I am actually rather glad that its resources are being expended on such a useless activity.
The Australian Bureau of Meteorology obviously has something they want to hide. Something about the climate.
The have no good reason to withhold this temperature data, and are obligated to make it public because the public is paying for the data collection and the salaries of those collecting the data. The is the Public’s data, not the property of The Australian Bureau of Meteorology.
Why isn’t The Australian Bureau of Meteorology being sued up one side and down the other over this obvious abuse of power?
True. I am willing to bet that BOM will not tell you that it is getting cooler in Canberra every night…..
at a rate of -0.05K/ annum
nogal
That is -2K in 40 years.
Wow. You would think somebody would have noticed that?
(Like I said in my report: minima are falling in the SH. The only logical explanation is that there is less heat coming from earth in the SH)
Bill
I would like to know. Are you saying that your station is also seeing that it is getting cooler in Canberra every night?
Nick and Bill really are clueless when it comes to statistics, and so much else.
They are also useless when it comes to actually getting data. I encourage them to put in their own FOI requests. And Bill should follow through when he manages to secure some, even if it is only a few month’s worth for Sydney, e.g. https://joannenova.com.au/2017/08/another-bom-scandal-australian-climate-data-is-being-destroyed-as-routine-practice/
After that fiasco, I won’t believe a word he says about anything. :-).
Paired t Tests really are perfect for establishing the statistical significance or otherwise of this type of data, when wanting to compare a series of values that are ‘paired’ in so much as they were recorded on the same day, but with different equipment.
These Paired t Tests can be undertaken on a month-by-month basis (ie. analysis can be limited to blocks of about 30 days), for the entire period of three years, and also for the period before and after any discontinuity, etc.
Since the Bureau started transitioning from mercury thermometers to probes in
automatic weather stations, officially beginning in November 1996, there has been very limited analysis published by BoM climate scientist Blair Trewin and more usually without stating the actual statistical test used, or the level of significance found.
With respect to the Brisbane Airport data, the Bureau is saying that according to its own analysis the difference between probe and mercury is not statistically significant, while not stating the test used or anything else. Extraordinary really.
More usually in science, statistical difference is stated and quantified. The Bureau
produces a lot of statistics, but it is unclear whether the Bureau employs statisticians.
And I’ve just got a note from TNT Radio. ..
“TNT Radio would like to thank you for your recent appearance on The Chris Smith Show.
Here is a link to your interview https://tntradiolive.podbean.com/e/mike-shedlock-dr-jennifer-marohasy-on-the-chris-smith-show-19-april-2023/
“the Bureau is saying that according to its own analysis the difference between probe and mercury is not statistically significant”
Can you quote what the BoM actually said? A link would be good.
I very much doubt they said that. A failure of a statistical significance test does not allow anything to be concluded. Their focus should be on whether the magnitude is within limits. I expect that is their focus.
The Bureau has explained:
“An assessment of the full 2019-2022 period finds no significant difference between the probe and mercury thermometer. ”
There is no mention of the actual test undertaken.
There is no link.
This is in response to The Australian newspaper asking for some comment on my findings, etc.
The newspaper has offered the bureau opportunity to explain all and everything about the Brisbane Airport data. But I gather all that has been received is a letter to the editor that I can’t find online, but I will attempt to upload a photograph of it here:
““An assessment of the full 2019-2022 period finds no significant difference between the probe and mercury thermometer. ”
There is no mention of the actual test undertaken.
Of course not. They are just correctly saying that the differences are small. They are making no statistical claims.
The BoM rebuttal also makes no claim of statistical testing. They are just saying that differences are within tolerance.
A complete study of all calibration reports and out of tolerance reports generated by the calibration agency is in order, including their standards that performed the calibrating. Saying everything was within normal tolerance and showing it to be traceable to normal tolerance are two different concepts.
We know what it is traceable to. They are two different instruments. Calibration talk is a red herring. All they are saying is that they measured air temperature at the same time with two different instruments in close proximity, and got very similar results. No significant difference.
”…and got very similar results.”
So what’s your definition of ”very similar” Nick??? How many fractions of a degree? I mean isn’t that important considering all the debating over the years we have had over fractions of a degree????
It isn’t my job to define it. The BoM said it is insignificant. For Brisbane it is less than 0.1°C, which I think well merits BoM’s term “insignificant”.
B.S. Using a discrepancy of around 0.1 degree C after ten years of measurement is ONE WHOLE DEGREE. That is significant in anyone’s book. Especially yours.
Prove it. Oh, that’s right, “Trust me” is just as good.
Dear Jennifer,
“The use of paired t-tests on time-correlated differences has always been and still is invalid.” Your statistical incompetence beggars belief.
I did not ask for Sydney data anyway, and had the IPA not paid for your requests I strongly doubt you would have spent your own money.
Blair Trewin would have compared whether mean values reported by the different instruments were different, which provided cycles were removed, would have been a valid test. He probably also examined residuals for normality, independence and equal variance. He may have also examined data distributions and other attributes (you know about those don’t you?).
Why tell porkies. At your age too? The song and dance about data that did not exist for Wilsons Prom, Cape Otway and Rutherglen. You never undertook regular weather observations using standard instruments anywhere, either. What about your secret friend at Goulburn? Why ignore site changes at Mildura, and the six (not four) Tmax values that were cooler by more than 0.5 degC than thermometer values in your data for Brisbane?
Why not simply place your data in the public domain?
All the best,
Bill Johnston
http://www.bomwatch.com.au
Bill, I have the distinct impression that you’ve now morphed from MarohasyWatch.com.au on to MarohasyStalking.com.au.
Thanks Mr,
As a paid-up member and supporter of the IPA I expect better than this.
Look at the graph again. All but several differences are within 0.6 degC of zero, which is the minimum uncertainty bandwidth of comparing two values.
Except for the fact that data are not homogeneous (evidenced by the step-change), and that differences are probably autocorrelated, the difference between instruments would not be significant.
She should place her data in the public domain.
All the best,
Bill Johnston
Not entirely irrelevant. Many, many years ago I worked in electronic component testing. We’d often upgrade equipment or software. We had several ways of ensuring the tests om the new equipment were no worse than the old. Standard devices which had been datalogged tested on old and new. Running several production batches in parallel and accounting for any differences.
When the new equipment had higher resolution than the old causing failures these had to be manually verified using oscilloscopes, and high specification meters and all the other things relevant to whatever it was we were testing.
When measuring temperature. humidity, windspeed and any other weather related metric I can see no reason or excuse for not carrying out a similar process and justifying why the new measurements are of a higher quality that the old.
Some statements are, in my opinion, in the famous last words category. “There are no mines there, we swept it yesterday!” boom! (or BOM!)
Let’s remind ourselves how the BoM plays its tricky little games.
a Climategate email from BoM’s David Jones to University of East Anglia’s Phil Jones –
from: “David Jones”
subject: RE: African stations used in HadCRU global data set
to: “Phil Jones”
Thanks Phil for the input and paper. I will get back to you with comments next week.
Fortunately in Australia our sceptics are rather scientifically incompetent.
It is also easier for us in that we have a policy of providing any complainer with every single
station observation when they question our data (this usually snows them) and the
Australian data is in pretty good order anyway.
Truth be know, climate change here is now running so rampant that we don’t need
meteorological data to see it.
Almost everyone of our cities is on the verge of running out of water and our largest irrigation system (the Murray Darling Basin is on the verge of collapse – across NSW farmer have received a 0% allocation of water for the coming summer and in Victoria they currently have 5% allocations – numbers that will just about see the death of our fruit, citrus, vine and dairy industries if we don’t get good spring rain).
The odd things is that even when we see average rainfall our runoffs are far below average,
which seems to be a direct result of warmer temperatures.
Recent polls show that Australians now rate climate change as a greater threat than world terrorism.
Regards,
David
Thanks for the reminder “Mr” – We must always keep the amazing Climategate revelations in mind. Here we have a top BoM scientist repeating ridiculous doomster talk like the Dr Prof Tim Flannery. And remember how all that sort of doomster rainfall talk assisted in the ~$20Bn spread of mostly useless seawater desal plants across the “Eastern States”. I remember how in Canberra – in the face of their abundant fresh water resources – the Chief Minister was asking his water experts if the ACT should build a seawater desal plant at Jervis Bay and pipeline the water uphill to Canberra!!! I kid you not.
If you carefully go through all the info from BEST, KNMI, MET, GHSCN, etc
you will not find any place or City where Tmin is going down, like I showed you on Canberra, in a comment upstream.
As someone already suggested, it looks like they cool the past to at least get a straight line on Tmin over time.
In the beginning of my own research I always wondered about that. Now I know. They think they know or they assume that man made warming caused by CO2 is ‘true’.
So if they find the Tmin trend is negative, they fiddle with the data to at least show the line is straight. They are correcting for what they think they know NOT to be true.
Obviously, if the theory of man made warming were true Tmin must be going up, pulling up Tmean.
Truth is that Tmin is going down, in the SH.
Go figure.
https://breadonthewater.co.za/2021/11/25/an-inconvenient-truth/
Just once I’d like somebody to find an error in the data that underestimates warming.
There is only one way to compare temperatures today with temperatures in the past. That is with the same equipment. I have no problem with the new devises but the temperatures recorded with them can NOT be compared to the temperatures recorded in the past. To make that comparison we need to use the same style equipment. End of story.
True. Is why I am not looking at data before 1974. And, in fact I like airports. Most of them had that new equipment with automatic recording installed by that time.
Do you see an increase in temp with the increase in jet powered aircraft? Max temps are more dramatic, but blowing hot gasses over concrete, asphalt, metal, brick, and earth will raise the air temp over them even after the jet has flown away. This should cause the Tmin to rise as the cooling starts from a higher point, but the night is the same length.