Ken Stewart
September 2013
Australia’s Bureau of Meteorology (BOM) was quick off the mark earlier this month when it proclaimed:
“Australia’s warmest 12-month period on record
September 2012 to August 2013: the last 12 months
The past 12 months have been the warmest on record for Australia. The average temperature across Australia for the period 1 September 2012 to 31 August 2013 was 22.92 °C. This is 1.11 °C above the 1961–1990 average, surpassing the previous record of +1.08 °C that occurred between February 2005 and January 2006.”
The satellite data for the mid-troposphere for Australia- Land has just been released by the University of Alabama- Huntsville (UAH). Unfortunately analysis of this data shows that the mean temperature for the 12 months to August 2013 was +0.668 Celsius, which makes this period the sixth warmest of the satellite era (since December 1978).
That’s 0.087 below the record set in the 12 months to June 2010.
Here is a graph of the 12 month running mean to August 2013:
Yes, it has definitely been warm. The Sub-Tropical Ridge being so far north, and the dry conditions in the northern inland, may have something to do with that too.
The Bureau expects temperatures for the rest of the year to be above average, and claim that
“If a mean temperature of more than 1.0 °C above average is maintained over the next one-, two-, three- or four-month periods, each of the 12-month periods ending September, October, November and December would exceed the previous record from 2005–06 for the warmest 12-month period.”
On the other hand, if UAH records data for the Australian region of +1.0C for the next four months, the 12 months to December will be +0.748, the warmest calendar year on the UAH record, but still in third position for the 12 month mean.
And there is no doubt that this could happen. But only in 2006 and 2007 were there a total of just three months out of 12 above +1.0C, not five in a row. I’m not holding my breath.
It depends which data you would trust, from satellites criss-crossing the globe 24 hours a day, or from 104 scattered stations recording a daily maximum and minimum.
Tags: bom, climate, temperature, UAH, weather
kenskingdom 
September 15, 2013 at 7:43 am
I find the reports mentioning ‘averages’ meaningless without further information such as you are providing.
Comparing an average temperature for the past 12 months against the 1961-1990 average to get an anomaly may have some relevance to the BOM and other meteorological folks. How many Australians understand that this is what the ‘x.xx degrees warmer and we’re all going to fry’ is based on, rather than an average over longer periods such as ‘since Federation’ or our own lifetime’s experience of temperature?
Methinks there should to be far more information provided with statements such as “Australia’s warmest 12-month period on record” as the means of comparison is as important as the figures being quoted (and sensationalised).
September 15, 2013 at 3:52 pm
The whole concept of an average of surface temperatures across a continent let alone the globe is very problematic. Averaging Darwin and Hobart is ridiculous. Using anomalies from the average for an agreed reference period attempts to get around this, but temperatures behave differently in different locations e.g. Alice Springs and Sydney, or Thursday Island and Glen Innes. They also behave very differently in different weather conditions e.g. humidity causes enormous differences. As well the mean of minimum and maximum datapoints doesn’t tell you much about what happened in between. Satellite data gives a much better idea of how the atmosphere over Australia has been behaving.
September 16, 2013 at 12:17 pm
When they compare temperatures, do they just use the ACORN stations only? There appears to be about 60 stations that go back to 1910, half a dozen around 1912/13 and 30 stations start after 1940.
January 1939 was extremely hot, especially in the eastern states, but there would only be about 75 stations where comparisons could be made.
As I have pointed out in previous blogs, the ACORN data has been adjusted (usually down) from the raw data. Is this why we have having such record weather?
I have also noticed that the last time we had a very low sunspot count was in 1910/11/12. Australia then had extremely hot weather in 1914/1915 with extreme drought. The past few years we have also had very low sunspot activity – then hot weather the past 12 months. Just coincidence?
September 16, 2013 at 3:17 pm
You have raised a very interesting issue. When they make claims about extremes and records, I believe they use the AWAP dataset with about 700-750 stations, of varying quality. These give an estimate of temperatures over the whole of Australia including remote areas with no thermometers. But these can’t (or shouldn’t) be used to compare with past weather events.
When they discuss climate trends, and compare a 12 month period with previous 12 month periods, they use ACORN. This is fine up to a point. However the Acorn record has its own problems as you say with limited data in early years. Indeed Learmonth has only provided data for 37 years.. As well the rounding problem affecting pre-1972 metric data, poor data quality, and questionable adjustments.
I cannot comment on the relationship between sunspot counts and weather conditions. If we had 1000 years of temperature records it might be clearer.
September 18, 2013 at 12:11 pm
[…] below, are the 12 month averages over Australia by satellite. Graphed at Kens Kingdom by Ken Stewart, with no doctorate in climatology and no government […]
September 20, 2013 at 12:50 pm
So 2010 is Australias hottest according to the UAH dataset which measures the lower troposphere and 2013 the hottest according to BOM records that measure surface temps.
Great point!
September 20, 2013 at 3:31 pm
[…] 19th, 2013 by Warwick Hughes Ken Stewart and Jo Nova have reported on the BoM claim. BoM mean T data for the 12 months September 2012 to […]
September 23, 2013 at 3:20 am
How precise is your land mask for Australa with UAH data? If there any sea surface temp in there? And when you say ‘Australia-Land’, are satellites really able to measure land surface temps? As I understand it, UAH (and other groups) try to extract an estimate of the lower troposphere about 4 kilometers high from a column of radiance much higher, even up to the stratosphere. You get different results depending on which satellite data set you use, so they clearly aren’t perfect.
I don’t know which metric would be best for ‘Australia-Land’, but it’s clear they both (surface and satellite) have problems. If BOM purport to be giving values for surface temperature from a home-grown data set, I’m not sure that the satellite data sets should be applied. Isn’t the correct terminology for the UAH record at the altitude you’re working with “lower tropospheric temperatures”?
September 23, 2013 at 2:33 pm
Hi Barry
Firstly, it is not “my” land mask, but Dr Christy’s. It is as precise as land data for other parts of the globe at http://vortex.nsstc.uah.edu/data/msu/t2lt/uahncdc_lt_5.6.txt . It includes the mainland as well as Tasmania, but not New Guinea or Indonesia, and I understand excludes ocean away from the coast line. No doubt there is some imprecision but not much. Incidentally, BOM’s Acorn annual data (when anomalies are recalculated using the same period as UAH, 1981-2010) matches UAH satellite data reasonably well. See my post at https://kenskingdom.wordpress.com/2013/03/17/how-angry-was-summer/
2. No, UAH does not measure land surface temperatures but temperatures of the lower troposphere (TLT) above the land area.
3. Yes they both have problems (and different uncertainties- UAH is +/- 0.1, BOM has undertaken to try to reduce uncertainty in its Acorn data in future to +/- 0.2.) However, satellite data give a good idea of continent wide temperatures- not precise surface temperatures, but as my post referred to above shows, matching quite reasonably.
I hope this answers your questions, please add any others.
Ken
September 24, 2013 at 12:20 pm
Thanks for the reply, Ken.
How are the uncertainties calculated? Just curious if you’re applying uncertainty on global temps for UAH, instead of for regional (Australia) They will be different, won’t they, because regional temps have more variability. A link to methods would be great.
September 24, 2013 at 12:31 pm
And thanks for the UAH link. I have it in my bookmarks and forgot they had added a column for Australian temps.
September 24, 2013 at 5:39 pm
Interesting point re UAH regional uncertainty- I’ll see if I can find out. Acorn uncertainty is referred to in the BOM replies to the review panel’s recommendations.
September 28, 2013 at 6:40 pm
Hi Barry
Dr Christy has advised the standard deviation of differences for the Australian part of the UAH dataset is +0.11 K. “So the 95 percent random error range would be about +/- 0.2 K for a particular month. For annual anomalies the error range would be half that.”
i.e. +/- 0.1.
I hope that answers your query.
See Christy et al The role of remote sensing in monitoring global bulk tropospheric temperatures, International Journal of Remote Sensing Vol. 32, No. 3, 10 February 2011, 671–685.
Ken
October 15, 2013 at 10:20 am
Thanks Ken. This makes it a statistical tie with other 12-month periods highest in the ranks in the satellite data. That’s the case for surface data, too.
October 15, 2013 at 12:25 pm
Statistically not different from other high periods, so not statistically higher either. You would agree that BOM should have claimed the 12 months period to August 2013 was equally as hot as two previous hot periods?
October 28, 2013 at 8:20 am
Ken, I haven’t heard back from BOM yet. While I’m waiting, where would I find the uncertainty estimate for Oz national data from the BOM? I presume you got that from one of their reports/papers?
October 28, 2013 at 10:47 am
See page 5 here:
Click to access ACORN-SAT_Bureau_Response_WEB.pdf
That is, uncertainty in their station data (not national estimates- my mistake) in the past is +/- 0.5 and BOM “undertake” to get it down to +/- 0.2. I find it intriguing that BOM can release statements about record temperatures while their daily and monthly data since February is still undergoing quality assurance and monthly data for most stations is only available at CDO up to February.
October 28, 2013 at 9:03 pm
It’s an interesting statistical feature that the uncertainty of individual measurements is greater than the uncertainty of a large sample of those measurements.
Reduction in uncertainty for large samples can be calculated:
a / N
(Hope the formatting works – if not, it’s ‘a’ divided by the square root of ‘N’)
where ‘a’ is the uncertainty for an individual measurement, and N is the number of samples. The values, then are;
1 / (365 x 104) = 0.005
This would not cover structural uncertainties, like the sparse gridding you have pointed out, so I can only guess that the error bars in the BOM’s graphs, which seem larger than the purely statistical result, factor this sort of thing in.
To my mind, this leaves as the major force of your criticism in the top post being that the UAH data tell a different story. From a purely statistical point of view, it may be that most recent 12 months is a clear record-breaker in the ACORN-SAT dataset, not a statisitcal tie. But hopefully the BOM will respond to my query.
October 29, 2013 at 9:38 am
Hi Barry,
I wouldn’t hold my breath waiting for a reply from BOM. 3 days used to be the turn around time, but if you haven’t had a reply still I’d say your query has gone into the black hole and a second request might be in order. If they suspect your query expresses some questioning of the received wisdom you will be told to go and write a paper and then ignored.
And we may see a record…
October 30, 2013 at 2:48 am
No, lools like they arem’t going to reply – they gave a 10 day maximum and ot’s past that. But on the other hand, it seems likely they made the right call for the ACORN-SAT data set.
October 16, 2013 at 12:01 pm
If the uncertainy intervals we’ve discussed are correct, I think that the BOM should have included them in their statement. The same should apply to your article.
I’ve just emailed BOM about this. I’ll let you know what they say if they reply.
October 28, 2013 at 10:35 am
I have been unable to discover the uncertainty values for annual data. I did re-read
Click to access CTR_050.pdf
where there are graphs on page 33 showing error bars that are more like 0.1C (+/-0.05) for annual temperatures. So I am curious where you got the 0.2 figure, Ken. The only thing I could ifnd that matched that was from here;
Click to access ACORN-SAT_IPR_Panel_Report_WEB.pdf
I thought that might be where you got the value from, as you said “BOM has undertaken to try to reduce uncertainty in its Acorn data in future to +/- 0.2.).”
If this is your source, it would be a mistake to translate tolerance for a single sensor to uncertainty for the national average, which would be a lower value (smaller errors on larger sample). Did you get the value from somewhere else?
September 24, 2013 at 1:02 pm
Thanks Ken for finding that ref to a Sevenson Screen in Springsure – I have added it to my blog –
September 24, 2013 at 4:28 pm
[…] 19th, 2013 by Warwick Hughes Ken Stewart and Jo Nova have reported on the BoM claim. BoM mean T data for the 12 months September 2012 to […]
October 1, 2013 at 11:32 am
Hi, Ken, I see the ABC scaremongers are at it again, Queensland’s hottest ever year, again.
October 1, 2013 at 4:51 pm
It has been very hot in many places, and I wouldn’t doubt that surface mean for Qld has been up there, even maybe a record. But many places have not seen anywhere close to records. I will post about Qld temps in a few days.
October 6, 2013 at 5:52 pm
Reblogged this on The Real World and commented:
Great post on the garbage being spewed forth by the Bureau of Masturbation.
October 6, 2013 at 9:02 pm
University of Alabama? Hurstville? They have “better” data on Oz weather than the BoM? FMD – buncha ****ing ignorant rednecks vomiting out their climate
denials. The earth still flat there Anthony? Creationism still the most popular course?
October 7, 2013 at 7:40 am
Dear Peter,
I am leaving this comment up (minus the obscenity) to illustrate the abysmal level of ignorance, abuse, and prejudice exhibited by many of the true believers such as yourself. Let’s see- how many boxes did you tick? “rednecks”, “denials”, “flat earth”, “creationism”- but no engagement with the topic. Unless you can make a civil and reasoned contribution, don’t bother commenting in future.
October 15, 2013 at 11:07 am
“It depends which data you would trust, from satellites criss-crossing the globe 24 hours a day, or from 104 scattered stations recording a daily maximum and minimum.”
I’ve read that you can get a fairly good correlation with the global record by using as few as 60, fairly well-spaced weather stations, (eg, this – blog, not study), so 104 for one country doesn’t seem like so little, even though it’s not nearly as well-covered as other countries, like the US.
Ken, have you tried taking different subsets of say 50 Australian weather stations as well spaced as possible and comparing the results to the main record? Would that have any merit?
There have been papers and blog articles comparing time series for coastal, inland only, airports only, rural and urban weather stations for the US record and global. The results have been illuminating.
I’ve bookmarked a few of the blog posts on these sorts of analysis, which you can check out if you’re interested.
http://moyhu.blogspot.com.au/2010/04/ghcn-results.html (global airport, rural, urban, NH, SH)
http://clearclimatecode.org/airport-warming/
http://residualanalysis.blogspot.com.au/2010/04/urban-heat-island-effect-probably.html
http://rankexploits.com/musings/2010/uhi-in-the-u-s-a/
I don’t have the know-how or patience to learn how to do this for the Australian temp record, but perhaps it will inspire you, if you haven’t already investigated this kind of thing.
October 15, 2013 at 11:10 am
Botched html code, sorry.
November 6, 2013 at 12:23 pm
Ken, I received a reply from the BOM. here it is:
“Dear Barry,
There is no formal uncertainty analysis for the area averages of Australian temperature data. This has been done for global data, based on spatial representation, and one could reasonably expect uncertainty in continental Australian data to be around the same order of magnitude, near or below 0.1deg C, since the Australian network density is greater than the global one.
The lack of a formal uncertainty is mostly due to the fact that structural uncertainties are present. The structural uncertainty makes estimates of uncertainty due to things such as spatial coverage (such as above) only partially representative of the true uncertainty. This makes representing uncertainty with a formal estimation (i.e. a single number) problematic.
In that light, a considerable amount of work has been done by the Bureau to test the sensitivity of the spatial averages to various different methods of data preparation and analysis. Again, such comparisons provide a partial representation of the true uncertainty, but the Bureau uses this type of cross-comparison to check record ranks specifically and to ensure that our analyses and public statements are robust.
An example of how this is applied can be found in our write up of summer records of 2012-2013 http://www.bom.gov.au/climate/updates/articles/a001-summer-heatwave-2013.shtml . We also note that our three datasets; ACORN-SAT, Trewin and AWAP, all show that the last 12 months have been record warm.
You can scroll down to the data section to find how cross comparison of ranks is typically performed.
This also links to the ACORN-SAT website. Under Methods and Development, there are some papers that explore the sensitivity of the data to various methodologies.
Regards,
Climate Analysis Section”
I had imagined that the structural uncertainties would be difficult to formalise. I’m away to Wagga Wagga for a couple of days, and will follow up the material they linked when I get back.
Regards,
Barry.
November 6, 2013 at 2:14 pm
Thank you Barry. That is very interesting and will take some time to digest. Like you, I’m pretty busy with other things so I don’t imagine I’ll get to look at it for a while.
Cheers
Ken
November 15, 2013 at 4:56 pm
Ken, I sent these graphs of NASA RSS data from Steven Goddard for Australia for the last 34 years to BoM and asked why they don’t correlate with any of the land or satellite measurements for the globe yet RSS and UAH generally correlate globally:
http://stevengoddard.wordpress.com/2013/11/05/rss-shows-no-warming-in-australia/
They replied that they were measuring different things which I knew but shouldn’t RSS still show comparable warming over that period?
At least similar to UAH?
November 16, 2013 at 9:10 am
Hi, very interesting graphs. I’m not sure what they represent and I can’t open the nc file unfortunately to check. I use UAH data as anomalies from 1981-2010, and convert Acorn data to the same climatology to compare so I can’t comment on whether it should be comparable with UAH. If RSS is looking at the same land mask as UAH they should be very similar.
November 19, 2013 at 1:31 am
RSS and UAH have trend differences over short duratons. ENSO events in RSS is also larger in amplitude. I don’t know if that would translate to differences over longer durations for regional trends, but it should be borne in mind that for some metrics, they are not so similar.
November 19, 2013 at 1:38 am
For example, here are the trends for each 1997 through 2012:
http://www.woodfortrees.org/plot/uah/from:1997/plot/uah/from:1997/to:2012/trend/plot/rss/from:1997/offset:-0.1/plot/rss/from:1997/to:2012/trend/offset:-0.1
November 19, 2013 at 8:14 pm
Yes Barry I am aware of the trend differences globally between UAH and RSS. I understand they use different methodology. I have no idea about regional differences, and can’t comment until I can look at the RSS data. My computer doesn’t recognise the format and I’m very wary about downloading so-called Microsoft approved programs, from sorry experience. I haven’t been able to devote any time to this or any other interests for some time due to other demands on my time.
November 19, 2013 at 9:16 pm
“I understand they use different methodology.”
Aye, and the exact same raw data, I believe.
Different baselining (climatology) should have no impact on the resulting trends. There is an annual cycle in satellite data that is most easily worked around by using annual instead of monthly data for trend estimates, so I’ve read hither and yon.
March 13, 2014 at 6:16 pm
Hi Ken,
Is the trend for Australian atmospheric temperatures from 1979 statistically significant?
March 16, 2014 at 1:55 pm
Sorry, I’ve been away for a few days. In short- yes.
January 26, 2016 at 10:45 am
Hi Ken, I’m interested in referring to one of the graphs from this page on my own website. What is the best way to contact you regarding getting said permission?
January 26, 2016 at 11:55 am
You’re welcome to use anything you like. It would be better to use more recent than 2 years ago.
January 26, 2016 at 10:45 am
[…] 2012/2013 in Australia was only anomalous relative to the previous year, and showed the graph from this site (which I’ve never seen […]