Ken Stewart, 28 January 2012
Introduction
In March 2011 I analysed trends in Australian Maximum Temperatures from 1910-2010, and now I am following up with Minima. Since that time I have looked at adjustments made by the Bureau of Meteorology (BOM) and in a paper co-written by Dr David Stockwell proposed an alternative network of climate sites we call the Minimally Adjusted Network (MAN). This network comprises sites included in BOM’s High Quality (HQ) network but makes adjustments by comparison of overlapping raw data only.
I have further revised this network by removing all sites where there is less than two years of overlapping data, where there is less than 80 years of data, where the record shows spurious data, or where metadata supplied by BOM indicates serious problems. I have deliberately included all urban sites including all capital cities except Darwin, usually at a nearby airport. The MAN dataset for Minima now comprises 89 sites.
I shall explain with one example what “Minimally Adjusted” means. Here is a graph showing the raw data available for sites near Cairns in North Queensland:
There is good overlap of 10 years from 1943 to 1952. The graph shows Cairns Airport minima are warmer than the Post Office, and in fact the mean difference is 0.29 degrees. Therefore the Post Office raw data may be adjusted up +0.29 degrees so that the two series are comparable, to give this graph of the spliced data:
No assumptions are made about the data, and no subjective decisions are made about whether to adjust and by how much, unlike the adjustments used in BOM’s HQ datasets. No other adjustments are made. If a site meets the criteria but the metadata indicates serious problems (e.g. Nhill) or the data appears to be spurious (e.g. Halls Creek), it is excluded. As more metadata becomes available it may be that further sites will be excluded.
To compare temperature across Australia I convert all data to anomalies from the 1961 to 1990 mean for each site. I then calculate the mean anomaly for each year for a selection of sites. While I show data from 1910 to 2011, much 2011 data has still not been “quality assured”, so many annual means were not available. For these sites I found the annual mean from the daily means, which is not the correct method, but is close.
Caveat: There is considerable uncertainty in all temperature data. BOM says that Automatic Weather Station (AWS) data has an accuracy of +/- 0.3C. Celsius thermometers (mercury in glass) have an accuracy of at best +/- 0.5C. Fahrenheit thermometers, certainly no more accurate, were used before September 1972 and temperatures recorded before then were converted to Celsius for modern comparison. The records show that in the Fahrenheit era, many observers used whole degrees even though BOM’s instructions to observers were to record temperature in tenths of a degree. That’s hard enough on a Celsius max/min thermometer. A number of quite recent observations appear to be recorded in whole degrees. The change to Celsius in 1972 may have led to an artificial warming relative to pre 1972. Certainly the Australian temperature record must be treated with great caution. Therefore my figures are subject to the same uncertainty.
Results
Here is the graph of Minima for all 89 MAN sites:
While Australian temperatures have been rising, it has been by no means a steady increase. There was an enormous jump up in the 1970s. The 4th order polynomial trend line shows the distinct fall and rise. Are we following a natural cycle? The next few years will be decisive.
What’s with WA?
Here’s Western Australia;
And here’s the rest of the continent:
Wildly different! It’s worth putting both together to show the difference more clearly.
WA is clearly a different climate zone.
City vs Country
Not such a great difference. Urban sites are warming slightly faster- UHI?- cooler up to 1950s, and in the 1990s, and are a little more variable, which is surprising.
Beach or Bush
Sites within 100km of the coast have much less variability than inland sites, due to the oceanic influence. Inland sites have greater extremes, but the overall trends are very similar, except that coastal sites are getting warmer, whereas inland sites’ extreme peaks have been decreasing since 1973. The extremes are at identical times, but the standout differences are 1973, 1976, and of course 2011. Note the late 1970s step up is very clear.
North and South
The boundary line is the South Australia/ Northern Territory border. First southern Australia:
Northern Australia:
Wildly different! Note the warmest year by far was 1998, and 2011 was colder than any year since the 1970s. Now let’s combine:
Note that northern minima- further from the South Pole- are warming faster than the southern. The north has much greater extremes and appears to have peaked.
Maxima and Minima
I have plotted Maxima and Minima only to 2010.
While the linear trend for Minima is greater than for Maxima, and after initially decreasing, Minima anomalies have been rising since the 1940s, in the past decade (“the hottest on record”), minima have flattened. The record means have been caused by much higher maxima. This is especially visible in a plot of the annual Max-Min range, 
which shows the linear trend steadily decreasing as is expected with the Greenhouse Effect. However, the range has increased steeply since the 1980s which is not what is expected. This is not the signature of Anthropogenic Global Warming, at least in Australia.
Conclusion:
The Minimally Adjusted Network provides an alternative network of 89 long term sites with few adjustments to the raw data. Uncertainties in the data accuracy of at least +/- 0.3C mean that the temperature record must be viewed cautiously.
There is little difference between Urban and Non-Urban minima.
Western Australia has a completely different climate to the rest of Australia.
Coastal and Inland trends are not greatly different, except for variability, until 2011.
Much can be learned from looking at Means, but an analysis of Minima yields further interesting information. Remember: Anthropogenic Global Warming due largely to the greenhouse effect should be visible in Minima. Temperature increase should be greater at night, in winter, and towards the poles. It is plain, however, that in Australia this is not the case.
Since the 1940s, very cold years have been getting progressively milder, but since 1973 very mild years have been getting progressively cooler. Further, the increase in minima is more pronounced in northern Australia than in the south- and while southern Australian minima have been steadily increasing, northern minima have a much greater variation and appear to be now decreasing. The Range bewteen Maxima and Minima does not show the greenhouse Effect signal. My guess is that El Nino/ La Nina conditions have a very large influence on northern Australian (largely Queensland) minima and consequently Australian minima.
The huge variations we see from the 1970s to about 2000 have decreased. The huge step up of the 1970s (The Great Climate Shift) is clearly visible in the minima, but the next few years will show if it has run its course. I expect it has.
And finally, 2011 was not particularly unusual- except in inland and northern Australia where it was the coldest since 1994.
Appendix: Minimally Adjusted Network
| Cairns Airport | Broome |
| Cardwell | Roebourne |
| Palmerville | Marble Bar |
| Georgetown | Geraldton |
| Richmond | Kalgoorlie |
| Boulia | Southern Cross |
| Camooweal | Merredin |
| Ayr | Wandering |
| Longreach Air | Rottnest Island |
| Barcaldine | Cape Naturaliste |
| Rockhampton Air | Cape Leeuwin |
| Sandy Cape Lighthouse | Jarrahwood |
| Gayndah Air | Bridgetown |
| Cape Moreton Lighthouse | Katanning |
| Amberley RAAF | Albany |
| Charleville | Esperence |
| Cunnamulla | Perth |
| Miles | Adelaide |
| Bollon | Innisfail |
| Alice Springs | Mackay |
| Yongala | Maryborough Qld |
| Rayville Park | Brisbane |
| Strathalbyn Park | Dalby |
| Robe | Hobart |
| Low Head | Gunnedah |
| Launceston Ti tree | Tamworth |
| Cape Bruny | Inverell |
| Mildura | Port Macquarie |
| Kerang | Broken Hill |
| Cape Otway Lighthouse | Newcastle |
| Laverton | Bathurst |
| Rutherglen | Orange |
| Omeo | Dubbo |
| East Sale | Sydney |
| Wilsons Promontory Lighthouse | Bowral |
| Gabo Island Lighthouse | Echuca |
| Tibooburra | Benalla |
| Wilcannia | Melbourne-Moorabin |
| Cobar | Maryborough Vic |
| Tenterfield | Ballarat |
| Glen Innes | |
| Yamba | |
| Jerrys Plains | |
| Mudgee | |
| Point Perpendicular | |
| Moruya Heads | |
| Wagga Wagga | |
| Deniliquin | |
| Hay |
January 28, 2012 at 7:21 pm |
Ken, You have spent a lot of effort on this. Thank you.
Some questions, if I may. Confessing that I’m not a fan of the “anomaly” method of expression, at which point did you relate temperatures to the 1961-1990 period? Did you process the whole set of stations first, one after another, then use these derived figures in later combinations of stations like north versus south?
If the anomaly adjustment is made later in the process, such as when all north and all south are combined, then one would expect each graph to average out to zero for the 1961 to 1990 period. This is not the presentation in, for example, the combined north Australia versus south Australia graph, where all values along the linear fit are positive (not half negative and half positive) in this term.
Therefore, the appearance of the graph depends on the stage at which the anomaly was calculated and applied.
The next matter is about errors. If the error bounds of the 1961 to 1990 anomaly period are not known, then it is hard to estimate the confidence that can be applied to the anomaly method.
The 1960-1990 base period is troublesome also because it contains the large shift of the 1970s, whose size probably varies from station to station a great dea; it also contains the start of the period of transition from glass thermometers read daily to thermistors read half-hourly (or a similar change in instrumentation. ) I think this is an understimated and understudied source of variance.
For these reasons, I’d be happier to see the base period for comparison chosen as 1910-2010, if you continue to use the anomaly method. I know that 1961-1990 has been used by global convention, but it has a problem when various authorities delete station data from this period without changing the calculation. The same could be said in criticism of the 100-year base period proposal.
These comments are merely food for thought as you contemplate the eaning of this mass of data in the future. They are not meant to be destructive criticisms.
January 29, 2012 at 8:36 am |
Gday Geoff.
I calculate anomalies for each site first, then combine to find annual means for each region. I use 1961-90 to be comparable with BOM say and others. Any series of years spanning a couple of decades in the past can be used. I find anomalies useful for comparing sites even though their raw temperatures can be many degrees different.
My analysis is subject to the same error and uncertainty as anyones, including BOM- +/- 0.3C at the very least, most likely much greater. Averaged over all sites the results are still valuable if you acknowledge the uncertainty. Trends can still be seen but may be swamped by error margins.
Ken
January 29, 2012 at 6:53 pm |
[...] How Cold Was 2011? An Analysis of Australian Minimum Temperatures 1910-2011 kenskingdom [...]
February 2, 2012 at 5:33 pm |
Well done Ken. This is a most logical, valuable and revealing investigation that you have carried out.
I agree with your method of selecting the stations in the MAN network, in comparison with the less well defined method used by BOM.
My investigations, albeit less comprehensive than yours, shows similar results: http://members.westnet.com.au/brigun/aust_temp_trends.html
Your average trend graphs in the various regions are most interesting and clearly demonstrate the geographical variations that occur. To me this shows how an “average Australian” trend is physically quite meaningless as it masks the variation that exists between regions. So just imagine the value of an “average World” trend, combining such diverse regions as central Africa and the Arctic!
Likewise, the combining of observed of daily maximum and minimum temperatures to get a “daily mean” masks the different trends (which you clearly demonstrate) between night and day temperatures. Ditto for annual versus monthly trends.
While the presentation of regional mean trend graphs are useful, I would also like to see what these trends are at individual locations. Would it be possible to see your trendlines for stations such as Richmond, Cape Leeuwin and Moruya Heads? My preferred approach is to concentrate on looking at the trends at individual stations – if a similar trend occurs at a number of stations this gives me increased confidence in drawing conclusions. But with an “average trend” only there is no indication of the variation between individual stations.
Keep up the good work Ken – it is valuable analysis and is much appreciated.
February 6, 2012 at 1:11 pm |
Brian, I could put up a post with the 89 individual graphs, time is a constraint at the moment however. Remind me again in the future please! I might put a couple up shortly if I get time.
Ken
February 6, 2012 at 1:24 pm |
Ken, I appreciate that 89 plots would take a lot of space and time. For that reason I suggested only Richmond, Cape Leeuwin and Moruya Heads be presented as examples of your analyses at individual stations – these three stations were chosen fairly arbitrariliy because of their geographical spread but also because these three records are largely complete over the past century. Just when you can spare the time. Brian
February 24, 2012 at 7:02 pm |
Thanks for your site—-have you looked at the bom site for air temperatures at the 12 tidal stations run by the BOM. 3 sites have a level trend , 5 have a lower trend and 4 have a rising trend. Two of the tasmanian sites have wood chip piles within 100 meters so they may be heated by the wood chip piles. The BOM has said that the sites were for tidal movements and the air temperature machinery is not up to Bom standards. Ifeel that if these facts were published ,that the carbon tax should be shelved. The Bom says 20 years is too short for a trend , but the Climate Commission put out the critical decade which did not worry about that constraint. —–trevor
May 14, 2012 at 2:14 pm |
[...] and inland sites have much larger swings than coastal and southern sites (see MAN analysis here and here), therefore any warming trend will be [...]