The Australian Temperature Record- Part 6: Victoria

Ken Stewart, June 2010

Each new state that I study brings a fresh surprise, and that certainly applies to Victoria.   My Great-Grandparents were farmers near Glenrowan in Victoria.  They lived near the Kellys.  Ned Kelly and his gang of bushrangers were horse thieves, bank robbers, and murderers, but have passed into history as folklore heroes.  The expression “Ned Kelly’s not dead yet!” is an Australianism meaning “What a rip off!”  This can be used in situations such as buying something and finding it has a huge mark up, say 133%.  The significance of this item of trivia will soon become obvious.

The Australian Bureau of Meteorology (BOM) has its headquarters in Victoria’s capital, Melbourne, and develops its climate Trend Maps and Time Series graphs from 100 sites nationwide, comprising the High Quality Australian Site Network.  13 of these are in Victoria. 

Here is a map of the sites:

Here’s the Trend Map

And Time Series Graph:

As you can see, BOM declares a warming trend of 0.09C per decade, or 0.9 degrees C for the last 100 years.

I am engaged in what I believe is the first ever independent study of the complete High Quality Australian Site Network, and as part of that project have analysed the data from the Victorian sites.  I averaged maxima and minima for all stations at each site, then compared with the High Quality means.  A spokesperson for BOM has asserted that: 

“On the issue of adjustments you find that these have a near zero impact on the all Australian temperature because these tend to be equally positive and negative across the network (as would be expected given they are adjustments for random station changes).”

Compare this with the results:

By these calculations (averaging the trend at each site) the raw trend is 0.35 degrees C per 100 years, and the High Quality state trend is 0.83C.  That’s a warming bias of 133%!

When the annual records for all 13 sites are averaged and graphed, the official trend is 0.85 C.  (BOM rounds to 0.9).

Graph of Raw and High Quality data for the State of Victoria:

Also note that the raw trend is practically identical to the table above, and the warming bias  for the state becomes 142.8%. 

This is a graph of the magnitude of the adjustments to the raw data from least to greatest.

Note that two stations’ trends are unchanged, one is cooled, and the rest have strong warming adjustments.  Omeo gets the gong for the greatest adjustment in the whole of Australia (so far).

Let’s look at these 13 records.

Three stations (Mildura, East Sale, and Cashmore) were listed in 1996 as being Urban, and therefore not supposed to be used in developing the temperature record.  Yet there they are.  As well, Laverton is now encroached by the outer suburbs of Melbourne and should also be regarded as Urban.

Because of the poor quality of records, I frequently had to splice data to see what was happening. My splices are not corrections but attempts to match separate records.  All increase the raw trend.

Mildura raw:

Note the Merbein CSIRO farm’s data is close to the airport’s, so the adjusted data is a good match:

Swan Hill raw:

The homogenised data has been adjusted down about 0.6, resulting in a steeper warming trend.

Kerang– shows Boort as well.  I used Boort less 0.1 to infill the splice.

But the adjustments increase the warming.

Nhill in western Victoria, shows cooling of about -0.2C.

But that has been turned right around!

A change of 0.95C in trend!

Ararat Prison and Ararat town data do not overlap, and there are large gaps:

Hard to join the two, so I checked Stawell.

There are overlaps with both datasets, so I reduced Stawell by 1.0.

 A good match but 0.3 above the prison, so I reduced the PO data by 0.3  and spliced with the prison.

So how did they make that adjustment?

Cashmore Airport, in SW Victoria:

To make a long period record, BOM needs to splice with other records.  As it happens, the Portland Airport was moved to Cashmore (about 14 km away) to get away from the aluminium refinery.  There are in fact five separate records for the Portland area.

The original Portland site with many gaps was superceded by Cape Nelson Comparison in the 1950s.  The Airport (about 1.5km from the old Portland site) recorded data for a few years with a good overlap with the lighthouse about 9km away.  This is the splice I made: old Portland less 0.3; Cape Nelson Comparison less 0.8 and Cape Nelson less 1.05 to match Cashmore.

I think that is a fair splice, but look at BOM’s adjustment:

Remember, Cashmore Airport is built from urban data, and shouldn’t have been used by BOM at all.

Cape Otway Lighthouse raw:

And adjusted:

Warmed again.

Laverton RAAF.   It was once out in the country.  Explosives factories were built nearby (out of town for obvious reasons) but in recent years Melbourne has expanded and Laverton is an outer suburb.

No problems with the adjustments:

Rutherglen Research Farm raw:

Nearby Wahgunyah, Viticulture College, and Post Office compared to the Research Farm:

which shows a good match between Post office and Viticulture College, and apart from 1922-1923-1924 the Viticulture record considerably above Rutherglen- and cooling.  So the flat Rutherglen record should be reliable.  However…


Cooling trend if you ignore 1935!  But Omeo has been adjusted:

That’s the biggest adjustment so far!

East Sale Airport raw:

No overlap at all.  But there are several stations in the Sale area:

Note the very close match of Maffra Forestry (16.6km away) with the other records, except for around 1970.  There was sure to be a discontinuity in the Maffra records, but to match the Airport I reduced the whole Maffra record by 0.3 to make a useable splice:

The splice gives a trend of less than 0.3C, but compare with the High Quality trend:

Wilson’s Promontory Lighthouse, the most southerly point of the Australian continent:

Slightly warmed, who knows why. 

Gabo Island Lighthouse, just off the far eastern tip of Victoria:


The only station out of the 13 to have its trend reduced!


There is a distinct warming trend in Victoria since the 1960s, which has been especially marked in the last 15 years.

The first half of the record shows a cooling trend.  BOM’s adjustments have attempted to remove this.

2007, not 2009, was the warmest year in the past 100 years.

Three stations identified as urban in 1996 have been included.

Many stations’ data have been arbitrarily adjusted to cool earlier years

Only one station has had its trend reduced.  Two are essentially unchanged.

Ten of Victoria’s 13 stations have been adjusted to increase the warming trend, to the extent that there is a warming bias of at least 133%, more likely 143%. 

These adjustments, and the Australian temperature record to which they contribute, are plainly not to be trusted.  

Ned Kelly’s not dead yet, and he works in the Bureau of Meteorology.   133%? I don’t buy it.

Progress report on the Australian High Quality Site Network:

Sites checked:             83  out of 100

Raw trend:                  +0.72 degrees C/ 100 years

High Quality trend:     +0.96

Average difference:    +0.24

Warming bias:           33.22%

There will have to be a massive cooling bias in the remaining 17 stations to return the overall Australian adjustments to neutral.

26 Responses to “The Australian Temperature Record- Part 6: Victoria”

  1. Niche Modeling » Australian Temperature Records in Question Says:

    […] Ken Stewart is engaged in the first ever independent study of the complete High Quality Australian Site Network, and as part of that project has analysed the data from the Victorian sites. He averaged maxima and minima for all stations at each site, then compared the result with the High Quality means. By these calculations (averaging the trend at each site) the raw trend is 0.35 degrees C per 100 years, and the High Quality state trend is 0.83C. That’s a warming bias of 133%! […]

  2. dribble Says:

    Thank you for this excellent article. One confusion I had was with the number of High Quality stations you quote for Victoria. There are more than 13, such as Geelong Airport, Hamilton airport etc which are shown on the BOM website but I assume that you are only checking those used in detemining the BOMs trend maps and time series graphs. Is this correct? Are you using only those stations deemed rural in your survey?

  3. Ibrahim Says:

    Old book 1913


  4. Annabelle Says:

    Excellent work!

    Now you will probably be slimed by Tim Lambert.

  5. Nick Says:

    What reasons do you have for claiming “these adjustments are not to be trusted” and that they are done “arbitrarily”?

    • kenskingdom Says:

      Read my posts, starting from Part 1. Check the evidence with your eyes. The adjustments are supposed to average out to neutral.


  6. AS Says:

    Hi Ken, any chance you can provide links to the data sets you’re using? In particular I’m after the “raw” (unadjusted) data. Thanks.

    • kenskingdom Says:

      Check each station including closed ones


    • kenskingdom Says:

      I will post showing exactly the steps involved in how to get at the raw data more precisely in a few days time.

      • AS Says:

        Thanks Ken. It would help to know which stations you combined (spliced) to get the fuller record. Take Swan Hill for example, there are 30 “nearby” stations listed in addition to Swan Hill. Did you combine ALL stations to create the complete record or only some? If some, which some and what were your criteria for selecting stations? Do your combined stations match those used by BOM? Also, did you statistically process the stations (eg area weighting) or simply combine the raw (unprocessed) data? Cheers.

        • kenskingdom Says:

          AS, I only spliced the stations right in the town (or up to 15 km away in the case of Cashmore). There were some cases in WA and Tas where I had to go further afield to see where the HQ figures could be coming from but generally there were only 2 (sometimes more) stations eg post office or similar and airport. They are listed in the graphs. If I had to do a manipulation of the data eg subtracting from the raw data to make it fit to another station to see how it matched with HQ I mentioned that in the notes to the graph. No area weighting, simply combined. Check the graphs.

  7. The Victorian Warmed Period | Watts Up With That? Says:

    […] school principal Kenskingdom was alarmed by this Bureau of Meteorology graph, showing a strong warming trend for Victoria, […]

  8. kenskingdom Says:

    David S wrote:
    You are being kind to them. I had a look at Laverton, as it was featured in a previous post on WUWT, and the population in the surrounding area had increased 10 fold since 1950. Yet there was no attempt to adjust the recent figures down, or the older figures up, in a steady fashion to reflect the UHI effect. If anything the true trend is even lower than you have identified.

    Moved as requested.

  9. Verity Jones Says:

    Excellent. What a marathon effort! I started to look at the Australian data used by GISS, but never really did that much with it, partly because it was so poor (lots of ‘dropped’ stations etc.).

    It is the same story as this all over the world with GISS (although changes tend to be less extreme). BOM, GISS, NOAA (can’t say with CRU as I haven’t looked) all seem to feel the need to homogenize stations such that all stations in an area have a homogeneous trend. Even my rudimentary knowledge tells me that not all sites will respond the same to longer term variations in the climate affecting weather and the predominence of weather patters. So certain sites will have more temperature inversions, wind direction from a moderating influence such as the sea, etc. OK enough ranting….. I share your frustration at this.

  10. Frank O'Dwyer Says:

    “By these calculations (averaging the trend at each site) the raw trend is 0.35 degrees C per 100 years, and the High Quality state trend is 0.83C. That’s a warming bias of 133%!”

    This is an amazing combination of mathematical and actual illiteracy. Impressive. The original claim is about the temperatures, not the trend.

    “On the issue of adjustments you find that these have a near zero impact on the all Australian temperature because these tend to be equally positive and negative across the network (as would be expected given they are adjustments for random station changes).”

    Do you see the word trend there?

    The point is the average adjustment should be zero. That doesn’t mean the resulting trend should be zero, or the same as the unadjusted trend (the unadjusted trend is pretty much meaningless).

    Of COURSE adjusting the temperatures will affect the trend since the purpose is to remove the noise and reveal the true trend.

    • kenskingdom Says:

      Thank you for your contribution to my literacy.
      So the average adjustments should be zero. Have you bothered to check if this is so? I have. Mildura, whose adjustments have not led to a higher rate of warming than the raw data, has an average adjustment of -0.17C. Some up, some down. Gabo island Lighthouse, which was the only one cooled, has an average adjustment of -0.07- pretty close to zero. Omeo, the leader of the pack, has an average adjustment of -0.34C. By the way, that’s per year. If you look at the graphs I think you will see what I mean- the difference between the raw mean and the adjusted mean at each year is pretty obvious. And sorry, but the temperatures year by year do produce a trend. And I agree with you, I have come to the conclusion that any trends found in the last 100 years are pretty much meaningless, but the trend announced by BOM and CSIRO is totally meaningless.

  11. Geoff Sherrington Says:

    Warwick Hughes paid for a large amount of BOM data in 1993. In the graph below for Omeo, I have subtracted the Tmean each year of Warwick’s data from other sources as noted on the graph.

    You can see that adjustments by others have produced differences as large as 2.1 deg C in a year.

    This creates a problem for a new researcher (say) wanting to calibrate some new proxy against the instrumented period. If the researcher did not know that many variations of temperature exist, his calibration could be shot to pieces by the wrong choice. What is worse, the zero value I have used for WSH 1993 in this difference graph has already undergone BoM adjustments in the style of Torok and others. However, it is near impossible to obtain raw data as shown on the data collector’s records.

    The importance of this is becoming increasingly pressing as KNMI, which has a magnificent site, continues to list data that an ususpecting person might choose because of the site’s reputation. There is a lot of fine print to be read on KNMI before one can choose an appropriate data string.

    It’s really up to the BOM to produce a definitive version at least of the Reference Climate Series of 100 or so stations, then to insist that the BoM series is the only valid one. It’s not really valid for NOAA, NCDC, GISS, GHCN, CRU, Albert E Newman or anyone else to take liberties with these data. There is a best version and there are other versions that should be dropped forever.

    Those who wish to work more on the Omeo data can use these numbers: (Sorry for the length, Ken).

    Year WSH cd BOM online Giss homogenise KNMI
    1879 12.32 12.40 10.46
    1880 12.60 12.60 11.22
    1881 13.35 13.35 13.45 11.94
    1882 12.59 13.00 12.1 11.20
    1883 12.10 12.10 12.13 10.67
    1884 11.95 11.95 12.02 10.53
    1885 12.75 12.75 12.53 11.35
    1886 11.90 11.90 12.14 10.51
    1887 11.55 11.55 11.57 10.12
    1888 11.40 11.40 11.44 9.99
    1889 11.70 11.70 11.82 10.33
    1890 12.15 12.15 12.27 10.73
    1891 12.15 12.15 12 10.72
    1892 11.80 11.80 12.04 10.40
    1893 11.90 11.95 11.85 10.53
    1894 11.50 11.50 11.43 10.06
    1895 12.05 12.05 12.24 10.73
    1896 12.60 12.55 12.62 11.16
    1897 12.10 12.10 12.11 10.73
    1898 12.55 12.60 12.67 11.14
    1899 12.75 12.75 12.33 10.98
    1900 12.40 12.35 12.46 10.96
    1901 12.25 12.25 12.3 11.26
    1902 12.05 12.05 12.2 11.05
    1903 11.45 11.45 11.52 10.44
    1904 11.25 11.25 11.14 10.28
    1905 10.80 10.80 11.03 9.83
    1906 11.55 11.55 11.55 10.58
    1907 11.25 11.25 11.14 10.27
    1908 11.55 11.55 11.58 10.56
    1909 11.00 11.00 11.22 9.41
    1910 11.85 11.85 11.92 10.28
    1911 11.60 11.60 11.52 10.00
    1912 11.70 11.70 11.73 10.15
    1913 11.60 11.60 11.53 10.02
    1914 12.75 12.75 12.68 11.14
    1915 12.05 12.05 12.17 10.44
    1916 11.45 11.45 11.53 9.88
    1917 11.50 11.50 11.42 9.88
    1918 11.60 11.55 11.69 9.96
    1919 12.25 12.25 12.13 10.65
    1920 11.50 11.50 11.58 9.90
    1921 11.85 11.85 11.98 10.27
    1922 11.65 11.65 11.61 10.08
    1923 11.60 11.60 11.62 10.02
    1924 10.85 10.85 11.09 9.26
    1925 11.10 11.10 10.93 9.47
    1926 11.70 11.70 11.73 10.08
    1927 11.20 11.20 11.19 9.61
    1928 11.75 11.75 11.73 10.18
    1929 11.40 11.35 11.57 9.74
    1930 11.80 11.80 11.68 10.71
    1931 11.20 11.20 11.28 10.08
    1932 11.30 11.30 11.39 10.20
    1933 11.25 11.25 11.14 10.14
    1934 11.60 11.60 11.7 10.48
    1935 11.03 10.95 11.2 10.11
    1936 10.05 10.05 10.08 8.98
    1937 10.45 10.45 10.53 9.36
    1938 11.60 11.60 11.56 10.53
    1939 12.10 12.10 12.27 10.18
    1940 12.50 12.50 12.5 10.38
    1941 11.40 11.40 11.29 9.46
    1942 12.10 12.05 12.02 10.09
    1943 10.65 10.65 10.88 8.72
    1944 10.72 10.75 11.18 9.35
    1945 10.90 10.90 10.88 9.34
    1946 11.10 11.10 11.15 9.55
    1947 11.40 11.50 11.55 9.93
    1948 10.80 10.80 10.83 9.26
    1949 10.80 10.80 10.91 9.25
    1950 11.65 11.65 11.56 10.13
    1951 11.40 11.40 11.53 9.88
    1952 11.25 11.25 11.23 9.66
    1953 11.15 11.15 11.33 9.80
    1954 11.40 11.40 11.38 9.86
    1955 11.30 11.30 11.43 9.76
    1956 11.10 11.10 11.18 9.58
    1957 11.25 11.25 11.07 10.09
    1958 11.20 11.20 11.52 10.08
    1959 11.80 11.80 11.73 10.64
    1960 11.30 11.25 11.22 10.18
    1961 12.35 12.35 12.38 11.18
    1962 11.10 11.10 11.19 9.98
    1963 11.25 11.25 11.22 10.09
    1964 10.85 10.85 11.08 9.67
    1965 11.00 11.00 10.7 10.39
    1966 10.85 10.85 11.13 10.28
    1967 11.05 11.05 11.15 10.47
    1968 11.80 11.80 11.75 11.19
    1969 11.50 11.50 11.56 10.87
    1970 10.85 10.85 10.68 10.22
    1971 11.25 11.10 11.13 10.49
    1972 11.35 11.35 11.27 10.72
    1973 11.90 11.85 11.9 11.22
    1974 11.30 11.30 11.46 10.68
    1975 11.65 11.65 11.53 11.08
    1976 11.20 11.20 11.34 10.56
    1977 11.41 11.50 11.5 10.90
    1978 11.35 11.35 11.49 10.73
    1979 11.55 11.55 11.41 11.59
    1980 11.72 11.85 11.74 11.82
    1981 11.70 11.80 11.94 11.77
    1982 11.55 11.55 11.45 11.57
    1983 11.45 11.45 11.56 11.47
    1984 10.45 10.45 10.66 10.48
    1985 10.85 10.85 10.77 10.88
    1986 10.75 10.75 10.45 10.58
    1987 10.70 10.70 10.72 10.72
    1988 12.05 12.05 11.87 12.05
    1989 10.85 10.90 11.06 10.92
    1990 11.43 11.40 11.31 11.47
    1991 11.15 11.55 11.77 11.58
    1992 10.70 10.70 10.57 10.76
    1993 11.30
    1994 11.20
    1995 10.80
    1996 10.70
    1997 11.70
    1998 11.90
    1999 11.85
    2000 11.80
    2001 11.75
    2002 11.55
    2003 11.80
    2004 11.90
    2005 12.15
    2006 12.15
    2007 12.7
    2008 11.7
    2009 12.6

  12. Roy Martin Says:

    Re: Wilson’s Promontory data you said: “Slightly warmed, who knows why.”

    Could be proximity to the ocean. I note while regularly scanning temperatures in Victoria, at BOM, that the stations at Cape Nelson, Cape Otway, Rhyll, and Wilson’s promontory are quite often one to two degrees C. higher than the nearest inland stations.

    The maps of ocean temperatures on the Weatherzone website (from NOAA) show adjacent SST of about 15 deg.C., which is higher than prevailing air temperatures. SST anomaly is currently +1 to +2 deg.C. There are considerable fluctuations in the SST in the region to the south of Australia, but I do not know if the current positive local anomaly correlates temporally with the higher temperature air Wilson’s Promontory.

    • kenskingdom Says:

      Yes I agree the proximity to the ocean ( and distance from it) appears to be a large factor in the trend in raw temps. The “Slightly warmed, who knows why.” comment is that the adjustements have slightly increased this trend as well.

  13. Geoff Sherrington Says:

    Proximity to Ocean, of itself, need not produce a trend over time. It will not produce a significant trend if the climate is close to an equilibrium. It might produce a trend if (e.g.) there was a fast regional cooling and land had a different thermal inertia to sea.

    This is one ofthe problems I have with a subset of rural stations I’m working on, where 7 coastal stations have barely changed temp in 40 years and 7 inland stations have shown a (linar) rise of about 1 deg per century, some rather more. Such a differential cannot go on forever, because it becomes impossibly large if you extrapolate for several decades and it is not there if you go back several decades.

    Ken, might I suggest that now you have reached this stage, you can divide your national rural stations into coastal and inland (more than 100 km) to see if there is a systematic difference? The more stations used, the more it helps unravel the problem, which at this stage cannot exclude instrumental and adjustment deficiencies.

    • kenskingdom Says:

      Gday Geoff
      Yes that’s one of the significant issues I have discovered, and it appears to be different in North Australia from Southern Australia regions. Points noted, it is odd. I will be looking at that shortly, as well as the Urban sites to see what’s been happening.

  14. Ed Moran Says:


  15. CC Says:

    I am a wind analyst and do much of my work in the state of Victoria. Just came across this conversation and thought I would add to it. Hopefully it is not dead and this is not a message in a bottle – but here goes.
    On a daily basis the wind undergoes a large coastal effect creating a massive shift between stable and unstable atmosphere. More so than in any other place I have studied. Its a phenomena that has caused many mistakes when analysing the wind. Temperature inversion (please google) is also very common in the region. Slight changes in temperature over massive areas (Australia and the ocean) will have an effect over this diurnal wind pattern. The wind and movement of air at differing temperatures will have a large effect on measured temperatures. In the state of Victoria, more than most places this will be evident. We can’t look at temperature without looking at weather patterns. Wind and temperature are forever caught in cycles of positive and negative feedback…..still working on that one….Also for what its worth, in my experience of the data looks funny…doesn’t mean its wrong (or right)

    • kenskingdom Says:

      No your comment is not lost. Changing wind, cloud patterns are behind most temperature change I believe. Thanks for your comment- I’d like to hear more from you!

Comments are closed.

%d bloggers like this: