Australian Temperature Data Are Garbage

From the Bureau’s hastily published “Fast Facts”:

“This means that each one second temperature value is not an instantaneous measurement of the air temperature but an average of the previous 40 to 80 seconds.”

That is complete nonsense.

At the end of each minute, the following data are recorded:

  1. Lowest one second reading of the previous 60 seconds
  2. Highest one second reading of the previous 60 seconds
  3. Reading at the final second of the minute.

Firstly, 40 seconds is not one minute, the integration period recommended by the WMO in 2014 and by the Bureau’s own officers in 1997.  Anything less than 60 seconds is not compliant.

Secondly, consider this plot, which is from actual 1 minute temperatures recorded at Hervey Bay Airport on 22 February 2017.  (Data purchased by me from the Bureau).

Fig. 1:

Hervey Bay 1 min 5 to7am 22 Feb

Sunrise was at about 5:40 a.m.  Temperatures do not increase until about 6:30 a.m.  Note the strangely low temperature- the daily minimum- which was reported as occurring sometime in the 60 seconds before 06:00:00.  The BOM would have us believe that each of the values in Figure 1, including the low of 23.2C, are “averages” of the previous 40 to 80 seconds.

Next consider what happens in that minute from 5:59 to 6:00, as per the following plot.

Fig. 2:

Hervey Bay 1 min 0559 to 0600am 22 Feb

We don’t know in which seconds the high and low readings for that minute occurred, so I have shown them for each of 59 seconds.  I have shown the 5:59 and 6:00 readings: both were 25.3C.

Consider how the value at 06:00 was obtained:

If by an “average” (however derived) of less than 60 seconds, the methodology is non-compliant.

If by an “average” of the previous 60 seconds, it must include values that contributed to the High of 25.4C and the Low of 23.2C.

If by an “average” of anything greater than 60 seconds, it must include values that contributed to both the Low and High values, and as well, values that contributed to the 5:59 reading- which is the same as the 06:00 reading.

Similar logic applies to the Low and High readings.

It follows that the intermediate instantaneous atmospheric temperatures that contributed to all three reported “average” values must have ranged from much higher than 25.4C to very much lower than 23.2C.

Look at Figure 1 again.  The air temperature at Hervey Bay on 22 February must have spiked down very much lower than the 23.2C plotted.


In the early morning there is very little near ground turbulence so temperatures do not fluctuate from one minute to the next by very much.  In How Temperature Is Measured in Australia Part 2 I showed that 91% of low temperatures vary from final second temperatures in the same minute by 0.2C or less.  A difference of 2.1C is extraordinary.  Fluctuations greater than that are difficult to believe.

However, in a comment at How Temperature Is Measured In Australia Part 1, Tony Banton, a retired meteorologist, says that the BOM explanation of cooler ground level air mixing upwards is correct.  If we accept that explanation, we must then face the problem of “comparability”.

In 61 seconds, the Hervey Bay AWS has reported temperatures of 25.3, 25.4, 23.2, and finally 25.3 degrees.  The BOM asserts that a liquid-in-glass thermometer will be able to respond as quickly and show similar temperatures- and remember, 23.2C was the morning’s official minimum.

My response: rubbish.  The data for 22 February at Hervey Bay show that no averaging is used at all, and the Low Temperature of 22.3C  23.2C is an instantaneous one second recording from a rogue downwards spike, whatever the cause, whether a natural event or other (e.g. electrical) factor.

Temperatures reported by the BOM are not fit for purpose of accurate reporting of maxima and minima, identifying records, or identifying warming or cooling by comparison with historic liquid-in-glass data.

Tags: , ,

10 Responses to “Australian Temperature Data Are Garbage”

  1. Geoff Sherrington Says:

    “Low Temperature of 22.3C is an instantaneous one second recording ” second last para.
    Did you mean 23.2?

  2. John in Oz Says:

    If readings from a LIG thermometer were plotted at one minute intervals I doubt the graph would look like the erratic results you plot above from the AWS.

  3. climanrecon Says:

    Ken, you may be interested in an analysis I’ve done (still ongoing) of the late 20th century climate shift in Australia. I’ve looked at how much AWS calibration shifts might have contributed to the sudden warming in the region around Deniliquin in NSW, and found nothing significant. Many stations in the area remained manual for long enough to check the AWS ones:

    • kenskingdom Says:

      Thanks, I’ll have a look when I get a chance. I’ve also looked in the past and the problem is the world wide step up in temperature from the mid 1990s to early 2000s makes any other contribution difficult to identify.

  4. climanrecon Says:

    Ken, another post that may interest you, confirming what you wrote some years ago about ACORN-SAT at Wagga Wagga:

    But a warning for you, I believe that the ACORN-SAT corrections for Rutherglen are largely … correct!

    • kenskingdom Says:

      Interesting analysis. Saw tooth patterns of discontinuity “corrected” by ham fisted adjustments to all previous data are not uncommon in ACORN. I wonder if any changes will be made in Version 2.

  5. climanrecon Says:

    Ken, I notice that the daily minima for Hervey Bay for the entire month of February are written with slanted numbers, possibly suggesting equipment malfunction (some later numbers are not slanted):

    Do you have a glitch that has ended up as a non-slanted daily min or max?

    • kenskingdom Says:

      Good morning. I have purchased 1 minute data for Hervey Bay (and several others) for 1/1/17 to 21/3/17. The “Quality of Air Temperature” column has “Y” values right through February. So no dodgy temps were flagged. The bloke from the Bureau certainly didn’t think there was a glitch.

Comments are closed.

%d bloggers like this: