Skip to content

Do Automatic Temperature Sensors Overstate Warming?

July 4, 2015

By Paul Homewood




AC Osborn reminds of this earlier post from Pierre Gosselin, over at No Tricks Zone.

Pierre’s post picks up on concerns that the automatic sensor system which we have nowadays could be picking up short term spikes, which the old mercury thermometers would not have had time to react to. This sort of bias would be particularly significant at airports, where such spikes could be large.

This is highly relevant to the recent debate over Heathrow. 


Reposted from No Tricks Zone:


Not only has siting of weather stations near urban heat sources have been a real issue for weather measurement stations worldwide, but so maybe has the recently implemented automatic electronic weather measurement instrumentation.

Reports are appearing that the new automatic system may be producing exaggerated temperature readings. For example this may be the case in Germany: read here and here.

Now we find another example, this one coming from the Alice Springs, Australia station.

According to the Australian ABC news site, the new electronic thermometer measured a scorching 46°C (an all time high) last Tuesday. However an adjacent mercury thermometer showed only 41.5°C, i.e. a huge 4.5°C less! It turns out that the 46°C reading was a “spike” that lasted only a minute before disappearing.

As a consequence, the ABC writes, the “Bureau of Meteorology (BoM) has withdrawn its advice Alice Springs recorded its hottest day on Tuesday, blaming a faulty thermometer for an incorrect temperature reading.”

The old record of 45.2°C was set 55 years ago, in 1960. The ABC quotes climatologist Joel Lisonbee:

It looks like we had an instrument fault with our automatic weather station at the Alice Springs Airport. […] We have some mercury and glass thermometers that did not show that spike to 46C. […]

They showed the maximum temperature yesterday to be only 41.5C.”

According to the ABC, the station is located right next to a “scorching” airport tarmac. So how could the new automatic thermometer produce such a faulty reading?

It seems that these new automatic systems are highly sensitive. As reported here at NTZ, one German weather instrumentation expert conducted an 8.5 year side-by-side comparison test of the new automatic electronic temperature measurement system and the former mercury glass thermometer. That test showed that the new automatic thermometers produced a mean temperature for the period that was a whopping 0.9°C warmer than the mercury thermometer. That result could possibly in part explain why Germany’s annual mean temperature jumped by a similar amount from 1985 to 2000, i.e. the period that Germany transitioned over to automatic measurement.

The Alice Springs inflated reading is an indication that the new measurement system indeed may be overstating temperature readings all over the world, thus adding uncertainties on top of those created by the urban heat island effect.

With the Alice Springs Station, the error was caught and the “record high” was withdrawn. Yet the question remains if this is the case all over the world. How many recent records are in fact not records at all, but rather are merely faulty readings produced from instrumentation and siting issues?

  1. July 4, 2015 5:38 pm

    My own weather station produces the odd spike (but can be programmed to ignore them).
    However I thought that was just a feature of my cheapo amateur equipment. I had no idea that professional equipment did the same!

    • July 4, 2015 6:54 pm

      I use a simple data-logger to track my local temperature, and I just set it to sample relatively slowly – say once a minute. Plenty of sensitivity to capture the general trend, but I also have a mercury thermometer right next to it just as a sanity check.

      I live in the American desert Southwest (Arizona) so it’s lots of fun capturing those amazing highs although this year spring was mild (spelled g-o-r-g-e-o-u-s!). And the summer is shaping up to be relatively mild as well; shucks I was looking for some amazing highs. Well, the season is still young and there’s always next year.

      Two things on “highs”. When I download my temperature data naturally it is a sine wave over 24 hours. I have wondered about using something like a Simpson integration over the daylight hours to obtain a “hot-day” index – units would be Tmptr / Time to establish a criteria for really hot days. A “hot” day would be a long day plus a day with not just an absolute high, but persistent high temperatures. I wonder how that would compare to days with absolute highs vs days with persistent highs.

      Rationale for wondering about that here in Arizona is it can be hot, but cloudy from the monsoon condition – it clears up and the sun comes out and scorches us, then disappears behind clouds and perhaps the wind kicks up and it rains. Thermometer may have hit 115 F, for half an hour and the rest of the day was far cooler.

      The second thing is also related to clouds and haze and that is glare. When it is dry and clear and hot without a cloud in the sky, you had better be wearing a hat and long sleeves; I also wear boots when I’m out in it on days like that. And when out in that kind of heat, find shade! So I wonder if there is a way to measure the glare-factor?

      And all the media talks about is this or that high for the day. It’s only part of the story.

      • July 6, 2015 3:51 pm

        The integrated temperature is the degree day value, used for A/C and heating in colder climes. These are widely published but not used to date in climate discussions so far as I know. Maybe they should be as they are far more important than mean temperature. In the UK the mean temperature is almost always asymmetric rather than a sine wave.

  2. July 4, 2015 6:16 pm

    Have been a met office observer for 50 years. Met Office will be removing Mercury and Alcohol thermometers in the next couple of years. Partly on health & safety grounds. Mercury and glass are now dangerous. UM I wonder what is in new sensors? My automatic does sometimes appear to have spikes but not more than 0.3 deg C

    • July 4, 2015 6:24 pm

      They are the worst kind, i.e. it is difficult to program the w.s. to ignore them automatically!

  3. Kelvin Vaughan permalink
    July 4, 2015 7:55 pm

    As a sunbather I know that just before a cloud starts to move across the sun the temperature rises quite a lot for a few seconds. There must be a diffraction effect.

  4. Ben Vorlich permalink
    July 4, 2015 7:56 pm

    one German weather instrumentation expert conducted an 8.5 year side-by-side comparison test of the new automatic electronic temperature measurement system and the former mercury glass thermometer.

    I my days as a component test engineer when upgrading ATE hardware or for new issues of software we had standard devices with a log of all parameters which were retested on old then on new and the results compared, then we’d test several batches on both and compare any miss matches pass/fail or fail/pass.

    Seems that Met Offices the world over need to sharpen up their procedures, because what the German did should be standard practice.

  5. Ian G permalink
    July 4, 2015 10:20 pm

    When Sydney had its highest temp a couple of years ago, the AWS rose 1C in four minutes and then dropped a degree in the next 5 mins. The highest actual recording was 44.9C leaving the old 1939 record of 45.3C intact.

    • AndyG55 permalink
      July 5, 2015 12:54 am

      The AWS summary the next day listed 45.3ºC.

      There has NEVER been a satisfactory explanation where the extra 0.5ºC came from.

      • Ian George permalink
        July 6, 2015 8:31 pm

        The AWS showed 44.9C at 2:49pm that day and 44.7C at 2:59C. The 45.8C was recorded at 2:53pm (the temp at 2:30pm was 45.1C).
        I still have the screen captures saved but, as you say, there has never been an explanation regarding the ‘spike’.
        I have noticed with my local station that every now and again there is a sudden surge and fall within minutes. I wonder if the BoM will acknowledge there is a sensitivity that affects readings?

  6. Jaime Jessop permalink
    July 5, 2015 1:15 pm

    Looking back at the Met Office’s tweets on July 1, they were reporting 35.7C at 2.43pm. 7 mins later, temp at Heathrow had apparently climbed to 36.2C. Then at 3.29pm they tweeted that the ‘record’ highest July temperature had been recorded at 36.7C. This did strike me as rather odd at the time.

  7. Billy Liar permalink
    July 5, 2015 9:40 pm

    I suspect the thermal inertia of the measurement system allows these ridiculous measurements. No temperature measurement should be recorded without the time constant of the sensor.

    If you use glacial ice as you temperature measurement system the time constant is of the order of several tens to hundreds of years – a lot of spikes get ironed out. If you use a mercury thermometer the time constant is probably of the order of minutes. Electronic temperature sensors have varying time constants from less than a second to a few seconds.

    The sensor time constant is critical information, it must be recorded with the temperature.

  8. Barry Armstrong permalink
    July 6, 2015 4:42 pm

    Many airports were and still are the site of weather measurements. Originally the airports were on the outskirts of larger centres such as Winnipeg where I live. The city has grown up around the airport over the decades What was farm land is now a mixture of residential and industrial use land with the resulting “heat island” effect now including the airport and its temperature readings.

  9. Ole Suhr permalink
    July 6, 2015 9:16 pm

    I worked as an electrical engineer making professional equipment for sound-systems for almost 10 years. There you had an exact similar problem: in the “old days” VU meters were used for measuring the electrical sound signals. VU meters are, however, rather slow (lots of inertia in the magnetic coils of the meters).
    Later the all electronic Peak Programme Meters were introduced. Much much faster – and (of course) giving entirely different readings because they were affected much more by short pulses.

    So how to compare PPM-readings to VU-readings ? Well, you really can’t.

    But what you CAN do, is to use the measurement results of your fast meter (PPM for sound – electronic thermometer for temperatures) to calculate which result you would have gotten using the “old” meters (VU – or mercury thermometers). That way you would easily be able to keep consistency in your temperature readings over time and independent of the thermometer type used.

    Why are the meteorological societies not doing this if they really want measurements to be consistent ?

    • July 7, 2015 7:14 am

      The temperature sensors, like the fast meters, are sensing shorter time span effects without so much damping. Where the system needs to be damped, to avoid oversensitivity, a common method is to take two readings before and after the timed value and average these; a simple and well known algorithm, about 5 minutes each side should get rid of instantaneous effects. Alternatively larger sheaths can be used to get the same damping as the mercury in glass system.

      Correcting readings is more dangerous as the corrections can have errors… as with temperature homogenisation!

Comments are closed.

%d bloggers like this: