By: Giles Harrison and Stephen Burt
Or so they sometimes say in the south of the United States. But without a reference ginger mill or ready access to Hades, how do we know how hot it really is, and how much can we trust the measurements of the record temperatures we had in July? The basics of air temperature measurement are simple enough – put a thermometer in the shade and keep air moving past it – but the details of doing this matter a lot. And perhaps in all the flurry about records, this detail isn’t so widely appreciated. For example, how many times have you heard a radio phone-in programme asking listeners for car or garden temperature readings to compare, or a tennis commentator mentioning the temperature on centre court at Wimbledon? For a thermometer anywhere in direct sunlight, sheltered from the wind, its temperature is just that of a hot thing in the sun. It’s highly unlikely to be a reliable air temperature.
Meteorologists have worked on this problem for a long time. The first liquid-in-glass thermometers appeared in Renaissance Italy in the 1640s, gradually becoming more reliable and consistent during the eighteenth century. Temperature measurements slowly became more widespread in Europe as thermometers improved, and became particularly well organised internationally in the eighteenth and nineteenth centuries. Some of the earliest reliable air temperature measurements began in national observatories making astronomical or geophysical measurements for which the temperature was merely needed as a correction factor, and many of these early “temperature series” still continue. The needs of modern climate science have made understanding these early meteorological technologies, and the exposure of the instruments, much more important.
Figure 1: Thermometer screens. (Left) Stevenson-type screen at the Reading University Atmospheric Observatory. (Right) Beehive screen at the meteorological site of the Universitat de les Illes Balears, Palma. Both sites also have nearby wind measurements.
To provide protection from direct sunlight, long-wave (terrestrial) radiation and other demanding environmental factors such as rain, while retaining airflow, thermometers are usually placed within a semi-porous shelter or shield, often referred to as a thermometer screen. Screens are almost always made from white material (externally at least) to reflect sunlight: many different designs are in use internationally. At a meteorological site they should be positioned for good airflow and arranged so that the hinged door to read the thermometer opens on the shady side. In later versions of the widely adopted thermometer screen originally designed by the lighthouse engineer Thomas Stevenson (1818-1887, and father of Robert Louis Stevenson), double-louvred slats are used to form the sides of the screen, to maximise thermal contact with the air passing through. Smaller cylindrical “beehive” screens based on the same principle containing smaller electronic sensors are now also widely used (figure 1).
The accuracy of the air temperature recorded by a screen depends on three main factors: how closely the in-screen temperature follows the air temperature, how quickly the sensor responds to changes in temperature, and of course the accuracy of the sensor used. A meteorological thermometer is typically a liquid-in-glass device (e.g. a mercury thermometer), or an electronic sensor, such as a platinum resistance thermometer. With their lower mass, the latter can respond more quickly than the former, so the World Meteorological Organisation (WMO) sets out observing guidelines on sensor response time, mandating that temperature measurements be averaged over 60 seconds. This helps ensure comparability of records between different instrument types (and thus historical records) and avoid spurious very short-duration maximum and minimum temperatures. Thermometers (whether liquid-in-glass or electronic) are calibrated by comparison against reference devices in laboratory experiments, and the corrections needed derived. With regular calibration checks to eliminate effects of drift, and many other precautions, measurements accurate to 0.1 °C become possible.
Figure 2: Temperature difference (Tdiff) between a thermometer in open air and screen temperature (Tscrn) at the Reading University Atmospheric Observatory, plotted against (left) screen temperature and (right) wind speed at 2m (u2), which is approximately at the screen height. (Modified from [2]).
The question of how closely the screen temperature represents the air temperature is much more difficult, as to assess it perfectly the true air temperature itself would be needed. Comparison against a reference temperature better than that of the screen is all that can be done, and the precision experiments necessary are difficult to maintain for anything other than short periods. Comparisons (or “trials”) between one design of screen and another are more common, and tend to be undertaken by national meteorological services. These of course only show how to account for changes in screen design, but not the fundamental question of how well air temperature itself is determined. Nevertheless, from the few investigations available, WMO states[1] that worst-case temperature differences between naturally ventilated thermometer screens and artificially-ventilated (aspirated) sensors and air temperature lie between 2.5 °C and -0.5 °C. With temperatures commonly reported to 0.1 °C, this seems astonishingly large! However, in a year-long study[2] at Reading University Atmospheric Observatory using a naturally ventilated screen with a careful procedure to overcome inevitable sensor breakages, differences as large as this were indeed occasionally observed, skewed to the same warm bias of the screen indicated by WMO (Figure 2). However, these large differences were exceptional, as 90% of the temperature differences were well within ± 0.5 °C. Figure 2 shows that the key aspect in reducing the uncertainties is the wind flow around and through the screen, because the largest temperature differences occur in calm conditions, both by day and by night. This was originally recognised by the Scottish physicist John Aitken (1839-1919, and more famous perhaps for his pioneering work on aerosols), who argued for forced ventilation through a thermometer screen[3]. Aspirated temperature measurements were hardly ever implemented until recent years, but improved technologies mean they are increasingly regarded as reference climate measurements, in the United States[4] and other countries, although, as yet, very few UK Met Office observing sites are equipped with aspirated sensors.
Ventilation is essential for rapid thermal exchange between the air, the thermometer screen and the enclosed temperature sensor itself, to try to ensure and maintain thermal equilibrium even as the air temperature fluctuates continuously. At low wind speeds, this is much less effective and the time taken for the thermometer screen to “catch up” with external air temperature changes can be quite long, as much as half an hour[5]. Further work[6] at Reading Observatory showed that this was improved to a couple of minutes for near-screen wind speeds of 2 ms-1 or greater, but that for wind speeds less than this, lag times increased considerably. Because winds are often light or even calm at night, this effect is more likely to affect a night-time minimum temperature than a day-time maximum. Some maxima or minima may therefore still be under-recorded in a poorly ventilated screen, in a sheltered observing site or in light wind conditions. For temperature measurements made in screens, the response time of the screen is greater than that of the sensor – sometimes many times so in light winds: for aspirated temperature measurements, in contrast, the sensor response time alone is the determining factor.
Figure 3: (left) Screen temperature (Tscreen) measured at Reading Observatory on 25th July 2019, and (right) screen temperature plotted against wind speed at 2 m (u2), using 5 min average values. The dashed red line marks Tscreen= 35° C, and the dotted blue line Tscreen= 20 °C.
Looking at the measurements made at the well-instrumented Reading Observatory for Thursday 25 July 2019 (Figure 3), the wind speed at 2 m (u2) is well correlated with the screen temperature. For the times when Tscreen was greater than 35 °C, the median u2 was 2.3 ms-1: in contrast, when Tscreen was less than 20 °C, the median u2 was 0.3 ms-1. This shows that, although the daytime maximum was well ventilated, this is not true of the nocturnal temperature minimum, which will have been less reliably determined.
The actual moment of temperature maximum is a very local phenomenon, amongst other things depending on airflow over the site, positions of heat sources and soil characteristics, urban heat island effects and, most commonly, the presence of cloud. For example, on 10 August 2003, when Reading recorded its hottest day to date at 36.4 °C, cloud materialised at Reading just before the time of the maximum in air temperature, and probably prevented a greater temperature being reached[7]. Even for the Reading Observatory thermometer screen on 25 July 2019, which was moderately well ventilated, temperature fluctuations lasting a few minutes, as might well have been generated beneath the broken clouds which were present, would be damped out.
The variations in maximum temperatures across nearby sites probably experiencing similar conditions on 25 July are interesting to compare (Table 1). Differences in radiative environment between extensive tarmac (Heathrow) and bleached grass surfaces (Kew Gardens) are perhaps not as great as might appear, as both had identical maximum temperatures. On the other hand, the more open instrument enclosure at Teddington (NPL) probably contributed to a slightly lower maximum temperature there than at other London sites.
Table 1. Maximum temperatures reported on 25 July 2019.
Reading 36.3 °C (from automatic system: maximum thermometer in screen 36.0 °C)
Heathrow 37.9 °C
Northolt 37.6 °C
Kew Gardens 37.9 °C
St James’s Park 37.0 °C
Teddington 36.7 °C
The median of these is 37.3 °C, with an inter-quartile range of 1.05 °C, so there is no doubt that temperatures were consistently that of an extremely hot UK summer day. Local factors, however, are evidently hugely important in determining which site “wins” the maximum temperature record. We now know that the new record UK screen temperature of 38.7 °C occurred at the long-running climatological site at the Botanical Gardens in Cambridge. From the arguments above, whether the air temperature there was indeed greater than that at Faversham in August 2003 (where the screen then recorded 38.5 °C, and was in many respects seriously anomalous anyway[8]) is rather difficult to say – neither site provided simultaneous wind data at screen height, for example.
An extreme “record” screen temperature value at any one site may consequently be of only limited quantitative usefulness, given local variability and inherent limitations in the measurement, although of course nothing here regarding the details of local measurements changes the robust result that globally, temperatures are rising. The maximum temperature continues to be of remarkably widespread interest, even if it isn’t well appreciated how it arises, how reliably it can be measured and whether – if only the newspaper headline writers knew it – that it could well be platinum rather than mercury which yields it.