Sunday 4 October 2015

Measuring extreme temperatures in Uccle, Belgium

Open thermometer shelter with a single set of louvres.

That changes in the measurement conditions can lead to changes in the mean temperature is hopefully known by most people interested in climate change by now. That such changes are likely even more important when it comes to weather variability and extremes is unfortunately less known. The topic is studied much too little given its importance for the study of climatic changes in extremes, which are expected to be responsible for a large part of the impacts from climate change.

Thus I was enthusiastic when a Dutch colleague send me a news article on the topic from the homepage of the Belgium weather service, Koninklijk Meteorologisch Instituut (KMI). It describes a comparison of two different measurement set-ups, old and new, made side by side in [[Uccle]], the main office of the KMI. The main difference is the screen used to protect the thermometer from the sun. In the past these were often more open, that makes ventilation better, nowadays they are more closed to reduce (solar and infra red) radiation errors.

The more closed screen is a [[Stevenson screen]], invented in the last decades of the 19th century. I had assumed that most countries had switched to Stevenson screens before the 1920s. But I recently learned that Switzerland changed in the 1960s and in Uccle they changed in 1983. Making any change to the measurements is a difficult trade off between improving the system and breaking the homogeneity of the climate record. It would be great to have a historical overview of such historical transitions in the way climate is measured for all countries.

I am grateful to the KMI for their permission to republish the story here. The translation, clarifications between square brackets and the related reading section are mine.

Closed thermometer screen with a double-louvred walls [Stevenson screen].
In the [Belgian] media one reads regularly that the highest temperature in Belgium is 38.8°C and that it was recorded in Uccle on June 27, 1947. Sometimes, one also mentions that the measurement was conducted in an "open" thermometer screen. On warm days the question typically arises whether this record could be broken. In order to be able to respond to this, it is necessary to take some facts into account that we will summarize below.

It is important to know that temperature measurements are affected by various factors, the most important one is the type of the thermometer screen in which the observations are carried out. One wants to measure the air temperature and therefore prevent a warming of the measuring equipment by protecting the instruments from the distorting effects of solar radiation. The type of thermometer screen is particularly important on sunny days and this is reflected in the observations.

Since 1983, the reference measurements of the weather station Uccle are made in a completely "closed" thermometer screen [a Stevenson screen] with double-louvred walls. Until May 2006, the reference thermometers were mercury thermometers for daily maximums and alcohol thermometers for daily minimums. [A typical combination nowadays because mercury freezes at -38.8°C.] Since June 2006, the temperature measurements are carried out continuously by means of an automatic sensor in the same type of closed cabin.

Before 1983, the measurements were carried out in an "open" thermometer screen with only a single set of louvres, which on top of that offered no protection on the north side. Because of the reasons mentioned above, the maximum temperature in this type of shelter were too high, especially during the summer period with intense sunshine. On July 19, 2006, one of the hottest days in Uccle, for example, the reference [Stevenson] screen measured a maximum temperature of 36.2°C compared to 38.2°C in the "open" shelter on the same day.

As the air temperature measurements in the closed screen are more relevant, it is advisable to study the temperature records that would be or have been measured in this type of reference screen. Recently we have therefore adjusted the temperature measurements of the open shelter from before 1983, to make them comparable with the values from the closed screen. These adjustments were derived from the comparison between the simultaneous [parallel] observations measured in the two types of screens during a period of 20 years (1986-2005). Today we therefore have two long series of daily temperature extremes (minimum and maximum), beginning in 1901, corresponding to measurements from a closed screen.

When one uses the alignment method described above, the estimated value of the maximum temperature in a closed screen on June 27, 1947, is 36.6°C (while a maximum value of 38.8°C was measured in an open screen, as mentioned in the introduction). This value of 36.6°C should therefore be recognized as the record value for Uccle, in accordance with the current measurement procedures. [For comparison, David Parker (1994) estimated that the cooling from the introduction of Stevenson screens was less than 0.2°C in the annual means in North-West Europe.]

For the specialists, we note that the daily maximum temperature shown in the synoptic reports of Uccle, usually are up to a few tenths of a degree higher compared with the reference climatological observations that were mentioned previously. This difference can be explained by the time intervals over which the temperature is averaged in order to reduce the influence of atmospheric turbulence. The climatic extremes are calculated over a period of ten minutes, while the synoptic extremes are calculated from values ​​that were averaged over a time span of a minute. In the future, will make these calculation methods the same by applying the climatic procedures always.

Related reading

KMI: Het meten van de extreme temperaturen te Ukkel

To study the influence of such transitions in the way the climate is measured using parallel data we have started the Parallel Observations Science Team (ISTI-POST). One of the POST studies is on the transition to Stevenson screens, which is headed by Theo Brandsma. If you have such data please contact us. If you know someone who might, please tell them about POST.

Another parallel measurement showing huge changes in the extremes is discussed in my post: Be careful with the new daily temperature dataset from Berkeley

More on POST: A database with daily climate data for more reliable studies of changes in extreme weather

Introduction to series on weather variability and extreme events

On the importance of changes in weather variability for changes in extremes

A research program on daily data: HUME: Homogenisation, Uncertainty Measures and Extreme weather


Parker, David E., 1994: Effect of changing exposure of thermometers at land stations. International journal of climatology, 14, pp. 1-31, doi: 10.1002/joc.3370140102.


  1. Kevin O'Neill, I agree with you that that paper is suspicious, but your comment is off topic below this post.

  2. Uccle has a lot to answer for. Located, as Eli recalls, in a valley in the middle of an industrial area it was subject until quite recently to really bad air pollution. The stratospheric ozone Dobson measurements were falsified by SO2 pollution and succeeded in confusing everybunny. Without a doubt one of the worst locations for a climate station in the historical network

  3. :-) Fortunately, we are talking here about the difference between two measurement set-ups. And about the difference between the small bias in the mean compared to the bias in the extremes. It has to be a completely different climate before that would be wrong for this combination of two instruments.

    What worries me more is that the instruments and screens are very well maintained at the main office by people with a lot of expertise. That can also influence the difference and could be very different elsewhere in Belgium. That is unfortunately typical for parallel measurements, you find them most near main offices of weather services.


Comments are welcome, but comments without arguments may be deleted. Please try to remain on topic. (See also moderation page.)

I read every comment before publishing it. Spam comments are useless.

This comment box can be stretched for more space.