Is the Meteorological Office’s Metrology as good as they claim?

Metrology is the science of measurement, including the theoretical and practical aspects of how to measure something accurately, precisely and repeatably. The Met Office is responsible for huge amounts of measurement over area, height and time and over a range of widely different aspects such as variously temperatures , humidity, wind speed, air pressure, solar insolation – the list goes on and on. I was recently contacted by a former Met office observer who offered me some extremely eloquent advice regarding the Met Office claims of being able to take readings to the 5th decimal place such as at Cavendish. By way of a “guest” post I detail his expert comments below. It is a technical piece but very enlightening and well worth reading.

“The Philosophy of Calibration.
The old question often said today is ‘Who examines the examiners?’   In measurements a similar question is asked, ‘How do I know that the instrument against which I calibrate my instrument is any more accurate than that of mine’?   It is a question of ‘Traceability’.

Standards.
In the Middle Ages most large markets had two lines scratched on a wall to define length. Unfortunately there was no consistency between the various lengths until the King, via his government, defined the measurement by means of the length of a piece of metal. This was portable and could be used to judge if fraud was suspected.  However carting the metal bar all over the country could easily damage it by dropping it, etc, so a second (and third bar) was made. These were called Sub-Standards and could be compared to the Standard to confirm their accuracy. There was Traceability.  The original bar was a Standard, and the other bars were called Sub-Standards. Whereas this was acceptable for a market trader, shop or fabric supplier, as the industrial revolution progressed then more accurate standards were necessary.  

The metre was historically defined by the French Academy of Sciences in 1791 as 1/10,000,000 of the quadrant of the Earth’s circumference running from the North Pole through Paris to the equator.    It was a convenient measure as it was of similar to the British yard.   The International Bureau of Weights and Measures in 1889 established the international prototype meter as the distance between two lines on a standard bar of 90 percent platinum and 10 percent iridium. By 1960 advances in the techniques of measuring light waves had made it possible to establish an accurate and easily reproducible standard independent of any physical artefact. In 1960 the meter was thus defined in the SI system as equal to 1,650,763.73 wavelengths of the orange-red line in the spectrum of the krypton-86 atom in a vacuum.

Obviously one requires a dedicated laboratory to measure to such accuracy.   It is also far more accurate than, say, that which is needed to measure my 12 inch metal rule; a Sub-Standard will suffice.   (Do not confuse the word substandard used in this context with something that is of lower quality often sold in disreputable shops.)   A Sub-Standard is an instrument, the accuracy of which can be traced back to a laboratory Standard measurement.
When I was at University in the 1970s the definition of an Ampere was the weight of Silver plated on a certain cell over a given time.  To do this we need
 [1] a resistance in series with the plating current
[2] A voltmeter in parallel with the resistance
[3] a timer
[4] a means of weighing the plated silver.
We need to know the accuracy of each of these devices if we are going to obtain a meaningful result.   This we can do by calibrating each device against a Sub-Standard of higher accuracy than that of the device.   The Sub-Standard can be traced back to a fundamental Standard.

Measurement of temperature.

[a] Mercury in Glass thermometers [MiG]

There are two types of MiGs.
[1] MiGs used to measure the temperature of a fluid or a gas in a closed container.
This type of thermometer has a mark somewhat between 6 and 10 cm from the bulb.   For consistent and accurate results the thermometer must be inserted into the fluid (or into the container) to this depth; failure to do so will result in incorrect readings.    Mercury (or whatever fluid is used by the thermometer) expands with an increase in temperature, but not only the mercury in the bulb, but also the small amount of mercury which is in the centre of the ‘tube’.

[2] MiGs used to measure air or gases.
Since the whole of the thermometer is subject to the gas to be measured the manufacturer has to take the expansion of the mercury in the whole tube.

[b] Platinum [Pt] resistance thermometers.
This type of thermometer relies on the character that metals increase in resistance as the temperature increases.    A naked wire will respond very quickly to small changes of temperature thus a manufacturer will encase the wire in a small container to ensure a more stable and protected product.   Nevertheless the response to a change in temperature will be detected by a resistance thermometer far more quickly than a MiG thermometer.

Calibration of thermometers.
Although Met Office thermometers are for air measurement they are usually calibrated in a fluid medium, usually well agitated water, as it is very difficult to obtain a constant temperature in air or gas.    Water is heated and is well stirred [but no cavitations] and allowed to cool slowly during which readings are taken of the thermometer and the Sub-Standard.   For readings below ambient we need to cool the fluid and allow it to warm slowly.
Sub-standard thermometers may be of several types.   The most convenient to use [in my opinion] is a Hewlett Packard digital crystal thermometer.   Crystals are remarkable things.   If they are cut in one direction the oscillations are not affected by temperature.   This is a useful characteristic for clocks, radio frequency controllers, etc.   Cut at 180° the crystal output frequency will change with a change in temperature in a very stable and repeatable manner.    In the UK the National Physical Laboratory [NPL] at Teddington has equipment that can calibrate a crystal thermometer.
A digital Sub-Standard must be at least 10 times more accurate than the thermometer under calibration, preferably more.   It is always necessary to read the declaration of accuracy of a digital instrument.   It is usually in the form of “ ±X%,  ±1 least significant digit [LSD]”.   Thus a reading of 100 on a digital instrument can be actually between 99 and 101 from the ± the last digit.    It is usual to allow the instrument under calibration to read a whole number and to then record the Sub-Standard, as the latter is more accurate.
An example is:-    
Read the thermometer/instrument under test as it passes a set marked value.    E.g. 20.0°C.   At that moment note the reading on the Sub-Standard.   For the example above we read 20.02°C on the Sub-Standard.   The ± LSD means that the test thermometer at 20.0°C can be between 20.01 and 20.03 ± the manufacturer’s accuracy of the Sub-Standard.

Measuring temperatures at Met Office weather stations.
The Met Office has clear instructions in choosing a site for a Stevenson Screen into which the thermometers are placed.   This short video indicates the gross errors of a certain screen.    
https://www.youtube.com/watch?v=QHi79lj9a-E
When I was a Met Observer in the 1970s all our thermometers were MiG.     Modern sites will have Pt resistance thermometers.   Whereas the chosen site was suitable for a MiG thermometer it may present problems for a Pt thermometer.    For example a site on an airport may present no problems for a MiG; the quicker response of the Pt may catch a quick burst of an aeroplane’s exhaust as it passes en route for take-off or landing.   This will result in a higher Maximum Temperature recorded for that day. {n.b Talkshop emphasis}
https://www.metoffice.gov.uk/binaries/content/assets/metofficegovuk/pdf/weather/guides/10_0230_fs_17_observations.pdf
The Allure of digital readings.
Consider a four digit multi-range voltmeter which has ranges of 1000/100/10/1 volts with a claimed accuracy of 5% ± 1 digit.
If we measure mains voltage on the 1000 V range, the meter reads 234V.    Taking account of the accuracy of ± the last digit, the voltage can be 233V or 235V.    We must add to this the accuracy of the measuring system (± 5% of 234 = 234 ± 11.7V)), i.e. our mains voltage is somewhere between 233 – 11.7 and 235 ± +11.7V.   That is then somewhere 221.3V and 246.7V.
If we were measuring on a PCB on the 1 volt range of the multi-meter and it happened to be 0.234V the same calculations would apply i.e. the voltage can be between 0.2213 and 0.2467.
If the meter was five readable digits I leave you to calculate the max and min readings in the case above.
Many people will read and record all the digits on an instrument whereas, in fact, the last two or three are of no consequence whatsoever.   A perfect example of this was the Met Office site where temperatures were recorded to 4 or 5 decimal places, giving the readings the sense of extreme accuracy, whereas they just showed the stupidity of the observer. {n.b. Talkshop emphasis}
The science of measurements is a complex subject that is not understood, nor appreciated, by most people.   It requires a modicum of common sense, something that seems not to be too common in the Met Office.

——————————————————————————————————————–

So what to make of the above response from the Met Office? The stupidity of the observer?

Source: https://tallbloke.wordpress.com/2024/12/31/is-the-meteorological-offices-metrology-as-good-as-they-claim/