WYE – 2nd Addendum – How Round Up kills data accuracy.

The Met Office runs its Location Specific, Long Term Climate Averages Stations data for the 30 year periods starting 1961 to 1990, 1971 to 2000, 1981 to 2010 and 1990 to 2020. The middle two periods formerly appeared on the web link above but are now omitted and only available by individual request.

Following my review of the WYE weather station, Dave Woolcock kindly supplied graphical analysis from the digital CEDA archives to show the temperature trend from 1959 to site closure in 2006. Wye actually had data to view available online (as copies of the original manuscripts) from 1924 so I opted to start transcribing these earlier readings to a digital format to allow Dave to do the analysis further back into history for longer term comparison. Then things became very interesting.

The historical context of Met Office temperature observations has to be taken in regard of the scales used, the measurement accuracy and observation methodology.

SCALES.

Up to 1961 the vast majority of temperature observations were noted using the scale devised by Daniel Gabriel Fahrenheit in 1724. This scale references (at mean air pressure) the freezing point of water as 32°F ( equivalent to 0°C) with the boiling point of water as 212°F (equivalent to 100°C). Thus the water properties referenced in the 100 degree range of Celsius are 180 degrees of the Fahrenheit units.

In simple unit terms, 1°C = 1.8°F and 1°F = 0.55°C.

HISTORICAL ACCURACY.

The Met office originally ranked weather stations by “Order” with “First Order” being regarded as the most “important” and lower rankings less so. This was reflected in the instrumentation standards. Lower order stations generally recorded temperatures to the nearest whole degree Fahrenheit. Higher rated sites recorded to the nearest 0.5°F and the very best “First Order” ones recorded to the nearest 0.1°F. The Met Office Surface Data Users Guide explicitly confirms these standards relative to the more recent Celsius recording.

1. 0.1C for all temperatures that are reported today
2. 1C for hourly temperatures received in SYNOP code prior to 1982
3. 1F for many maximum and minimum temperatures prior to 1961
4. 0.1F for most air temperatures prior to 1961 and for some maximum and minimum temperatures prior to the 1920s
5. 0.5F for maximum and minimum temperatures at a few stations prior to 1961

The following lines also appear giving further information

“There is a convention, followed by the human observer, that where a value is read as half way between two points on the observing scale it is “rounded to the odd“

“All temperatures in MIDAS have been converted to Celsius and are stored with a precision of 0.1C.” 

Observation Methodology

Accurate reading of Liquid in Glass Thermometers is by no means as straightforward as reading the digital screen on a modern Platinum Resistance Thermometer Data logger. Observers attempting to read as small a unit as 0.1°F( 0.055°C) needed specific training to avoid parallax errors as shown below.

A long term observer told me of various techniques he used to avoid such errors, without which, even a change of footwear could adversely affect accuracy. A great deal of diligent effort was required.

Relevance to Wye.

In starting to convert the pre 1959 readings at Wye from Fahrenheit to Celsius I initially referred to my findings at Swanage and Faskally. Here the transcribers had inconsistently employed the Met Office rounding system which is to round down any second decimal point (100ths) below 0.05 and round up any second decimal point readings from 0.05 and above.

This rounding is particularly awkward when 1°F = 0.55°C. Thus 33°F (1 degree F above freezing) equals 0.55°C and is rounded up to 0.6°C. Conversely 31°F (1 degree F below freezing) converts to -0.55°C which is rounded up (?) to -0.6°C. The 2°F range from 31 to 33 perversely becomes recorded as a 1.2°C differential when it is in fact only 1.1°C. The round up principle effectively hits every other degree conversion from Fahrenheit and means that only certain fixed points appear in scale converted archives. Below is an example of Swanage records showing 0.6°C and -0.6°C following conversion.

Unfortunately, as was shown at Faskally, not all transcribers seemed to have got the memo on this rounding protocol and “improper” conversions were found.

The above small differences may seem trivial though clearly others certainly consider accuracy very important. “Every part per million and every fraction of a degree temperature increase has a real impact on our lives and our planet. Prof. Celeste Saulo, WMO Secretary-General”

Considering my own attempts to convert readings for Wye I wondered how the more accurate Fahrenheit readings to the tenth of a degree would be dealt with by rounding. Heathrow (formerly known by the Met Office as London Airport) was also known to read to the tenth of a degree Fahrenheit. To my dismay the digital archives showed the exact same conversion points as noted at Swanage i.e. starting from zero being 0.6, 1.1, 1.6. 2.2°C and so on. This suggested that the original Fahrenheit readings had been subject to rounding, converted to Celsius and then hit a second rounding. Let that sink in………

A diligent and well trained observer has taken accurate historic readings for these to be, for want of a better word, “botched” by a likely bored intern/transcriber.

This is an example effect of this treatment: 34.4°F is originally recorded. This is rounded down to 34°F then converted to Celsius as 1.1°C. Alternatively 34.5°F is originally recorded, this is rounded up to 35°C which converts as 1.65°C again rounded up to 1.7°C in the archives.

Thus in the above example, a reading difference of just 0.1°F ( 0.055°C) ends up as a recorded difference of 0.6°C………………. an eleven fold error factor. And this happens throughout the whole temperature range.

Below is the manuscript for Heathrow covering January 1949

And below is what the above period suitably “botched” looks like in the digital archive.

What was that important statement? – “Every part per million and every fraction of a degree temperature increase has a real impact on our lives and our planet. Prof. Celeste Saulo, WMO Secretary-General”

The probable reason why the Met Office Location Specific Long Term Climate Averages only start from 1961 is that any digitally archived readings prior to that date have been subjected to such rounding mutilation in being entered into the public digital archives as to be practically unusable for genuine later period comparison.

If anyone can offer a better explanation for this crude butchery of initially high quality data I would welcome their views. However, the availability of Wye’s original Fahrenheit readings enables an alternative reconstruction of the earlier period’s climate record using the Fahrenheit readings directly. On production of rolling monthly averages these can then, and only then, be converted over to celsius accurately to the second decimal place to compare with subsequent similar monthly figures derived from celsius.

A hopefully easy, if laborious, fix for the Wye data but one that begs the obvious question……..what figures do the Met Office and associated “Climate Scientists” actually use? If anyone genuinely knows and can evidence their claim I would be very interested indeed.

Source: https://tallbloke.wordpress.com/2025/09/29/wye-2nd-addendum-how-round-up-kills-data-accuracy/