St Catharines Point – The required depth of data analysis. – A Guest posting.

There are still 10 Met Office UK operational weather stations for me to review in order to complete the surface Station Project for existing sites. Many former sites have already been reviewed and there are several more to cover as part of the process of arriving at an appropriate set of reliable weather stations to use for the creation of a new and genuinely representative historic temperature record for the UK. This report is by way of a guest post by a reader who has provided significant support and guidance and helped to “Unstall” my difficulites wih St Catheines Point .

Andrew is very well versed in meteorology and data analysis and as an especially highly cited academic in the field of science is an expert in the rigours required to put forward a sound, testable and reliable database.

The Isle of Wight is a useful location to represent the south central coast of the UK in that all of the island is a coastal climate. There is insufficient “inland” area to significantly modify the maritime effects. In contrast to Kent for example where I was able to demonstrate significant differences between coastal versus inland readings at Margate and Goudhurst, this scale of difference is unlikely to be found on Wight. There is also the benefit of several sites on the island dating back into the 19th Century from which to form an individual island history.

I had personally “stalled” on analysing the data files for St Catherines Point as these were quite alien to me – at which point Andrew took over and rationalised it all plus supplied a deep dive into site provenance. From personal experience I can vouch for this being a very laborious but necessary process. So over to Andrew:

“St Catherines Point

The site
The digital archives cover the period from December 1924 to December 1960. It is clear that there were two sites. The first site was located at 50°35’ N; 1°17’W at an altitude of variously 120 or 130 ft (the height given usually refers the altitude of the barometer). This was called St Catherines Point, Niton or Niton Undercliff (see September 1950 or January 1951). A National Grid Reference of 40/50LY55 is given.

On certain monthly returns it is referred to as the Royal Navy Signals Station (May 1951). A careful search of local websites and the Ordnance Survey indicates that this corresponds to the Coastguard Station located on Castlehaven Lane. The probable location next
to the Lane is 130 m from the beach.
At the end of May 1951, the weather station was moved to its current location at the St Catherines Lighthouse (50°34’ N; 1°18’W, Altitude 51 ft). Daily readings were taken from July 1951. The new location is 50 m from the beach. There are many photos showing the current location of the Meteorological instruments.

The lighthouse was built in 1838. Both sites were of great interest to the Germans (photo). There wasa bombing raid on June 1 st 1943, but this didn’t interrupt the daily readings. The sites have historical significance in that Marconi installed his radio equipment at Knowles Farmjust to the west of the lighthouse in 1899. As a result, many radio and signals stations were setup
along the South Coast.
https://iowrs.org/about/marconi-on-the-wight/

{Area of “Niton” site – exact location not pinpointable}

The readings

Dry bulb and Wet bulb readings were taken 3 times a day:

1924-1925: 0700, 1300, 1800 GMT
1926-1944: 0700, 1200, 1800 GMT
1945-1951: 0900, 1500, 1800 GMT

The Min and Max readings were taken twice a day at 0700 and 1800 GMT (so there are two sets of Min and two sets of Max).

From 1949, the evening min and max were read at 2100 GMT. The difference in timing between the 1800 Dry bulb readings and the TN/TX was probably due to the need to be able to describe the cloud types and weather conditions in daylight.

At the lighthouse from July 1951 until December 1956, readings were taken at 0900, 1500, 1800 GMT with the Min and Max read at 0900 and 1800.

From 1957, the procedure changed. There is only one Dry bulb reading at 0900 and the Min and Max are read once daily at 0900.

The Data

{example of Andrew’s derived data file}

There are very few missing days. Apart from June 1951 when the change of site took place, there are only a few periods of a week or so were either the min or max thermometers were broken and they had to wait for a replacement (Min: 8.10-25.10.1941 and August 1942, 6.10-1.10.1951; Max: 5.1-18.1.1952).

The thrice-daily dry bulb readings are usually recorded to the nearest 0.1°F until 1944 while the Min and Max are recorded to the nearest whole degree. Some min/max readings are given to 0.1°F or 0.5°F. From the calligraphy, it would appear to depend on the observer!

From 1945, almost all readings are to the nearest degree. A consequence of this is that the RH is only an approximate value
and is often only noted as say ‘8’ indicating ~80%.

I have typed out the three daily Dry bulb readings (columns D to F) and the 4 minima and maxima (columns G to J) (2 from 1957) plus the daily rainfall for interest (W). Values that I had difficulty reading, or were corrected by the observer, or post submission by the MO (mostly from 1945 onwards) are highlighted (blue italic for the min, orange italic for the max).

I would systematically cross-check the sums for the month. In column K: Yes, if correct, no if not. Some of the yes’s are in
italics; the total may be out by only one or two. Sometimes it is obvious that the sums were not correct (out by 10 or 30)! I have not made any attempt yet to correct the values.

After converting columns G (min) and J(max) to °C, I then imported the MIDAS data (columns O and P). The MIDAS data is only available to December 1956 (It starts again in 1964 – I have no idea why there is a gap). There are some glaring errors in the MIDAS data that will need to be checked (e.g. 16.2.1950; 16.6.1953).
It was immediately clear that the MIDAS data is an extraction of the lowest of the two daily minima (G, I) and the highest of the two daily maxima (H, J), even though the observers and the MO only used one min (G) and one max (J) to calculate the monthly means.

A clear example is 31.1.1956 where the cold came in and dropped the T during the day from 43°F at 0900 to 29°F by 1800. The
differences are given in columns S&T. The sum of the differences (MIDAS TX –Raw) =2362.3. Dividing
by 11673 (n° of lines) gives +0.2°C difference.

For the Min, (MIDAS TN –Raw) = -2339.9. Dividing by 11679 (n° of lines) gives -0.2°C difference. So the differences effectively cancel out. Given that minima in winter can occur after 0700, it will be a challenging task to provide true minima from the data. A similar argument applies to the maxima.
As a test, I shifted the 0700 GMT TX readings (column H) which were not thrown back as clearly indicated in the observer’s recordings by the carry-over of the last day of the month to the followingmonth. The sum of the differences of the TX reduces to 1623.7 (11689 lines) giving an average difference +0.1. “

——————————————————————————————-

Andrew’s extensive data files will be an invaluable starting point for the recreation of a temperature history for Wight as an test precursor to a national system. My next post will be to consider the option of “site pairing” to transparently demonstrate how “infilling” missing data can be acheived using the examples of two historically good cquality and neighbouring sites at Wye and East Malling.

Source: https://talkshop.blog/2026/11/08/st-catharines-point-the-required-depth-of-data-analysis-a-guest-posting/