Normalised UK regional temperature

Posted: March 9, 2015 by tchannon in Analysis, innovation, methodology, weather

The fuss about extreme rainfall last year tripped me into looking for myself. This led to an innovative analysis of Met Office areal time series for precipitation. There was little interest shown but also little criticism. I’m bringing up Windows 8.1 64 here, same hardware, testing various codebases.
As a wonder-if… the Met Office publish areal series for air temperature, Tmean, Tmax and Tmin. Daft idea, pull one file and eyeball, looks the same data format as rainfall. Do the lazy thing, copy code to a new directory, few trivial edits and hit go. It works. The results look sane.

Tmean for East Scotland, one of 68 plots. The four PDF, Tmean, Tmax, Tmin and Precipitation are linked later. Zoom to any scale works on what are postscript vector data, details can be seen.

A take-home from seeing the results is the episodic nature of weather. Mostly it is bouncing around as weather does but also there are sustained periods with less noise and perhaps floods or droughts, warm or chilly. The temperature data says we have recently had cool and then warm episode. Where this is notable it seems to last for around a year, as-if anything is a definite rule.

Another feature I’ve noticed is how an overall average tends to misleading for the UK because there are climatic regional effects, what applies to the south and east does not apply to the north-western Atlantic coast. and so on.


The drought of 1976 was regional, temperatures do not show the same effect, nor is this obvious if the whole UK is considered.


Southern England Tmax does show some effect without a clear sustained high bias. If it is dry it will tend to heat up quicker.

We see the same thing on a grander scale, this winter for example has been cold for the US north east but spring arrived very early for the US north west. Similarly Southern and South-Eastern Europe has been getting deluged in snow, state of emergency declared in parts of Bulgaria whereas southern England has had a fairly mild almost snow free winter. Much the same seems to apply more globally, as though there are paired regions.

This is characteristic of flicker noise, 1/f, chaotic.

[UPDATE] Doug Proctor asked in comments if sunshine data is available, yes but from 1929

UK-plot-smallPDF of sunshine data here.


Please read this section before looking at any of the plots because I mean innovative, you have to get your head around the meaning, not immediately obvious.

What is done
Areal data is computed by the Met Office from point station data to try and give a regional time series. All the usual caveats about stations changing apply.
The published time series Jan 1910 to Feb 2015 has the annual cycle removed by harmonic analysis, the result is then bent by a mathematical function to best fit Normal data, all done by computer code which optimises adjustment parameters for best fit (all my own code). Plots are then done.

The ranking tables give a clue on why getting your head around this is not so trivial


Tmean North & NE England, 2006 was hotter than 1938 but ranks lower because March (1938) is a colder time of year than July (2006), ie. the bias from annual cyclic variation has been removed. The statistical adjustment has no effect on ranking.

The statistical adjustment is attempting to equalise the significance of hot and cold to the same whereas in reality the data is asymmetric. This is of limited utility since in effect outliers, or extremes are part of weather. This goes off into special analysis most often done in hydrology to do with predicting civil engineering requirements, how large a flood drain is needed, how large a dam and so on.


Tmean (2MB)
Tmax (2MB)
Tmin (2MB)
Precipitation (2MB)

Law bending

This uses a hyperbolic function. See near end of May 2014 article here, and a region map if you need it. Code is unaltered apart from text and range of annual cycle.

As usual if you want data then ask, say what you need.

Post by Tim

  1. A C Osborn says:

    Should we be able to expand the first graph?
    There does not seem to be much “warming” going on in that graph, are the other regions the same?

  2. You might also be interested to read my latest article which is also on normalisation of trends:

    recent temperature trends are not abnormal
    [mod: fixed url, goodness knows what happened. –Tim]

    The summary:-
    “It has frequently been stated that 2oth century warming was “unprecedented” or “cannot be explained”. This article sets out to test this assertion on CET the longest available temperature series. I find the CET data rejects the hypothesis of ‘climate change’ (>58%) & current ‘global warming’ (>72%) and that overall global temperature has not changed significantly more than would be expected. I do however detect a marginally higher trend over a 50year period ending 2009 with about 2.5σ and a 35% chance of occurring normally within the dataset. However this is inconsistent with an established trend as progressively shorter periods toward the present time tend toward lower trends (40yr: 1.7σ, 30yr: 1.3σ, 20yr: 1.6σ, 10yr: -0.9σ).”

  3. tchannon says:

    ACO, wasn’t, large files where I have no particular reason to focus on that parameter and region.

    Added one anyway.

  4. Doug Proctor says:

    A z-score (aka, a standard score) indicates how many standard deviations an element is from the mean.

    Is the data there that lets this be done for Hours Bright Sunshine in each region to look for changes in cloud cover?

  5. tchannon says:

    That’s the general idea Doug, trying to get away from physical units. Validity dubious.

    Ooo.. look, whoever did that PDF has live links to the data, click, sunshine. Downside is it runs from 1929. Hold on… bit hairy at the moment with various scheduled data archiving/backups firing… edit data URL, save, click. ooo look, datasets appeared and running… annual cycle not as strong as temperature one, (first stage is compute those, slow process), and stop. Click, plotted UK and running, at least it didn’t crash, lets look, good, plot simply starts, 1929, CDF is good, law is good, annual, ouch, need to change scale, ranking for UK, most sun 1947.63 and least 1983.38., hows it doing, almost, blog backup kicks in, plots finished, do a combine into one, click. Scan file, all works, doesn’t look particularly edifying.

    I’ll take a few minutes to edit names and titles, sort the annual scale and rerun the compute/plot, first stage doesn’t alter.

    Edit plot title ==> sunshine hours.
    Find an annual file, about +-70 so reckon pd.yscale = {-10, 10} needs 100.
    Change final file name to sunshine, notice spelling mistake, oops, areal ain’t spelt like that.

    Click, slow process, mountain of code.

    Quick check, yep, +-100 is fine for annual, looks sane.

    Now what? I’ll prepare bitmaps and upload as an article addendum. Not had time to actually look yet.

  6. tchannon says:

    Midlands is vaguely CET, quick hack

  7. I ran some control charts on winter precipitation, sort of ‘move along nothing to be seen here’.

    Perhaps I should run some more on the various claims?

  8. Ben Vorlich says:

    From 1962 until about 1974 I used to spend my late summers early autumns (mid August until mid October) Grouse beating and Deer Stalking (looking after the ponies) in Perthshire. The weather was pretty variable year on year sometimes we got soaked most days and some years were drier. What I do remember is that one of the last years, and I can’t remember which year it was perhaps 72 or 73, there was no rain all summer and autumn. We took the ponies into areas on the estate where any other year we couldn’t have ventured for fear of getting bogged down.

    In 1975 I went to work in Kent, and it hardly rained from then until until Dennis Howell fixed the problem! What does stick in my memory is that the weather in Perthshire wasn’t as sunny as Kent (weekly phone conversations with my mother) and that a colleague and I talked about how there’d been no real drought for them during the one I’d experienced. Confirming that there are significant regional variations across the UK year on year.

  9. A C Osborn says:

    tchannon says:
    March 10, 2015 at 2:11 am

    Midlands is vaguely CET, quick hack

    Is that 10y for real, it seems to spend most of it’s time outside the data?

  10. tchannon says:

    Ben Vorlich,
    You had a chance to really get to know the land and the weather, oh, and biting little beasties.

    Mentioning the pony range is the kind of telling detail which gives clues on unravelling the past. Long distance calls, expensive, we remember a different era yet today telephone is taken for granted. One that amuses me but gets blank stares is how back then there were more household cars than telephones, both about communication, one less important.

    Kent is about as different as it could get.

    Blocking weather doesn’t last all year, but maybe if can be repeating. The other explanation is the disturbed zone staying north or south.

  11. tchannon says:

    ACO, twin y-axis.

    An arising question is the validity of applying a linear process to normalised data and whether the data was linear in the first place.

  12. […] article showing these plots, has references, here Various Talkshop articles have appeared eg. here and the technique works on rain, sun and temperature […]