Notes on various analyses of data for site integrity etc.
A.
NOTE REGARDING THE MINIMUM TEMPERATURES (MAY ALSO APPLY TO MAXIMA AS WELL)
(written/uploaded March, 2000)
[ I was concerned after 3 months intercomparison with readings taken a few
streets away (Roman Hill in the COL bulletin for those interested), that my
night minima in particular were on the cold side. The following is a summary of
the analysis ]:
Using a sample size of 77 events, the average difference is 0.5 degC (my site
colder), with a sd=0.47. 47% of occasions were 0.3degC or less of difference,
and 34% were 0.2degC or less of difference, i.e. around third of occasions.
Investigating those occasions within 0.2degC (i.e. close agreement), then all
but one occurred after nights of either (a), overcast, wet conditions or (b)
strong winds, or both. In other words, when the lower atmosphere would be
fairly homogeneous due to the wind/overcast skies. So, the larger differences
at other times are probably due to genuine differences between our
sites rather than any instrumental problem.
On the nights of the strongest winds, then the difference was 0.0 or 0.1degC
which is gratifying. Looking at occasions of *greatest* difference (arbitrarily
over 0.9degC/2 sd or greater), then there are 8. There was no clear pattern,
although several occurred after clear nights with very light winds. However,
the sometimes large differences are of concern, so after particularly
clear/still nights I will take particular note of the actual temperature at
reading (which we don't routinely note), and check the minimum temperature
against the dry bulb.
Overall though, I think the differences are genuine - the screen is mounted on
a bracket on the end of the garage wall - north-facing but in the most
shaded/coldest part of the garden. Only in the summer half year will the area
get reasonable sunshine. So I think the temperature is 'genuine' but obviously
with a cold (microscale) bias relative to a 'standard' exposure.
I further note that the Beaufort Park site is even colder than mine, but is
of course more 'rural' and open than either of the sites used in this
intercomparison.
B.
Analysis of data: Bracknell (Tawfield) and Bracknell (Beaufort Park)
(written/uploaded September 2003)
Period covered: November 1999 to December 2002 (but missing data – see below)
RAINFALL:
(period covered Nov.1999 to Dec.2002, with one month missing = 37 months)
Over this period, Tawfield (TF) has "collected" 2963mm and Beaufort Park (BP) 2835mm
TF > BP by 128mm (or 4.5% excess over BP total).
Excluding the last two months ( when we suspect that there were large errors due to failures at BP), the figures are TF = 2631mm and BP = 2544mm
TF > BP by 87mm (or 3.4% excess over BP total).
Looking at the individual month-by-month figures, 22 months (60%) of occasions are within 4mm of the BP total and 29 months (78%) within 9mm of same. The sample is skewed negatively, a bias towards TF being wetter than BP. Of the remaining, 4 'outliers' were 15 or more mm adrift (mostly TF > BP). However, all occurred in the last few months of 2002, when we suspect that the instrumentation was faulty.
Conservatively then, I regard us (TF) as catching, on average & on a month-by-month basis,
some 4% more than Beaufort Park. This will not be a genuine climatological difference: probably poor exposure, with the buildings, garden walls etc., focussing the rain into the rain-collecting area. (I haven't done a seasonal analysis because there is not enough data.)
TEMPERATURE:
(period covered Nov.1999 to Jul.2002 = 33 months)
(a): MAXIMA:
(i): SUMMER:- Mean (24hr) maximum values were around 1degC too high in the period May to August, with an average over the four months of 1.1degC (TF > BP) and a peak in June of 1.4. Given the timing, and the fact that the differences are around zero or slightly –ve at other times of the year (<=0.5degC), it looks as if solar heating of both the old screen (only single louver) and back-radiation from the garage played a large part. Also, the sheltered 'corner' position is a factor.
(ii): WINTER:- using the four months November to February, the average difference is 0.41degC (TF < BP) confirming the 'enclosed garden' factor – not warming sufficiently on sunnier days.
(b): MINIMA:
(i): SUMMER:- Mean (24hr) values are just under 1 degC too high in the same summer period of May to August, with average over the 4 months on 0.8degC (TF > BP), and a peak in June and July of 0.9degC. Again, this suggests the heat released by the garage (after a day of strong heating) in particular is of significance. At other times of the year, the readings (averaged) are generally within 0.3degC of BP values.
(ii): WINTER:- using the same four months (November to February), the average difference is 0.06degC – we are slightly warmer than BP, but not dramatically so.
Summary:
> About 4% more rainfall than Beaufort Park averaged over this period. As this represents about 0.4mm in 10mm, or 4mm in 100mm, the difference is better than I dared hope, and means I can use BP averages as a good proxy for our site.
> Higher maxima in summer by about 1degC.
> Maxima at other times within 0.5degC of BP values.
> Higher minima in summer by about 0.8degC.
> Minima at other times within 0.3degC of BP values.
Since the above, I have reconstructed the screen (near-standard construction, but same location), and we have a new set of thermometers in use.
During the hot ("record-breaking") spell (5th to 12th August, 2003), for maximum temperatures, we were averaged – 0.1degC [ BP-TF ] over these 8 days, with a maximum difference of +0.7. On the hottest day (10th August), Beaufort Park recorded 36.1degC and we recorded 35.6degC, not too bad an agreement.
There did not appear to be a systematic error, with 3 (out of 8) colder than BP and the remainder (5), warmer than BP. On three occasions we were +/- 0.2degC of each other, with only one event (the 0.7 mentioned above), >0.5degC.
As regards minima during the period, the strong positive bias (TF > BP) is still there, and we will not be able to remove this as the screen will have to stay in it's position mounted on the North Wall of the garage.
I hope that by re-building the screen, I may have eliminated some of the bias shown up in the former analysis – time will tell, but not against Beaufort Park figures unfortunately.