On 18 Jan, 19:39, Trevor Harley wrote:
On 2008-01-18 18:12:59 +0000, Paul Hyett said:

... can be a total PITA, as you can never tell what time of day or
night they will occur!

It happens much more frequently than you might think, and has a
significant effect on your mean max and min. Looking at my records
just now over 20 years (1988-2007) -

- a higher night max (i.e. max 09-09h 09-21h) occurred on average 35
days per year (and an average of 8 days in December, a little over 5
in both November and January, so over the Nov-Dec-Jan period an
average somewhat more than once per week) - but only 6 days in 20
years in July

- a higher night min (i.e. min 09-09h 21-09h) occurred on average 58
days per year (and an average of 11 days in December, 10 in January
and 9 in November, so over the Nov-Dec-Jan period on average every 3
days) - but only 5 days in 20 years in July

- Difference between 09-21h/09-09h average max is 0.10 degC over the
year as a whole, 0.37 degC in December;

- Difference between 21-09h/09-09h average min is 0.28 degC over the
year as a whole, 0.81 degC in December.

The difference in the means is significant (0.20 degC or more, i.e.
the difference you'd expect between normal calibrated instruments) in
mean maximum for three months of the year but for mean min *six*
months of the year. It dwarfs decadal variations from e.g. climate
change and for minimum temperatures is comparable with urban heat
island effects in a large conurbation.

And these figures are for 09-21/21-09h. Choosing 06-18/18-06 shows
even larger effects, mainly because the peak frequency of winter
minima is between 0600 and 0900 UTC. Arguably the method with least
possible objection is 00-00h, but this still leads to a high frequency
of max or min at 0000 or 2359 (and unless you have an AWS it's a mite
unpopular ... )

It reinforces how important it is, when undertaking serious
climatology, to (a) agree a standard for terminal hours (i.e. 09-09h
or close to it in UK and Ireland) and (b) stick to it across the
network and for long-period sites. Changing terminal hours will dwarf
many other genuine climatological variations.


Stephen Burt
Stratfield Mortimer, Berkshire