"Stuart Brooks" wrote :

For many years figures from Met Office sites were based on 0900-2100 max
and 2100-0900 min (these are still calculated from NCM reports) precisely
to avoid singleton cold nights counting twice. Average differences between
monthly averages worked out from 09-21/21-09 stations and 0909 stations
were as much as 0.7C for December minima, though depending on the
character of the month the figure could be even higher.

The standardisation on 0909 was presumably to make it easy for once a day
stations, though there is no ideal time of day for a reset that would be
valid all year.

In the Monthly Weather Report, MO synoptic (and auxiliary synoptic)
stations were tabulated with 09-21-09 terminal hours from 1945 to
1983 inclusive, but used 09-09 from 1984 until it was so rudely
terminated in 1993.

Before 1945, synoptic stations (or whatever they were called then) used
terminal hours of 18 and 07 (or in some instances 06), but climatological
stations have always been predominately 09-09 though there were more
exceptions in earlier years (some 07-07, 06-06, 18-18, for instance).

Clearly, 09-09 has to be accepted for conventional practice, although
my opinion is that 09-21-09 gives a better feel for the climate of a
particular place. However, there are are still disadvantages with
09-21-09 .... you have to make special arrangements (or be very
careful) when tabulating highest maxes and lowest mins for the
month because these can occur at night (for maxes) and by day
(for mins). Also, a night minimum can easily occur between
sunset and 21z ... something I've noticed very frequently here at
Whipsnade during this particular winter.

Philip