(For some reason I can't post a reply to the previous thread on this
subject, so here's the thread as a new post)

From: "Roy Avis"

So in the Report output of my Vantage Pro 2 I have the option to
produce the overall mean
temperatures by either:

1. The Integration Method (I presume this just adds all the reported
temperatures up and divides
them by the number of reports), or

2. By the simpler (and I presume the Met Office official method) of
adding all the Max & Min
temps and dividing them by two.

I am sure I have stated neither of these method in a true scientific
manner (maths was never my
forte!!) ----- but!!

Which one should I consider to be the true representative - for both
(or either!) accuracy and
for comparison with past data - or are we (like sunshine hours)
creating more rods for our back
with all this computer stuff?

And - Help! When we say Mean do we mean Average - I always get
confused!

Best Regards
Roy Avis

-----

The terms 'mean' and 'average' in statistics are synonymous i.e. they
can used interchangeably. There's a slight preference in climatology
for 'monthly averages' and 'long-term means' but there's no real
difference in what is meant.

When it comes to daily mean (or average) temperatures, what you
calculate/use depends mostly upon what you're trying to do:

- Defining a mean daily temperature as the average of all samples
during a day is without doubt the more accurate statistical
representation of the day, provided the samples are reasonably
frequent (say hourly, or more frequent). An AWS with, say, 5 or 15 min
samples gives a very good representation of the true daily mean over
the sampling period. For accuracy this is the best method.

- The climatological standard in the UK and many other countries is
the 1/2 (max+min) method. This has the advantage of being quick and
easy to calculate, and is suitable for the vast majority of the
historical record where only daily max and min temperatures will be
available. Over a period of time it does give a figure very close to
the 'true' 24 hour mean, and you can verify this for yourself using
your own records. Individual days will, however, often give very
differing results. For purposes of historical comparison (comparing
with long-term means or extremes) this is the currently preferred
method, but as we increase number of sites with a true 24 hour sampled
record and the length of those records then this method will I'm sure
become standard - possibly for the 1981-2010 normals but more likely
the 1991-2020 period (which is the next 'WMO standard' averaging
period).

You should also consider the effect of terminal hours. A daily mean
calculated over 00-00h will be significantly different from that
calculated over the standard 09-09 GMT climatological day, so if you
are using 00-00h 'mean of all samples' and comparing these to long-
term averages based upon 09-09h 1/2 (max+min) you'll introduce two
separate sources of error. The errors are significant, and are greater
than the typical calibration error of standard sheathed-pattern
thermometers.

--
Stephen Burt
Stratfield Mortimer, Berkshire