I bet if you extended the analysis back a few years you would find 2009 and 2010 stacked to the nines with them , with 2010 a reverse 2015 more or less .
Originally Posted by: Spring Sun Winter Dread
I'm fiddling around to try and generate the data via Excel. It's interesting, though, as I can't get the data to match.
The Met Office provide data for daily CETs and that's easy enough to import and play with.
As an example, look at the 1st January. For the 61-90 reference period, you have (from 1961 to 1990):
9.3,7.9,7.5,7.5,7.3,7.1,6.8,6.7,6.4,6.2,6,4.8,4.3,4.1,4,3.8,3.8,3.5,2.8,2.5,2.3,2.1,1.7,1.1,-0.8,-0.9,-1.3,-1.6,-3.8,-4.9
30 data points, and easy enough to work out the mean (3.54).
That matches the black mean line on the MetO's own chart:
https://www.metoffice.gov.uk/hadobs/hadcet/graphs/2023/daily_meantemp_cet_2023.png However, for the percentiles it gets more awkward, as 30 doesn't go into 20 - you have to interpolate. Excel does that for you, and using the appropriate formula you get:
5th percentile: -2.8
95th percentile: 7.7
However, the Met Office graph instead has about -2.6 and 9.0 - quite different, and clearly wrong.
Perhaps the data is actually for the whole range of daily data, which goes back to 1772... but guess what, Excel borks on any dates before 1900!
Let's see what happens if I add 400 years to all the dates...
EDIT: Well, that was fun! It comes out as:
5th percentile: -2.6
95th percentile: 9.145
....and that matches pretty well. Looks like those HADCET charts are comparing the entire 252-year record, so in that case... a bottom 5th percentile is truly amazing. No wonder we don't get them very often!
Edited by user
21 December 2023 14:26:51
|
Reason: Not specified