Figure 1, overlay plots of pairs of almost raw datasets, in each case the Met Office areal mean temperature data for England and Wales, less annual, against gridded UAHTLT V6 beta 2, UAHTLT V5.6, RSSTLT V3.3, Hadcrut4 V126.96.36.199, the latter is unfair because it is gridded at 5 degrees instead of 2.5 degrees. In all cases the geographic area overlap is very approximate.
Click image for larger but preferably download this PDF (106KB) since as a vector plot zoom/enlargement and pan allow examination in great detail.
Figure 1 is revealing. Obviously this is close if not accurate, as least for the satellite data. The curious data shape 1978 .. 1986 is shared, a hallmark. Rather nicely the data matches well with no adjustment (annual and simple mean removed from areal).
An unpleasant surprise is UAH V6 beta 2 being a worse match than V5.6 or the similar RSS.
Under protest because I don’t like straight lines, the linear trend is the same for Hadcrut4 as Areal data. All satellite data has a substantially smaller trend than Areal.
Figure 2, not quite sure if the lines are right, facility of some software not been used before. usually I work out slopes myself.
[updated, Hadcrut4 plot]
The satellite data follows a different function, both skew and kurtosis are different. Experimentally compensating this brings the dataset r2 closer to each other.
I don’t know how, if at all, satellite sensing takes into account weather related lapse rate variation, probably a substantial effect over near maritime regions with their varying weather conditions.
Recently I have been quiet at the Talkshop but working hard evolving software, posting material on my own blog. Some of this is a side effect of adding UAH V6 data to an existing large work.
I expect detail work could improve the result. If anyone wants to do this I can provide data extracts.
Post by Tim