In article ,
Alastair McDonald writes:
But isn't the CET just a massaged set for a (small) region of the
UK? OTOH, it is longer but is it accurate?

Enormous effort was devoted to making it as accurate as possible.
Obviously the earliest part of the series is less accurate than the
rest. Gordon Manley's two seminal papers describing how he derived the
values can be found online. See:

http://www.rmets.org/sites/default/files/qj53manley.pdf
http://www.rmets.org/sites/default/files/qj74manley.pdf

As the Wikipedia article about the series says:

"The earliest years of the series, from 1659 to October 1722 inclusive,
for the most part only have monthly means given to the nearest degree or
half a degree, though there is a small 'window' of 0.1 degree precision
from 1699 to 1706 inclusive. This reflects the number, accuracy,
reliability and geographical spread of the temperature records that were
available for the years in question."

and

"Although best efforts have been made by Manley and subsequent
researchers to quality control the series, there are data problems in
the early years, with some non-instrumental data used. These problems
account for the lower precision to which the early monthly means were
quoted by Manley. Parker et al. (1992) [1] addressed this by not using
data prior to 1772, since their daily series required more accurate data
than did the original series of monthly means..."

So I think, even if you're very cautious, you can rely on the values
from at worst 1772 to be high accurate.
--
John Hall "He crams with cans of poisoned meat
The subjects of the King,
And when they die by thousands G.K.Chesterton:
Why, he laughs like anything." from "Song Against Grocers"