"Terry Pinnell" wrote in message
...

Thanks Martin, appreciate the advice. That Hungarian site's forecast
gives a third set of temps, although they are much closer to BBC's
than to Accuweather's. I wonder how the latter can get it so wrong?


.... we won't be able to tell who is 'right' and who is 'wrong' until
after the event of course. The problem with *many* of these
model-driven sites is that they often use global models (designed to
cope with the broadscale features across the entire planet) and
regression routines to implicitly forecast the temperatures at local
ground level, derived from a model variable that is not necessarily
representative of the local topography / surface type. A better
technique is to use such models to power a local site-specific
statistically updated model which then becomes (properly formulated)
'tuned' to the local conditions and explicitly forecasts the local
'screen' (roughly 1.5m) temperature.

If you go to the national site, you are more likely to pick up such
output; if you use someone like the BBC (in fact taken from the Met
Office global model) or AccuWeather (probably based on NCEP though I'm
not sure of that), then we're dealing with global model stuff which
may or may not be correct.

Martin.


--
Martin Rowley
E:
W: booty.org.uk