date: Fri, 10 Feb 2006 14:35:22 +0000
from: Jonathan Gregory <j.m.gregoryatXYZxyzding.ac.uk>
subject: Re: a thought on deltaQ2x
to: Tim Osborn <t.osbornatXYZxyz.ac.uk>
When I tried to do the extrapolation from constant 2xCO2 and 4xCO2 back to
origin, it didn't really give sensible answers either. Possibly the forcings
were too low then as well. This could happen if the effective climate
sensitivity was changing, perhaps. In a standard "Gregory" plot of the HadCM3
4xCO2 run, for instance, climate sensitivity increases so the line of F vs T
flattens out, and if you extrapolate back to the origin you consequently get
a forcing which is too small. Maybe this kind of thing happens in all models.
Given F = Q - lambda DT(t) = Q0(t) * f - lambda DT(t) where f is the forcing
factor, what about doing a multiple linear regression of F on Q0(t) (the
"profile") and DT? Of course it's probably rather degenerate.