date: Mon, 19 Jul 2004 09:26:46 -0600 from: Tom Wigley subject: Re: your ppt to: Sarah Raper OK, I agree that there may not be a true sensitivity. I am still looking at this through MAGICC eyes. For MAGICC to simulate an OAGCM reasonably well on multiple time scales we need only specify two constants, the true sens and RLO. So, in this sense, no matter what the AOGCM does and no matter whether eff sens changes in time, there really is a single constant sensitivity that can be applied usefully. The issue is how to back out the best value for this. You seem to have done this very well already. The results with PCM are pretty amazing over different forcings. If MAGICC, with a single set of two sens params that are kept constant, can emulate an AOGCM over a range of forcing cases then what more do you want? Tom. PS If you can look over the new GSIC ms soon I will be very happy. ================== Sarah Raper wrote: > > On 19 Jul 2004, at 14:45, Tom Wigley wrote: > >> Sarah, >> >> I presume you are still in Germany, but sending to both places >> anyhow. > > > > Yes > >> >> In your ppt, you have spelled sensitivity wrong in slides 9 thru >> 13. Also a few other typos that you can find yourself. > > > Oh, others too? Thanks for pointing out, will try and correct > >> >> Here is my take ... >> >> Even with constant true sensitivity (which is what is put into >> MAGICC), the effective sens varies with time (unless there is >> a lucky land-ocean flux). This is a consequence of the fact >> that land and ocean sensitivities are different is it not? > > > I will say this but its more complicated because lamda not only > different for land/ocean but it is local. > >> I think >> you need to say this up front. So the message seems to be that >> the effective sensitivity is *not* a good way of defining sensitivity. >> People who think the *true* sensitivity changes with time are >> wrong -- it is only the effective sensitivity that changes with time, >> and these changes are entirely predictable (at least qualitatively). > > > Yes, and no. The 'true' sensitivity is not useful (I think) because it > is hypothetical and has an ocean in equilibrium. Not really relevant > for transient experiments of a few centuries. I think it is the ocean > and perhaps atmosphere dynamics which effect the regional temperature > response and hence the local lamda*deltaT, and hence the effective > global sensitivity. Lamda may also change locally as well, I don't > know. It is a real problem for the HadCM models which don't seem to > approach an equilibrium. And that equilibrium is likely not relevant > for 21st century warming > >> >> So the *effective* sensitivity should be dead (or at least looked >> at very carefully) -- but *not* the true sensitivity. > > > I think we need a definition of sensitivity to use in MAGICC which can > be got from CMIP2. This is what I am hoping to get agreement on. > >> >> So the Q remains as to how to get the true sensitivity out of an >> AOGCM. After a long time the eff sens tends to the true sens, >> but examining the full time series as in some of your plots allows >> one to extrapolate to the true sensitivity more efficiently. > > > I don\t think there is a 'true' sensitivity, just want a practical > working one! > >> >> Tom. > > > PS am going through the GSIC paper > >> > > >