cc: mann@psu.edu, Keith Briffa , Tim Osborn date: Fri, 12 Aug 2005 08:59:56 -0600 from: Caspar Ammann subject: Re: [Fwd: Storch drift] to: Stefan Rahmstorf Stefan, this is very important news indeed. The runs will get a huge hit from this. The only way a coupled model can get a continued trend (without invoking an energy leak somewhere) is when there is a terrible deep-ocean spin up available even for their present day initialization, not to speak about the subsequent shock to pre-industrial conditions. Did you really say 1.5 degrees? Wow, that is quite a bit. Seems to me they must have used Levitus ocean data with an atmospheric restart file, then hit it with the solar/GHG changes. It seems rather large of a drop to come from a fully coupled stage. 1.5 degrees is about 30% too large to be exclusively from the atmospheric composition and solar irradiance, thus my suspicion regarding levitus. Now it would be important to know what happend because some people are using the run as a possible real-world scenario (although Hans in talks does not claim so). Caspar PS Now, bare in mind that the Science paper applies to the reconstruction, and for the general discussion the influence of spinup should not make that big of a difference (other than inflating the difference of the coldest period to the calibration period, which creates some issues discussed by Mike previously). Michael E. Mann wrote: > > > ------------------------------------------------------------------------ > > Subject: > Storch drift > From: > Stefan Rahmstorf > Date: > Thu, 11 Aug 2005 15:37:27 +0200 > To: > mann@psu.edu > > To: > mann@psu.edu > CC: > Gavin Schmidt , Keith Briffa > , t.osborn@uea.ac.uk > > > Hi Mike, > > here is some interesting new info on the drift problem in the VS04 > runs. Irina Fast and Gerd Bürger submitted a comment about this to > Science some months ago; it was rejected and they did not pursue it. > I'm trying to encourage them to resubmit this elsewhere. I do not have > the ms. but have seen several graphs. There are two key points. > > 1. The ECHO-G run started at year 900, the VS04 paper of course shows > only results starting from year 1000. I've seen the full run now. > Between 900 and 1000, the NH temperature drops by about 1.5 ºC! That's > how severe their initialisation problem is. From my experience of how > the THC responds after such step-function changes in forcing, the > strong warming from 1050-1150 in VS04 could well be a rebound effect > from the 1.5 ºC cooling that precedes it, since the THC tends to > oscillate on such a time scale when forced rapidly. > > 2. Irina has run ECHO-G initialised with modern climate and then > switching to pre-industrial conditions similar to the run shown by > VS04, but without any further variability in the forcing. Thus, this > shows the pure drift from initialising this run - this is what Tim has > been estimating in MAGICC. The actual drift in ECHO-G is even larger > and more persistent than what Tim found: there is a cooling between > the years 1000 and 2000 of over 0.6 ºC, and this is an almost linear > trend over the whole time. I.e., not just drifting during the first > few centuries, but over the entire 1000-year period. > > Cheers, Stefan > -- Caspar M. Ammann National Center for Atmospheric Research Climate and Global Dynamics Division - Paleoclimatology 1850 Table Mesa Drive Boulder, CO 80307-3000 email: ammann@ucar.edu tel: 303-497-1705 fax: 303-497-1348