date: Sat, 17 Jul 2004 20:42:33 -0600 from: Tom Wigley subject: more GSIC to: Sarah Raper , Sarah Raper Sarah, You are right that all the GSIC will melt even for zero warming from today. Our old model was better in that regard because it had a different eventual melt for different warmings. But this is a tricky thing for me even to conceptualize. My intuition says that if we stabilized temperature at today's level then not all of the GSIC ice would melt -- but how much would remain? It also seems reasonable that there is some warming amount that would ensure that virtually all the GSIC ice would melt --- 3degC, 5degC, ??? So the next step would be to try to get some realism here, but I really have no idea what would be realistic. However, I think what we have is fine out to 2400 for most cases. For zero warming from now, sea level rises at about 5cm/century initially and then more slowly later -- so it would take 1000 years or more for all the GSIC ice to melt. We never get far enough for things to look silly. This melt rate (%cm/century) is proportional to (0.15 + T(1990)), so the 0.15 has an important effect. Climate sensitivity has an effect too (in MAGICC) since this influences T(1990) -- a model artifact. (The TAR says that the 0.15 comes from Zuo and Oerlemans, which I have added to the refs.) So this leads to a question -- at present the GSIC model, embedded in MAGICC, uses T(1990) from MAGICC. Perhaps it would be better to use an observed value (like the 0.65 I use in the text example)? For most cases this would have little effect, but it has a noticeable effect for cases with very little post-1990 warming. Arguing against this, to add another add hoc correction like this would just make things messier. I'm just thinking out loud here. As it happens, the analytic result I gave was for zero warming from 1990, not (as the text implied) for linear warming. The linear warming result is quite complicated -- but all I needed to illustrate the points made in the paper was/is the zero warming case. I have corrected this. I have also done the A proportional to V**n case, and dumped the n=1 results. The refs you gave were just what I needed, and I have cited these. n.NE.1 has little effect out to 2100, but does lead to greater melt by 2400. I have revised the text to cover all this, but need to re-do the diagrams. I hope to send this new version to you tomorrow. I think it is a significant improvement even though the results are pretty much the same. Some of the bibliographic holes are filled -- I can imagine that someone like Oerlemans would criticize anything that didn't liberally cite his work (so there are now more Oerlemans refs). Best wishes, Tom.