From	David Rind	Date	Thu, 13 Jan 2005 17:00:26 -0500
To	Stefan Rahmstorf
CC	David Rind, Tim Osborn, Jonathan T. Overpeck, Keith Briffa, Eystein Jansen, fortunatjoos@email.arizona.edu
Subject	Re: 6.5.8 revisions
Here are my responses to Stefan's comments. While I could have made each of these points in the document itself, it is already sufficiently long that Jonathan had me cut it before most of you guys saw it. At 8:53 PM +0100 1/13/05, Stefan Rahmstorf wrote: Hi folks, on the topic of climate sensitivity. I just lost a long mail on it due to a software crash, so sorry if I'm brief now. I think it makes no sense for the purpose of the IPCC to discuss a climate sensitivity to orbital forcing - if such a thing can be defined at all. The first-order idea of orbital forcing is that in annual global mean it is almost zero - and in any case the large effect orbital forcing has on climate has very little to do with its global mean value. Hence, we'll confuse people by discussing it in this way, and even citing numbers for it. For the purpose of IPCC, I think climate sensitvity should refer to climate sensitivity wrt. greenhouse gases. The point here is that climate can be forced by other factors than simply a global, annual average radiation change, which is the metric now being used. The orbital forcing induced changes are wonderful examples of this, hence the paleoclimate chapter is a perfect place to discuss it. Variations in seasonal and latitudinal forcing clearly have had a major impact on climate, including forcing of ice ages, yet the annual average radiative change is small. The importance of this with respect to IPCC is that other climate forcings can also affect the seasonal and latitudinal distribution of radiation - aerosols, land surface changes, and even solar radiation (considering cloud cover distributions) - hence they too may have a disproportionate influence compared to their annual global average magnitude. What is said in this subsection is simply that this one metric clearly fails with respect to the major variations in paleoclimate, and as a general rule, there should be room for an expanded concept (which may then have utility for current and future climate forcing as well). Also, it is questionable to discuss climate sensitivity for uncoupled models, especially for glacial times - Ganopolski et al. (Nature 1998) have shown that glacial climate looks very different with mixed layer ocean vs. coupled. I think for a 2007 IPCC report we shouldn't be discussing old uncoupled runs when coupled model results are available. (And it is a little odd that the above paper, the first coupled model simulation of glacial climate, cited over 150 times so far, is ignored here in the discussion of the last glacial maximum - if you do a search on the Google Scholar engine for the key words "Last Glacial Maximum", you'll find it's the second-most cited paper on this topic after the Petit et al. Vostok data paper.) In fact, most if not all of climate sensitivity measurements have been done for what Stefan calls "uncoupled models", atmospheric models coupled to mixed layer ocean models. The results from all prior IPCC reports give sensitivities from precisely these types of models - for the basic reason that almost no one has ever run a coupled model for 2CO2 to equilibrium. The other disadvantage of coupled models in this regard is that their control run, if simulated long enough, often does not reproduce the current climate in important respects - one is then getting a climate sensitivity with respect to something far removed from the current climate, so what good is it? The fact that models coupled to a dynamic ocean and those coupled to mixed layer oceans may get different responses - and one can see from the numbers that the responses are actually fairly similar in general - can be related to the ocean dynamics changes; as the text notes, that is considered a feedback in this subsection, and therefore an appropriate part of the climate sensitivity calculation. I still think it makes no sense to say that climate sensitivity depends on the sign of the forcing. Talking about greenhouse gases: whether you will do an experiment going from 280 ppm to 300 ppm, or the other way round from 300 ppm to 280 ppm, should give you the same climate sensitivity. Perhaps you mean that going from 280 to 300 will give a different result compared to going from 280 to 260, but then you're really comparing different mean climates. I think this "directionality" of climate sensitivity is not a good concept. It's not the forcing per se that's the issue here, it's the feedbacks that potentially can alter the climate sensitivity to the sign of the forcing. It has been suggested in the past that climate sensitivity is larger to cooling perturbations then to warming ones, and we ourselves have found that result in some earlier model runs. The standard reason given is that with a cooling climate perturbation, sea ice can expand further equatorward, to cover a broader area, and intersect more solar radiation - therefore providing a more positive feedback to the cooling. In a warming climate, the sea ice retreats and intersects less radiation - but the sunlight-weighted area is smaller in the regions it is retreating to, so its positive feedback to the warming is not as large. However - water vapor works the opposite way. Given the exponential dependence of water vapor on temperature, in a warming climate the added temperature would allow for a greater water vapor change (increase) than would occur with a cooling climate of the same magnitude. Hence the water vapor feedback should be greater in a warming climate. So the answer is - nobody knows. Jim Hansen did a survey of people at GISS recently to see what the general opinion was for a paper he's working on (and sending around). Since paleoclimates have suffered both positive and negative forcings (in the examples given in this section), and since we don't know the answer to this question, we can't really say whether the sign of the forcing is important or not. So I've left it as an open question, with the possibility that it might matter. Relating forcing to response, the sensitivity from the models is then on the order of 0.6°C/ Wm-2 (or higher, depending on the model used); the sensitivity from the observations, if taken at face value, would be considerably less. I still don't understand how you get this conclusion. This would mean: if you take models with those estimated forcings and run them, they should show a big mismatch with the proxy data. As far as I can tell from the diagram by Mike Mann attached, combining models and data, only the Von Storch simulation (not shown on this one) does show such a mismatch. (And that uses 1.5 times the Lean solar forcing.) If you look at the various model simulations done for this time period, the only way the models can reproduce the "observed" cooling relative to the present is by using only a subset of the forcings. When you use all the forcings, you get a much higher number. You can do the math yourself: with a "best-guess" radiative forcing change of 2.4Wm**-2, models with a sensitivity of 0.6C/Wm**-2 will get a temperature change of some 1.5C, which over the course of 300 years shows up in GCMs. For example: Cubasch et al (1997), using just solar forcing in the ECHAM 3 model came up with cooling of 0.5C; if you add a reasonable response to the approximately 1.5-2 W/m**2 forcing from trace gases plus aerosols, you get an additional 1C cooling (given the sensitivity stated above). Counteracting that could be land surface changes - but counteracting that are undoubtedly the reduced pre-industrial tropospheric ozone, plus any additional volcanic cooling (a la Crowley). So assuming those sort of cancel, we have a 1.5C cooling for the MM time period from solar plus anthropogenic, similar to what we get in the GISS model (as noted in our 2004 paper). That can be compared with the Mann et al reconstruction - and you can see from your figure that for the 1700 time period relative to the 1990s, the cooling is about 0.5C. Similarly, Fischer-Bruns et al. (2002) with the ECHAM 4 model, using solar forcing of -0.1% for the MM, and volcanic forcing greater than today (like Crowley) got a cooling of 1.2C. The Zorita et al study also got a large magnitude cooling when using all the forcings. BTW, neither ECHAM 3 nor ECHAM 4 has a large climate sensitivity - it is of the order of 0.6C/Wm-2, as referred to in the comment above. Note that none of these models are shown in your accompanying figure, and all are GCM studies. How did the Crowley and Bauer studies that are shown in the figure (using EB or EMIC models) get the smaller cooling magnitudes indicated there? Only by using a subset of the forcings - Crowley basically threw out the solar changes (and had a lower sensitivity model), Bauer et al. used a large aerosol effect and still needed a large deforestation warming to bring her results in line with the Mann et al. reconstruction (in fact, it was done specifically for that reason). None of these runs used the tropospheric ozone reduction that we have evidence did occur. My impression is that these studies took the observations as given and were asking the question of what forcings would be needed to reproduce them. That is an interesting question, but it obviously does not validate the observations. The specific comment you refer to above relates to the discussion in the previous paragraphs, which detail the radiative forcings and all the different model responses. It is a fair representation of the current status, however unsettling that is. But in the current incarnation of this subsection, we do not use it to imply a low climate sensitivity - we simply say that given the uncertainties in forcing and response, we cannot use this time period to better understand climate sensitivity. And I think that's accurate. David -- /////////////////////////////////////////////////////////////////////////// ///////////////////////////////////////////////////////////////////////////