The assessment of equilibrium (or effective - let's not involve ourselves with the difference here) climate sensitivity depends not only on the empirical and semi-empirical studies that have been discussed at this blog in recent days, but also on the climate models. As the second order draft of AR5 shows, most of the models sit within the IPCC's preferred 2-4.5°C range (represented by the shaded area). There are some models that run hotter, but I don't think anyone takes these seriously.
When it comes to the empirically constrained evidence for ECS, in the draft second order draft of AR5, the IPCC shows separate subfigures for each of the different types of study, which makes comparisons somewhat tricky. I've therefore redrawn the PDF graphs (but only the PDFs - it's also hard to make direct comparisons to studies in which only a range is stated) on the same set of axes, colour coding them by type. I've added some recent results as well.
Purples are estimates based on satellite data (including Forster and Gregory), greens on intrumental data (including the recent Lewis and Masters papers), and blues on paleoclimate data. The grey band is, once again, the 2-4.5°C range of the IPCC. As noted, the latter relies mainly on GCM estimates. The black curve is the Skeie et al paper, which kicked off all the recent interest in ECS and which has not yet passed peer review. The very tall peak is Lindzen and Choi. Those authors did not include a PDF in their paper, but Nic Lewis provided me with an estimate based on their data. Because the result is so tightly constrained, the exact shape is relatively unimportant.
We know that some of the papers that the IPCC is citing should not be included because they have an unnecessary warm bias due to their use of uniform priors in climate sensitivity. Who can forget Steve Jewson's words?
Uniform priors are a fundamentally unjustifiable methodology that gives invalid quantitative results. If these papers are cited in the IPCC, the risk is that critics will (quite rightly) heap criticism on the IPCC for relying on such stuff, and the credibility of IPCC and climate science will suffer as a result.
Among the paleo estimates, Libardoni and Forest and Olson et al are affected by the use of uniform priors in climate sensitivity, and because the data so ill-constrains their results the effects will be significant. On the other hand, the Skeie and Aldrin results are much better constrained by the data and so the effect of the bias should be relatively insignificant. It's also worth noting that correcting the priors in the latter two papers would only make their estimates of ECS lower still. On this basis, here is the spaghetti graph with heavily biased papers paleo papers removed.
Purists may prefer to see what it looks like with all the flat-prior papers removed and the relevant graph can be seen here.
There is clearly a discrepancy between the paleo and other estimates, with the paleo estimates spanning much wider ranges - these studies are very badly constrained indeed. It seems to me that the best way to deal with this is to recognise that the range of overlap is the area that can explains both sets of results and that this should be treated as the best estimate of ECS. On that basis it's hard to justify a range that spans much more than 1-3°C.
If there are other PDFs that people think can be included, by all means send them over and I can add them in. In the meantime WUWT has a compilation of low-sensitivity studies, which includes all the studies that only publish a range rather than a PDF.