Bamber and Aspinall (2013) carried out an expert elicitation of ice sheet experts. From this they could derive uncertainty ranges of the rate of mass loss from each ice sheet in 2100. In order to calculate the total 21st century contribution to sea level rise we need to integrate the rates. This is what I do in the two tables to the right.
Assumptions: A linear increase in the rate of loss in 2000 to the 2100 rate from BA13.
 The 2000 rate of loss I adopt from Shepherd et al. 2012.
Why I like this elicitation?There are several known unknowns or physical processes that are not well represented by ice sheet models. Potentially these structural uncertainties may be very important. We cannot get the full uncertainty from ice sheet model ensembles that do not incorporate these processes. I cannot see an alternative to relying on expert judgement to figure out how much and in which direction these processes will affect the result. An expert elicitation is probably the best way to map out the additional uncertainty from these known unknowns. see also the comment on BA13 by Cooke et al. here.
[ Update: PostAR5 studies show that the highend tail in the elicitation have some support from process models. See here.]
Input: two supplementary tables from BA13 with the rates of mass loss in 2100 (in mm sea level equivalents):
Why an expert elicitiation? Large structural uncertainties...
 modelling iceocean interaction calving immature
 few long term records for validation.
 ice hydrology / ice dynamics models immature
Experts know the models and they have an idea of what processes are poorly represented. They can have an informed opinion on what that means for projections. It is a useful tool to assess uncertainty ranges.  Results
Note BA13 report 20102100, whereas I report 20002100. BA13 also uses a slightly greater initial rate than Shepherd et al. from the elicitation. The 95th percentile value of 910 mm is very consistent with the 840 mm from the BA13 abstract. The other major contributors If you want to get global sea level rise then you have to add the other main contributions. Here are some estimates for a mid range scenario: This adds up to 42 cm which may have an uncertainty of ~ ±20 cm (but judge for yourself by visiting the links). To these 42 cm we have to add the ice sheet contribution. If we take the mean(/median) from the first table above as our best guess then we get 80 cm (/72 cm).
A high end estimate of global sea level rise. If we want to explore the very likely upper limit (95%) then it is clear that it must be greater than 91+42 = 133 cm, as this does not allow for any uncertainty in the 42 cm. You should also note that the 42 cm were for a midrange scenario. For a highend forcing scenario we might expect about a third more. So based on this I conclude that for intense scenarios like RCP85 and A1FI then the very likely upper limit is atleast 180 cm [based on this crude calculation: 1.3*(42+10+91) = 185 cm ]. I would even argue that it is hard to fully exclude 2m as the 91 cm from table 1 depends on assumptions of covariance and performance weighting in Bamber and Aspinall (2013), and an assumed constant acceleration. Some experts will probably think that such a high loss is unlikely, but the very likely upper range is designed to explore the limits of plausibility under an extreme warming scenario.
