In November, I wrote a post describing research on Antarctic glacial melting by Catherine Ritz, Tamsin Edwards, Gaël Durand, Antony Payne, Vincent Peyaud, and Richard Hindmarsh (“Potential sea-level rise from Antarctic ice-sheet instability constrained by observations”). I had asked one of the authors, Tamsin Edwards, to address a few questions about the study. I also asked glacier expert Richard Alley a few questions. Alley got back to me right away, but Edwards was unable to do so, so I wrote up Alley’s commentary here, with the intention of covering Edwards’ response at a later time. Over the weekend, Edwards responded to my questions as well as many of Alley’s comments, and thus, this post.
In my original post, I wrote,
The study asked how much Antarctic ice sheets might contribute to global sea level by 2100 and 2200 AD. The results contradicted some earlier estimates which are on the high end, but conformed very closely to the current IPCC estimate, raising that number by a negligible amount.
To this, Edwards responds,
Our likely range (central two thirds of the distribution: 4–21 cm by 2100) is a few centimetres higher than the IPCC’s estimates of the likely range for A1B. Our extremely unlikely threshold (1 in 20 chance of exceeding 30 cm) is lower than some previous estimates of the upper bound and is also at the low end of the IPCC’s estimate of “not more than several tenths of a metre”.
I asked Edwards if it was correct to cay that the study’s results conform to expectations based on the prior summary of research from the IPCC (with a minor adjustment), but that the results also contradict some earlier higher-end estimates of Antarctic contribution to sea level rise. Her response was that it would be correct “…to conclude that our results do not contradict estimates of large potential sea level rise from instability in the long-term. Palaeodata provide information on millennial timescales about how much ice is potentially unstable, while our study focuses on how quickly that ice can be lost over the next 200 years. For example, we say, “These constraints are not absolute bounds—greater deglaciation has occurred in the past over longer time scales—but appear to limit the amount of ice that can be lost in two centuries.”
I also asked about the interplay between ice melting vs. falling off (as ice bergs, etc.) into the sea. She told me that, “this can only be evaluated with a process-based model, of course, so this is one of the strengths of our work over previous papers that extrapolated from past observations (and therefore could not account for this). Our results are also consistent with high resolution models that represent these processes in more detail.”
In my previous post I quoted Richard Alley as noting that not all of these non-melting mechanisms were accounted for. Alley had told me,
…the model does not allow loss of any ice shelves, does not allow grounding-line retreat from calving of icebergs following ice-shelf loss, and does not allow faster retreat from breakage of cliffs higher than those observed today, especially if aided by meltwater wedging in crevasses. The model restricts grounding-line retreat to the rate given by thinning of ice during viscous flow of an unbuttressed but still-present ice shelf, with a specified upper limit enforced on the rate of that retreat.
Fundamentally our study aims to represent the aggregate effects of multiple mechanisms, not to simulate each of the individual mechanisms themselves. We then use a wide range of possible representations to sample the uncertainties.
For example, regions predicted (by other studies) to be vulnerable to ice shelf collapse are given a “MISI onset” date, after which the grounding line is forced to retreat. This means the actual ice shelves in the model are, essentially, irrelevant: removing them has little effect because it is “over-ridden” by the forced retreat. The same applies to iceberg calving – we represent its effects in moving the grounding line.
Alley had said, “the model also does not allow retreat up a sloping bed under forcing.” To which, Edwards replied, “We do allow retreat along regions of up-sloping bed. I’m not sure how long a distance Richard would think was sufficient. Also, our aim was to estimate sea level rise due to MISI (a hypothesis specifically about down-sloping beds).”
Alley also noted that the model used in the study had an enforced upper limit that would not allow a very rapid retreat. To this, Edwards provided this response:
We used grounding line retreat rates of up to 3 km per year everywhere in Antarctica and tested rates up to 5 km per year – much higher than observed in the Amundsen Sea Embayment. Our projected ice losses were somewhat restricted by the limit on unbuttressed thinning (and also, in the ensemble, by testing with observations). When we turned this limit off in two of the ensemble members with the highest sea level rise, the results were only 15 cm higher at 2100; when we turned off the observational testing, we predicted the chance of exceeding half a metre increased to only 2%.
Edwards notes that cliff failure may produce higher rates of ice loss, and
by hacking ice off even faster and without the theoretical and observational constraints we used. But it was described by the authors – Dave Pollard, Rob DeConto and Richard himself – as “somewhat speculative”. There are no observations that confirm or quantify it, so we don’t think there is yet sufficient evidence to override the information we do know. It’s also not included in state-of-the-art models (with which, as I said, our results are consistent), such as the high resolution BISICLES: to my knowledge people do not see this as a limitation.
Edwards pointed out to me that the use the term “implausible,” meaning unlikely, but not impossible, and that unexpected processes may at some point emerge.
We look forward to further papers that either confirm our results or else provide strong evidence that faster ice losses are likely over the next two centuries: for example, moving cliff failure from “somewhat speculative” into “current understanding” and estimating the probability of such an “ice swan” occurring over substantial regions on this time scale.
On a finer point of detail, Edwards took the opportunity to clarify what might seem a fine point, but one that is very important, in the research. She notes that the Guardian writeup noted that the study involved 3000 slightly different versions of the model. However, the total range of the variables were wide, but with individual similar models being only a little different from each other.
I’m not entirely sure how to interpret these apparent differences. I strongly suspect, as I wrote here, that a full understanding of the mechanisms of non-melting deterioration of ice sheets will result in higher rates of contribution to sea level rise (which is probably the main variable of concern here). And, I think Alley agrees with this. But Edwards is making the case that these factors have essentially been covered in the reported results, though allowing for the possibility that there are processes that may surprise us. I used the analogy (to which Edwards refers) of an ice sculpture swan falling apart. We can hope that the swan is understood, and that future melting of major ice sheets do not turn out to be a black swan rather than a mere ice swan.