|I thank the authors for considering my comments on the initial submission. I think that manuscript has been improved, but I still have some concerns, as I explained below. Some of them might be a matter of style, and thus are debatable, but the authors and the editor may want to think about them once more.|
Improvements: The main message of the manuscript is now clearer. The study defines a transient sensitivity and attempts to estimate from observations and from model runs. These estimation shows a discrepancy, and the authors conclude that the models may be underestimating future sea-level rise.
1) One main concern is again related to the length of the manuscript, whereas at the same time it compresses important information too strongly. I do not really understand why the authors want to cram all that information, as the length of the manuscript is by far not close to usual limits.
For instance, the the main point of the study is the disagreement between the estimations of the transient sensitivity from observations and models. The reader would think that the methodology, including a clear description of the uncertainties, biases, etc. is crucial. Yet, the method section devotes just one sentence to describe the estimation of uncertainties (We use Monte Carlo sampling...)- Well, Mane Carlo sampling can be accomplished in many different ways, for instance to preserve autocorrelation of the regression residuals, with replacement or without replacement. The reader does not even know how many samples enter the regression - it seems that just 3 data points (?). If this is true, how is Monte Carlo re sampling really accomplished ?
2) Perhaps my most important concerns is indeed related to that very limited number of samples. The study claims that although the physical processes relating global mean temperature and sea level rise would in principle render that link non-linear,, the data indicate that the link is close to linear., at least in the range of the observed changes. I struggle to understand this claim when the available number of samples is just 3. Even when looking at Fig 1 , I would even go so far to say that a regression line drawn using the first two observational data points, PI and TG would not hit the third data point Sat even considering the uncertainty ranges.
I wonder if the claim of linearity would hold in the perfect world of climate models. For instance, if we also include the corresponding pre-industrial and 20th century data points derived from one one model run, would we see a linear relationship through the whole period from pre-industrial to 2100 ? If the linear assumption of this study holds, this should be the case. Actually it should be the case for all individual model runs, as each run would be a surrogate for observations. If this assumption does not hold for the individual model runs - for instance if the the scenario data points fall bellow or above the regression line drawn with the PI and 20th century data points, then the linear hypothesis would not be correct, and the comparison shown in Figure 1 would not be indicative of an under or overestimation by the models.
This would be one possibility., perhaps there are others. What I mean here is that the manuscript leaves open some avenues to support or reject the main linear hypothesis , and I do not see why they are not pursued further in this study.
3) The linear hypothesis raises some additional points. The authors agree that the two main process causing sea level rise (thermal expansion and land ice melting) have a vastly different temperature sensitivities. Then the question arises as to why the linear link between temperature and sea level rise would hold. Is it because the sea level rise is still too small to show that non-linearly ? is it because of these two processes only one has been dominant so far ? There are estimations of the contribution of these processes to 20th century sea level rise. Frederike et al 2020, (doi: /10.1038/s41586-020-2591-3) found that the main mechanisms during the 20th century was land-ice melting and that that contribution has grown larger through time. . It is plausible to assume that for the prindustrial period the main mechanism was water expansion. Thus, why is the link still linear ?
I am of course aware that these questions are not easily to solve, but why not include here a a first step ?
In summary, essentially the very short manuscript is based just in a few regressions - with s very small sample size. The process leading to those results is not clearly explained and the consequences of those results are not really explored.