Review of “Assessment of the Greenland ice sheet atmosphere feedbacks for the next century with a regional atmospheric model coupled to an ice sheet model” by Sebastien Le clec’h and others.
Summary:
The authors have dramatically improved the quality of the paper since the last version and most of my comments have been sufficiently responded to. Since the paper has changed a lot, I still have a number of additional minor comments that the authors should consider before publication. Because Figure 9 does not appear to be displayed in the manuscript, I feel I have to tick major revisions. If this problem should be resolved and the figure is checked by at least one of the reviewers, I don't need to see the manuscript again.
General comments:
It is not clear to me why the sea-level contribution (and some other quantities like surface elevation, ice thickness, ice masks ...) should be averaged over 10 year periods in this reporting. In consequence, a forcing period of 100 years (2000-2100) is practically reported as a 90 year difference in the results, which is confusing. The sea-level contribution is physically a time integrated quantity and does not exhibit inter-annual variability that has to be averaged out. In most cases (instead where strong inter-annual variability exists (SMB, temperature)) I would omit the averaging and calculate direct differences. If not, the labels in the table would have to be adjusted to the centre of the averaging period in all tables and plots instead.
In the reorganisation of the manuscript (that I appreciate), the 2W experiment is now discussed first as the reference and differences to NF and PF later. Since you introduce 2W as the standard experiment, I would plot and discuss the differences NF-2W and PF-2W (instead of e.g. 2W-NF). It just changes the sign, but seems more consistent with the rest of the document. I think it would also make sense to explain the model setup/coupling method in that order, starting with 2W.
Minor comments
I believe the affiliation count is incorrect with Fettweis and Wyard linked to Brussels and Ritz linked to Liege.
P1 L9 Start a new sentence after MAR: "They are fed ..."
P1 L16 Remove "important". There is already an "important" in the line before.
P1 L18 Remove "tend to" before "favour" to make this statement less vague.
P1 L20 Does it "reduce the SMB signal" or rather re-distribute the additional mass. Clarify.
P1 L22 This should probably be "ice volume above floatation".
P2 L1 The results this conclusion is based on have not been described so far in the abstract. This must arise from a comparison to the uncoupled experiments. Suggest to mention those before this conclusion.
P2 L14 Replace "polar" by "Arctic".
P2 L18-21 Complicated sentence, consider splitting in two and revising.
P2 L23 Add "changes" before "in ice-covered area" to make the relation clearer.
P2 L23 I think with "albedo feedback" you mean here the change in land surface type from ice to tundra, which leads to warming and further ice sheet retreat. This is typically a slow process, I would call "planetary albedo feedback". There is also a more immanent "melt-albedo feedback" as melting snow at the surface absorbs more short-wave radiation, leading to more melting and increasing albedo. These feedbacks are quite different in nature and time scale and should ideally be distinguished.
P2 L27 Replace "predict" by "project".
P2 L29 In the manuscript you use different ways to order the references. Here it goes from newest to oldest. The TC guideline is open ("In terms of in-text citations, the order can be based on relevance, as well as chronological or alphabetical listing, depending on the author's preference."), but I would at least try to be consistent in all cases throughout the manuscript to not confuse the reader.
P3 L32 The approach of Edwards et al. was only used to correct for the SMB-height feedback. Here it sounds like it was also used for downscaling ("An alternative approach [to Franco]"). Please reformulate.
P4 L9 Replace "surface energy balance" by "surface mass balance".
P4 L27 Replace "second" by "third".
P4 L31 "to the other, uncoupled experiments".
P5 L11-13 "and 24 vertical levels" misses a verb. Suggest "MAR has a horizontal resolution of ... and 24 vertical levels ..."
P5 L17 Maybe "covered by at least 0.001 % tundra and at least 0.001 % snow" to make clear both surface types are represented. Also add an explanation why that is done.
P5 L21 Add why these parameter changes were needed? What was improved?
P6 L12 Regular grid? What projection? Refer to Bamber?
P6 b_melt is only defined for grounded ice, I suppose? What is done for shelf melting? Clarify.
P6 L19 Remove "also" before "affects".
P6 L27 Consider defining "The shallow ice approximation (SIA)" instead.
P7 L17 Section title should be "Initialisation procedure". "Spin-up" should be reserved for a consistent long-term transient run as often done with ice sheet models.
P7 section 2.2.2
At the time of the first review of this manuscript, the GMDD paper of the same author describing the initialisation procedure was not available. It was therefore not possible to understand the procedure from the limited information given in this manuscript. I have therefore asked for more detail to be included. Now that the GMDD paper is published, it would be enough to give a broad overview of the method here and refer to the other paper. At the moment, a large part of 2.2.2 is a copy-and-paste from the other paper with a fair amount of technical detail. I don't think this is a problem in terms of plagiarism, but I would encourage the authors to rewrite and shorten this part to summarise the most important aspects.
What I miss so far (also in the GMDD paper) is a clear idea of the basic principle of the method. While similar to PD2012, the main difference is that beta is modified in function of the thickness ratio, rather than the thickness difference. This has not been clearly state and is lost in the complicated formulation. I also find the formulation through U^{corr} very confusing, since the modelled velocities are never directly corrected nor compared to observations. When the equations 4,5 and 6 are put together, one can arrive at a simple expression for the way beta is adjusted in the method (see below). I think it would add substantially to the process-understanding to include it here and/or of still possible in the GMD paper.
In the method the basal drag coefficient beta is updated iteratively by multiplying the old beta with a factor rbeta = beta_{new}/beta_{old} (was U^{slid}/U^{corr}_{slid} in Eq 6 before).
By combining Eq 6,5 and 4 one can show that the inverse of rbeta is beta_{old}/beta_{new} = rH + U_{deformation}/U_{sliding} * (rH-1),
where rH = H^{G}/H^{obs}, is the ratio of modelled and observed ice thickness, and U_{deformation} and U_{sliding} are the modelled velocities due to deformation and sliding, respectively. This means that the adjustment of beta is in the end a function of the thickens ratio with stronger adjustment in regions dominated by deformation.
P8 L5-31 It seems to me that you start by skipping step 1 and then end by skipping step 2. If that is the case it would be clearer to exchange 1 and 2 and start the description with what is now called step 2.
P8 L20 Maybe add that beta_new = beta_old in the first iteration for clarity.
P8 L28 I don't think NB_cycle is used afterwards. Maybe it is possible to avoid using the symbol all together.
P9 L7 I would say the "impact" of this condition is strong because it keeps ice from building up where there is none in reality. What you want to say is that it has a negligible impact on the projected SL contribution.
P9 L8 Remove "quite" before "negligible".
P9 L23 This sentence repeats most of what is already in the sentence line 20. Remove?
P9 L27 I don't think the SMB can be said to be consistent with the 5 km topography.
Maybe "generate a 5 km resolution SMB for the Bamber et al. (2013) topography"?
P9 section 3 You chose to show results for 2W first and discuss differences to the other experiments afterwards, I agree with that. But wouldn't it then make sense to also explain the model setup/coupling method in that order, starting with 2W?
P10 L18 It is not clear what "aggregated" means. Some process to go from 5 km to 25 km, but what is it exactly? Please explain that better.
P10 L28 First sentence of 4.1.1 difficult to read. Move "in the 2W experiment" before (Eq. 1).
P11 L5 You say there is no inter-annual variability anymore after 2100, but where does the +-13 Gt/yr come from then? I suppose MAR still has inter-annual variability, but not the GCM boundary condition. This should be clarified.
P11 L8 I have problems with the formulation "two distinct patterns". You chose to analyse the results separated in two different regions. If you would have made four regions, you would probably observe four different distinct patterns. Can this be formulated better?
P11 L26 There should also be a *shift* of the ELA, which should be important to note here.
P11 L26-27 The relationship between the first and the second part of this sentence has escaped me.
P11 L29 The ice thickness anomaly at what time? Clarify.
P11 L29 Reformulate "two distinct patterns".
P12 L8-9 Fig 5a and 5b look nearly identical, because you plot the integrated total SMB (5a) and integrated total ice flux divergence. It would be much clearer to compare integrated SMB anomalies (and integrated flux divergence anomalies) instead.
P12 L10 Replace "ice melting" by "runoff".
P12 L15 I think "ice dynamics" is a too broad term here, you may want to say that "ice flow" is impacted by changes in the geometry instead.
P12 L19 "by the combination" of what. There is only one thing mentioned in the following: larger surface slope. Clarify.
P12 L23 What do you mean by "fully consistent with the decrease of ice thickness"? The velocities are dependent on both thickness and surface slope (and maybe on changes in basal sliding), as you state yourself in the next sentence. This is confusing.
P12 L29 Is it really that clear cut? Changes in surface slope above 1500, ice thickness changes below? Maybe the wording should be modified to "dominated by" or "governed by", allowing for transitions between the two.
P13 L30 "in the same region with full ice cover."
P13 L31 Remove "artificially".
P14 L7 "positive elevation-SMB feedback".
P14 L8 "and thus increased runoff".
P14 L21 You should explain why the SMB is lower in PF compared to 2W here. The SMB change in PF is only related to elevation change but in 2W wind changes are dominant.
P14 L29 Could you explain why there is mainly negative values in PF? It seems strange to me that there are not more positive values. Or are the positive changes so small that they are not visible on the graph. If so, it would be good to mention that.
P15 L21 I am not sure the attribution to model resolution is as clear cut as it is written here, it seems like a speculation rather than a result for me. Using a different model as such could be an explanation as well I wouldn't exclude. Or is there any other information that convinces you about the resolution dependence, maybe your own experience running MAR at a lower resolution? Please add if that is the case.
P15 L25 Albedo feedbacks have not been really discussed so far. Suggest to remove specific mention here.
P15 L8 You have to make clear somewhere (earlier) that *the ice sheet mask (in the ISM) is free to evolve in all three experiments*, but that MAR does not see mask changes in PF and NF, because it is calculated on a fixed observed ice sheet geometry. I.e. make a clear distinction between the (modelled) ice mask in the ISM and in MAR.
P15 L31 It seems that at the very margin the albedo effect is over-compensated by the cooling due to wind anomalies. Make that clearer. Also note that the melt-albedo feedback (as distinguished above) should be active in the ablation at any case.
P16 L1 "A widely used method to estimate the GrIS contribution to global sea-level rise".
P16 L3 I would say you do something altogether different here by coupling two models. Reformulate.
P16 L4 "by GRISLI, and MAR sees those changes in the topography", or similar.
P16 L9 A note on notation. These SMB symbols read like "MSK minus NF". Maybe use underscores instead.
P16 L12 "this error has a similar magnitude compared to errors made when ...".
P16 L14 I have mentioned this in my earlier comments and without insisting on it, I would like to reiterate it here: I think it would be interesting to see what mask changes would be produced by using the SMB changes in NF and PF to modify the ice sheet mask offline.
P16 L18-20 This is already a long sentence and may be better to split, but I would suggest to add "two uncoupled experiments" before PF.
P16 L21 remove "first" before step.
P16 L27 I agree that extending your experiments by using additional and different GCMs would be an important next step in its own right. I would mention that first, not at the end of the paragraph. How the difference between 2W and PF would play out in those runs is certainly interesting, too. I would speculate that the stronger the forcing, the more important the differences between coupled and uncoupled experiments are, similar to the finding of stronger relative error for longer simulations (i.e. stronger forcing).
P16 L30 "Another question ..."
P16 L32 "Another consequence is ..."
P17 L5 I think repeating the exercise with a different RCM would also be important, since they may have different sensitivities.
P17 L6 "Another limitation is ..."
P17 L9 Replace "we choose to" by "we have chosen in this study to".
P17 L14 I am not convinced it would be a good compromise to accept unrealistic model drift in the ISM in exchange of a better absolute SMB. I would say there is simply no way around further improving the initialisation, which may also require higher resolution of the ice sheet model.
P17 L18 This is not a full sentence. Reformulate.
P17 L33 I agree, but aside from improving the model in this regard, you would also need to force the model with appropriate ocean forcing, which is so far excluded in your setup.
P18 L4 ""it seems to be preferable" is a bit weak a statement. I would say "it is essential".
P18 L17 Add "taken" before "constant in time".
P18 L27 Can you find a better description than "GRISLI-like models"? Maybe "Large-scale GrIS models".
P19 L4 I don't think the models used in Edwards et al were more sophisticated. In fact , GRISKI was one of them.
P19 L10 Apparently the use of AD is discouraged. Use CE instead or just write "year 2000 to 2150".
P19 L16 I hope "In turn, changes in the shape of Greenland modify the ice velocities." Is not a conclusion of this study. That's ice sheet modelling 101.
P19 L18 "accounting for the feedbacks".
P19 L19 "sea-level rise of 20.4 cm in 2150 against 7.9 cm only in 2100" only illustrates that the forcing increases, but not that the feedback does. You could use changing differences between 2W and NF for that.
P19 L25 Problematic conclusion, because RCM-only experiments (like you) also ignore other factors like ocean forcing. So they overestimate the SMB component, but that may be compensated by missing outlet-glacier changes. These details do matter.
P19 L30 You don't mention the code for the GRISLI model. How can the readers obtain the code of the ice sheet model?
P20 L6 It is not really my role to comment on the Acknowledgements, but in appreciation of your own work and that of the other reviewers, I think it is more than "the writing" that has improved in the manuscript. Suggest to write "helped to improve the manuscript".
Tables:
P26 Table 1 No need to average (see general comment). I suggest to calculate 2050 2100 and 2150 differences to the year 2000 instead.
P27 Table 2 Same comment as for Table 1. No need to average. Then remove "Mean" from the caption.
For a better overview these results (SLR) could be joined with table 1 and the two SMB values reported in the text.
Change label in first row to SMB_(MSK_NF).
Same comment as above: these SMB symbols read like "MSK minus NF". Maybe use underscores instead.
Figures:
Figure 1
Caption: What ice mask definition is used for the integral over the "entire GrIS", Clarify.
Over which periods are the regression lines defined? Specify in the caption.
What does the black dashed line stand for? If it is for 0, a solid thin line may be better and should extend all the way to the left y-axis
Figure 2
Why are b and c given on the MAR grid? Shouldn't they be given on the GRISLY grid for consistency?
In this figure as well as many other, anomalies are denoted x -- y. Not sure if the symbol "--" stands for minus. If so, why not use the minus sign "-" instead?
You should add a note that the colour key is non-linear.
The dashed grey surface elevation contours are not visible on a A4 printout of the paper (same for most of the other 2D plots). Consider using a darker colour (black). On the contrary, the lat/lon lines are much less important and could be less pronounced.
Figure 3
Also here, It seems strange to report surface elevation as a time average, when you could as well just plot the 2150 result. Even stranger for the ELA! Just plot the 2000, 2100 and 2150 results in line with your legend in the figure.
It looks like the 2000 and 2100 lines of the ELA are identical between experiments. You should mention that or that the lines are not (hardly) distinguished to avoid confusion.
Use colours for the ELA lines that are not part of the colour key.
Figure 4
Why not use the same colour scheme as in Figure 5a/b? That would make comparison easier.
Figure 5
It would help the interpretation a lot to compare integrated SMB *anomalies* (Fig 2a) and integrated ice flux divergence *anomalies* instead. Now the two plots are on first view identical and it is hard to visually extract the difference.
Explain in the text what the integrated ice flux divergence is when introducing EQU 2.
Add to caption that the grey shade is the land-sea mask.
Figure 6
Since you introduce 2W as the standard experiment, I would plot NF-2W and PF-2W instead (opposite sign).
Add to caption that the grey shade is the land-sea mask.
Remove the log10 label from the plot (applies also for other plots).
Figure 7
Plot labels a,b,c, ... and refer to those in the caption.
The Joughin et al (2010) data set has since been updated to full coverage over the GrIS. Consider using this improved data (ref below).
The inset panels for observations are way to small. Include as additional full size panels on the top. You can gain extra space by removing redundant colour keys (the middle two are repeated).
Remove "surface" before "velocity".
JOUGHIN, I., SMITH, B., & HOWAT, I. (2018). A complete map of Greenland ice velocity derived from satellite data collected over 20 years. Journal of Glaciology, 64(243), 1-11. doi:10.1017/jog.2017.73
Figure 8
Same here, suggest to plot NF-2W differences instead.
Figure 9 !does not appear to be in the manuscript. I see only the caption!
Figure 10
Same comments as for Fig 8
Figure 11
Shouldn't red in a) and blue in b) be identical? For some points that does not seem to be the case.
Again, I doubt the averaging is really needed. If maintained, the labels should be adjusted to the mid-point of the averaging period.
Except for a few points, PF-NF is only negative. Why?
Figure 12
Please use a different set of colours in b and a to avoid confusion.
The averaging is really not needed here.
The differences are very small, but why are NF and PF different in (b) from 2W at the beginning?
Supplement
Figure S1 Match figure title with caption (anomaly vs error).
Figure S2 Mention that the scale is non-linear.
Figure S3 Suggest to repeat and update the caption from S2 for convenience of the reader.
Mention that the scale is non-linear.
Figure S4 Mention that the scale is non-linear.
Figure S5 Not clear what "loosing 100 % of the ice cover" means. I think you mean the pixels that have 0% ice cover in the end, even if they had less than 100 % to start with. Maybe "pixels loosing all of the initial ice cover"?
Figure S6 Remove first part of the caption which is scrambled.
Shouldn't the annual snowfall anomaly (a) be equal to the sum of the 4 seasons? Or are you plotting the snowfall anomaly in m/season instead of m/yr? Same problem may apply to S7 and S4, but I don't know if plotting ratios solves it for you in the latter.
Figure S7 Suggest to repeat and update the caption from S6 for convenience of the reader.
Figure S8 Mention that the scale is non-linear.
The red solid line is not visible for me on this plot.
Figure S10 I would e.g. call this a difference not an anomaly
I think it should be a reference to Eq. 2 instead.
Generally throughout the manuscript. I would use "differences" between different experiments and "anomalies" within the same experiment or when a differences reveals an anomalous signal. "Errors" can be used when a clear target reference exists.
Figure S11 Consider plotting an updated velocity field with full coverage:
JOUGHIN, I., SMITH, B., & HOWAT, I. (2018). A complete map of Greenland ice velocity derived from satellite data collected over 20 years. Journal of Glaciology, 64(243), 1-11. doi:10.1017/jog.2017.73 |
- Printer-friendly version
- Supplement
