Review of “Effects of extreme melt events on ice flow and sea level rise of the Greenland Ice Sheet” by Beckmann and Winkelmann

The authors present a set of ice-sheet model simulations to 2300 that explore the impact of extreme events of varying frequency and intensity relative to a simulation with a baseline climate forcing. They construct the baseline temperature forcing using regional climate model estimates of Greenland surface temperature and an emulated global mean temperature time series. They then calibrate a positive degree day model for this temperature time series in order to attain surface mass balance forcing to 2300. They construct nine scenarios for combinations of periods of 5, 10, and 20 years and relative intensities of 1.25, 1.5 and 2, relative to the running decadal mean temperature in the baseline forcing. Running the ice sheet model under these forcing scenarios, they find that including extreme warm events can have a significant impact on long-term mass loss for events with high frequency and intensity. They find a 14% increase relative to the baseline scenario for the most extreme scenario that includes ice dynamics and SMB-elevation feedback. For a case that considers only surface mass balance (i.e., neglecting ice dynamics changes), they find a 16% increase in mass loss.

Overall, I think the study is interesting and the paper is well written. I have two primary concerns with the paper, however, which cause me to recommend major revisions.

The first concern is that the initial condition is not a good representation of the present day ice sheet, with many major outlet glaciers over- or under-estimating observed velocities by a wide margin. It is not possible to discern from the figures just how far from the modern state the initial condition is, but some major outlets seem to differ from observed velocities by >100% (estimating from Fig S2 by eye). Some misfit statistics are reported in the text, but these are skewed by the very large slow-flowing part of the ice sheet and so the reported average misfit of 9 m/yr is not terribly relevant. It is unclear why this initial condition was used, when there are other PISM initial conditions that look much closer to observations. The present initial condition makes it difficult to interpret the results, as I would not expect this model configuration to respond to external forcing in the same way as a configuration that is closer to observations. I recommend improving the initial condition (possibly just using one that has been published) and repeating these simulations and analysis, or at least somehow demonstrating that the initial condition does not significantly bias the results relative to a more realistic initial condition.

My second concern is that the experimental design is to essentially add extra temperature forcing, which leads to the conclusion that extreme events are important. But applying these extreme events to the baseline temperature forcing time series results in a stronger average temperature forcing than the baseline. Thus, there is no way to determine how much of the excess mass loss is really due to the extreme events and how much is due to this increase in average temperature forcing. It seems that the proper methodology would be to ensure that the baseline and extreme scenarios have the same mean temperature forcing over some long-term average (probably a few decades to a century). This would much more clearly show the impact of variability vs mean forcing.

I also found the Discussion section to be rather limited in scope. I have added a few suggestions below of topics to enhance the Discussion. A number of more specific edits, comments, and questions that are also listed below. There are a number of mis-referenced figures, especially in the very long supplement, so that should be checked carefully during revision. 

Specific comments:

L 73: “Consider changes in ocean melt or sliding due subglacial or subglacial processes” needs revision

L 88: submarine melting is kept constant, but how is it calculated, or what dataset is used? Is the melt rate constant for each glacier for all time, or does each cell have an associated melt rate that is applied when the glacier terminus is in that cell? Are there different treatments for floating and grounded ice? Please provide more information about this.

Is there any calving law or criterion used here?

L 123–124: By this logic, Humboldt Glacier should have a fairly good match to observations because its width is large compared with the model resolution. However, the fit is very poor there.

L 125: There should be similar statistics reported for just the fast-flowing part of the ice-sheet (e.g., where speed > 100 m/yr or some other reasonable threshold), where the velocity and thickness are much more relevant than over the ice-sheet interior. 

Specify which version of BedMachine is being used. Presumably v3? Citing the paper rather than the dataset (ie., the NSIDC page) is a bit ambiguous because the Morlighem et al (2017) should be cited when using v4 or v5 as well.

Fig 1B: Subplot title missing a letter?

Figure S1: please add a panel showing thickness misfit as a fraction of observed (BedMachine) ice thickness.

Figure S2: Color bars on all plots are too narrow, resulting in very large areas of saturated color that make it impossible to judge the fit to the observations. Please use wider color bars; 10^4 m/yr would be a more reasonable upper limit for panels (a) and (b).  Also consider using a signed log-scale (e.g., -10^3, -10^2, …, 10^2, 10^3) for panel c to aid with visualization. There should also be a plot that shows the misfit as a percentage of observed velocity. Some of the velocities at these large outlets (notably Humboldt, NEGIS, most of the NW sector, and potentially others, but hard to tell on this color scale) are very far from the observed velocities, which will significantly bias model results in these regions. This makes interpreting the results rather difficult, as the modeled ice-sheet state is quite far from the true modern state.

L 138–140 and Fig S3: The agreement between the modeled and observed mass balance from 1972–2017 seems overstated to me, given that the slope of the observed mass balance is almost twice the slope of the modeled mass balance from 2000–2017.

L 157–162: Difficult to understand. I don’t understand how the anomaly years contain the monthly anomalies, for instance. Please revise these lines for clarity.

L 164: Is Figure S9 the correct figure to reference here? I don’t see how it relates to the text here. Seems like it should be Figure S5

L 181: Should be I1.5f5?

Figure S10: There is only one tick on the vertical axis here, which makes it impossible to determine the vertical scale.

Section S2.1: “ Figures S5 and S6 show that the extremes would increase…” These don’t look like the correct figures. Should be S8 and S9? Also, the brown curve is not defined in S8 and S9.

Figure S12: Why are the two MAR curves here so different over most of the century? I don’t fully understand what is meant by: “SLR from the original MAR data set (Miroc5) of 1km resolution was derived from the ∆SMB”, so perhaps that can be phrased more clearly, with a reference to another figure if relevant.

L 205: It would be helpful to remind the reader in this sentence of what the scenarios are.

Figure 2 and in general: It seems strange that only extreme warm events are included in these scenarios, rather than including both extreme cold and extreme warm events. By including only warm events, you’ve essentially just increased the decadal (or multidecadal to centennial) average temperature by a few degrees, which will of course lead to correspondingly more mass loss.  It seems that the proper comparison would be to make temperature time-series that have the same multidecadal average, so that the impact of variability is actually quantified, rather than to add extra temperature forcing to a baseline temperature time series as is done here.

Figure S13: The vertical axis label should be dST/dz, correct?

L 210: In the SMB-only experiments, does ice thickness change due to SMB? Or is ice thickness held constant in time? Or is advection active, but velocity is held constant? Please add a bit more detail about this set of experiments.

Figure S14: Text seems to reference something that isn’t present in the figure: “he

corresponding ice sheet extent in 1971 (i) and the emerging ice retreat in years 2100 (ii), 2200 (iii) and 2300 (iv) are given in light blue and

red shading, respectively.”

L ~245: Is this shown in a figure or table anywhere?

L267: typo: Mirco5

L 300: Could it also be that CW is the only one that continues to accelerate because Jakobshavn remains a marine-terminating outlet, and that’s not the case for most other large outlets? From Fig1, it looks like the only other outlets that remain in contact with the ocean are Petermann, maybe NEGIS, and maybe Humboldt.

What basal friction law is used here? I see that you use an exponent of 0.6, but what is the form of the law? That could have an effect on the slow-down you observe while driving stress decreases.

The Discussion section is very short and the Conclusions section reads like it should be in the Discussion. Consider expanding the Discussion and including more of a summary of your findings in the Conclusions. Particularly, the discussion should touch on the impact of the initial ice sheet state on these results, as the spun-up initial condition is quite far from the observed modern ice sheet state (Fig S2). This initial condition should be compared with other model initial conditions for Greenland, at least with the initial condition from Aschwanden et al. (2019). Another topic to touch on in the Discussion is that temperature extremes will in reality increase the flux of meltwater to the bed and thus affect ice dynamics through subglacial hydrology, which is not accounted for in these simulations. Finally, some discussion of the full dynamics runs vs the runs without SMB-elevation feedback would be good.

There is no equivalent of Fig 4 given for the dynamic case without SMB-elevation feedback. Overall, it seems like those runs were ignored compared with the SMB-only and full dynamics cases. There should be another subsection analogous to 3.2 in which the full dynamics and no-feedback runs are compared in more detail.

I have rated Presentation Quality as "Fair" because there I think the manuscript relies too heavily on the numerous figures in the Supplement, while there are only a few figures in the main paper.

The following is a review of, “Effects of extreme melt events on ice flow and sea level rise of the Greenland Ice Sheet” by Beckmann and Winkelmann.

This manuscript presents a study of the sensitivity of a Greenland Ice Sheet model to an increased frequency of extreme temperature events in century-scale future projections. The authors design a suite of experiments, in which they the frequency and intensity of events in simulations through the year 2300.  To initialize the experiments and spin up the dynamic state of the present-day ice sheet, PISM is run through the last glacial cycle.  Then, a number of dynamic and surface mass balance (SMB) parameters are calibrated to best fit total ice sheet mass change over the recent historical period.  The authors present global mean sea level contribution results with consideration to SMB forcing only, SMB and dynamics without surface elevation feedback, and a fully dynamic ice response. These experiments illustrate a strong intensification of ice elevation feedbacks after 2100 in response to more frequent high-temperature events, where more frequent and intense events promote increases in interior ice velocities and overall ice sheet mass loss. 

These types of experiments are an important contribution to the characterization of uncertainties in future projections of Greenland Ice Sheet mass balance, as very few studies have investigated the response of ice sheet dynamics to shifts in natural variability.  For this reason, I find that the authors offer valuable insight into the sensitivity of ice sheet dynamics and elevation feedbacks, and into the significance of simulating atmosphere-surface-ice sheet interactions properly. However, I am not convinced that the authors sufficiently introduce, present, and discuss their study, and I think there are a number of improvements needed to better communicate their results to the general cryosphere science community.  Therefore, I support publication of this manuscript but after major revisions.

I have a number of general concerns, which are listed below:

• This study is introduced as a follow-on to Delhasse et al., 2018. While the Delhasse paper is certainly part of the justification for the experiments presented here, it pertains to a much different question, related to atmospheric circulation and blocking patterns that drive extreme melt events on Greenland Ice Sheet.  Because the study presented here does not investigate any shifts in the spatial regime of SMB, but only tests increases in continental temperature anomalies, I urge the authors to revise 1) either how the manuscript is introduced and conclusions presented so that they more accurately support experiments designed to study the effects of extreme increases of temperature of the ice sheet or 2) the experiments themselves so that they use forcing in some way derived from the Delhasse et al. paper.  The current version of the manuscript is framed as a study of both the quantification of how the SMB changes from increased blocking events alters ice dynamics and of how the frequency and intensity of thermal warming events affects dynamic response. Unfortunately, I am not convinced that it fully investigates either, and the current message seems scattered.  See more specific comments below.

• If the main goal is to test the effects of extreme melt events on the ice sheet dynamics, and the experimental design is designed to test the effect of variability (as argued in e.g. lines 343-345), then the authors should consider imposing variability in a way that does not perturb the mean SMB forcing. That is, they could compare a run forced with temperature change spread over the entire year and compare that to the response to the change concentrated within one month (July).  These simulations would more pointedly explore the effect of extreme events on ice dynamics (as opposed to testing the response to just adding more overall accumulated warming throughout the simulation).  If the main focus is to instead investigate the ice dynamic response to SMB change resulting from increased warm events that may be missing from future projections, then is no need to design experiments that contain a varying regular frequency of extreme events, but instead the authors could investigate the ice dynamic response to something akin to the magnitude of SMB change that Delhasse et al., 2018 suggest could result from a persistent blocking pattern (i.e. 2000–2016 climate).  Addition of a distinctly stated scientific question with targeted experiments would improve clarity on which are the intended goals of the study.  This may be able to be accomplished with the current simulations, but reorganization and reframing of the text to support the current experimental design would be needed.  Doing so would strengthen the manuscript greatly and make it more accessible to a broader audience. 

• The manuscript is written for a specific audience and assumes the reader has an extensive knowledge of ice sheet atmospheric and dynamic modeling. While this is acceptable to a certain extent, I would like to see the authors rework the manuscript so that it can be accessible to a broader audience in the cryosphere science audience. Currently, I think that even ice sheet modeling experts will need to read the manuscript multiple times to really grasp what the results are suggesting.  Extension of the discussion to help lead the reader through the implication of the results, especially with regards to pointing out the extensive impact that surface elevation feedback has on the simulations would strengthen the manuscript.  Also, because elevation feedback is so important, a description of how the model simulates such feedback should be included in the introduction or supplement in some way.  Finally, including some equations that describe the physical response of ice to a change in surface slope (in the methods or supplement) would serve to support an extended discussion and the current description of ice dynamic response within the results.

• There are a number of inconsistencies in figure legends and within the text, with regards to how simulations and forcing are references. For instance, please keep acronyms, and especially their capitalization consistent throughout the manuscript and the supplement (e.g. GrIS and MIROC5). Also, especially in the supplement, since there is very little text currently describing the figures, please either expand the captions or add some additional text sections to specifically describe what each figure is, and possible a take home message for each.  Finally, within the text, just MIROC5 is often used to describe the forcing, but my understanding is that it is actually MAR-MIROC5. The correct description of the forcing should be used every time, so that it is clear to the reader what product is actually being used.

Specific comments/questions/suggestions:

Abstract: Please mention within the abstract which future scenario is being used for the projections. 

Abstract: Since the future SMB is self-imposed by the experimental design (hypothetical changes in intensity and frequency for events), total numbers reported in the abstract in terms of sea level contribution do not have significant meaning to the reader, and are a bit misleading.  Perhaps, you could report percent contribution from ice dynamics specifically with respect to its forcing (e.g. percent change in warming)?  Or perhaps another diagnostic that is more appropriate and better represents the study results? 

Abstract: Mentioning that surface elevation feedback is very important would also be appropriate here since it is a major finding of the study.

Line 15: sea level rise -> global mean sea level rise

Line 30: Isn’t this true only if 1 heatwave can happen per year?

Line 31:  Please define for the reader what your criteria is for an extreme melt event. 

Lines 45-46:  Please reword this sentence.  It is a bit awkward and unclear what is meant.

Lines 61-63:  Hanna et al., 2008 is an older reference.  Perhaps this statement could be expanded to include references that pertain to conclusions about the newest CMIP simulations, as well as other natural climate states that modulate melt (i.e. Delhasse et al, 2021)

Line 67: To best introduce the reader to the past work on this subject, please expand this statement and summarize the results from Delhasse et al., 2018 here (i.e. their results suggest that current projections neglect changes in extreme events, and so underestimate future reduction in SMB).  In addition, please frame how the experiments here relate to the Delhasse study.  For instance, Delhasse et al., 2018, specifically look at blocking patterns and therefore the spatial impact they have on Greenland SMB.  However, the study presented in this manuscript does not consider spatial patterns in temperature and SMB.  Please lead the reader through how the two studies connect, despite such disconnects.

Line 74: “do not consider changes” -> “do not consider ice response to changes in ocean-induced melt or basal sliding”, or something similar that is more descriptive of the processes you refer to here.

Line 85: Please describe/cite the methods used for moving the grounding line and the calving front.  Is the calving from model independent from the retreat of the model extent on land?  Or are the plots of ice retreat (e.g. Fig. 1) just showing where ice thickness is <1m on land area that is above sea level?

Line 87: Please note what is the strongly negative SMB defined as. Is this negative SMB prescribed at the points in question? How does that work with the PDD scheme used for the experiments, and does it result in steep slopes and gradients at the edge of the ice sheet?

Line 88: Please include how these melting rates are defined, and in the cases where the grounding line retreats and there is new interior floating ice, how are the submarine melting rates set?

Line 96: What is the reasoning behind the very fine vertical resolution?  Does the thermal model require this resolution?

Line 97: Does the model treat 1m as ice as still part of the ice sheet, but these areas are then excluded from the analysis (and plotted as non-ice in your plots)?  Please specify this in the text for clarity.

Line 105: MAR3.9 “forced with” ERA-40 and ERA-Interim. Please note the year span that each reanalysis (40 vs. Interim) product is used.

Lines 111-112:  Please specify at what point of the initialization (what year) this is done.

Line 115:  Are these the values for the entire spin-up, or are they imposed at a certain time?  Please note that in the text.

Line 134: What are the criteria used to determine that these are agreeing well?  The trends are very different between the two starting in 2000, which - if I understand correctly - may suggest the model is not really responding dynamically to the shift in climate in 2000.  Also, for the SMB comparison in Fig. S3, is part of the discrepancy between PDD and Mouginot SMB because here it is shown with the control subtracted, but emphasis was placed on a SMB match without subtracting the control (i.e. results show in Fig. S12)?  Overall, it is unclear to me what the results in Fig. S3 mean, so perhaps you could use some more text to acknowledge some of the mismatch and offer explanation / argument on why that is, and justify why it is acceptable for your experiments.

Line 141: Please quantify “slight”.

Line 159: Is this MIROC5? My understanding is that it is MAR-MIROC5.  Please note this accurately throughout the text and the supplement.

Line 164: Is this Fig S5?

Line 166: Is this Fig S12?

Lines 163-166: This paragraph is overall confusing, perhaps because the figures names do not seem to line up with the text.  However, I suggest that it be rephrased to better explain what is meant by “the calculated SLR in this case is closer to the original MAR results than when using a 2D temperature field” and how it pertains to the particular Figure being referenced.

Line 172: Please note here that the changes are being made to the forcing only during July.

Line 198: In this figure you show the closest fit over the whole ice sheet, but what is the match like spatially after this tuning?  Are the gradients in SMB comparable at all? This is a pertinent question since these gradients will be driving the future slope change of the ice sheet and therefore its dynamic response.

Line 199: Looking at S12, it looks like the best match was 5 degrees, but also with consideration to average temperature (as opposed to 2D)? Please specify. The S12 caption also mentions that the result is time-dependent, but it is unclear what that means.  I urge the authors to include much more description and discussion of this method.  Currently, it is difficult to follow how the different standard deviation options are derived and then judged. A description of this method is important, because results likely depend greatly on the resulting SMB that is derived by the method.

Lines 202-203: What are these values typically set to in past studies?

Line 207: Please specify surface temperature (to distinguish from ice temperature).

Line 208: Please do not refer to model output as data.  This should be updated throughout the manuscript and in the supplement.

Line 214: Is atmospheric lapse rate the same as the temperature lapse rate above in line 208?  If so, please use consistent wording for clarity.

Line 216: Please describe what the conditions are for the control run and how it is created somewhere in the methods or supplement.

Line 222: Please quantify slight here, and reference the plot which shows this.

Line 230: This conclusion seems a bit obvious, that if you greatly reduce the total SMB, Greenland will lose more mass.  As mentioned above, it would be very interesting if you added experiments that allowed you to determine if extreme events themselves, with respect to their frequency and intensity, cause a different response than an equivalent amount of SMB change applied throughout the year. 

Lines 243-248: Please reference figures and tables in this paragraph to help lead the reader through these results.

Lines 261-264: This is true, you are definitely testing something different, and here, you make the point that your experiments are not imposing SMB changes that are as extreme as Delhasse et al.’s permanent blocking conditions.  It is difficult to make the connection on why that comparison is important to the manuscript, because your experiments are so different.  Could you expound upon that in the text, and offer some more discussion around why this is an important point with respect to your study and results?

Lines 271-274: Please make a specific mention of the ice elevation feedbacks here, since it is driving most of your results.

Lines 340-346: As mentioned above, it is clear that this study is much different from the Delhasse et al., 2018 study.  However, it is not clear what value comparison against their results brings to your manuscript. In addition, the last two sentences of this paragraph are awkward and difficult to interpret.  Please rephrase these sentences for clarity and to help the reader understand the value in comparison against the past study.

Lines 350-352: Please add some text to explain to the reader what the results of these comparisons mean or suggest. What should be the reader’s takeaway?

Discussion: Elevation feedback should be added to the discussion, beyond within the bounds of limitations, and the authors should add text interpreting the fact that ice dynamics without feedbacks do not really change much between the various experiments or over time.

Line 365: Can you offer any reasons why your simulations show more sensitivity?  Have you tested whether it may be related to the refreeze factor or other tuning?  Are there comparable estimates of this feedback for the Aschwanden projections or others that use PISM?

Lines 370-371: Note here that the tuning was done for total ice sheet-wide SMB.

Lines 377-378: With the current experimental design, I would argue that instead of showing the importance of including extreme (short-term) events, your results show how important it is to capture feedbacks between atmospheric circulation, ice dynamics, ice surface change.  Such results suggest that the use of better surface models and a coupled ice-atmosphere setup for projections may be imperative to properly quantifying ice sheet response to future climate. Perhaps this is just a matter of how the text is currently worded, and I misunderstand your meaning here.  Please think about rephrasing and expanding the conclusion section to frame the conclusions and explicitly tie them to your results for the readers.

Figure 1: It is unclear how the ice front changes in the model, and what is being shown in this figure.  Are places <1m thickness indicated as bedrock?

Figure 1: In panel (b), what does the reddish color along the edge of the ice mean?

Figure 2: This please indicated MAR-MIROC5 instead of Miroc5, if that is indeed the case.

Line 383:  This seems to be the same link repeated.  Please add a link to the MAR-MIROC5 results here.

Figure S1: Please saturate the colors on this figure so that it is easier to see smaller changes.

Figure S2, panel (c): This figure looks over-saturated.  It may help to expand the color bar bounds.

Figure S3: Note that MAR should not be referred to as data.  Also, please add further explanation of what the orange line is.

Figure S5: These differences look to directly correspond to where the dynamic speedups of the simulations are found (i.e. Fig. S22), suggesting that your results may be much different if you were to take 2D fields into consideration.  In the text (line 355), you mention that this treatment could potentially add a bias to the results.  I would argue that your results strongly suggest that the experimental design adds a bias.  Please update your wording to reflect this in the Limitations section of the text.  Do you have any thoughts on the magnitude of the total bias that this introduces?  It is likely related to your highly sensitive elevation feedback in the area.  In this case, it is not clear how meaningful your total number of sea level contribution are. Can you offer further justification for the experimental design (i.e. computational resources or other technical complexities)?

Figure S8:  Please explicitly define std_extreme. 

Figure S11 panel (b): How much does this relationship vary over the ice sheet?  It likely varies seasonally - does this represent only Julys or summers?  

Figure S11 panel (c): Could you offer fit/regression statistics for the line?  What do the number of points represent here (i.e. what is your temporal resolution)?

Figure S12: With total sea level on the y axis, would sea level equivalent be more appropriate of a label than SLR?

Figure S15: The caption says 2300, but should be 2100

Figure S16: The caption of this figure is a bit misleading, because it sounds like you are comparing against Delhasse et al., 2018.  However, as far as I can tell, you are only plotting your results, and the reader will need to have the knowledge of the other manuscript to make a meaningful comparison.  Is there a way to add Delhasse et al., 2018 results here? This figure is also difficult to digest and needs much more explanation and interpretation in the caption or within the supplement text. For example, please explicitly clarify what each legend symbol means.

Figure S21: This figure needs a better explanation of what it is showing, or a rewording of the caption for clarity. 

Figure S23: Please define what you mean by flux?  Is it ice flux through define gates on the ice sheet?  Is it equivalent to total ice discharge?

Below, I offer some other specific technical suggestions:

Line 66:  “are” -> “would be”

Line 70, 76, 103, 150:  GRIS -> GrIS, please be consistent.

Line 131:  “These are” -> “This is”

Line 381: pism->PISM

Fig. 2, S4, S6, S7, S12, S26, Table S2: Miroc5->MIROC5

References:

Delhasse, A., Fettweis, X., Kittel, C., Amory, C., and Agosta, C.: Brief communication: Impact of the recent atmospheric circulation change in summer on the future surface mass balance of the Greenland Ice Sheet, Cryosphere, 12, 3409–3418, https://doi.org/10.5194/tc-12-3409-

2018, 2018.

Delhasse, A, Hanna, E, Kittel, C, Fettweis, X. Brief communication: CMIP6 does not suggest any atmospheric blocking increase in summer over Greenland by 2100. Int J Climatol. 2021; 41: 2589– 2596. https://doi.org/10.1002/joc.6977