the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 Jul 2024
Investigating the surface mass balance of the Laurentide Ice Sheet during the last deglaciation
Abstract. In spite of decades of research, the role of climate feedbacks in the Pleistocene glacial cycles is still not fully understood. Here, we calculate the surface mass balance (SMB) of the Laurentide Ice Sheet (LIS) throughout the last deglaciation using the isotope-enabled transient climate model experiment (iTraCE). A surface energy balance framework is used to calculate yearly melt, and a parameterization of the refreezing of snow melt and liquid precipitation is incorporated. The SMB calculated from iTraCE overestimates the total ice mass loss rate in comparison to the ICE-6G reconstruction from the Last Glacial Maximum (LGM; 21 ka) until about 15–14 ka; subsequently, the fully forced climate model experiment better fits the ICE-6G ice volume loss rate. We find the melt rate for the LIS to be primarily set by the small residual of large net shortwave and longwave radiative fluxes. The melt, and hence the SMB, are very sensitive to small changes in the albedo and downwelling longwave radiation. By increasing albedo by a mere 1.9 % or by decreasing downwelling longwave radiation by only 1.45 % (well within the uncertainty range of these variables), the large overestimation of the rate of mass loss deduced from the SMB compared to reconstructed rates of mass loss from 19–15 ka can be eliminated. The inconsistency of the climate model-derived, offline SMB calculation and the ice mass reconstructions exists irrespective of the role of ablation caused by ice flow, which cannot be calculated using this analysis. The extreme sensitivity of the melt rate suggests that General Circulation Models (GCMs) still struggle to reliably calculate the SMB, presenting a significant roadblock in our attempt to understand the Pleistocene ice ages.
- Preprint
(10173 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-1998', Anonymous Referee #1, 02 Sep 2024
Review ofInvestigating the surface mass balance of the Laurentide Ice Sheet during the last deglaciationby Koepnick and othersGeneralThis paper presents estimates of the surface mass balance (SMB) of the Laurentide ice sheet (LIS) through its last deglaciation (21-12 ka ago). A benchmark is provided by the glacial isostatic adjustment (GIA) from ICE-6G, which puts thresholds on mass loss rates. An important discrepancy is found, which cannot be explained by the non-consideration of ice dynamics because the SMB model overestimates mass loss. Various potential reasons and sensitivities are discussed. The paper is interesting, well written and the figures are clear. The analysis is generally meaningful, but important information is lacking to get a clear picture of potential sources of error, see below.Major commentsl. 51: "In another study, Ullman et al. (2015) found that the SMB of the LIS for key time slices, when forced by an AOGCM (Schmidt et al., 2014), was positive throughout much of the deglaciation, therefore suggesting that ice flow and dynamic discharge was mostly the cause of mass loss until about 9 ka." This is inaccurate. When solid ice discharge is nonzero, SMB must be positive for an ice sheet in balance. When SMB decreases, but still remains positive, the ice sheet will lose mass. This is for instance the case for the contemporary Greenland ice sheet. Exactly as you state one sentence later "Moreover, this study implied that the sign of SMB is not a good predictor of glacial growth or decay."l. 68: "...independent geophysical constraints as represented by ICE-6G". It is great that these constraints are independent, but how accurate and suitable for this goal are they? One weakness is that GIA depends on the ice volume history used, and represents mass changes from both surface and ice dynamical processes.l. 120, Equation (1): (a) The equation contains the geothermal heat flux, but this flux is usually neglected in the surface energy balance as it is so small. Moreover, at the surface of a thick ice sheet the bedrock is even further away, further reducing this flux. You mention this later, but the text and even figure devoted to GF in l. 126-130 is too elaborate given its insignificance (and uncertainty).l. 120, Equation (1): (b) What is however missing from the equation is the subsurface (or ground) heat flux, usually denoted by G. It is the conductive heat flux along temperature gradients just below the surface. This flux cannot be neglected and plays an important role in the modulation of surface melt, see e.g. a recent paper by Van den Broeke on the various energy contributions to melt in Greenland and Antarctica (doi: 10.1371/journal.pclm.0000203). As the only non-latent heat transport process, the subsurface heat flux also is important in the subsurface refreezing process. Why was this flux not included?l.121-125: Radiative fluxes are the most important drivers of surface melt. By using the net surface radiative fluxes from CLM, you commit to CAM's radiation schemes and CLM's (snow/ice) albedo scheme. Please describe these schemes here. What albedo values are used for ice and snow, are impurities considered, impact of clouds and snow wetness/grain size etc.? The modern CLM has an elaborate snow albedo scheme.Section 2.3: Other important information is missing in this section. What was the time step used for the melt calculation? Many processes associated with melt over polar ice caps are highly nonlinear, so using e.g. daily averages of energy fluxes to calculate melt will lead to large uncertainties, especially in regions where melt is non-continuous.l. 145, Equation 2: Upon refreezing, large amounts of latent heat are released in the snow/firn. This will reduce the subsequent refreezing capacity. Is this accounted for?l. 150, Equation 3: this way of defining SMB includes refreezing or 'internal accumulation' and is formally referred to as 'climatic mass balance' (see glossary of mass balance here: https://wgms.ch/downloads/Cogley_etal_2011.pdf). Fine to define SMB this way (many do it) but for clarity it's good to show that you're deviating from the formal SMB definition.Same line: I find the notation of the mass fluxes confusing. If 'P' stands for precipitation, I interpret P_s as solid precipitation (snowfall) and P_l as liquid precipitation (rainfall). The SMB equation then becomes, with P_s - SU being snow accumulation:SMB = P_s + P_l - SU - RUwhere SU is sublimation and RU is runoff, which can be written asRU = (ME + P_l) (1 - f)where f is the refrozen fraction. Substitution givesSMB = P_s - SU + f P_l - (1-f) MECautionary note: you use 'SMB' both for ice sheet integrated mass change (kg/yr, Fig. 1) as for specific mass loss (m/yr, Fig. 2). Avoid the term 'net melt', instead use runoff or the likes.Figure 2: Are these fluxes averaged during melt?Figure 3: Not sure if I understand the signs of all these fluxes. Should netLW not be negative and SHF predominantly positive? In meteorology, SHF is defined positive when warming the surface.Minor and textual commentsl. 3: "...the isotope-enabled transient climate model experiment (iTraCE)." Is the fact that the model is isotope enabled relevant for this work? If so, please state that here and explain why. Also applies to l. 66.Figure A1: Please include ice thickness over Greenland also.l. 115: "Snow accumulation is denoted by PS". Do you mean snowfall? Snow accumulation is usually defined as snowfall minus sublimation.l. 116: "liquid rain accumulation". This is unclear, rain is always liquid and rain does not tend to accumulate. Did you perhaps mean rainfall?Figure 1: In y-axis labels adding "w.e." is not relevant, because you provide integrated mass fluxes in kg/yr. On the other hand, in Fig. 2 (units m/yr) adding 'w.e.' is relevant, but not done...Figure 1: How deep does the ICE-6G curve dip below the x-axis in the BA? Ah, I see this is presented in Fig. A2.Reference list: the reference list was messy and hard to read, as it contained a mix of names with/without first names and did not start with last names.l. 192: "Given the abrupt warming during the BA period, it is possible that ice flow and calving account for the difference between SMB and the estimated trajectory of the ice mass rate of change based on ICE-6G." Would be good to mention some processes that explain why strong warming could lead to enhanced ice flow and calving.Figure 2 caption: centured -> centeredl. 310: " which would be out of the scope of this present study". Still it would be interesting to provide a first order-of-magnitude comparison with RCM produced LIUS SMB.l. 332: " XX references"l.385: " by an error in one of these two estimates". Or in both.Citation: https://doi.org/
10.5194/egusphere-2024-1998-RC1 -
RC2: 'Review of "Investigating the surface mass balance of the Laurentide Ice Sheet during the last deglacation"', Anonymous Referee #2, 11 Sep 2024
Triggered by the somewhat provokative abstract, I've read the manuscript by Koepnick et al.. To my regret, I adviced the editor that this research does not meet the standards to be publishable in Climate of the Past. Below I'll motivate my advice.
The final conclusion of the authors is "that General Circulation Models (GCMs) still struggle to reliably calculate the SMB", and this is the correct conclusion for the results presented in the manuscript, but not for ESMs (GCMs is an outdated term) in general. I'm sorry to say that the authors have based their analysis on a model simulation that is not state of the art for modelling the SMB with an ESM. If the authors wanted to know, they could take a look at the recent research by Miren Vizcaíno, for example, to see how well an ESM can model the SMB of an ice sheet, or read the numerous papers by various research groups running coupled atmosphere-ocean-ice sheet models on their extensive efforts to get realistic SMBs within their coupled model environments. In the ITraCE simulation, Brady et al obviously did not. Otherwise they would have calculated the SMB on the fly, but they didn't.
So the conclusion of the manuscript is obvious, namely that if a model doesn't aim to model SMB, it won't get the SMB right. I don't see the need to publish this. It is common sense that if a model is not set up to model a particular property right, there is no chance that by some magic that property will be modelled correctly.
The authors give the impression that they have estimated the SMB from the output, and I am willing to believe that the authors think they have estimated the SMB to the best of their ability. The latter, if true, is rather worrying because the manuscript gives the impression that the authors do not know what is essential to model the SMB correctly. This impression arises from the pointless discussion on whether or not to use the geothermal heat flux as an estimate of the ground heat flux of a glaciated surface (the ground heat flux is dominated by other processes), the funny unit error in Figure 1 (if you're talking about kg, you don't need to specify that it's in water equivalents. You do if you're using volume or thickness), the fact that the authors present year-averaged, ice-sheet-averaged fluxes in Figure 2, and, lastly, the failure to realise that before you start playing around with small changes in albedo, you need to show that albedo makes sense at all. Given that there is negative SMB all the way up to the ice divide, I'd say it doesn't.
The authors also seem unaware of the complexity of modelling the SMB of an ice sheet IF you have a proper (online) estimate of meltwater runoff and hence SMB. Ablation zones are generally narrow and steep, especially for land-terminating ice edges (ok, the dying glaciers we have around the world contradict this statement, but yes, these glaciers are dying). When working with lower resolution input data, as is the case here, these ablation zones are usually missed, leading to an overestimated integrated SMB. So if you find an SMB that is still too low, then something is really wrong.
Finally, the authors' assumption that the SMB should at least not be more negative than the long-term mass loss is inadequate. I assume the authors are aware that the Greenland Ice Sheet, although losing mass, still has a positive SMB of about 30% of its accumulation input. Yes, the Laurentide Ice Sheet was land-terminating at its southern margin, but marine-terminating everywhere else. Yes, ICE-6G does not provide a mass budget for the Laurentide Ice Sheet, but I don't see what's complicated about investigating the existing modelling literature for the typical mass budget of ice sheet models representing this period, by personal communication if the papers don't reveal it. I'm sorry, but this is poor research practice IMHO.
So what can be done to improve this manuscript?
In iTraCE, the surface energy balance is not derived for glacier surfaces, or at least not correctly. So simply reconstructing the melt from the surface energy balance (Eq. 1) does not work, as the authors rightly conclude. I don't see the added value of improving the analysis of the results presented here, to conclude again that if a model doesn't try to model SMB, it doesn't model SMB correctly.
In my opinion, the authors need to start from scratch. If the authors want to continue to use the iTraCE simulations to estimate the SMB, they need to develop a model or method to estimate the SMB that acknowledges the shortcomings of the GCM data. There are several studies by palaeo-ice sheet modellers using GCM data that have done this before. However, the result of such research is a new paper, not a revised version of this paper.
However, if the editor decides that this manuscript should be given a second chance, I would suggest that
- The authors first discuss the (summer) surface energy balance (including summer albedo and near-surface temperature) and evaluate whether the modelled fluxes are realistic. Modern Greenland and Antarctica can provide some clues as to what might be expected, taking into account orographic and (Milankovisch-driven) isolation differences. In this evaluation, all energy fluxes (including LHF) are expressed in W/m2 and the 'flux convention' is used, i.e. fluxes are positive when directed towards the surface. [And remove the geothermal heat flux, don't embarrass yourself].
- Next, the modelled SMB is analysed.
- The authors make a realistic estimate of mass loss to the ocean as function of the time, so that a proper "observational" integrated SMB time series is used.
- The authors compare the modelled SMB patterns with SMB patterns used in other studies, generated by other methods.
- When the authors then try to correct the SMB by increasing the albedo, both the new summer albedo fields and the new SMB patterns are shown, at least for some key moments during the transient simulation. The authors should be aware that increasing the albedo decreases the surface temperature, which increases SHF, LHF and LWup (so, e.g. LWup becomes less negative). On instantaneous data, one can make some first order estimates of how large these feedbacks are to arrive at relatively correct updated melt estimates. However, doing this on monthly data is questionable - but ignoring these feedbacks would be even more questionable. So any result of such an analysis should be accompanied by a proper idea of the uncertainties involved.
- The authors remove Figure 2 and its subsequent analysis as it stands. You can look at papers such as https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020GL090653 to see how you can work out what processes are driving the melt. I know there is more than one way to partition the energy sources of the melt, but annual, ice-sheet averages are not among the methods that give meaningful results.
- The authors revise the wording of their conclusions, acknowledging that models that are not set up to model the SMB cannot be expected to model the SMB correctly. This research doesn't say anything about the performance of GCMs (well, use ESMs, not GCMs) that do try to model the SMB correctly.
Citation: https://doi.org/10.5194/egusphere-2024-1998-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 222 | 76 | 32 | 330 | 10 | 10 |
- HTML: 222
- PDF: 76
- XML: 32
- Total: 330
- BibTeX: 10
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1