the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Late Pleistocene glacial terminations accelerated by proglacial lakes
Abstract. During the glacial cycles of the past 800 thousand years, Eurasia and North America were periodically covered by large ice sheets. While the Late Pleistocene glacial cycles typically lasted 80 – 120 thousand years, the termination phases only took 10 thousand years to complete. During these glacial terminations, the North American and Eurasian ice sheets retreated which created large proglacial lakes in front of the ice sheet margin. Proglacial lakes accelerate the deglaciation as they can facilitate ice shelves in the southern margins of the North American and the Eurasian ice sheets. Ice shelves are characterized by basal melting, low surface elevations and negligible friction at the base. Here we quantify the effect of proglacial lakes, and the combined effect with glacial isostatic adjustment (GIA) on Late Pleistocene glacial terminations.
We find that proglacial lakes accelerate the deglaciation of the ice sheets mainly because of the absence of basal friction underneath ice shelves. If the friction underneath grounded ice is applied to floating ice, we find that full deglaciation is postponed by a few millennia, the Barents-Kara Sea region does not fully deglaciate, and there are no extensive ice shelves. Additionally, the large uncertainty in melt rates underneath lacustrine ice shelves translates to an uncertainty in the timing of the termination of only a few centuries at most.
Proglacial lakes are created by the depression in the landscape that linger after the ice sheet has retreated. The depth, size and timing of proglacial lakes depend on the bedrock rebound. We find that if the bedrock rebounds within a few centuries, instead of a few millennia, the mass loss rate of the ice sheet is substantially reduced. This is because fast bedrock rebound prevents the formation of extensive proglacial lakes. Additionally, a decrease in thickness is partly compensated by the faster bedrock rebound, resulting in a higher surface elevation with lower temperatures and higher surface mass balance delaying deglaciation. We find that a very long bedrock relaxation time does not affect terminations substantially, but will lead to a later inception of the next glacial period. This is because initial inception regions, such as North-Western Canada, remain below sea level throughout the preceding interglacial period.
- Preprint
(4310 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on cp-2023-42', Niall Gandy, 08 Aug 2023
Review of “Late Pleistocene glacial terminations accelerated by proglacial lakes” by Scherrenberg et al.
Signed: Niall Gandy
This manuscript presents ice sheet simulations of the past 800 ka BP, using various idealised modelling approaches to better understand the affect of proglacial lakes during deglaciation phases. The authors show that the presence of proglacial lakes accelerate deglaciation, and they attribute this primarily to the negligible basal friction beneath ice shelves. Further simulations also demonstrate the sensitivity of the ice response to the rate of bedrock rebound.
Overall, I think the work presented here will be a significant contribution to understanding palaeo ice sheet behaviour, with the possibility of encouraging further model development and motivating future experimental design to include lake feedbacks. Indeed, this is a mechanism largely ignored in future simulations, and important progress made here could link with future projections. The work is clear and concise. The precision and clarity of the figures are excellent, with the exception of some minor points made below.
The manuscript would benefit from a clearer justification for the experimental design (and perhaps some more analysis), along with some minor points listed below. As such, I recommend publication after minor revisions.
Main points
Experimental design: I appreciate the novel approach of the experimental design, where you have made certain adjustments to better explore the ice sheet behaviour. However, the detailed justification for this is not explained sufficiently in the text, and I am left not fully convinced that your experiments explore the behaviour you intend. For example, the “Rough Water” experiment is designed to show the effect of negligible friction beneath ice shelves, but various feedbacks (surface profile, buttressing effects, a different GIA response to more grounded ice, ect) could be confusing the results. The experimental ethos could be explained in more detail. You could also undertake some offline ice shelf mass budgeting (at each timestep what is the flow over the grounding line, what is lost to surface melt, sub shelf melt, and calving) to disentangle the behaviour.
Specific points
18: linger > remain?
67: I think it is common to conflate susceptibility to MISI/PLISI and sub-shelf melting/calving. Could you clarify the mechanistic difference for the reader before this sentence?
80: Could you comment on the suitability of a Hybrid model to simulate PLISI and grounding line migration? How is this parameterized?
89: In some ways the lacustrine environment might be quite different from the marine environment; the thermal structure may be different, and lakes could become chocked with icebergs. From a practical modelling perspective it is reasonable that you treat marine and lacustrine the same, but you could discuss this further in the text.
93: Do you know (or could you know with some offline calculations) what proportion of the ice sheet margin is missing lakes because the model cannot simulate above sea level lakes?
101: merge > merged
108: Not unfeasible! Millennial scale coupled climate-ice sheet simulation studies do exist, but it is understandable why this is not a reasonable modelling choice here. Please clarify.
126: The use of brackets to describe the reverse behaviour is a tad tricky to follow.
164: A set 30% adjustment assumes that distal ice sheets fluctuate in unison with simulated ice sheets. Is this reasonable?
Figure 4: This figure is challenging to follow, particularly panels b, d, and f. Are the points of deglaciation onset numerically defined? The onset of glaciation curves are difficult to read; I would suggest either removing or replotting. Ideally we shouldn’t need a paragraph (lines 172-176) just to describe how to read a figure, not yet describing the results or discussion the implications.
190: Can you comment on the mechanism for the higher sensitivity of the Eurasian ice sheet?
211: The importance of the simulated ice shelves may depend on their spatial extent and if they are constrained laterally. It would be good to see a figure of simulated ice sheet location and morphometry.
216: I don’t follow the logic here?
Discussions: This section is very limited, it could be incorporated into the Results section.
315: It would be good to develop this point a little further. What are the potential effects of lakes on future Greenland? And do models represent this?
423: It would be preferable for the simulations to be reproducible without contacting the author.
Citation: https://doi.org/10.5194/cp-2023-42-RC1 - AC3: 'Reply on RC1', Meike D.W. Scherrenberg, 29 Nov 2023
-
RC2: 'Comment on cp-2023-42', Fuyuki Saito, 18 Oct 2023
This paper presents a series of numerical simulation of ice sheets over the northern hemisphere for the past 800 thousand years, forced by climate forcing based on GCM. The main focus of the present paper is the effect of proglacial lakes on glacial terminations, and the authors show that proglacial lakes accelerate the deglaciation of the ice sheets due to the absence of basal friction underneath ice shelves. Moreover, they show that the time scale of bedrock isostatic response to the variation of ice sheets is a key factor for deglaciation, through the formation of proglacial lakes. I think this paper is fairly well written with some exception below, and can be accepted with minor revision.
Description of the ice-sheet model: Section 2.1 describes the ice-sheet model briefly, but I feel it too short. A complete repeat of the previous papers are not necessary, but at least the description of the methods relating to the sensitivity experiments of the present paper should be included in detail.
(i) how to deal the grounding line migration in the model. it is all right without any special treatment, but mention explicitly.
(ii) how to deal the formation of proglacial lakes. I suppose that the depth of lake is the difference between the background sea level and the bedrock.
(iii) explicit equation of basal sliding, and the difference in how to apply between base-line and rough-water experiments.
Other points:
Abstract: There are no description of the method used in the present paper, even no explanation that the discussion is based on a series of numerical simulation. This must be clarified.
BMB experiments: Typical mass balance terms, both surface and base, over ice shelves in the baseline experiment should be mentioned, which will help to get typical ratio of surface/basal mass balance terms.
Figure 2. Plotting two points corresponds to PI and LGM may help.
Figure 3. Better to mention in the caption that this is the result of baseline experiment.
Figure 4 and main text. Please note the definition of the onsets of (de)glaciation in this paper. Why some points (e.g., 240ka in the total and Eurasia, 290ka in North America) are not detected as the onsets?
Appendix Eq.(1) Please clarify that f_{snow} is no more than 1. I calculated by hand, and observed that f_snow become 1 when T - T0 is around -10K, and exceeds 1 when -11K. I am not sure my rough computation is correct, but please clarify the actual computation of f_{snow} (Of course it is all right if T - T0 is always more than -10K).
Eq (3). Units are not consistent. Need unit of the coefficient 0.015 before Melt_{prev}. Again, it is hard to check whether the albedo is between 0 and 1 by this equation.
Eq (4). Need units of the coefficient 0.012.
Eq (21). Need unit(scale) of P_{PI,corr} to compute log. Or, is the first term e means exp() function, not e times ()?
Citation: https://doi.org/10.5194/cp-2023-42-RC2 - AC1: 'Reply on RC2', Meike D.W. Scherrenberg, 29 Nov 2023
-
EC1: 'Additional review comments from Editor on cp-2023-42', Lev Tarasov, 24 Oct 2023
As you draft your response to reviewers, I have identified 4 other
major issues that need to be addressed.1) Experimental design. The current design leaves too many
uncertainties unaddressed in part stemming from model limitations
somewhat buried or not even described in your much too brief model
description (as already raised by one of your reviewers).1a) As far as I can tell, you do not use a soft/hard bed mask
for Eurasia, even though it has long been known that this can have a
large impact on ice sheet response and geometry (eg cf Tarasov and
Peltier, 2004). So a significant fraction of your sensitivity to
basal drag changes from pro-glacial lakes could be due to this easy to
address design limitation as regions that are soft-bedded will generally
have less basal drag (when warm-based).1b) I suspect you have no basal hydrology (again model description is
way to brief on relevant aspects of the model configuration) and
I also suspect you are simply taking the adjacent water depth to compute
effective pressure for the Budd sliding law. On a 40km grid, that choice
is hard to defend. At the very least, try an easy to implement leaky
bucket approach ( eg as in PISM or https://tc.copernicus.org/preprints/tc-2022-226/)
for comparison.1c) The assumption that lake depth is given by bed depth below
contemporaneous sealevel is another major limitation, eg long known in
the geological community (or cf Tarasov and Peltier 2006 from a
modelling perspective) from the critical dependence on controlling
sill elevation for eg Laurentide Lake Agassiz. Surface drainage
solvers have been in some published models for almost two decades.
I can understand adding one is non-trivial, but at the very least
more thought needs to be put into addressing the some of the uncertainties
arising from this model limitation.2) Scientific precision/accuracy/transparency
Eg:
"We find that the modelled sea level matches the reconstructions well"
Your model fails to capture any signal of last glacial inception, so
blanket use of "matches the reconstructions well" is inaccurate.No where do you show your LGM ice extents (only 11 ka), instead your
only LGM ice extent map shows a cited map from Abe-Ouchi et al (2015).
This may suggest to the reader that you do not want them to see what
our LGM ice extent looks like.or
"our simulations tend to have slightly too long interglacial periods
compared to reconstructions" From what I can read off of your plot in
figure 3, your last interglacial is @ 50kyr too long, that is not
"slightly".3) Adequate referencing of past relevant litterature, eg role
of GIA in deglaciation of ice sheets (eg Tarasov and Peltier,
Ann. Glac., vol. 25, 58-65, 1997), impact of pro-glacial lakes on
ice extent (eg Cutler et al, 2001, Geology), or relative sensitivity
of Eurasian versus North American ice sheets to orbital/CO2 forcing
(eg Tarasov and Peltier, JGR 1997).4) claims in conclusions that have not been shown, eg:
"3.4 Glacial isostatic adjustment ... The North American ice sheet may
not even fully deglaciate during some interglacial periods. This is
because proglacial lakes are not created when the bedrock uplift is
too fast". This is not shown in your results (which would require a
further sensitivity experiment with lakes turned off for soft/hard GIA
experiments).Furthermore, ice volume is a very limited metric.
Lake calving will more directly affect ice extent than ice volume
(though the two are obviously related). Without relevant map-plots (at
least in a supplement), the reader is unable to adequately evaluate
the extent to which you results match your claims.Citation: https://doi.org/10.5194/cp-2023-42-EC1 - AC2: 'Reply on EC1', Meike D.W. Scherrenberg, 29 Nov 2023
-
EC2: 'Comment on cp-2023-42 responses to reviews', Lev Tarasov, 14 Dec 2023
While revising your submission and responses, please ensure the following
points from you initial responses to the reviews are addressed (your responses in bold).While technically possible, calculating the missing lakes is not trivial and
computationally heavy.#
Compared to the ice dynamics, the surface drainage and lake solution can
be quite cheap (in the GSM (glacial systems model), it's less than 10% of
overall GSM compute load).3)We have conducted a few additional sensitivity experiments. We have
conducted the Baseline simulation with a respective 50% increase and
decrease in the till friction angle#
Ensure you justify your chosen sensitivity ranges, so that the reader
can judge whether you've covered maximum plausible bounds, eg as a case in point:To address these issues, we propose an alternative experiment to quantify the effect
of this simplification. We select a large region in North America (south of the Hudson
Bay to southern margin of the domain) where the water level of potential proglacial
lakes is set to 50m above present-day sea level.#
Just choosing 50m is running blind. Refer to previous studies that explicitly
model surface drainage to come up with appropriate upper and lower bounds
for lake surface elevation relative sealevel.
>89: In some ways the lacustrine environment might be quite different from the marine environment;...We will mention these differences in the introduction and discussion sections.
#
Just mentioning differences might not be enough. Experimental uncertainties
that are easy to address (eg reducing calving and submarine melt coefficients
to reasonable lower bounds) are best done or at least arguments to need be
provided as to why they are not relevant.We will change this in manuscript to explain that it is technically
possible to run coupled climate-ice simulations, but at a large computational cost.#
Depends on which climate model is used. For EMICs, this is more accurately
moderately large cost.Our ice sheet extent at LGM matches reconstructions reasonably well, compared to
other freely-evolving paleo-ice-sheet models (e.g., Berends et al., 2018. Willeit et al.,
2019).#
Neither of the above make use of available geological constraints
for LGM extent. For North American and Eurasia ice extent,
geological reconstructions are more appropriate. Current most
appropriate for comparison would therefore be Dalton, et al,
QSR 2020. I can provide latlong rasterized isochrones if you can't
deal with the shapefiles attached to the above (lev@mun.ca).Citation: https://doi.org/10.5194/cp-2023-42-EC2
Data sets
IMAU-ICE output M. D. W. Scherrenberg, C. J. Berends, R. S. W. van de Wal https://surfdrive.surf.nl/files/index.php/s/pxv9RF6pHwCzJbf
Model code and software
IMAU-ICE model code and Matlab scripts for figures M. D. W. Scherrenberg, C. J. Berends, R. S. W. van de Wal https://surfdrive.surf.nl/files/index.php/s/pxv9RF6pHwCzJbf
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
447 | 177 | 27 | 651 | 31 | 21 |
- HTML: 447
- PDF: 177
- XML: 27
- Total: 651
- BibTeX: 31
- EndNote: 21
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1