the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Cloud- and ice-albedo feedbacks drive greater Greenland ice sheet sensitivity to warming in CMIP6 than in CMIP5
Stefan Hofer
Trude Storelvmo
Xavier Fettweis
Abstract. The Greenland Ice Sheet (GrIS) has been losing mass since the 1990s as a direct consequence of rising temperatures and has been projected to continue to lose mass at an accelerating pace throughout the 21st century, making it one of the largest contributors to future sea-level rise. The latest Climate Model Intercomparison Project 6th phase (CMIP6) models produce a greater Arctic amplification signal and therefore also a notably larger mass loss from the GrIS when compared to the older CMIP5 projections, despite similar forcing levels from greenhouse gas emissions. However, it is also argued that the strength of regional factors such as melt-albedo feedbacks and cloud-related feedbacks will partly impact future melt and sea-level rise contribution, but little is yet known about the role of these regional factors in differences in GrIS surface melt projections between CMIP6 to CMIP5. In this study, we use high-resolution (15 km) regional climate model simulations over the GrIS performed using the Modéle Atmosphériqe Régional (MAR) to physically downscale six CMIP5 RCP8.5 and five CMIP6 SSP5-8.5 extreme high-emission scenario simulations. Here, we show a greater annual mass loss from the GrIS at the end of the 21st century, but also for a given temperature increase over the GrIS, when comparing CMIP6 to CMIP5. We find a greater sensitivity of Greenland surface mass loss in CMIP6 centred around summer and autumn, yet the difference in mass loss is largest during autumn with a reduction of 14.1 ± 4.8 mmWE for a regional warming of +6.7 °C, 12.5 mmWE more mass loss than in CMIP5 RCP8.5 simulations for the same warming. Assessment of the surface energy budget and cloud-related feedbacks suggests a reduction in high clouds during summer and autumn – in addition to enhanced cloud optical depth during autumn – to be the main drivers of the additional energy reaching the surface, subsequently leading to enhanced surface melt and mass loss in CMIP6 compared to CMIP5. Our analysis highlights that Greenland is losing more mass in CMIP6 due to two factors; 1) a (known) greater sensitivity to greenhouse gas emissions and therefore warmer temperatures, 2) previously undocumented cloud-related surface energy budget changes that enhance the GrIS sensitivity to warming.
- Preprint
(25715 KB) -
Supplement
(57822 KB) - BibTeX
- EndNote
Idunn Aamnes Mostue et al.
Status: final response (author comments only)
-
RC1: 'Comment on tc-2023-24', Anonymous Referee #1, 23 Mar 2023
Mostue and co-authors explore summer and autumn differences in surface mass and energy budget (SMB and SEB) components as well as cloud cover based on output of a regional climate model forced by several global climate models from two generations under the high-end baseline scenario. They start by relating anomalies in near-surface temperature with individual SMB and SEB components, showing similar behavior and magnitude to SMB, melt, runoff, longwave down and longwave up. However, the set of models used from the most recent generation of global climate models projects more warming than previously. Finally, it is shown a contrasting relationship between the anomaly of near-surface temperature and cloud cover anomaly between the two generations of global climate models. The authors hypothesize that the cloud cover decrease allows more absorption of solar radiation by the surface, generating enhanced surface melt and runoff in the ablation zone in summer extending to autumn.
This piece of work explores relevant scientific aspects, but the methods and set of variables used are not sufficient to prove the robustness of the results. The level of detail provided in the Section 3.1 and 3.2 is commendable, but presents an unnecessary detailed picture between near-surface temperature with SMB and SEB components. The most relevant part starts with the lower panels in Figure 2, serving as motivation for the rest of the manuscript. Even though the authors show no relevant changes in cloud optical thickness, it would be worthwhile to explore changes in cloud microphysics and its relationship with cloud cover. In addition, summer and autumn precipitation should be included in the analysis, given its role to surface albedo.
My comments are provided by line number (LN) or specific figure below.
The Introduction is short and does not summarize/highlight what the scientific community has recently done concerning the impact of clouds and surface albedo on SEB and SMB components over the Greenland ice sheet.
LN22: the accelerating mass loss pace since the mid-1990s is not only a consequence of increased temperatures from anthropogenic greenhouse gases, but rather a consequence of a superimposed effect with extraordinary atmospheric conditions in recent summers (Bennartz et al. 2013, Fausto et al. 2012, Tedesco et al. 2011, Tedesco et al. 2016). Consider rephrasing.
LN28: the SMB definition should not be part of the Introduction, but in the Methods section, naming individual components and explaining how do you define accumulation and ablation zones.
LN31: in addition to solar radiation, consider the role of sensible heat flux to darken the surface (Wang et al. 2021)
LN41: the authors should address the fact that as a consequence of more open waters, CMIP6 projects more precipitation and more rainfall in Greenland than CMIP5 (McCrystall et al. 2021). This point can also be later discussed as a factor contributing to decreasing albedo, as also shown by Box et al. (2022).
LN50: state that a surplus in SEB is energy available for melt and not necessarily surface melt
LN64: the last paragraph of Section 2.1 could be moved to the Introduction, where a few of these references could better distilled
LN72: it would be relevant to explain here why only RCP8.5 and SS5-8.5 is chosen for the study, as Hofer et al. (2020) made use of all the projected scenarios
LN75: it is also unclear why the period 1961-1990 is chosen. I would assume the last 3 decades (1991-2020), responsible for the accelerated mass loss, a better period for comparison with future projections
LN82: it should be indicated how the ice cover mask (more than 10\% ice cover) can influence the following results
LN83: it is unclear why a twenty-averaged period for ~4ºC is chosen for the dissemination of certain the results
LN85 how can you gain insight of changes caused by rapid Greenland warming using a twenty-year averaged period?
LN115: could you present the same charts (Figure 1 and 2) but for the differences between CMIP5 and CMIP6, making use of statistical inference to state the robustness of the mentioned differences?
Figure 1: legends and axis labels missing. Also, consider making the season as a subtitle of the subplot as in Figure 2
LN139: start the sentence with "In SON" instead of "Here". Otherwise, it is not clear to which season this sentence belongs
LN147: in LN139 you explain that more SW$_{net}$ is due to darkening and here is due to SWD. Please, rephrase it.
Figure 2: legends missing and temperature unit incomplete
LN165: why do you assume that no differences in cloud optical depth means no differences in cloud microphysics? Isn't this statement contradicting Hofer et al. (2019)? Could you elaborate your thought?
LN181: The twenty-year averaged cloud cover anomaly is compiled by a wide variety of circulation patterns. Only high frequency of a certain circulation pattern would depict the topography influence on the cloud cover composite. Thus, there is no information enough to infer the likelihood of circulation-driven cloud cover change.
Figure 4, 5, 6 and 7: use statistical inference to indicate the level of confidence in changes between CMIP5 and CMIP6.
Figure 5 and 6: consider two different color maps to stress the fact that colors shading in summer is not comparable with autumn. Perhaps, relative changes (e.g., ratio) instead of absolute changes could be here considered
LN245: precipitation has so far been discarded of the analysis, but here it would be interesting to assess if precipitation, more specifically liquid precipitation, could play a role in the snow darkening and surface runoff
Technical corrections
LN9: spell the name of the regional climate model correctly
LN11: indicate the corresponding level of uncertainty
LN27: spell the surname of the main author correctly
LN32: spell the surname of the main author correctly
LN51: downwards instead of "down towards"
LN51: LWU is defined as LWD
LN55: introduce SWD at the beginning of the sentence
LN56: suggested place to define SMB instead of doing it in the Introduction
LN59: spell the surname of the main author correctly
LN71: spell the name of the regional climate model correctly
LN154: Figure 2 c and d, instead of "a and b"
LN156 Figure 2 c instead of "a"
LN179: total instead of "toal"
References
Bennartz, Ralf, et al. "July 2012 Greenland melt extent enhanced by low-level liquid clouds." Nature 496.7443 (2013): 83-86.Box, Jason E., et al. "Greenland Ice Sheet Rainfall, Heat and Albedo Feedback Impacts From the Mid‐August 2021 Atmospheric River." Geophysical Research Letters 49.11 (2022): e2021GL097356.
Fausto, Robert S., et al. "The implication of nonradiative energy fluxes dominating Greenland ice sheet exceptional ablation area surface melt in 2012." Geophysical Research Letters 43.6 (2016): 2649-2658.
Hofer, Stefan, et al. "Cloud microphysics and circulation anomalies control differences in future Greenland melt." Nature Climate Change 9.7 (2019): 523-528.
McCrystall, Michelle R., et al. "New climate models reveal faster and larger increases in Arctic precipitation than previously projected." Nature communications 12.1 (2021): 6765.
Tedesco, Marco, et al. "The role of albedo and accumulation in the 2010 melting record in Greenland." Environmental Research Letters 6.1 (2011): 014005.Tedesco, Marco, et al. "Arctic cut-off high drives the poleward shift of a new Greenland melting record." Nature Communications 7.1 (2016): 11723.
Wang, Wenshan, et al. "Greenland surface melt dominated by solar and sensible heating." Geophysical Research Letters 48.7 (2021): e2020GL090653.
Citation: https://doi.org/10.5194/tc-2023-24-RC1 -
RC2: 'Comment on tc-2023-24', Anonymous Referee #2, 17 May 2023
Review Cloud- and ice-albedo feedbacks drive greater Greenland ice sheet sensitivity to warming in CMIP6 than in CMIP5.
I've read the manuscript with interest, and the study is publishable after some questions are properly addressed. It analyses in detail the differences between projections for the Greenland Ice Sheet from CMIP5 and CMIP6 models. This has not been done before, and the study provides interesting new findings. However, addressing the issues below can strengthen the study and can take away potential concerns of readers.
Main issues:
1: The focus on the radiative terms only.
With some references, the authors justify why only the radiative terms of the surface energy balance (SEB) are discussed in this manuscript. I'm not a prior convinced that this is justified. The sensible heat flux (SHF) is a significant contributor to melt in the ablation zone (you can find many papers about that), and the ablation zone is where the differences are made. I, therefore, ask the following:
- a) The authors repeat the analysis as shown in figure 2a and 2b and figure 5 for SHF, LHF (latent heat flux) and GHF (ground heat flux - the residual most likely). If these terms are indeed insignificant, as the authors argue now, these figures may be added to the SOM and single references like "we've studied these terms too and they are insignificant" will do in the main text. However, if SHF/LHF/GHF changes are not negligible, their discussion needs to be included in the text.
It might also add more clarity to the cloud arguments, as SHF, in contrast to LWD, is not influenced by changes in cloud cover. Conversely SHF is also influenced by surface warming too, I'm not sure a prior if changes in SHF are clearer than the LWD change. Well, the authors have to find out.
- b) Consequently Equation 2, the SEB, should be adjusted to
ME = LWD - LWU + SWD(1 - \alpha) + SHF + LHF + GHF [W /m2]
Please remove the \epsilon \sigma T^4 term from the equation, as this equation is never used in the manuscript (and please remove the crosses as that denotes the outer product of matrices), this relation can be mentioned in the running text; and please correct the units. Finally, the equation as a whole is the SEB, not the right-hand side.
2: The role of melt water buffering by firn:
In brief, the manuscript now states this: The CIMP6-CMIP5 change in SMB for a given warming (fig. 1a) is occurring in fall (fig. 1c), while the fall SEB is virtually unchanged (fig 2b, 5-right row). However, the CMIP6 fall albedo of the ablation zone is lower, leading to more melt which, in absence of firn, runs off directly (fig 6 - right column). Correct, however, by zooming in on the fall, it ignores the relevance of the melt-refreezing-runoff pattern in summer. Furthermore, I'm highly puzzled by the lack of refreezing increase during any phase of the warming. This runoff increase is the overarching feature of all RACMO simulations and very visible (as far as I know) in MAR simulations driven by reanalyzes. Therefore, I ask the following.
- a) A description of the exact run protocol of MAR needs to be added in the methods section. Furthermore, (as understanding the firn response is important in this study), expand the description of the firn model with details relevant for this study. For example, how thick can the firn column become in MAR? Hence, the authors need to be able to address to which extent coding choices have impacted the modelled melt water buffering capacity in a transient climate?
The authors now state in lines 248-254 that the faster warming in CMIP6 induces that the percolation has more remaining melt water buffering capacity when the simulation crosses the 4 K warming point. I'm not sure if that is correct, as many MAR projections available nowadays are not run in single linear mode, thus one single MAR realization that started in 1950 and ended in 2100. Those many MAR projections that are in the community are run with a "each year initialized separate"-protocol; thus, that the weather of that year is repeated until the MAR firn column, and hence the SMB and its terms, have become in equilibrium.
- b) If the latter protocol has been used, the 'faster warming' argument is invalid. Still, a detailed explanation of figure 6, left column, needs to be given. If the first protocol has been used, it is worth to show the difference in firn air content (or another firn state metric) to highlight that this different firn state strongly contributes to the pattern visible in the left column of figure 6. Furthermore, and that is really important IMHO, the conclusions should then be: CMIP6 melt more due to stronger warming and, our new point, decreasing (high) cloud cover, both only partly mitigated by more remaining melt water buffering capacity due to the faster pace of warming.
- d) The authors should discuss in more detail why not part of the melt increase is buffered by runoff (figure 1), contrary to findings in preceding studies (like Noël 2021, doi: 10.1029/2020GL090471). From Figure S7, bottom row, I would expect a clearer visible increase in refreezing. Or is the increase in refreezing matched by the increase in rain? If so, please consider adding rain and runoff in Figure 1. Furthermore, is runoff indeed zero in the interior of Greenland? From Figure S7 it is not 100% clear.
- d) The discussion of the very different impact of clouds changes over the ablation zone, compared to the percolation/accumulation zone, could be much stronger and clearer if in the SEB analysis (figure 2) the ablation zone and accumulation zone are separated. Most likely a static separation mask is much easier than a transient mask. A good mask is to take the ablation zone outline for 4 K warming, as that is the "warming" point in time that is analyzed most in the paper.
3: Don't leave uncertain things you could verify in the model data:
On numerous points (about 10-20, I lost count) the authors are unsure as they use "we argue" (line 181), "we expect" (line 124), "can possibly be explained" (line 248), "we suspect" (line 267) or "we suggest" (line 270), "likely due" (line 277) while the answer can be found in the model data the authors should have. Go and check your ideas in the model data and write with certainty when it is true and remove the statement if it is untrue. I don't see a valid reason to be unsure. When some of those "unsure statements" cannot be verified and are retained in the manuscript, please address the reasons in the reply to this review.
4: Demonstrate that these results are not coincidental by the CMIP5 & CMIP6 model selection but a genuine difference between CMIP5 and CMIP6:
- a) Hofer et al 2020 gives the arguments for the model selection. This information needs to be summarized in this paper as it should not be necessary to read Hofer et al 2020 to understand this model selection.
- b) In the discussion, it needs to be showed (as good as possible) that these 5 & 6 models are representative for change in the modelled cloud climatology over Greenland. It (representativeness for cloud cover changes) is not mentioned in Hofer et al, 2020. I know this can be a lot of work (as modelled cloud cover over Greenland from ~30 models needs to be compared), so I can understand if the authors use existing studies to demonstrate this - if these are available. Nonetheless, the authors make implicitly this generalization, however, it should be justified.
Minor general point on units:
- a) I would prefer that K (Kelvin) is used instead of degrees Celsius.
- b) Albedo is unitless, and percentual changes of albedo are meaningless. Please correct this in Figures 7, S9 and S10, and section 3.4. Especially the numbers in lines 274-275 are wrong wrt units.
- c) Use hPa and not Pa in lines 184-185.
- d) Equation 1 has units Gt yr, but no result is shown with that unit. mmWE yr-1 or mmWE season-1 is used everywhere. So why not for Equation 1 too? At the other hand, Gt yr-1 is a unit easier to interpret for a larger audience, so it is worth to consider to use this unit more often in the manuscript - like in the running text.
Minor comments:
13: Please remove these two numbers (4.8 & 12.5 mmWE) as they make no sense for anyone without more context. Alternatively, use Gt yr-1 here. But in case of the latter, Gt yr-1 should be used more often in the manuscript - as already stated above.
19: "Undocumented" Please rephrase by e.g. "unnotified". Those changes have been documented before - as they were in the output files of many simulations - but nobody has written a paper about it before.
24: Add Noël and van Kampenhout, 2021 (cited above) here.
32: "Broek" is "Broeke". Check also other the references on typos.
51: The second LWD must be LWU.
77: "In turn" is IMHO not the right connection word here.
83: Add a bit more detail - is this a 4 K a (near) surface (T0m, T2m) of, e.g., 500 hPa, warming? Is the year chosen for the GrIS as a whole or for each grid point separately? And what does it imply for the "consistency of the results" as the JJA and SON 4 K warming frames does not refer to the same years and thus firn state?
103: The exact locations of the model edges, expressed in lat & lon, give a reader little clue. Add a figure of the domain extend (in the SOM), or write it down in words, e.g. "extends till Svalbard", "extends xxx km" to all sides of Greenland", ...
106: Please add the typical elevations of the lowermost model layers.
113: "near surface temperature anomaly [C (or K)]"
118: Consider not to start a new paragraph here.
Figures 1, 2, 3: Labels have turned into white in the official pdf. Ensure this is ok in the revised manuscript. Furthermore, if you consider a SMB anomaly for a season, the unit is mmWE season-1, not mmWE. I'm not aware of a nice abbreviation for season (like yr and s), so possibly the authors may conceive one.
171: Please state more clearly at the end of the paragraph that despite the changes in Figs 2d, 3b, 4-"right", there is no discernable change in Fig 2b that explains Figure 1c.
241: "we do not see the same buffering effect". Please mention that this is due to the absence of significant melt in fall in the percolation zone. There is no water to buffer.
Figure 8: I like the idea of an explaining figure. This one, however, is very unclear, thus fails to meet its aim. It should show that that in summer, the LW effect dominates over the SW effect in the percolation and accumulation zone due to the high albedo of snow, while SW dominates in the ablation zone due to the lower albedo. Improve the figure or remove it.
Citation: https://doi.org/10.5194/tc-2023-24-RC2
Idunn Aamnes Mostue et al.
Idunn Aamnes Mostue et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 276 | 98 | 10 | 384 | 35 | 5 | 4 |
- HTML: 276
- PDF: 98
- XML: 10
- Total: 384
- Supplement: 35
- BibTeX: 5
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1