the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Brief Communication: annual large-scale atmospheric circulation reconstructed from a data assimilation framework cannot explain local East Antarctic ice rises’ surface mass balance records
Abstract. Ice cores are influenced by local processes that alter surface mass balance (SMB) records. To evaluate if large-scale atmospheric circulation explains contrasted SMB trends at eight East Antarctic ice rises, we assimilated ice core SMB records within a high-resolution downscaled atmospheric model, while incorporating radar-derived SMB constraints to quantify local observation errors. The reconstruction captures the diverse variability from SMB records but may over-fit by introducing unrealistic wind spatial heterogeneity. While local errors are quantified, they might not cover all uncertainties. Moreover, small-scale wind circulation, unresolved in the reconstruction, could significantly affect local ice core SMB signals.
- Preprint
(5596 KB) - Metadata XML
-
Supplement
(601 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-3140', Anonymous Referee #1, 17 Dec 2024
Review
Brief Communication: annual large-scale atmospheric circulation reconstructed from a data assimilation framework cannot explain local East Antarctic ice rises’ surface mass balance records
By Cavitte, Goosse, Dalaiden, GhilainThis short manuscript describes the suitability of simulating the near surface wind pattern over the Dronning Maud land region from downscaled fields from an atmospheric model, in which ice core and ice penetrating radar results are assimilated.
The manuscript is reasonably well written with clear figures. The topic is appropriate for The Cryosphere and presents interesting results.
I do have some comments that I feel need to be addressed before publication.
My main concern is the clarity of the manuscript, where especially the method is not clearly described. As far as I understand, downscaled fields from an ensemble of models is taken. Then these are modified by assimilating the ice cores including information from the ice penetrating radar. This then results in time series of SMB and 10-m wind fields. I might be wrong, since it is not fully clear what is actually done, in what order, and what results from it. Also how the 10-m wind is related to or constrained by the SMB field is not clear to me, especially how the 10-m wind field is changed after assimilation of ice cores in the SMB field.My second main point is the use of the term 'large-scale atmospheric circulation' when the 10-m wind field is described. In atmospheric terms, the large scale atmospheric circulation usually refers to the circulation that is not, or only limitedly, influenced by the surface. Usually the circulation at 500 hPa or 300 hPa is taken to represent it, not the 10-m wind since that is clearly influenced by interaction with the surface. Given the presented results you actually mean the near surface wind field on a larger spatial scale than purely local. Your title and manuscript should more clearly reflect what is actually presented.
In relation to this, I am also curious why the SMB patters are related to changes in the flow pattern instead of the flow pattern themselves. Most logical steps would be to first try to explain the SMB pattern with the flow pattern, followed by SMB changes with changes in the flow pattern. Both are now a little bit mixed in this manuscript, with the title being different from the described research goals.Other comments:
L9-10: rephrase this sentence: the processes that make up the surface mass balance are (reasonably) well understood, as are the processes that result in the spatial distribution. The actual absolute amounts are not well constrained, resulting in uncertainties in the details of the spatial patterns.
L17: add 'limited' before 'grid resolution'
L20-21: rephrase this sentence: especially seasonal resolution is limited. The actual resolution very much depends on the accumulation rate. In L187 you also state that ice cores are rarely sub-annually resolved.
L24-25: rephrase sentence: it now reads as if the observations underestimate the SMB at the observation site. Since the ice cores are actual observations, they cannot underestimate the mass balance at that site. It might result in an underestimation of the SMB of a larger region, but that is not what I read in this sentence.
L27: replace 'offers thus' with 'thus offers'
L33: rephrase sentence: not sure what you mean by 'a same area as grid points in a model'.
L41: clarify: 'recent compilation' of what are you referring to?
L42-43: Suggest to replace 'These eight ice core SMB records have an annual
resolution with differing trends in close proximity (∼100-300 km between ice rises, Fig. 1a).' by 'These eight ice core SMB records are in close proximity (∼100-300 km between ice rises, Fig. 1a) of each other, and have an annual resolution with differing trends.'
L44-51: Clarify the method described here.
L44: Explain on first usage of this term what you mean with 'model prior'. My guess is that it is similar to or basically a 'first guess' field.
L46: I guess you mean that you downscale to the RACMO2.3 grid, not the resolution, which can be two different things.
L49-51: please rephrase sentence: it is not clear to me what you wish to say with this sentence.
L55: what model do you mean with 'the model'
L58-59: Due to my lack of knowledge of DA methods, this sentence is very unclear. What are or do the particles represent? Are they SMB fields for each year over the 165 year period? Times 10 then results in 1650. And what is the role of the prior in this?
L88: replace 'can thus provide' with 'thus can provide'
L91: rephrase sentence: better to state that melt in this region is limited and therefore the SMB is dominated by the snowfall and can be represented by it.
L105: replace 'a majority' with 'the majority'
L103-1012: Do I conclude correctly that although RACMO for individual ice core sites does not correlate with the original ice cores or the reconstruction, over DML in general, there are regions where the temporal variations in the reconstructed SMB does correlate significantly with RACMO SMB?
L112: What heterogeneity do you refer to?
L113: remove 'must'
L113-114: Why would the 10m wind field reconstruction show physical validity of the reconstruction? The 10m wind field is, if I understand correctly, not independent. Furthermore, you do not look at the wind field pattern itself, but at changes in the pattern. See my comment above. I sugest to first look at the pattern itself, before looking at changes in the patterns.
L116: Why is comparing changes a validity test? It only works if you know what the pattern should look like, and you do not know that.
L124: replace: 'Figure 3 shows the difference in wind strength between the youngest and oldest 10-year intervals of the reconstruction, so 2002-2011 versus 1987-1996.' with 'Figure 3 shows the difference in wind strength between the 2002-2011 and the 1987-1996 10-year intervals of the reconstruction.'
L130: rephrase sentence: how do you know it is not realistic? It is different from RACMO. And rephrase 'large-scale', see my comment above.
L143: replace 'that' with 'than'
L180: check sentence, the word 'from' does not make sence to me.
Figure 1:
Caption: add the source of the 10-m wind field.Figure 2:
Please check how you refer to RACMO2.3p
In the legend of panel a. and in panel b. it states RACMO5. Either use R5, as stated in the caption, or RACMO2.3p.
Also check line 4 of the caption where le should be replaced by the equal to / smaller than sign.
Perhaps add in the description of panel b, that you present the temporal correlation.Figure 3:
Change caption in: Difference in (a-b) SMB and (c-d) wind circulation, average over 2002-2011 minus average over 1987-1996 10-year intervals of the reconstruction.Citation: https://doi.org/10.5194/egusphere-2024-3140-RC1 -
RC2: 'review#2', Anonymous Referee #2, 22 Jan 2025
In this short communication, the authors use height ice core surface mass balance (SMB) records from a previous study, annually resolved over the period 1987-2011, from ice rises located in coastal Dronning Maud Land, East Antarctica. They use a downscaling method that associate circulation patterns with surface fields (snowfall and surface wind) from the regional atmospheric model RACMO at 5.5km resolution. They apply this downscaling to the 10 ensemble members of the CESM2 Earth System Model. They use all years of this downscaled ensemble as prior for a data assimilation method, as well as a radar-based estimation of the representativeness error.I think this article is interesting because the authors are honest with their results and with the limitations of the reconstructions. I would like to point out that it's not often that authors publish negative results, and I think this type of study is very useful for the scientific community.
However, I think the article would benefit from clarifying its objectives.
Major comments
In the introduction, the authors state that "To reconstruct SMB beyond direct instrumental measurements, which only cover the last decade or two (Wang et al., 2021), ice cores are the main in-situ observations (Lenaerts et al., 2019)." This sentence suggests that the article is about SMB reconstruction over a longer time period than the last decade or two. Then the time period of the ice rise records seems to be 1987-2011, which is covered by direct instrumental measurements.
Then, the authors show that ice rises SMB records are not correlated to RACMO2 SMB, which means that either :
- the ice rises SMB variability is not representative of a large-scale signal, or
- RACMO2(ERA-Interim) SMB variability is wrong.
If I understood well, the authors have explored option (2), with the hypothesis that the link between large-scale circulation and SMB in RACMO2 is correct, but that it may be the large-scale circulation forcing RACMO2 that was wrong. Then they use the relationship between SMB and large-scale circulation to test whether they can obtain a SMB map more consistent with the ice rises SMB variability.
Can you tell me if I understood well or not? In any case, I think the article would need a clarification of the hypotheses that are being tested.
I think the authors should also give there hypotheses on why RACMO2 SMB variability might be wrong. Are they suggesting that the ERA-Interim forcing reanalysis might have a biased atmospheric circulation?
Conversely, they use the relationship between large scale circulation and RACMO2 SMB to downscale a large ensemble CESM2, which means that they trust the downscaling but not the ERA-Interim large scale circulation. It would be good to clarify if this is based on previous studies or it is a working hypothesis.
Finally, I was troubled by the emphasis on "test(ing) if using the representativeness error derived from radar data improves the reconstruction of SMB" which at the end did not seem to be a center point of the article.
Minor comments
Abstract
"To evaluate if large-scale atmospheric circulation explains contrasted SMB trends at eight East Antarctic ice rises"
- -> "contrasted SMB trends" with regard to what?
1 Introduction
"Regional climate models with a high spatial resolution (∼a few km2), such as the polar-oriented Regional Atmospheric Climate Model version 2.3 (RACMO) (Van Wessem et al., 2018), struggle to capture the mean SMB state in the ice sheet interior, while they have a reasonably good fit with coastal mean SMB (e.g. Agosta et al., 2019)"
- -> I don't see this information in Agosta et al., 2019, can you clarify where it comes from?
- "Fig. 1" Nice figure.
- "Fig. 1: Background is the RAMP RADARSAT mosaic (Jezek et al., 2013) with superimposed the Reference Elevation Model of Antarctica (REMA) v2 elevation for each site (units are meters and referenced to the WGS84 ellipsoid)."
- -> the colorbar is missing for REMA elevation maps , and I recommend to use a continuous colormap instead of the rainbow colormap.
- "Fig. 1: (b) Mean 10-m winds over Dronning Maud Land for 1987-2011."
- -> from which dataset?
2. Method
2.1 Data assimilation
"Particle weights are assigned based on the agreement between model results and observations, with closer matches receiving higher weights. "
- -> provide the exact metric : extract the model at the nearest grid cell? And RMSE accross model-obs?
- -> Then can you provide the exact formula of weighted mean using this weigth?
- "Our observation error is the root-mean-square combination of the instrumental error and the representativeness error (Supplementary Table S1)."
- -> I don't see in the short method description where the observation error is taken into account
2.2 SMB observations
- -> add more details abount SMB observations, notably the time period, time resolution and associated uncertainties with regard e.g. on ice core dating.
2.3 Downscaled model ensemble
"This downscaled product has an improved spatio-temporal distribution of snowfall compared to global climate models, such as the CESM2 simulations on which it is based, and can thus provide information on large-scale atmospheric circulation patterns associated with local snowfall changes."
- -> Is the improvement quantified in this study or in a previous study? In both case, can you give the reference to this quantification?
3 Results
- "We highlight that the number of particles retained by the filter is quite low (about 7% on average). Retaining too few particles implies that the reconstruction may be based on too few samples and the estimate may become less accurate (filter degeneracy)."
- -> The final number of particles renaited should be added in the Method section.
- Figure 3:- -> I suggest that you add the difference in SMB field between the reconstruction and RACMO for the full period, and the same for winds, to assess the difference between the reconstruction and RACMO2 SMB (on top of the difference in changes in SMB)
"Because the ice core records are rarely sub-annually resolved, due to the nature of the measurements (if we want to go back further than 10 years), we would need to turn to other types of SMB records such as automatic weather stations. The issue in that case is that such records are currently not long enough to retrieve meaningful SMB trends"
- -> Many automatic weather stations are already assimilated in reanalyses, though not for SMB. Can you clarify what you want to recontruct or constrains with ice cores? Do you want to reconstruct circulation patterns beyond the reanalyses era? Do you want to use SMB to better constrain reanalyses?
"ML and AI emulators"
- -> replace abbrevations by full words
" downscaled annual wind simulations are available at xxxx."
- -> typo
Citation: https://doi.org/10.5194/egusphere-2024-3140-RC2
Data sets
Gridded surface mass balance derived from shallow radar stratigraphy over eight ice rises along the Dronning Maud Land coast and one site in the Dome Fuji region, Antarctica Marie Cavitte https://doi.org/10.14428/DVN/J34MQO
MASS2ANT Snowfall Dataset (Downscaling @5.5km over Dronning Maud Land, Antarctica, 1850-2014) Nicolas Ghilain, Stephane Vannitsem, Quentin Dalaiden, Hugues Goosse, and Lelsey De Cruz https://zenodo.org/records/4287517
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
200 | 59 | 84 | 343 | 28 | 13 | 12 |
- HTML: 200
- PDF: 59
- XML: 84
- Total: 343
- Supplement: 28
- BibTeX: 13
- EndNote: 12
Viewed (geographical distribution)
Country | # | Views | % |
---|---|---|---|
United States of America | 1 | 116 | 32 |
France | 2 | 29 | 8 |
China | 3 | 17 | 4 |
Germany | 4 | 15 | 4 |
Netherlands | 5 | 15 | 4 |
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
- 116