the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Jul 2025
Seasonal mass balance drivers for Swiss glaciers over 2010–2024 inferred from remote-sensing observations and modelling
Abstract. Reliable estimates of glacier mass balance for an entire mountain range provide valuable insights into the impact of glacier melt on regional water resources. Here, we derive daily mass balance estimates for every glacier in the Swiss Alps over the period 2010–2024. To do so, we leverage a glaciological model and remote sensing observations, i.e. geodetic volume changes and observations of the snow-covered area fraction (SCAF) of glaciers during summer, together with machine-learning techniques for extrapolation purposes. This allows reproducing the seasonal variability of glacier mass balance for glaciers without in situ observations and determining daily glacier mass balance across Switzerland. Over the study period, the Swiss glaciers lost almost 25 % of their 2010 ice volume, which corresponds to a wastage of − 15.2 ± 1.6 km3 of ice. The highest winter snow accumulation is inferred to occur in central and western Switzerland, with up to 1.5–1.9 m w.e. by the end of April, whereas the lowest winter accumulation is detected in Valais and ranges between 0.9–1.2 m w.e. Furthermore, winter balances are found to show better correlation in space compared to long-term annual balances, which range between − 0.6 and − 1.5 m w.e., indicating different dominating mechanisms. Finally, we assessed the spatio-temporal variability of seasonal mass balance to gain in-depth insights into the relation between glacier mass balance and the driving climatic factors in the Swiss Alps.
Competing interests: At least one of the (co-)authors is a member of the editorial board of The Cryosphere.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(5359 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- CC1: 'Comment on egusphere-2025-2929', Argha Banerjee, 12 Aug 2025
-
RC1: 'Comment on egusphere-2025-2929', Anonymous Referee #1, 26 Aug 2025
In this study, the authors present the application of a method developed in Cremona et al. (2025) to predict daily mass balances for all the glaciers of Switzerland for the period 2010-2024. The model is calibrated with different datasets (geodetic mass balance and SCAF) and forced with high resolution meteorological fields provided at 2 km resolution by MeteoSwiss. Mass balance results are aggregated at seasonal scale (winter) and annual scale to discuss the meteorological drivers and the spatial patterns of glacier mass balance at the scale of Switzerland.
First of all, I am impressed by the ambition and quality of this work within a series of development that add value to field measured mass balance measurements, combined with remote sensing data (Cremona et al., 2023, 2025; Landmann et al., 2021). However, I feel that this specific article requires some revisions to be a very useful contribution to the literature. The authors pursue two objectives: expending a previously developed method to a regional scale and investigating the drivers of mass balance, but none of them is fully achieved.
Regarding objective one (application and improvement of Cremona et al. (2025)), I have a number of general comments:
- The uncertainties are generally poorly treated. A sub-section of the method explaining what are the sources of uncertainties is missing. Uncertainties and errorbars are sometimes shown or quoted in the text (e.g. fig 6, L7 and 391, L385), but it is never clearly stated what they refer to (one or two sigma? what is actually included in the uncertainties?). Even more worrying is a very strange statement in the discussion on L374-375 that states that the uncertainties originate solely “only account for differences in the geodetic ice volume changes used during calibration (see Sect. 3.2.3), as this is identified as the main impacting factor (Cremona et al., 2025)”. First of all, there is no mention of uncertainties from geodetic volume estimates in Cremona et al. (2025), so the reference is wrongly cited. Second, I would be very surprised that the volume change estimates from very high resolution photogrammetric DEMs are the main source of uncertainties compared with the rest of the modelling chain, including the density conversion assumption, calibration and extrapolation of the melt and precipitation factors.
- The machine learning extrapolation based on XGBoost is difficult to grasp for non-specialist readers. It would be good to illustrate the model performance to predict annual mass balance expressed in m w.e., because the only evaluation metric shown in the text is the RMSE of c_prec which is not very intuitive. Afterwards melt parameters are calibrated a second time to match the geodetic estimates (L240-241). I do not fully understand why this correction is necessary and what is the actual impact of this two-step correction/calibration.
- I do not fully understand the calibration procedure, which I find quite confusing. My understanding of Cremona et al. (2025) is that annual glacier-wide mass balances from GLAMOS are used to calibrate the multiple melt parameters. But in this manuscript, it is mentioned that the melt factors are calibrated on geodetic estimates only (L218-220), and I don’t understand how a single observation is used to calibrate two or three parameters. Can you clarify this discrepancy? It is also important to highlight that your method does not rely on field measurements for calibration if this is actually true. I am also wondering if there isn’t a kind of circular reasoning when comparing your annual mass balance estimates with GLAMOS because the later ones are also calibrated with the same DEMs and model used in this study.
- I was confused with the manuscript structure, especially the “study site and data” and “methods” section. The long description of the DEMs and their origin in section 2 is not well placed and should be in the same sub-section as current 3.1. In my opinion, the method section should start with the model description, then the calibration strategy and the geodetic data processing. This is more a personal opinion, but it might help differentiating this work from (Cremona et al., 2025)
Regarding objective two:
- Why not looking at summer mass balance? -> ok not to do it, but it should be explicitly acknowledged
- Many other “processes” could be explored with your simple mass balance model, in particular to explore further extremes as in Cremona et al. (2023). Can you quantify the effect of the lengthening of the melt season? The effect of early summer heatwaves? The intensification of melt? Your time series is rather short and might not allow to discuss these issues, but it would still be interesting to touch upon these questions.
Specific comments:
L3-4: “daily mass balance estimates” are not really presented in the manuscript text (higher validation resolution is seven days). Consider removing from the abstract.
L32: “combine” is ambiguous. In practice geodetic estimates are used to calibrate a model.
L121: filtering at one standard deviation is quite extreme as it removes more than thirty percent of observed data in the case of a Gaussian distribution. Classical hypsometric filtering are much less aggressive with three or five sigma thresholds (McNabb et al., 2019)
L122: local or regional hypsometric interpolation? What happens to the elevation bands with no observation?
L130-140: a reference to Piermattei et al. (2024) should be added, as they face the same issue with the temporal stamping of SwissTopo DEMs
L165: albedo should be unitless
L166-167: what is the advantage of using two melt models and taking the average estimate?
L242: the “ensemble approach” is not well defined. What does it refer to? What is the ensemble size?
L264: NMAD has a precise definition which is different from the one given here (Höhle and Höhle, 2009). You should try another term like Standardized MAD.
L268 and figure 5: can you colorcode data points with their elevation or whether they belong to the ablation or accumulation area? Why restraining to 90 days? Winter observations might be particularly useful to validate the model performance.
L284: start a new paragraph for annual values
Fig. 7: It would be better not to use a diverging colorbar for panel a, and to center the colorbar of panel b on zero (conventionally positive mass balance are in blue and negative ones in red)
L343-344: “which is within the respective uncertainty ranges” -> which uncertainty ranges are referred to? See my general comment.
L358-362: this interpretation could be developed further
Fig. 12: could you add a panel that shows the total annual mass balance for glaciers surveyed by GLAMOS only?
Conclusions: if I understand correctly your method does not need field measurements. Could you comment on its applicability to unsurveyed glaciers? What would be the limitations? Mereological forcings? DEM availability?
Cremona, A., Huss, M., Landmann, J. M., Borner, J., and Farinotti, D.: European heat waves 2022: contribution to extreme glacier melt in Switzerland inferred from automated ablation readings, The Cryosphere, 17, 1895–1912, https://doi.org/10.5194/tc-17-1895-2023, 2023.
Cremona, A., Huss, M., Landmann, J. M., Schwaizer, G., Paul, F., and Farinotti, D.: Constraining sub-seasonal glacier mass balance in the Swiss Alps using Sentinel-2-derived snow-cover observations, J. Glaciol., 71, e25, https://doi.org/10.1017/jog.2025.1, 2025.
Höhle, J. and Höhle, M.: Accuracy assessment of digital elevation models by means of robust statistical methods, ISPRS J. Photogramm. Remote Sens., 64, 398–406, http://dx.doi.org/10.1016/j.isprsjprs.2009.02.003, 2009.
Landmann, J. M., Künsch, H. R., Huss, M., Ogier, C., Kalisch, M., and Farinotti, D.: Assimilating near-real-time mass balance stake readings into a model ensemble using a particle filter, The Cryosphere, 15, 5017–5040, https://doi.org/10.5194/tc-15-5017-2021, 2021.
McNabb, R., Nuth, C., Kääb, A., and Girod, L.: Sensitivity of glacier volume change estimation to DEM void interpolation, The Cryosphere, 13, 895–910, https://doi.org/10.5194/tc-13-895-2019, 2019.
Piermattei, L., Zemp, M., Sommer, C., Brun, F., Braun, M. H., Andreassen, L. M., Belart, J. M. C., Berthier, E., Bhattacharya, A., Boehm Vock, L., Bolch, T., Dehecq, A., Dussaillant, I., Falaschi, D., Florentine, C., Floricioiu, D., Ginzler, C., Guillet, G., Hugonnet, R., Huss, M., Kääb, A., King, O., Klug, C., Knuth, F., Krieger, L., La Frenierre, J., McNabb, R., McNeil, C., Prinz, R., Sass, L., Seehaus, T., Shean, D., Treichler, D., Wendt, A., and Yang, R.: Observing glacier elevation changes from spaceborne optical and radar sensors – an inter-comparison experiment using ASTER and TanDEM-X data, The Cryosphere, 18, 3195–3230, https://doi.org/10.5194/tc-18-3195-2024, 2024.
Citation: https://doi.org/10.5194/egusphere-2025-2929-RC1 -
RC2: 'Comment on egusphere-2025-2929', Anonymous Referee #2, 28 Sep 2025
In this study, the authors estimate daily mass balance for ~1,400 Swiss glaciers between 2010 and 2024. The authors use geodetic volume changes and observations of the snow-covered area fraction (SCAF) to calibrate a glacier evolution model with two different ablation models. The framework explicitly accounts for debris cover. Calibration is performed for 87 glaciers with available SCAF data, using a precipitation correction factor for winter mass balance and two calibrated parameters for summer mass balance in each ablation model. A machine learning model is then trained on these glaciers, using topographic features as predictors to estimate the calibrated precipitation correction factor, which is subsequently extrapolated to all remaining glaciers without SCAF observations. Model outputs are evaluated against seasonal glacier-specific mass balance records and point mass balance measurements. The authors report that Swiss glaciers lost −15.2 ± 1.6 km³ of ice during the study period and emphasize the importance of winter accumulation in explaining interannual spatial variability in glacier mass balance.
Overall, the study represents a valuable contribution to regional glacier mass balance modeling. It successfully integrates remote sensing data, machine learning, and glacier evolution modeling, and will be a useful resource for the glaciological community. However, the manuscript also contains several shortcomings that need to be addressed before it can be considered for publication. Detailed comments and suggestions are provided below.
Major comments:
1) Temperature lapse rates: There are several issues regarding the treatment of temperature lapse rates in this study.
a) Initial lapse rate calculation: The authors calculate the initial lapse rates from the temperature and elevation of the cells next to the glacier (L170, L215). I am concerned that if the neighboring grid cells are not ice-covered but instead rock, forest, or other surfaces, their microclimate and hence surface temperature differs from that of glaciers. Consequently, the derived lapse rates may be inaccurate. I recommend calculating lapse rates directly from pressure levels at glacier grid cells, as for example done in Draeger et al. (2024) or Marzeion et al. (2012). Additionally, the manuscript does not report the initial lapse rates used. Please provide a range and compare them to commonly used lapse rates in global glacier modeling (e.g., the constant value of −6.5 K km⁻¹ in Schuster et al., 2023).
b) Lapse rate adjustments: The authors introduce a new third calibration step, where temperature lapse rates are adjusted to further optimize the model’s reproduction of observed SCAF (L211–212), which is new compared to Cremona et al. (2025). It is unclear to me why the remaining bias in the third calibration step is attributed to the lapse rate rather than to near-surface temperature values. Previous glacier models have often applied a separate temperature bias correction in addition to lapse-rate correction (e.g., Rounce et al., 2020). To correctly separate the sources of bias, I recommend calculating more accurate lapse rates from pressure levels and framing the adjustment as an additive temperature bias correction instead of a temperature lapse rate adjustment. Conceptually, this distinction is important: lapse rate bias represents the bias in the vertical temperature gradient, while temperature bias reflects errors in near-surface temperature. This reframing would not change results, as the lapse rate correction is also applied additively (see minor comment on assumed typo in Equation (4)), but it would clearly disentangle the two effects and improve conceptual clarity.
2) Extrapolation: The extrapolation relies only on topographic features and does not incorporate climatological variables, which could in my view introduce substantial uncertainty. L383–385 states: “Considering that the Swiss-wide average precipitation correction factor is 1.8 and the winter mass balance 1.27 m w.e., uncertainties in winter mass balance introduced during the extrapolation correspond to 20–25% on average.” This is considerable, and it would be worth testing whether including climatological predictors per glacier (e.g., average mean summer temperature, average total winter precipitation, average temperature range, etc.) could improve the performance. Moreover, since SCAF observations are mostly available for larger glaciers (L59), it would be very valuable to include an analysis of extrapolation performance for smaller glaciers. As the training set is biased toward large glaciers, there is a risk that mass loss from smaller glaciers is not correctly modeled (underestimated?) when extrapolating parameters. For instance, you could compare the test-set performance between small and large glaciers.
3) Evaluation: I am wondering why the authors have not compared their estimates against the seminal dataset by Hugonnet et al. (2021), which provides geodetic mass balance for all glaciers worldwide, including the Swiss Alps, for 2000–2019. This dataset is currently the most widely used for calibrating regional and global mass balance models. It would be valuable for the glacier modeling community to compare the mass balance results in this study with theirs and to discuss differences in the treatment of digital elevation models, etc.
4) Area changes: It is unclear how exactly the area changes are calculated. The authors update the area-elevation distribution of their model domain“by attributing relative annual area changes equally to all bands below the mean equilibrium line altitude” (L201). Does this mean that each elevation band below the mean ELA changes by the same relative area? If so, this may not reflect reality, as lower-elevation bands near the terminus typically lose mass and area fastest. I am wondering how glacier retreat occurs in your model under this assumption? Applying uniform relative area changes could artificially retain lower bands, potentially exaggerating glacier-wide negative mass balance and contributing to the reported bias of −0.05 m w.e. (L253). Additionally, Cremona et al. (2025) report a 0.04 m w.e. difference in seasonal mass balance for a 0.5% yearly glacier area change (L377–379), which may explain part of the observed bias.
Regarding the “mean equilibrium line altitude” (L201), are you assuming a constant ELA throughout the study period? How was it calculated – for example, using median elevation as a proxy or derived from yearly mass balance values per band? Please clarify the rationale for assuming a constant ELA. I also suggest considering a sensitivity study to assess how changes in ELA would influence glacier-wide mass balance.
5) Albedo parameterization: In general, net radiation, particularly shortwave radiation, is the dominant term in the surface energy balance of glaciers (Hock, 2005). Therefore, the albedo parameterization is critical. The authors use the parameterization by Brock et al. (2000), which is based on a single glacier in Switzerland with data from 1993–1994. However, albedo can vary significantly between glaciers, even within the same mountain range, and also shows strong temporal variability (Williamson et al., 2025). Moreover, the authors state in L309–311: “The high correlations [of winter precipitation anomaly with winter mass balance, and summer temperature anomaly with summer mass balance] confirm the predominant influence of temperature and precipitation on determining seasonal mass balance variations.” While temperature and precipitation are indeed important, the authors have not analyzed other drivers, such as albedo variations, even though they acknowledge their significance (L313). For example, Williamson et al. (2025) found that 31–41% of increased melt in Western North America and the Canadian Arctic is attributable to albedo decline. I therefore suggest: (1) Including a sensitivity study on the influence of the chosen albedo parameterization on mass balance and (2) analyzing correlations between albedo variations and mass balance to assess its potential impact. This would provide a more balanced assessment of the factors driving seasonal and annual glacier mass balance and avoid over-attributing effects to temperature and precipitation alone.
6) Description of data and methods: In the current draft, the data and methods sections are intermixed. For example, important methodological details, such as volume–area scaling, are presented in the Data section, which makes it difficult to follow and locate information. For better readability, I suggest clearly separating the Data and Methods sections.
Minor comments:
Abstract: It would be good to mention that the authors are using a simplified surface energy balance melt model.
L55: The authors state that “the last complete assessment of geodetic mass balance covering all glaciers of the Swiss Alps spans the period 1980–2010 (Fischer et al., 2015)”. However, the dataset by Hugonnet et al. (2021) is not mentioned. I recommend that the authors acknowledge and discuss this dataset (see Major Comment 3).
L64: Typo
L 68: Typo: should read “Fig. 2.”
L 102: The manuscript uses the term “validation.” Model validation usually refers to assessing model structure and assumptions during the development stage, whereas “evaluation” refers to assessing model performance afterwards. I argue that what is performed here is “evaluation,” and suggest using this term consistently throughout the paper.
L129-130: Typos
L168: Equation (4) seems to contain a typo. I assume you mean T(z) = T + (z – z_ref) * deltaT/deltaz instead of multiplication.
L169-170: Please clarify how the temperature lapse rate was calculated, including the method used (linear regression?), the number of neighboring grid cells considered, the elevation range they cover, and whether extrapolation beyond this range was applied to higher or lower elevation bands.
L177-178: The manuscript states: “For glaciers larger than 2 km², we use the melt factor provided by Rounce et al. (2021).” It is unclear what is meant by “melt factor.” Are you referring to the sub-debris melt enhancement factors from Rounce et al. (2021)? If so, please use the correct terminology consistently throughout the manuscript to avoid confusion with the melt factor (MF) in your model.
L 179: “At each elevation band, the average debris melt factor is calculated.” Please provide more detail on this calculation. The description of the methodology for dealing with debris cover is currently vague and would benefit from clarification.
L180: “Only if the elevation band is snow-free.” Please explain how you distinguish between snow-covered and snow-free conditions.
L195: Consider rephrasing “define c” as “calculate c”.
L 203-226: It is unclear how SCAF modeling is performed. Please provide a brief overview of the methodology in addition to referencing Cremona et al. (2025).
L208: Please provide a reference for this statement.
L239: In addition to absolute error measures, please also provide relative error metrics for the machine learning model (e.g., mean absolute percentage error). Also, please specify the sizes of the training, validation, and test sets.
L247: The phrase “To address the uncertainty” would be better phrased as “To evaluate the model,” since the uncertainty itself is not being addressed here.
L384: The authors mention the average precipitation correction factor. It would be helpful to also report the mean and ranges of the other calibrated parameters in the manuscript.
Figure 4: Please explain all abbreviations (ALE, RHO, etc.), as these have not been introduced in the manuscript.
References:
Brock BW, Willis IC, Sharp MJ (2000). Measurement and parameterization of albedo variations at Haut Glacier d’Arolla, Switzerland. Journal of Glaciology. 2000;46(155):675-688. https://doi.org/10.3189/172756500781832675
Cremona, A., Huss, M., Landmann, J. M., Schwaizer, G., Paul, F., and Farinotti, D. (2025). Constraining sub-seasonal glacier mass balance in the Swiss Alps using Sentinel-2-derived snow-cover observations, Journal of Glaciology, 71, e25, https://doi.org/10.1017/jog.2025.1
Draeger, C., Radić, V., White, R. H., & Tessema, M. A. (2024). Evaluation of reanalysis data and dynamical downscaling for surface energy balance modeling at mountain glaciers in western Canada. The Cryosphere, 18(1), 17–42. https://doi.org/10.5194/tc-18-17-2024
Hock, R. (2005). Glacier melt: a review of processes and their modelling. Progress in Physical Geography: Earth and Environment, 29(3), 362-391. https://doi.org/10.1191/0309133305pp453ra
Hugonnet, R., McNabb, R., Berthier, E. et al. (2021). Accelerated global glacier mass loss in the early twenty-first century. Nature 592, 726–731. https://doi.org/10.1038/s41586-021-03436-z
Marzeion, B., Jarosch, A. H., & Hofer, M. (2012). Past and future sea-level change from the surface mass balance of glaciers. The Cryosphere, 6(6), 1295–1322. https://doi.org/10.5194/tc-6-1295-2012
Rounce, D. R., Hock, R., & Shean, D. E. (2020). Glacier Mass Change in High Mountain Asia Through 2100 Using the Open-Source Python Glacier Evolution Model (PyGEM). Frontiers in Earth Science, 7. https://doi.org/10.3389/feart.2019.00331
Schuster L, Rounce DR, Maussion F. (2023) Glacier projections sensitivity to temperature-index model choices and calibration strategies. Annals of Glaciology. 64(92):293-308. doi:10.1017/aog.2023.57
Williamson, S.N., Marshall, S.J. & Menounos, B. (2025). Temperature mediated albedo decline portends acceleration of North American glacier mass loss. Commun Earth Environ 6, 555. https://doi.org/10.1038/s43247-025-02503-x
Citation: https://doi.org/10.5194/egusphere-2025-2929-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 994 | 110 | 23 | 1,127 | 14 | 20 |
- HTML: 994
- PDF: 110
- XML: 23
- Total: 1,127
- BibTeX: 14
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors build on their previous work to reconstruct the daily time series of all the Swiss glaciers. At the heart of it is a degree-day mass-balance model, which is calibrated using decadal-scale geodetic data and satellite right snow-covered area fraction during the summer months. The change in melt rate due to the presence of debris cover and the changes in glacier geometry are also parameterised using simple models. The model output is validated at annual, seasonal, and subseasonal time scales for several glaciers using in situ observations. While this is an interesting piece of work, leading to a powerful dataset of glacier-specific forcing and the reconstructed mass balance response for all the Swiss glaciers, and is a natural continuation of the work published previously by some of the present authors, I found some major limitations in the persent manuscript. These are listed below.
1. The presentation of the method leaves a lot of gaps, possibly because of the body of past work by the authors which is assumed to be known to the reader. However, for a common reader, that makes it quite difficult to figure out the details of the methods. The issue with referencing highlighted later makes it even harder a task. In fact, this slowed me down quite a bit as I was going through the first three sections, and I could not focus at the results and discussions as much as I would have wanted to. Here are example,
Sect 3.1: The part of the third paragraph on assigning a date to each DEM is hard to follow. In the example discussed and in Fig 2, why do you omit the DoDs of 2009-2018 & 2008-2018 (blue) and 2008-2015 (red)? The three blue DEMs (2009, 2018 & 2023) seem to be producing inconsistent thinning estimates: ’09-’18 ~ 0.9, ’18-’24 ~ 0.9, but ‘09-‘24 ~1.8?
Sect 3.2.1: The melt factor is not defined with an equation. What is “debris-coverage” - the fraction? the total area? How is it kept constant as the hypsometry evolved? How do you, if at all, validate the performance of the subdebris-melt module?
Sect 3.2.2: Why not validate the performance of Eqs. 5&6, along with a glacier-specific c, using the observed geodetic balance and area-change data, wherever these are available? Please see the limitation of these method pointed out in the comments in the attached pdf. How are the issues of a different scaling exponent, stagnation of tongue, etc., known for the debris-covered glaciers taken care of in this evolution model?
Sect. 3.2.3: Why do you consider only the difference in various geodetic estimates as the only source of error? What happens to the glaciers with only one or two geodetic data points? Is the final prediction uncertainty comparable with the spread in the geodetic data, or is it much less than that? If it is the latter, then how can you explain that?
Sect 3.2.4: You seem to extrapolate only cprec . However, it is not apparent from the heading or the first several sentences of the section. Also, the rationale for doing this is not given. The best-fit values/range of cprec remained hidden for me even as I searched both the papers (this and the JoG one). Therefore, I do not have any clue if an RMSE of 0.37 is good or bad. Moreover, the consequent uncertainty in the computed winter and annual balance is not discussed. There is a fair chance that this could be very significant.
Sect 3.3: The logic behind the choice of the 10 glaciers is not explained (L250).
There is no validation attempted for the daily-scale product. It is hard to imagine that constants like c_prec, which are derived using a seasonal-scale mass balance, will capture the variability on a daily scale. While the model resolution is daily, robust estimates may require some averaging.
Strangely enough, the MAD is given as %-age for shorter time scales, while the actual values are given in the seasonal and annual scales. The scatter in fig 5 does not scale with the value, and the noise appears additive (as opposed to multiplicative). The biases are omitted for the subseasonal case as well. In fact, In all three cases, the scatter plots show that there is a significant variation in MAD and bias from one glacier to another. Its consequence for regional to catchment scale estimates remains to be assessed.
It is unclear if the subdebris melt estimates were included in the validation. If so, what is the corresponding MAD?
2. The limitation of the uncertainty analysis is another factor that, in my view, weakens the paper. Only using the spread of the geodetic balance values – that too without taking into account their uncertainty – is done without any serious justification. The reference to the previous JoG paper actually increases the confusion – I did not find any clear computation of the uncertainty in mass balance there (see comment in attached pdf). The MAD is not the same as the uncertainty.
The Fig. 5 of the JoG paper may suggest SCAF has less scatter for the larger glaciers like Aletsch, which is not unexpected. However, it also had the largest MAD for annual balance, which was about twice as large as the quoted mean. The Fig. 4 of the present paper also suggests a possibly larger scatter of winter balance for some of the smaller glaciers. However, the scatter for some of the smaller and larger glaciers is similar for the annual balance. Trends like this need to be investigated carefully to avoid uncertainties and biases present in the fitted model.
In fact, the biases on individual glaciers, including some of the large ones, may be significantly higher than the mean quoted in the text (L257). The biases are not discussed much except giving a mean bias, which is going to be small because of the oscillating sign. Could you check the mean absolute bias? A clear systematic bias also shows up in Fig5 , with the model overestimating melt wherever the observed melt is less (more negative) than -2 m/y. This bias and its effect on the model output, which are completely ignored, are likely to be significant, particularly on the extreme years, and demand a thorough analysis.
It was not demonstrated if that the subdebris melt and glacier-geometry evolution, two pieces that were added to the model here, actually improves/changes the estimates. How much they increase the uncertainties needs to be considered as well, given the known limitations of the specific schemes used (see attached pdf).
It may be true that the set of independent geodetic mass balances for a glacier may have a large spread. However, that may not help in getting an accurate measure of uncertainty, as these data sets may have different error bars, and sometimes there may not be enough measurements. Another alternate approach, which we had taken in Banerjee et al., 2022, JoG, is to add appropriate noise in a Monte Carlo and compute the mass balance for each case to generate a large ensemble to obtain the error bar.
Another related question: could you add the modelled mean balance for Aletsch in Fig. 2b, with an uncertainty band, and see how it compares with the spread of the input geodetic values, which vary by almost 100% . Since this is the basic input calculation are based on, the uncertainties in output should be comparable to this spread. If the procedure yields a lesser spread, the questions would be why and how. That’s where the uncertainty band in Fig6 may require a careful revisiting.
The model uncertainty as a function of time scale of prediction, starting from days to decades, would be a good addition to the paper.
3. The discussions can be more mindful of the model assumptions, and the inferences have to be substantiated using the model output. For example, to describe the modelled variability, you refer to things like avalanching, wind-driven redistribution, Saharan dust, etc., none of which are present in the model! Also, while you consider the role of P&T, you do not look into radiation and SCAF to understand the variability of observed mass balance, particularly that of the annual and summer balance. According to me, some really interesting features in our output (e.g. see the comments to Fig. 10, in the annotated PDF) are overlooked. I get the feeling the powerful, detailed data set of forcing and response that you produce can be exploited a bit more.
4. The writing and referencing leave scope for improvements. I have pointed out several instances of complicated sentences or incomplete information in the annotated PDF attached, which should help illustrate the point.
The authors may want to look beyond a set of papers familiar to them and look harder for the most appropriate references. While I could not check all the references, I did find a few surprising inclusions. Here are a few examples:
L19 van Tiel et al. (2025), which deals with buffering of runoff in an extreme year, may not be the most suitable reference for general properties of runoff from glacierised catchments. I remember referring to the excellent paper by Hock, Jansson, & Braun (2005) while discussing runoff variability.
L22 From the titles of the three references cited, while not totally unrelated, they do not seem to be the most appropriate ones on the topic of “ increasing interest in accurate monitoring of glacier mass balance and runoff on the regional scale”.
L25 "Dussaillant et al., 2018; Denzinger et al., 2021; … " how were these papers chosen, when you want to introduce something as basic as geodetic mass balance? While these are interesting global and local scale studies, they may not serve as a basic introduction to the method and are also not the most relevant ones for your study area, as far as my limited understanding of the topic goes.
L225 Cremona 2025 JoG does not really show that as far as I can tell. See the comment in the annotated PDF.
Sect 4.4 A quantitative comparison with only one of the previous reconstructions (where some of you are coauthors) is a missed opportunity I feel. Can you not compare with the other existing studies? Are there other reconstructions available for the Swiss glaciers – at least at annual scale, or for specific glaciers or regions – beyond van Tiel 2025? Apart from van Tiel 2025, you only mention Dussaillant 2025, but do not compare with their results. As it stands, the claim made in L355 is unsubstantiated. Are there not other reconstructions, say, based on remote-sensing proxies like snowline etc.? They are many in the Himalayan literature that I am more familiar with. I am half sure they may be there Swiss Alps. In general, there may be a need to connect better with the existing literature and work by other groups, through improved referencing, and by incorporating results from such studies.
Please see the annotated PDF attached for more detailed comments, some of which are referred to above as well.
– Argha Banerjee