the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Oct 2025
Air temperature partitioning of snow accumulation, erosion and melt: a regime shift occurring on Mt. Ortles (Eastern Italian Alps)
Abstract. Glacier mass balance measurements and models are key tools for understanding the glacier response to climate change and specific processes occurring at the glacier surface. Snow accumulation and wind-driven erosion are among the most difficult processes to measure and model in high-altitude alpine terrain and on glaciers, due to their high variability in space and time, and to the scarcity of in situ observations. Here we use a unique dataset of nivo-meteorological and mass balance observations collected between 2011 and 2015 at 3830 m a.s.l. on Mt. Ortles (Eastern Alps) to investigate snow accumulation and erosion processes. We applied the physics-based snow cover model SNOWPACK, constrained by field data, to reproduce the local mass balance and to explicitly simulate snow erosion by wind. The model reproduces the observed seasonal and annual mass balance variability with good accuracy over the four-year study period. Results indicate that wind erosion is the dominant ablation process at the study site, removing 21 % of the snowfall, whereas melt plays a minor role. Erosion is most effective in winter, during or shortly after snowfall events, and its efficiency is controlled by air temperature, with dry snow being much more susceptible to erosion than wet snow. Sensitivity experiments to air temperature perturbations demonstrate that wind erosion provides a negative feedback to the mass balance, because increasing temperature accelerates snow metamorphism and makes the snow surface less erodible. However, a further 1 °C warming would promote a transition from an erosion-dominated to a melt-dominated mass balance regime. Our findings emphasize the importance of accounting for wind erosion in projections of glacier mass balance under climate change. They also highlight the relevance of snow erosion for the interpretation of ice core records, because long-term variations in snow erosion may have affected the formation and preservation of the seasonal paleoclimatic signal.
- Preprint
(2462 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 19 Jan 2026)
- RC1: 'Comment on egusphere-2025-5186', Anonymous Referee #1, 17 Dec 2025 reply
-
RC2: 'Comment on egusphere-2025-5186', Luis Durán, 29 Dec 2025
reply
Figure 2. I recommend restructuring this figure into three separate panels.
First, a combined plot displaying temperature and relative humidity, using a dual-axis format, provided that relative humidity data are available.
Second, a precipitation plot using stacked bars, incorporating a distinct color scheme to indicate precipitation events occurring when the mean temperature is below 0 °C.
Third, I suggest including a wind rose diagram, divided into 12 directional sectors and approximately four wind-speed bins, to provide a clearer representation of wind patterns.Line 124. Air temperature and relative humidity are known to be strongly anticorrelated, and analyses based on air temperature are therefore often assumed to implicitly account for humidity effects. However, in the present context relative humidity may exert an independent control on mass balance and snow dynamics, particularly through its influence on sublimation and ablation processes. It is therefore not immediately clear why relative humidity was not included in the analysis. If its influence is negligible compared to that of temperature and wind, this could be explicitly mentioned.
line 195. States: “The snow accumulation was simulated using the Solda precipitation, multiplied by a correction factor to account for the increase of precipitation with elevation (vertical precipitation lapse rate).”. Could you please give some basic information or a reference about this method?
line 200. When you state that “the multiplicative factors for precipitation and wind speed were adjusted iteratively to minimize the RMSE between measured and modelled mass balance at the simulation site”, you are implicitly assuming that these calibration factors do not compensate for other structural deficiencies or errors in the snowpack model itself.
Could you clarify whether this assumption has been evaluated? In other words, how do you ensure that the tuning of precipitation and wind-speed multipliers is not merely correcting for biases arising from other processes (e.g., snow densification, redistribution, sublimation, melt parametrization, or energy-balance components) rather than improving the physical representation of precipitation and wind inputs?
line 207 Maybe this was mentioned or implicitly assumed, but what is the treatment of bidirectional sublimation in the model? Is the contribution of sublimation—both mass loss (sublimation) and mass gain (deposition/subsublimation)—considered negligible compared with other processes, or is there a reason why it cannot be distinguished from wind-driven snow erosion in your results? In other words, how do you ensure that sublimation/deposition is not being misattributed to wind erosion in the mass-balance interpretation? A basic clarification might be needed here.
Line 318. If I understand this figure correctly, this represents the frequency of each threshold wind speed for dry and wet snow. It seems hard to interpret since higher winds are less frequent than lower winds, and there might be a high correlation between high speeds and lower temperatures. It is hard for me to interpret the peak at 9 m/s for dry snow mentioned in Line 312. However, this peak is largely influenced by the higher occurrence of moderate wind speeds in the dataset, rather than indicating an intrinsic preference of the event for this velocity. To better isolate the physical relationship between wind forcing and event occurrence, the event frequency should be normalized by the wind speed distribution. Figure 7 partially solves this issue. A more informative representation is therefore the conditional probability of the event given wind speed, which quantifies the likelihood of the event occurring for a given wind regime and removes the bias introduced by the uneven frequency of wind speeds. Maybe something in this direction should be done or a proper explanation in the text.
Following this line of reasoning, an additional informative visualization would be a two-dimensional scatter (or binned) plot using wind speed and air temperature as the horizontal and vertical axes, respectively, with the size (or color intensity) of each marker representing the frequency of wind erosion events. This representation allows the combined influence of thermal and dynamical conditions to be assessed simultaneously, highlighting preferential regimes where wind erosion is most likely to occur and revealing potential interactions between temperature and wind forcing that are not evident in one-dimensional distributions.
Line 363. I understand that Table 2 is in some way examining the performance of SNOWPACK during this period. The analysis seems to assume that the model reproduced the snowpack accurately, yet I wonder if a higher frequency validation could be shown to support this assumption apart from Table 2.
Line 359. This paragraph provides a very clear explanation of the study’s scope and limitations. These ideas might have been helpful at the beginning of the paper. I may be mistaken, and they may already appear in the introduction, but they are particularly clear here.
This may be a minor point, but the figures have a spreadsheet-like appearance that could be distracting to the reader, especially Figure 13.
Citation: https://doi.org/10.5194/egusphere-2025-5186-RC2 -
RC3: 'Comment on egusphere-2025-5186', Luis Durán, 30 Dec 2025
reply
Figures 3 and 4: here no distinction is made between wet and dry snow erosion whereas later importante differences are shown. Could it not be misleading to calculate these correlations without making the distintition.I have some particular comments to add to my previous list. I hope they are helpful.Fig 2: Maybe too simple for the information available. More information could be included especially about the year to year variability. For example max and min values could be shown with a shaded area or swarmplots could be employed. Also other variables such as radiation and relative humidity were recorded by the AWS and used within the model but are not shown here.Section 4.1 and Table 2: It seems that the model has some problems reproducing the measured SWE especially during summers. This is later explained in the Discussion but maybe a sentence could be included here also. As of now one can see this discrepancy in the Table but no mention is found within the text until the Discussion.Lines 271 and 272: Maybe a better description of the statistical goodness of these fits is necessary. Also it is said that "the dry snow fit lies significantly above" but no indication of this significance is made in terms of statistical tests.Line 292: "dry snow erosion accounts for 91% of the total modelled erosion" Could you please provide information about how much time you have wet or dry snow conditions? Right now it is unclear whether this difference is just because of erosion being less likely to occur with wet snow or also because dry snow conditions are significantly more common in your studied period and site than wet snow conditions.Line 297: same as the previous comment. It would be useful to know how much of this could be because of dry snow conditions being far more common than wet in your site.Fig 7: again this Figure may be affected by the different amounts of wet vs dry snow conditions in your datatset. Maybe you could include a version where the values are normalized by the number of hours with wet or dry snow conditions in each case?Fig 8: in a similar manner as previous comments maybe this Figure would be easier to interpret if you represente it by stacked area instead of lines. That way it would be easier to interpret the sum of wet and dry cases as a general case distribution.Fig 11: This Figure could be made much smaller without hindering its readability. Its data is also already displayed within Table 3 (except the surviving snow percent that could be easily added as a new line at the end). Maybe it would be better to reduce or delete it in favor of Table 3.Line 485: Again it would be best to know the amount of times where wet and dry snow conditions are met within your data. As now this sentence could be influenced by your site characteristics (not much wet snow conditions) rather than it being more difficult for erosion to take place with wet snow.Citation: https://doi.org/
10.5194/egusphere-2025-5186-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 174 | 95 | 20 | 289 | 23 | 20 |
- HTML: 174
- PDF: 95
- XML: 20
- Total: 289
- BibTeX: 23
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comment
The authors investigate the snow cover at one AWS location at Mt. Ortler (on the glacier) for the 4 year period 2011-2015 using the SNOWPACK model. Main focus of the study is the assessment of wind erosion of snow depending of snow conditions and potential future changes. While the general topic is very interesting and the approach taken understandable, several aspect remain unclear to me. In the following, a list of main and specific comments.
Main comments
Regime shift
In my understanding, the described shift from an erosion- to a melt-dominated regime at Mt Ortler is not a new phenomenon, but is happening since the accumulation areas of the glaciers are decreasing and the glaciers are retreating. This shift seems to now also have arrived at the investigated site fairly high up close to the peak. In case I understand this correctly, why is it important that this regime shift also has arrived at the investigated site?
Ice core, paleo-climatic reconstructions
Reading the manuscript, it seemed that the authors are very closely linked to the ice core community. A lot of parts of the manuscript are designated to the description of ice cores and glacier mass balances in general. Especially the introduction focuses a lot on theses topics. However, in my opinion, this study is not about ice cores or paleo-climate investigations and also not about glacier mass balances. Snow simulation at one point are conducted, which in this case was on a glacier. I think that similar investigations on snow erosion also could have been conducted at a non-glacieted site. The focus of the introduction should be more shifted to snow modelling and the modelling of snow redistribution processes, I think.
New insights, innovative aspect of study
The authors state that already in the 80s and 90s it was reported that temperature regulates the susceptibility of the snowpack to wind erosion. As far as I know, following this type of knowledge snow erosion processes where implemented into SNOWPACK. So the model includes a temperature-dependence of wind erosion of snow. So when your results point out that dry snow erodes more than wet snow, it is because that it is defined like this in the model. Please point out what are the new insight from your study. To me the most interesting aspect is the quantification of the negative temperature feedback on the snow erosion (more melt due to higher temperatures, but less loss due to erosion). This aspect is a bit lost in all the other information, I think. Maybe it is possible to more focus on this aspect?
Precipitation correction, model adjustment
It is not entirely clear to me what correction and adjustment steps with regard to the precipitation were taken. As you take precipitation from a valley station nearby (~1500 m below the site) you first correct for undercatch (according to Kochendorfer?), then adjust lapse-rate based and then multiply everything with a constant factor, right? Please clarify. Please also justify why you selected the constant correction factors for precipitation and wind to calibrate your model. Please also discuss the issue of equifinality. Are there other combination of the two correction parameters that also result in good results? If I understand correctly, you correct the measured wind speed with a factor of 0.7. Why should the at the site measured wind speed constantly have been too high so you have to correct them with this factor? Optimization of a highly complex snow model as SNOWPACK only based on six snow depth measurements between 06/2012 and 09/2014 seems questionable to me. Can it be that you adjust for the wrong reasons and compensate for errors, particularly when only adjusting two constant multiplicative factors. Please consider to snow precipitation and wind measurement before and after correction and if possible also show the full four years simulated including the time steps with observed values using for model optimization. Please also consider to show results of modeled snow layering. Due to your modeling approach you have very detailed insights and the exemplary presentation of an erosion event could be very insightful.
Selection of site
You only have very little data (four years of temperature and wind from one site more than ten years ago). Please justify better why you select this site and data to investigate temperature-dependence of snow erosion. Aren't there better sites and data sets to investigate this aspect? I think that a continuous time series of SD and/or SWE would be very beneficial for the evaluation of the model results.
Warming/Cooling
The authors choose a very simple a basic approach to assess the impact of temperature changes by simply adding/subtracting 1/2/3 °C to/from the temperature time series. This results in physically in-consistent model input and I am not sure this is a good idea when modelling the show cover with a physically-based snow model. To me the combination of the very simple temperature change approach for only four years of data and the SNOWPACK model needs to be justified better. I am not sure if the robustness of the provided numbers can be guaranteed. I am not sure if the results from 4 years can be generalized considering the complexity of the area. As long-term historical and future climate data is available, I find it difficult to understand why no such data has been used. Please argue why the selection of this simple approach is justified.
Specific comments
Line 5 'unique dataset': As far as I know, a climate station on a glacier is not unique. There are other sites with longer measurements and more measured variables. Why are measurements of snow depth once per year useful to investigate snow erosion? Wouldn't continuous measurements of SD and SWE be required?
Line 33-34: I am not sure I understand why 'elevation' represents a positive feedback mechanisms? How is snow accumulation due to avalanches a negative feedback mechanisms? Maybe the wording 'feedback' is a bit misleading here. High elevations, topographic shadowing and accumulation of snow due to avalanches favor the formation of a glacier, but I would not call it feedback. Maybe it is possible to re-phrase these sentences to avoid misunderstandings.
Fig 1: From what year is the glacier extent? Glaciers seem still fairly large compared to the current state. Please consider to update to a current state.
Fig 1: The map takes the entire page. Please consider to optimize the figure size, e.g. by including panel b into a.
Fig 2: As it is only 4 years of data, please consider to show the entire time series of the measured data (not monthly values only). Also consider to mark the periods which were gap-filled. Please also clearly show in the figure that precipitation is coming from another location.
Line 188: I am not sure where all the other meteorological variables used to force SNOWPACK at the selected site were coming from. Please clarify where e.g. humidity and the radiation components where measured.
Section 3.1: I think that the description of the AWS site fits better to the data description. Please consider to move.
Line 219: Why not using liquid water content directly to assess if the snow was wet or dry?
Fig 3: I am not sure what to think about the fact that a very low correlation of 0.09 is considers highly significant. Is it possible to plot the values, e.g. for SWE_eroded and T_Ereosion, so the reader gets a better visual impression of the correlation?
Fig. 7 and 8: Figures take a lot of space and it is all white on the right side. Please try to optimize the figure quality.
Line 359: Which snow depth sensors?