the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Separating snow and ice melt using water stable isotopes and glacio-hydrological modelling: towards improving the application of isotope analyses in highly glacierized catchments
Abstract. Glacio-hydrological models are widely used for estimating current and future streamflow across spatial scales, utilizing various data sources, notably streamflow and snow/ice observations. However, modeling highly glacierized catchments poses challenges due to data scarcity and complex spatio-temporal meteorological conditions, leading to input data uncertainty and potential misestimation of snow and ice melt proportions. Some studies propose using water stable isotopes to estimate water shares of rain, snow, and ice in streamflow, yet the choice of isotopic composition of these water sources significantly impacts results. This study presents a combined isotopic and glacio-hydrological model to determine seasonal shares of snow and ice melt in streamflow for the Otemma catchment in the Swiss Alps. The model leverages available meteorological station data (air temperature, precipitation, and radiation), ice mass balance data and snow cover maps to model and automatically calibrate the catchment-scale snow and ice mass balances. The isotopic module, building on prior work by Ala-Aho et al. (2017), estimates seasonal isotopic compositions of precipitation, snow, and ice. The runoff generation and transfer model relies on a combined routing and reservoir approach and is calibrated based on measured streamflow and isotopic data.
Results reveal challenges in distinguishing snow and ice melt isotopic values in summer, rendering a reliable separation between the two sources difficult. The modelling of catchment-wide snow melt isotopic composition proves challenging due to uncertainties in precipitation lapse rate, mass exchanges during rain-on-snow events, and snow fractionation. The study delves into these processes, their impact on model results, and suggests guidelines for future models. It concludes that water stable isotopes alone cannot reliably separate snow and ice melt shares for temperate alpine glaciers. However, combining isotopes with glacio-hydrological modeling enhances hydrologic parameter identifiability, in particular those related to runoff transfer to the stream, and improves mass balance estimations.
- Preprint
(18158 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-631', Anonymous Referee #1, 11 May 2024
This work presents an interesting modeling effort in a Swiss glacierized catchment, which involved a water stable isotope module. Although the authors provided detailed information regarding the data, model structure, and sensitivity analysis, the novelty of this work needs to be further emphasized. Specifically, the authors should explain how the assessments in this study can inform hydrological modeling in glacierized basins. In the introduction, the authors listed a couple of modeling challenges for glacierized basins, which is good. My major suggestion for the authors is to improve their results by providing more evidence on how the built isotope-glacio hydrological model or the assessments conducted in this study can help to address such modeling challenges. Currently, the paper reads more like a description of the model and data. How the model and their findings will contribute to the research community is not well evaluated, which falls below the standard of an original research paper. Additionally, the authors concluded that the isotope-glacio hydrological model didn't guarantee better separation results of the water shares and showed similar performance to a normal glacio-hydrological model. Even though the authors tried to emphasize that combining isotopes with glacio-hydrological modeling enhanced hydrologic parameter identifiability, this aspect was not adequately assessed in the results. Given the added model parameters and model configuration complexity, why an isotope-glacio hydrological model is evaluated in this study? What are the benefits of integrating isotopes into a well-verified glacio-hydrological model in the study area? Perhaps, the authors might consider adding a benchmark model for improved assessments on the influences (or benefits) of combining isotopes. Other comments are as follows:
1. As listed in lines 82-85, some processes were added in the model. However, the impacts of the additional processes on the model simulations were not well assessed. I would suggest adding a reference model in which these processes were not involved for the assessment.
2. Reference literature for many equations was missing. For example, Eqs. 3, 4, 5-7, 9, and 23.
3. Eq. 28, please add calculations for ϕRain and ϕRos.
4. The model calibration procedure should be better described. Is this a stepwise calibration or a simultaneously multiple-objective calibration? How were the competitions between the performance of discharge simulation, isotope simulation, and snow simulation dealt with in the calibration?
5. In Fig. 4, the coefficients of determination of the linear curves appear to be very low. Do the lines make any sense for the modeling? Was any trend-test conducted?
6. In Fig. 6, sub-daily fluctuations of daily discharge can be seen from the plots. Are these daily data (one point per day)?
7. In Section 5, please add discussions on the model's performance in simulating snow, ice, isotopes, and discharge, with respect to comparisons with peer research.
8. In Fig. 10, what are the reasons for the much higher isotopic composition of measured discharge than the pre-event? Is it due to evapotranspiration or contributions from other sources? From the plots, a very low fraction of pre-event can be observed in the discharge. If the fraction of pre-event is low, is this plot good evidence to illustrate the response of discharge to rain events?
9. In lines 703-709, as commented before, what are the benefits of involving isotopes for hydrological modeling in this study area?
10. In Section 5.4, why not give an estimate for the fractional contribution of groundwater to discharge? Without a contribution number, it is a very weak discussion on the role of groundwater.
11. Sensitivity of model parameters to the assumptions should be added. Also, please add a sensitivity analysis of model parameters to the model configuration of isotopes, as well as the influences of added processes on model parameter identifiability.
12. Perhaps the conclusions can be further improved by emphasizing their novel findings better. "(iii) assess how water isotopes can be used to estimate the proportions of different water sources at the outlet." This might need to be better interpreted.
13. In Table A1, why are there multiple temperature lapse rates?
Citation: https://doi.org/10.5194/egusphere-2024-631-RC1 -
AC1: 'Reply on RC1', Tom Müller, 26 Jul 2024
Comment on Egusphere-2024-631 - Anonymous Referee #1
General comments CommentThis work presents an interesting modeling effort in a Swiss glacierized catchment involving a water stable isotope module. Although the authors provided detailed information regarding the data, model structure, and sensitivity analysis, the novelty of this work needs to be further emphasized. Specifically, the authors should explain how the assessments in this study can inform hydrological modeling in glacierized basins.
In the introduction, the authors listed a couple of modeling challenges for glacierized basins, which is good. My major suggestion for the authors is to improve their results by providing more evidence on how the built isotope-glacio hydrological model or the assessments conducted in this study can help to address such modeling challenges.
Currently, the paper reads more like a description of the model and data. How the model and their findings will contribute to the research community is not well evaluated, which falls below the standard of an original research paper. Additionally, the authors concluded that the isotope-glacio hydrological model didn't guarantee better separation results of the water shares and showed similar performance to a normal glacio-hydrological model. Even though the authors tried to emphasize that combining isotopes with glacio-hydrological modeling enhanced hydrologic parameter identifiability, this aspect was not adequately assessed in the results. Given the added model parameters and model configuration complexity, why an isotope-glacio hydrological model is evaluated in this study?
What are the benefits of integrating isotopes into a well-verified glacio-hydrological model in the study area? Perhaps, the authors might consider adding a benchmark model for improved assessments on the influences (or benefits) of combining isotopes.
Answer
We thank the reviewer for their careful read of our manuscript and detailed comments. We understand that the reviewer’s main suggestion is to i) further emphasize the novelty of this work and ii) to emphasize the usefulness of the isotopic module to inform a glacio-hydrological model.
We agree that the paper is complex and addresses multiple topics, which may be difficult to follow for the reader. Although we developed a complex model, the focus of the paper is not directly on assessing the benefits of informing a glacio-hydrological model with isotopes, but rather to illustrate the complexity and evolution of the isotopic composition of different end-members in a glaciated catchment. This will be made clearer in the revised version.
Our main finding is that attempting to use isotopes to separate the shares of snow and ice in a glaciated catchment in the Alps is not advisable and that a “simple” mass-balance model may provide better estimates (see Fig. 11). Again, this will be made clear in the revised version.
Since we developed a complex model, we also attempted to address the benefits of the combined isotopic and glacio-hydrological model, but the improvement in terms of hourly discharge estimations is clearly limited. This is indeed not clearly illustrated in the paper, and we will provide more information. In particular, we will show how the hydrological transfer module and, hence, the modelled discharge and the model parameters are affected if the model is calibrated against discharge data only, or with discharge and isotopes data simultaneously. Nevertheless, as already attempted in Figure 10, we will show that isotopes may inform glacio-hydrological models: not so much in term of discharge volumes, but in terms of the model structure and in terms of how the model reacts to rain events for instance. This is useful to improve glacio-hydrological models focusing on the response to (extreme) rain events or how rain-on-snow affects the hydrological response.
Finally, we will emphasize the main reason why an isotopic module may not provide additional information for discharge simulation with glacio-hydrological models: because snow and ice isotopic ratios have similar values during a large part of the melt season. This finding is transferable to many similar alpine systems; and this finding is important: Even if many studies have reported very similar isotopic ratios of snow and ice, the idea that they can be used to improve hydrologic modelling in this context persists in the literature.
Furthermore, will better argue in the revised version that, in exchange, our hydrological and isotopic model may provide more useful information for snow-dominated catchments where ice is not present, as already proposed by Ala-Aho. For this reason, we detailed in the discussion the sensitivity of the isotopic module to different parameters. This discussion may be useful for readers focusing more on snow isotope modelling and also provide complementary analysis to the model proposed by Ala-Aho.
In short, the paper addresses 3 main topics: 1) the complexity of snow and ice isotopes in a glaciated catchment ; 2) the limits of the isotopic module compared with a well-informed glacio-hydrological model without isotopes but may provide some interesting insights in the response to rain events ; 3) the paper discusses limitations of the isotopic module, which may be useful for snow-dominated catchments and may contribute to the further development of such models.
In the new version of the manuscript, we will highlight those three main topics in the discussion more clearly so that the paper appears more readable and has a clearer structure.
Detailed comments Comment 1
As listed in lines 82-85, some processes were added in the model. However, the impacts of the additional processes on the model simulations were not well assessed. I would suggest adding a reference model in which these processes were not involved for the assessment.
Answer 1
Indeed, this was not clearly shown in the paper. We will add a section to compare model results when the model is calibrated against discharge only or using discharge and isotopes.
Comment 2
Reference literature for many equations was missing. For example, Eqs. 3, 4, 5-7, 9, and 23.
Answer 2
Equations 3 to 9 were developed for the model needs and are not referenced from previous literature. As they may appear difficult to “read”, we illustrated their behavior in Appendix (Figure D1). This will be more clearly described in the new manuscript.
Comment 3
Eq. 28, please add calculations for ϕRain and ϕRos.
Answer 3
This will be added.
Comment 4
The model calibration procedure should be better described. Is this a stepwise calibration or a simultaneously multiple-objective calibration? How were the competitions between the performance of discharge simulation, isotope simulation, and snow simulation dealt with in the calibration?
Answer 4
The model calibration is stepwise, each module is calibrated separately using multiple-objective functions. This is detailed in section 3.1.7, 3.2.4, 3.3.6. Our goal was not to develop a fully simultaneously calibrated model, but rather to assess the added benefits of the hydrological transfer and isotope module when estimating discharge volumes and snow and ice-melt fraction. For this reason, the calibration was performed separately for each module and there is, therefore, no competition or weights between the calibration of the 3 modules.
We understand this approach may appear unconventional, but we believe it better illustrates the benefits of the each module that way. We will make this clearer at the beginning of section 3.
Comment 5
In Fig. 4, the coefficients of determination of the linear curves appear to be very low. Do the lines make any sense for the modeling? Was any trend-test conducted?
Answer 5
Indeed, the lines are only here to show that no trend with elevation can be found here. We believe they help the reader to better visualize the results. These lines are not used in the modeling.
Comment 6
In Fig. 6, sub-daily fluctuations of daily discharge can be seen from the plots. Are these daily data (one point per day)?
Answer 6
No, the data are hourly. This is stated earlier in the manuscript. We will make it clearer in the figure legend.
Comment 7
In Section 5, please add discussions on the model's performance in simulating snow, ice, isotopes, and discharge, with respect to comparisons with peer research.
Answer 7
We agree we can provide a comparison for discharge. However again, the goal of the paper is not to create yet another glacio-hydrological model, but the model is rather used as an example to illustrate snow and ice-melt processes and to assess the use of isotopes in glaciated catchments. We will nonetheless try to better contextualize our results with peer research.
Comment 8
In Fig. 10, what are the reasons for the much higher isotopic composition of measured discharge than the pre-event? Is it due to evapotranspiration or contributions from other sources? From the plots, a very low fraction of pre-event can be observed in the discharge. If the fraction of pre-event is low, is this plot good evidence to illustrate the response of discharge to rain events?
Answer 8
The isotopic composition of measured discharge increases during rain events due to rain which has a much higher isotopic composition than the pre-event discharge (composed mainly of snow and ice-melt). During rain events, the isotopic composition of the discharge increases due to the rain mixing with snow and ice-melt. This plot shows that, especially during the early- to mid-melt season (July), the discharge is affected by the rain water during more than a week, which appears in the isotopic signal but is hardly visible in the discharge volumes.
We will attempt to indicate the isotopic composition of rain during those events in the figure and improve line labels and the text of the legend. We will also review the discussion to make our point clearer.
Comment 9
In lines 703-709, as commented before, what are the benefits of involving isotopes for hydrological modeling in this study area?
Answer 9
Here, as discussed in the answer to the general comment, we will provide a figure to compare the model results with and without involving isotopes in the calibration. We will also try to better separate the findings of this study in the discussion, which may confuse the reader as it is now.
Comment 10
In Section 5.4, why not give an estimate for the fractional contribution of groundwater to discharge? Without a contribution number, it is a very weak discussion on the role of groundwater.
Answer 10
Ok, we can provide a fractional estimate and compare it with the literature, thank you for the suggestion.
Comment 11
Sensitivity of model parameters to the assumptions should be added. Also, please add a sensitivity analysis of model parameters to the model configuration of isotopes, as well as the influences of added processes on model parameter identifiability.
Answer 11
Since the paper's goal is not to propose a “new” glacio-hydrological model, we do not want to further extend the manuscript by providing a full detailed sensitivity analysis of all parameters. We restrain the sensitivity analysis to the evolution of the snow isotopic composition, which is one of the paper's novelties and may be useful for further development of isotopic models. In addition, using isotopes to restrain model parameters would be interesting only if the model was calibrated in a simultaneous multiple-objective calibration using all data, which is not the case. An interesting paper by He et al. (2019) already provides such an approach, which is not our aim here.
Comment 12
Perhaps the conclusions can be further improved by emphasizing their novel findings better. "(iii) assess how water isotopes can be used to estimate the proportions of different water sources at the outlet." This might need to be better interpreted.
Answer 12
We will improve the conclusion after improving the paper structure with the main outcomes, as discussed previously.
Comment 13
In Table A1, why are there multiple temperature lapse rates?
Answer 13
The temperature lapse rate function is estimated using a “gaussian” like shape which needs 4 parameters. It is indeed not adequate that all variables have the same name, we will fix this.
Citation: https://doi.org/10.5194/egusphere-2024-631-AC1
-
AC1: 'Reply on RC1', Tom Müller, 26 Jul 2024
-
RC2: 'Comment on egusphere-2024-631', Anonymous Referee #2, 15 May 2024
The work presents glacio-hydrological model incorporating stable water isotope data in the simulations. The work is done at a glacierized catchment in the Swiss Apls, and uses a multitude of data sources for model calibration, which is performed in different steps. They found out that the time-variable snowmelt isotope values converge with ice melt isotope values late in the season, brining about the typical problem in isotope-based mixing model analysis in glacial catchments. However, the tracer-aided model can provide additional insight to the runoff generation processes and snow vs glacial melt contributions.
I find that the work is a interesting and ambitious case study leveraging of multiple data sources for model calibration. I liked the fresh approach of using transfer module approach, circumventing the need for explicit conceptual storages for soil and GW most often used in hydrological modeling. Though also that approach resulted in distribution parameters to calibrate, and the stepwise calibration approach was a bit difficult to follow. Because of the explicit model description and multitude of data sources, I found the paper fairly long, and suggest below some parts that could be considered leaving out. I recommend the authors address my comments below publishing the work.
L135: specify how measured? average snow depth reduction in the ablations stakes? Where the density 900 assumption comes from?
L161: Shaded location may not be enough, if average is 7, highs can get much more than that. Did you examine the potential evaporation effects? No paraffine oil of other fine tubes to prevent evaporation?
L209: where is this Fig. D1d?
L221: is the snow redistribution scheme based on some existing work? I find no refs here. Wind speed is not a factor?
L278: how were the other two field datasets used in the PEST inversion?
L302: how good is the regression goodness? Typically you would miss the extreme values, the lows in particular
L328: I’m not convinced by the assumption. Why would a think snowpack be less ripe by assumption? It can be equally isothermal and retain liquid water as a smaller pack. Do you have any data to back up this assumption? See my later comment on this
L333-338 :The rules seem a bit subjective. Why did you not compare your simulated i_sp with the depth (or SWE even better) averaged snow pit d2H? Or use the snowmelt samples you had? Instead of this more “soft” calibration.
L338 – L 345: this is very difficult to understand without seeing the timeseries data
L365: not sure I understand this, not familiar with the kinematic subsurface saturated flow concept: do you mean that all water delivery from the hillslope is darcian-type GW flow, with instantaneous recharge. No overland flow?
L447: do you propose that snow surface samples are representative of the snowfall precipitation over winter? That would be the logic behind the Beria et al statements, where the isotope variability in snowfall events over winter is higher than variability in snowmelt.
L452: Interesting data, sublimation can be causing this. There is pretty high variability in your snow surface d-excess values, considering your sampling strategy. If the surface snow is deposited from the same storm, and experiences the same atmospheric conditions after deposition, the d-excess values should be pretty similar. I recommend to discuss what pre and post depositional processes might cause this.
L453: to me it looks like there is a positive correlation between H2 and elevation, contrary to what one would expect.
L473: why did you not calibrate one parameter set for both years? This would be the typical approach where parameters are assumed invariant in time.
L484: can you justify how SWE simulations were deemed good? RMSE of ~100 mm w.e. seems pretty high? What was your catchment average SWE, so you could put the error to perspective
Fig 8a: add in the legend “snow melt composition” not to confuse with snowpack
Fig8: what is the physical process that would explain the snowmelt isotope values to become gradually more depleted at times your 2021 simulations?
L534: this seems counter intuitive, that a bigger snowpack would mix less with rain, ie. deliver more rain water throught the snowpack, that a smaller pack. I think liquid water retained in the snowpack would be important to consider here. Snowpack with 2000mm w.e. is massive, and conceptually difficult to see that storms of your magnitude (mostly 10 mm/day) would seep through the snowpack. One modeling Rule of thumb for snow water retention is 5% of w.e., which in your snowpack would be 100m of water. Is there any other explanation why you get less enrichment in thick snowpacks?
L565-575: good discussion about the uncertainties in sublimation.
L588: including the objective functions, and in particular how the snow extent maps were numerically compared with simulations, would be an important addition to the methodology.
Chapter 5.3 This is interesting discussion, but for the benefit of shortening the paper, could be left out without compromising the main findings in the paper.
L695-702: the degrees of freedom caused by the model complexity are apparent in this discussion. The stream isotope response can be explained by many independent processes. This is not a critical comment as such, but you system is very complex to conceptualize even with isotope data and models.
L704: in the objective function?
L751: what do you mean by outlet snowmelt d2H? The stream value, of d2H leaving the snowpack (the meltwater)?
L791: also the model would be good to validate with either snowpack or snowmelt samples.
L801: not clear why you did not collect the bulk snowpack samples in this case: they are easier to get as a by product of SWE measurements than a snow pit profile.
Citation: https://doi.org/10.5194/egusphere-2024-631-RC2 -
AC2: 'Reply on RC2', Tom Müller, 26 Jul 2024
Comment on Egusphere-2024-631 - Anonymous Referee #2
General comments Comment
The work presents glacio-hydrological model incorporating stable water isotope data in the simulations. The work is done at a glacierized catchment in the Swiss Apls, and uses a multitude of data sources for model calibration, which is performed in different steps. They found out that the time-variable snowmelt isotope values converge with ice melt isotope values late in the season, bringing about the typical problem in isotope-based mixing model analysis in glacial catchments. However, the tracer-aided model can provide additional insight to the runoff generation processes and snow vs glacial melt contributions.
I find that the work is a interesting and ambitious case study leveraging of multiple data sources for model calibration. I liked the fresh approach of using transfer module approach, circumventing the need for explicit conceptual storages for soil and GW most often used in hydrological modeling. Though also that approach resulted in distribution parameters to calibrate, and the stepwise calibration approach was a bit difficult to follow. Because of the explicit model description and multitude of data sources, I found the paper fairly long, and suggest below some parts that could be considered leaving out. I recommend the authors address my comments below publishing the work.
Answer
We thank the reviewer for their detailed comments, in particular regarding all issues linked to the isotope modelling framework, which nicely complement the comments of the reviewer 1. The paper is indeed relatively long and the approach may appear rather unconventional. Our aim was not directly to propose (yet) another model, but rather to use a detailed model to illustrate how water isotopes evolve in a glaciated catchment. We also discuss where they may be more useful. As also discussed in response to review 1, the paper addresses 3 main topics which will be better separated in the new version of the manuscript:
1) the complexity of snow and ice isotopes in a glaciated catchment; 2) the limits of the isotopic module compared with a well-informed glacio-hydrological model without isotopes but may provide some interesting insights in the response to rain events ; 3) the paper discusses limitations of the isotopic module which may be useful for snow-dominated catchments and may contribute to the further development of such models.
We will improve the structure and length of the paper to make it more readable. We will also address all detailed comments of the reviewer as follows below.
The explicit description of the model (in the methodology) is indeed long but is required for a detailed understanding of the paper. We will try to encourage the readers to skip this part if they are not interested in the model details, and we will reference some parts of the methodology in the results and discussion.
Detailed comments Comment 1
L135: specify how measured? average snow depth reduction in the ablations stakes? Where the density 900 assumption comes from?
Answer 1
We mean here that we measured at each ablation stake the decrease in elevation in centimeters, 3 and 7 times during the season in 2020 and 2021 respectively. The data at each stake are then used for the model calibration without averaging. We will make this phrase clearer. The ice density of 900 comes from common literature and will be provided here.
Comment 2
L161: Shaded location may not be enough, if average is 7, highs can get much more than that. Did you examine the potential evaporation effects? No paraffine oil of other fine tubes to prevent evaporation?
Answer 2
No parafine was used here. Our results also showed no deviation of the samples from the local meteoric water line which indicates limited evaporative losses. In addition, von Freyberg et al. (2020) showed a deuterium difference of about 1‰ for a two to three weeks period at such temperatures. We therefore believe that no significant evaporation should have occurred.
Comment 3
L209: where is this Fig. D1d?
Answer 3
Not sure about this question. It is at the end of the paper in Appendix.
Comment 4
L221: is the snow redistribution scheme based on some existing work? I find no refs here. Wind speed is not a factor?
Answer 4
No, we made our own function, which is a very rough but simple approach but which conserves mass over the catchment.
Comment 5
L278: how were the other two field datasets used in the PEST inversion?
Answer 5
The three dataset were calibrated simultaneously using a multi-objective function. We will make this clearer and provide the weights of each function.
Comment 6
L302: how good is the regression goodness? Typically, you would miss the extreme values, the lows in particular
Answer 6
This is provided in the results section 4.4 and Figure 7. Indeed, the low values have a relatively strong influence on the regression which was also shown in the sensitivity analysis in Figure 9 (red area).
Comment 7
L328: I’m not convinced by the assumption. Why would a think snowpack be less ripe by assumption? It can be equally isothermal and retain liquid water as a smaller pack. Do you have any data to back up this assumption? See my later comment on this
Answer 7
We mostly answered this essential comment in the comments 17 and 18 below.
Here, we would like to stress that we are talking about the mixing of the isotopic signal and not directly of the water retention of the snow. This assumption was based on the work of Juras et al (2019) who showed that rainwater flowing through a thicker, colder, snowpack was isotopically less affected by the snowpack isotopic composition than water flowing through a shallower, ripe, snowpack (meaning there is less mass exchange between rain and snow in this case).
There is, indeed, no clear evidence that a thicker snowpack is less ripe. We can only say that a 2000 mm snowpack only occurs at high elevations where temperature and melt remain very low and where ripe conditions are unlikely in summer. Finally, the physical basis of this function is arguable for sure, but the calibration of this function led to satisfying results and this formulation was therefore retained. It has the advantage of being simple, while estimating the snow energy state hourly at each grid cell seemed too complex.
We attempted to discuss this essential topic in subsection 5.2.3. We will improve those explanations in the revised manuscript, and also avoid any confusion regarding the mixing the isotopic signal (and not the release of water).
Comment 8
L333-338 : The rules seem a bit subjective. Why did you not compare your simulated i_sp with the depth (or SWE even better) averaged snow pit d2H? Or use the snowmelt samples you had? Instead of this more “soft” calibration.
Answer 8
We unfortunately did not have enough data for a proper calibration against the snow pit or snowmelt measurements. This would have been indeed more straightforward, but those data are hard to acquire, and we lacked data during the whole summer. We nonetheless verified that our calibration fitted the average value of the snow-pits. Indeed, the modelled snowpack δ2H at the snowpits were around -119‰ for both pits compared to two mean measured snowpits δ2H of -125‰and -117‰.
In addition, this “soft” approach also provides some advantages which may be further investigated in future studies. It does not require snowpit samples, which are difficult to acquire and may vary significantly due to local snow conditions. By using stream samples, we can much more easily obtain a catchment-averaged value of snowmelt and we can acquire data during the whole season and especially during the end of the melt season when snow samples are almost impossible to acquire at high elevation. We will add this comment in the results. The lack of data for calibration is already discussed in the discussion where we also encourage to validate this approach with depth-averaged snowpit data.
Comment 9
L338 – L 345: this is very difficult to understand without seeing the timeseries data
Answer 9
Indeed. We were reluctant to provide data results in the methodology. We will improve this in the new manuscript, possibly adding a simplified sketch of the time series.
Comment 10
L365: not sure I understand this, not familiar with the kinematic subsurface saturated flow concept: do you mean that all water delivery from the hillslope is Darcian-type GW flow, with instantaneous recharge. No overland flow?
Answer 10
Yes, we assume that all water infiltrates in the sediments, which are very coarse. However, the groundwater flow in those sediments is fast due the slope and high hydraulic conductivity. Even if no overland flow occurs, water is rapidly conveyed downslope. No overland flow can be observed after rain events, only tributaries coming from lateral glaciers can be observed. This was observed and discussed in previous research (Müller et al., 2022). We will make a reference to this.
Comment 11
L447: do you propose that snow surface samples are representative of the snowfall precipitation over winter? That would be the logic behind the Beria et al statements, where the isotope variability in snowfall events over winter is higher than variability in snowmelt.
Answer 11
At this stage in the results section, we only comment our results. But snow surface samples do represent at least one “layer” of snow which accumulated during the winter. It is, however, not homogenous as snow melts faster in the lower part of the catchment. Also, due to snow sublimation and vapor deposition, vapor exchanges within the snowpack occur. As a result, surface snow samples may have deviated from the composition of the snow event. Yes, surface snow samples represent more or less the variability of the composition of snow events in winter, which then homogenizes due to vapor exchanges and during melt. This is to our understanding also what is discussed in Beria et al. We will provide a brief explanation in the new manuscript.
Comment 12
L452: Interesting data, sublimation can be causing this. There is pretty high variability in your snow surface d-excess values, considering your sampling strategy. If the surface snow is deposited from the same storm, and experiences the same atmospheric conditions after deposition, the d-excess values should be pretty similar. I recommend to discuss what pre and post depositional processes might cause this.
Answer 12
Indeed d-excess appears variable and not trend can be observed. This is probably because the snow at different elevation represent different snowfall events, and the rate of sublimation/deposition may vary spatially. We briefly tackle this process in the discussion (L754) and also direct the reader to more focused research on the topic (e.g. Sprenger et al. 2024 https://doi.org/10.5194/hess-28-1711-2024). We will add a sentence in the revised document to discuss this shortly more.
Comment 13
L453: to me it looks like there is a positive correlation between H2 and elevation, contrary to what one would expect.
Answer 13
As the coefficient of correlation is very low (R2 = 0.26) for dH2 of snow surface and elevation, we find it difficult to confirm a clear trend, also for the reasons discussed above. It is also contrary to what one would expect, as precipitation dH2 typically decreases with elevation (see Beria et al). One explanation could be that at high elevations, surface snow represents a snowfall event from the end of the winter period, which typically has a higher temperature and thus dH2 value, while, at lower elevations, the end of winter snowfall has already melted and the surface snow actually represents snowfall from the mid-winter season, which has a typically lower dH2 value. But this cannot be confirmed with our data.
Comment 14
L473: why did you not calibrate one parameter set for both years? This would be the typical approach where parameters are assumed invariant in time.
Answer 14
Since both winters and summers showed different weather conditions (one drier and one more humid summer for instance), model parameters may vary as a simple temperature-index model cannot account for such conditions. In addition, here the goal was to obtain a precise estimation of the isotopic signal for each year of the study and not to calibrate the model for projection in the future. As such, a yearly calibration leads to better results.
Comment 15
L484: can you justify how SWE simulations were deemed good? RMSE of ~100 mm w.e. seems pretty high? What was your catchment average SWE, so you could put the error to perspective
Answer 15
Ok, we will put this number in the perspective of other research and compare it to the total snow melt of each year (1860 and 1527 mm). Note that the RMSE is ~100mm, but the mean error is less than 10 mm. The RMSE error seems to be mainly due to the local conditions at each snow depth measurement point. Also, our model remains simple, with no spatially resolved snow redistribution process, which probably results in a smoother SWE accumulation than in reality. Also note that our model needs to correctly simulate snow on the steep hillslopes which, as a tradeoff, may also increase the model error on the glacier, where snow depth data were obtained. This will be added to the manuscript.
Comment 16
Fig 8a: add in the legend “snow melt composition” not to confuse with snowpack
Answer 16
Yes indeed !
Comment 17
Fig8: what is the physical process that would explain the snowmelt isotope values to become gradually more depleted at times your 2021 simulations?
Answer 17
This comment also relates to your previous comment #7 about our function to mix ROS with the snowpack.
We believe that the increase of snowpack and snowmelt isotopic ratios in summer is due to rain-on-snow (ROS), which partially integrates/mixes with the snow. Juras et al (2017), in particular, also showed, using an artificial setup, that snowpack samples showed a considerable increase in dH2 ratios after a rain event (a sprinkling event). This suggests that mass exchange between the snow and rain occurs, therefore increasing the dH2 composition of the snowpack. This observation is an essential and a highly difficult process to simulate with a simple model (without simulating the snow internal temperature for instance). For this reason, we dedicated a subsection (5.2.3) to this process. We agree that the model may need further development, but we at least showed and discussed the importance of ROS to simulate snow isotopic composition. Finally, we want to stress that the increase in snowmelt isotopic ratios during the summer may be due to both ROS and snow fractionation. In our results, it seems that ROS influences snow composition more than fractionation, especially when the snowpack is shallow. This is shown in the sensitivity analysis in Figure 9.
Comment 18
L534: this seems counter intuitive, that a bigger snowpack would mix less with rain, ie. deliver more rain water throught the snowpack, that a smaller pack. I think liquid water retained in the snowpack would be important to consider here. Snowpack with 2000mm w.e. is massive, and conceptually difficult to see that storms of your magnitude (mostly 10 mm/day) would seep through the snowpack. One modeling Rule of thumb for snow water retention is 5% of w.e., which in your snowpack would be 100m of water. Is there any other explanation why you get less enrichment in thick snowpacks?
Answer 18
Please also refer back to our answers to comment 7 and 18.
Yes, this is definitely a weakness of the model, which we borrowed from Ala-Aho. Here, they assume that the isotopic signal of the rain is completely integrated in the snowpack (and the rain water is then released with a composition similar to snow). We tried to improve this simplification in our model with a function that only partially mixes the isotopic signal of the rain with the snow.
We want to make some aspects clearer. Here we talk only about how the rain isotopic composition mixes within the snowpack through mass exchange processes. In our model, snow water retention is not clearly modelled, and all rain passing through the snowpack is released (but the release rate depends on a transfer function which is a function of snow depth). However, while traveling through the snowpack, the isotopic composition of the rain water will partially mix with the snowpack. Juras et al. showed that in a cold (-1°C) but thicker snowpack (500 mm), the rain isotopic composition seeping through the snowpack retains its original composition much more than in a shallower isothermal snowpack. They attribute this effect to the formation of preferential flow in the non-ripe, cold, snowpack, leading to less mixing. Our function attempted to mimic this effect in a very simple way, using SWE as an indicator of snow ripeness. This function is debatable, but SWE was (and is in other studies) readily available, while estimating the snow energy state is complex. We discuss this essential topic in subsection 5.2.3, which we will improve. We will, in particular, avoid any confusion regarding the mixing the isotopic signal (and not the release of water).
Comment 19
L565-575: good discussion about the uncertainties in sublimation.
Answer 19
Thank you !
Comment 20
L588: including the objective functions, and in particular how the snow extent maps were numerically compared with simulations, would be an important addition to the methodology.
Answer 20
This is rapidly described in section 3.1.7 Calibration. Ok we will improve this in the methodology.
Comment 21
Chapter 5.3 This is interesting discussion, but for the benefit of shortening the paper, could be left out without compromising the main findings in the paper.
Answer 21
Ok, thank you for your comment. The paper tackled essentially two topics: to illustrate snow and icemelt isotopes processes in a glaciated catchment on one hand; and how isotopes can inform a glacio-hydrological model. This discussion mainly inform the latter. But as discussed with reviewer 1, we will rearrange (and shorten) those two topics for clarity.
Comment 22
L695-702: the degrees of freedom caused by the model complexity are apparent in this discussion. The stream isotope response can be explained by many independent processes. This is not a critical comment as such, but you system is very complex to conceptualize even with isotope data and models.
Answer 22
Yes indeed, and this is why we provided figure 9 in particular. We will try to improve to some extent this by also comparing the model results with or without istopic data calibration.
Comment 23
L704: in the objective function?
Answer 23
Yes, we will be improved as discussed in the previous comment.
Comment 24
L751: what do you mean by outlet snowmelt d2H? The stream value, of d2H leaving the snowpack (the meltwater)?
Answer 24
The stream value yes !
Comment 25
L791: also the model would be good to validate with either snowpack or snowmelt samples.
Answer 25
Yes, we will provide this comparison with the data available (see also answer 8).
Comment 26
L801: not clear why you did not collect the bulk snowpack samples in this case: they are easier to get as a byproduct of SWE measurements than a snow pit profile.
Answer 26
In our approach, we needed the isotopic composition of precipitation to estimate the snowpack isotopic composition spatially and during the whole season. I am not sure if we could obtain the same seasonal modelling of snow isotopes based on snowpack data, which becomes hard to acquire during the late season. The model may be calibrated with snowpack samples only, but this approach may be as much (or more) time consuming, and we did not acquire many snowpit data unfortunately (see also Answer 8). Ideally, using both precipitation samples and snowpack data is clearly advisable. We will add a comment on this in the paper.
Citation: https://doi.org/10.5194/egusphere-2024-631-AC2
-
AC2: 'Reply on RC2', Tom Müller, 26 Jul 2024
-
RC3: 'Comment on egusphere-2024-631', Anonymous Referee #3, 17 Jun 2024
General comment
This manuscript reports an interesting combination of a glacio-hydrological and isotopic model to estimate the seasonal shares of snowmelt and ice melt in the stream waters of a high-elevation catchment in the Swiss Alps. The authors well described their dataset, the model with the various modules and the sensitivity analysis, and provided some useful recommendations for future studies including the collection of samples for isotopic analysis from different water sources in glacierized catchments. Overall, I think this is a valuable research paper that deserves to be published. However, some modifications are needed (please see the specific comments) before the acceptance of this paper. Among these changes, in agreement with reviewer 1, I think the authors should better emphasize the advantages of integrating isotopic observations and an isotopic model into the glacio-hydrological model that was used. Indeed, some results and the discussion highlight more the challenges of using the isotopic tracers than the usefulness of their integration into the glacio-hydrological model.
Furthermore, since the manuscript is quite long, I recommend to the authors to better emphasize the novelty of the manuscript (compared to other research papers) in the abstract and the conclusions, to shorten some paragraphs in the conclusions and better highlight the key findings.
Specific comments
- Lines 50-53: These basic sentences about fractionation can be skipped because they are not meaningful for the introduction. Given the topic of the manuscript, if the authors want to define isotopic fractionation, I think they should provide an example regarding the snowpack instead of vapour masses and precipitation.
- Equation 1: The sentence at Lines 54-55 is enough and does not require Equation 1.
- Lines 288-289: Is there any consideration based on the uncertainty in the isotopic analysis or is it just a simple preference for δ2H?
- Lines 443-444: ‘likely due to the preferential elution of solutes in the snowpack (Costa et al., 2020)’ belongs to the discussion.
- Line 447: ‘As suggested in other studies (Beria et al., 2018)…’ belongs to the discussion.
- Lines 450-452: This sentence also belongs to the discussion.
- Lines 461-462: This sentence also belongs to the discussion.
- Lines 497-499: This sentence also belongs to the discussion.
- Section 5.2.1 and Figure C2: It is interesting to note that ice melt had a relatively small spatio-temporal variability, despite the larger variability observed in ice surface samples. This is in agreement with the isotopic composition of ice and meltwater samples collected over a glacier surface in the Italian Alps (Zuecco et al., 2019). In Figure C2, I wonder whether the first ice melt samples collected in July 2019 were affected by mixing with snowmelt or recent rain water.
- Section 5.6 and Conclusions: By reading these two sections, I wonder whether stables isotopes of hydrogen and oxygen represent a real added value for the model application and for improving our understanding of hydrological processes in glacierized catchments. It looks like that the huge effort and the still-present challenges make the application of isotopes not that appealing compared to other tracers (e.g., major ions, trace elements, other isotopes, artificial tracers) that could better help discriminating the end members in stream runoff.
- Lines 787-820: Given the length of the manuscript and of the conclusions (quite long), I suggest organizing this text using bullet points and reducing the paragraph starting at Line 799. Bullet points and a shorter text should help the reader to understand the novelty and the take home messages of this manuscript.
Technical corrections
- Line 145: ‘Snow profiles for isotopic analysis’ instead of ‘isotopic snow profiles’. Please change the term at Line 147, as well.
- Line 245: It should be Walter et al. (2005) instead of Todd Walter et al. (2005).
- Line 265: It should be Walter et al. (2005).
- Line 288: Please remove ‘water’ before ‘stable’.
- Line 334: Please replace ‘remain below’ with ‘more depleted than’.
Citation: https://doi.org/10.5194/egusphere-2024-631-RC3 -
AC3: 'Reply on RC3', Tom Müller, 26 Jul 2024
Comment on Egusphere-2024-631 - Anonymous Referee #3
General comments Comment
This manuscript reports an interesting combination of a glacio-hydrological and isotopic model to estimate the seasonal shares of snowmelt and ice melt in the stream waters of a high-elevation catchment in the Swiss Alps. The authors well described their dataset, the model with the various modules and the sensitivity analysis, and provided some useful recommendations for future studies including the collection of samples for isotopic analysis from different water sources in glacierized catchments. Overall, I think this is a valuable research paper that deserves to be published. However, some modifications are needed (please see the specific comments) before the acceptance of this paper. Among these changes, in agreement with reviewer 1, I think the authors should better emphasize the advantages of integrating isotopic observations and an isotopic model into the glacio-hydrological model that was used. Indeed, some results and the discussion highlight more the challenges of using the isotopic tracers than the usefulness of their integration into the glacio-hydrological model.
Furthermore, since the manuscript is quite long, I recommend to the authors to better emphasize the novelty of the manuscript (compared to other research papers) in the abstract and the conclusions, to shorten some paragraphs in the conclusions and better highlight the key findings.
Answer
We thank the reviewer for their careful reading of our manuscript and detailed comments, and especially their suggestion to better structure and shorten the manuscript. As also answered to the reviewers 1 and 2, the focus of the paper was not directly to assess the benefits of informing a glacio-hydrological model with isotopes, but rather to illustrate the complexity and evolution of the isotopic composition of different end-members in a glaciated catchment. In the new version of the manuscript we will try to better separate the main messages of the paper with a clearer structure, especially for the discussion part. In particular, the paper attempts to cover 3 main topics: 1) the complexity of snow and ice isotopes and their evolution through a melting season in a glaciated catchment; 2) the limits of the isotopic module compared with a well-informed glacio-hydrological model without isotopes but may provide some interesting insights in the response to rain events ; 3) the paper discusses limitations of the isotopic module which may be useful for snow-dominated catchments and may contribute to the further development of such models.
As suggested by this reviewer, we will emphasize this in the abstract, introduction and conclusion to better highlight the key inputs of the paper. We will also address this reviewer’s specific comments as discussed below.
Specific comments Comment 1
Lines 50-53: These basic sentences about fractionation can be skipped because they are not meaningful for the introduction. Given the topic of the manuscript, if the authors want to define isotopic fractionation, I think they should provide an example regarding the snowpack instead of vapour masses and precipitation.
Answer 1
The target readers of the paper may include glacio-hydrological modelers who may not be fully familiar with isotopes; we, thus, believe a short theoretical background on isotopes is needed, and the reason why snow is more depleted in heavy isotopes than rain. We will improve this with an example more linked to the snowpack. We will move the theoretical information about isotopes in the “experimental method” part of the document to better structure the manuscript.
Comment 2
Equation 1: The sentence at Lines 54-55 is enough and does not require Equation 1.
Answer 2
Ok, this can be skipped !
Comment 3
Lines 288-289: Is there any consideration based on the uncertainty in the isotopic analysis or is it just a simple preference for δ2H?
Answer 3
The analytical error for both d2H and d18O is very limited. However, slight evaporative fractionation may have occurred for samples acquired using the automatic sampler in the river, which samples were not protected against evaporation using oil for instance. For those samples, evaporative fractionation would have influenced more the value of d18O than d2H. For this reason, too, d2H was retained. We think this information is too detailed for the paper and can be skipped.
Comment 4
Lines 443-444: ‘likely due to the preferential elution of solutes in the snowpack (Costa et al., 2020)’ belongs to the discussion.
Answer 4
Ok we agree. This sentence may not be important and will be removed from the results.
Comment(s) 5
Line 447: ‘As suggested in other studies (Beria et al., 2018)…’ belongs to the discussion.
Lines 450-452: This sentence also belongs to the discussion.
Lines 461-462: This sentence also belongs to the discussion.
Answer 5
Ok, we will move the theoretical information about isotopes to the “experimental method” part of the document to better structure the manuscript. We will revise this section to focus on the results only.
Comment 6
Lines 497-499: This sentence also belongs to the discussion.
Answer 6
Ok, this will be moved to the mass-balance discussion section.
Comment 7
Section 5.2.1 and Figure C2: It is interesting to note that ice melt had a relatively small spatio-temporal variability, despite the larger variability observed in ice surface samples. This is in agreement with the isotopic composition of ice and meltwater samples collected over a glacier surface in the Italian Alps (Zuecco et al., 2019). In Figure C2, I wonder whether the first ice melt samples collected in July 2019 were affected by mixing with snowmelt or recent rain water.
Answer 7
Thank you for the reference which can be mentioned here too. We will double check the location of the sample having a value of -134 ‰ in our database, and provide a comment about this “outlier”.
Comment 8
Section 5.6 and Conclusions: By reading these two sections, I wonder whether stables isotopes of hydrogen and oxygen represent a real added value for the model application and for improving our understanding of hydrological processes in glacierized catchments. It looks like that the huge effort and the still-present challenges make the application of isotopes not that appealing compared to other tracers (e.g., major ions, trace elements, other isotopes, artificial tracers) that could better help discriminating the end members in stream runoff.
Answer 8
Yes. This is one of the messages and outcomes of the paper: water isotopes of snow in a glaciated catchment may vary strongly spatially and temporarily, but even more importantly, may completely overlap the ice composition, at least in temperate glaciers in the European Alps. As a result, snow isotopes application should be considered with care and not be confused with ice in mixing models. They may still be useful to analyze the transfer of rain through such a system, as discussed in Figure 10. We agree that other (conservative) tracers may be needed depending on the application. As mentioned in the general comment, we will try to better structure and bring out this message in the discussion and conclusion.
Comment 9
Lines 787-820: Given the length of the manuscript and of the conclusions (quite long), I suggest organizing this text using bullet points and reducing the paragraph starting at Line 799. Bullet points and a shorter text should help the reader to understand the novelty and the take home messages of this manuscript.
Answer 9
Thank you, yes we will try to shorten the conclusion, focusing on the may outcomes of the paper in a more straightforward way.
Technical corrections Comments
- Line 145: ‘Snow profiles for isotopic analysis’ instead of ‘isotopic snow profiles’. Please change the term at Line 147, as well.
- Line 245: It should be Walter et al. (2005) instead of Todd Walter et al. (2005).
- Line 265: It should be Walter et al. (2005).
- Line 288: Please remove ‘water’ before ‘stable’.
- Line 334: Please replace ‘remain below’ with ‘more depleted than’.
Answer
We thank the reviewer for these technical corrections which will all be addressed in the new version of the manuscript.
Citation: https://doi.org/10.5194/egusphere-2024-631-AC3
Data sets
Water stable isotope, temperature and electrical conductivity dataset (snow, ice, rain, surface water, groundwater) from a high alpine catchment (2019-2021). Tom Müller https://doi.org/10.5281/zenodo.7529792
Stream discharge, stage, electrical conductivity & temperature dataset from Otemma glacier forefield, Switzerland (from July 2019 to October 2021) T. Müller and F. Miesen https://doi.org/10.5281/zenodo.6202732
Weather dataset from Otemma glacier forefield, Switzerland (from 14 July 2019 to 18 November 2021) Tom Müller https://doi.org/10.5281/zenodo.6106778
Model code and software
Combined isotopic and glacio-hydrological model developped for the Otemma glacierized catchment. T. Müller https://doi.org/10.5281/zenodo.10736126
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
340 | 150 | 35 | 525 | 15 | 15 |
- HTML: 340
- PDF: 150
- XML: 35
- Total: 525
- BibTeX: 15
- EndNote: 15
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1