the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Exploring the Use of Multi-source High-Resolution Satellite Data for Snow Water Equivalent Reconstruction over Mountainous Catchments
Valentina Premier
Carlo Marin
Giacomo Bertoldi
Riccardo Barella
Claudia Notarnicola
Lorenzo Bruzzone
Abstract. Seasonal snow accumulation and release are so crucial for the hydrological cycle to the point that mountains have been claimed as the "water towers" of the world. A key variable in this sense is the snow water equivalent (SWE). However, the complex accumulation and snow redistribution processes render its quantification and prediction very challenging. In this work, we explore the use of multi-source data to reconstruct SWE at a high spatial resolution (HR) of 25 m by proposing a novel approach designed for mountainous catchments. To this purpose, we exploit i) daily HR time-series of snow cover area (SCA) derived by high- and low-resolution optical images to define the days of snow presence, ii) a degree-day model driven by in-situ temperature to determine the potential melting, and iii) in-situ snow depth and Synthetic Aperture Radar (SAR) images to determine the state of the catchment (i.e., accumulation or ablation) that is needed to add or remove SWE to the reconstruction. Given the typical high spatial heterogeneity of snow in mountainous areas, HR data sample more adequately its distribution thus resulting in a highly detailed spatialized information that represents an important novelty. The proposed SWE reconstruction approach also foresees a novel SCA time-series regularization from impossible transitions. Moreover it reconstructs SWE for all the hydrological season without the need of spatialized precipitation information as input, that is usually affected by uncertainty. Despite the simple approach based on a set of empirical assumptions, it shows good performances when tested in two different catchments: the South Fork catchment, California, and the Schnals catchment, Italy, showing a good agreement with an average bias of -40 mm when evaluated against a HR spatialized reference product and of 38 mm when evaluated against manual measurements. The main sources of error introduced by each step of the method have been finally discussed to provide insights about the applicability and future improvements of the method that may be of great interest for several hydrological and ecological applications.
Valentina Premier et al.
Status: closed
-
CC1: 'Comment on tc-2022-146', Pau Wiersma, 29 Aug 2022
Dear ms Premier,
Congratulations on this great work, I believe it is an important contribution to the field of SWE modeling. I find your solutions very original and at the same time intuitive and straightforward. The manuscript is very readable and detailed, most of the questions that came up while reading were answered later on. Having said that, I do have a number of comments, which I have gathered in the attached document. I hope they are clear and that they will be useful to you, and my apologies if I interpreted the manuscript incorrectly.
Kind regards,
Pau Wiersma
- AC1: 'Reply on CC1', Valentina Premier, 13 Sep 2022
-
RC1: 'Comment on tc-2022-146', Anonymous Referee #1, 13 Sep 2022
I would like to congratulate the authors to this very interesing work, as they have presented a novel methodology and useful contribution to snow science. The presented approach might open new doors for reconstructing snow water resources also on larger scales, by employing multiple data sources. Especially, the usage of SAR backscattering signals to both, detect snow ablation, and indirectly correct daily high resolution SCA maps is an original methodology. The simple, yet efficient basic concept of the reconstruction approach, together with the parsimonious use of in-situ data is highly appealing. In my opinion, the organization of the work is good and the presentation of the results is rather clear. However, many sentences could be restructured to make for an easier read. I have made various comments on the manuscript that I hope will help to improve the paper.
There are still some important details that remain unclear to me, especially concerting the determination of the catchment state:
- It would be helpful to see a more detailed description about how the ablation state is derived from Sentinel 1 data and how this would translate to the three snowpack phases described in Marin et al. (2020) (i.e. moistening, ripening, runoff). In line 185 only a “relevant drop” in backscattering is mentioned. Does this mean the catchment state ablation already starts when any liquid water is present in the snowpack? Since it seems possible (at least with S1A and S1B) to identify the snowpack runoff phase and a separation to previous moistening and ripening phases, the use of a DD-style melt model would be much more justified - as this is practically eliminating the need to track energy states (cold content) in the snowpack modelling. Please clarify if ablation is based on a drop in backscattering (melting phase) like line 185 suggests, or on the minimum of backscattering (runoff phase), as implied line 174. If the latter is the case, there is more physical grounds to employ a DD-style melt model and this should be brought forward in the text. However, if the ablation state is corresbonding to the moistening phase (i.e. a mere drop in backscattering), this decission should be also explained in more detail.
- If I understood correctly, the catchment states ablation and equilibrium can exist both at the same time-step (different pixels have different states), but in accumulation all pixels have this state. Line 104 and 189 contradict themselves in this regard. Please clarify.
- Although bringing the reconstructed SWE time-series in context to the catchment discharge might provide some insights into the estimated snow cover dynamics, however, the way this information is presented in Figure 13 and interpreted in the text (line 409 -411) is not suitable for this purpose. There are basically two statements emerging from this analysis: i) there was more snow in one year than the other, ii) SWE decreases and subsequently discharge increases at some point. As it can be seen in Figure 13 there are very different discharge responses in the spring freshet between the two years. I do not see this analysis to be much helpful in the current state and would advise either to remove it from the manuscript, or expand the analysis (giving more information about precipitation, hydrological characteristics of the catchment, and changing the units in the figure (e.g. to mm)).
- The authors discuss various sources of uncertainty in the methodology and state that due to a number of preprocessing steps a formal sensitivity analysis is difficult to perform. However, it might be still very valuable for the reader to get a feeling about how possible errors might translate to the SWE reconstruction. Please consider the possibility to provide a simplified version of part of the problem, by e.g. perturbing the values of tSD and tSA (and perhaps keeping the states constant during this time) and showing the consequences in terms of peak SWE. This could also help to underline the statements in line 480f.
Additional Comments
Abstract/1 Introduction:
- 1 reconsider recasting the first sentence.
- 3 how about ablation processes, or are you specifically talking about peak SWE?
- 11 At this point, the reader might not follow what you mean with “time-series regularization from impossible transitions”. Mentioned again at 105, 111, and 127 before finally explained in section 2.2 at 194.
- 17 I’m not particularly fond of using present perfect (throughout the manuscript). But that might be personal preference. Please consider using present tense (or simple past) consistently.
- 21 not only on local hydrology, as many regions of the world rely e.g. on the spring freshet hundreds of km downstream
- 23 „from at least 50%“please check again with Vivironi et al (2003). Although snowmelt is a major contributor of mountain water resources, as far as I know, the numbers given by Vivironi and others include mountain waters in general.
- 27 precipitation variability is affected by orography, interpolation is affects by sampling density among other factors, and observations can be erroneous due to e.g. undercatch
- 65 “accumulation and melt = „accumulation and ablation“ (i.e. in this sense ablation includes erosion and evaporation etc.)
- 66 consider topography vs geomorphology (throughout the text)
- 80 they range from DD to complete energy balance models (as used in Bair et al 2016), and these do not require calibration
- 114 (Italy) like (USA)
2 Proposed approach to HR SWE reconstruction
- 120 first sentence is obsolete iMo.
- 132 please specify „too vast“.
- 136 Maybe this would be clearer: „in detail, the catchment state is characterized by the change in SWE, but is also associated with possible changes in SCA“. Or similar.
- 141 „extension“ = extent
- 143 „…dSWE < 0 due to melt water drainage“.
- 145 snow depth, not height. Anyway, better to talk here in terms of mass/SWE. e.g. „..if the snowpack is melting only partially“.
- 167-168 maybe better placed in the discussion section
- 169 potentiality = potential to detect the presence of a melting snowpack…
- 174 most important in terms of what? Certainly not in terms of melt water production as melt rates increase towards later season. Peak SWE is not necessarily the peak of melt water runoff. Please clarify.
- 181 replace “quotes” with “elevations”, “elevation bands” or similar (throughout the manuscript)
- 189 ablation and equilibrium classes can exist at the same time, but 104 states that the state is assumed homogeneous for all pixels of the catchment. Please clarify!
- 196 contamination = obstruction?
- 276 which variable is used as external drift, elevation?
- 296 contemporary = simultaneously ?
- Figure 3, I don’t see this figure referenced anywhere in the text,
3 Study Areas and Dataset Description
- You acknowledge forest canopy as an important source of uncertainty but do not give any information if, and how much of the area is forested.
- 321 Are manual SWE observation only available for one of the two seasons?
- 336 S1 is not introduced in the text (unlike sentinel-2)
- Figure 4: resolution should be improved; it is very hard to read.
- Figure 4b: Why show all SWE observations if you only use a few of them (e.g. in Fig 10)? Or is the mean performance metrics based on all of them? Please clarify.
4 Results
- Figure 6: image resolution should be increased. “Trend” could be recast as “time-series”
- Figure 7: rather small, also image resolution could be increased, caption does not mention ASO
- Figure 8 / Table 1. Although it can make sense to specify total catchment wide SWE in Gt, I would much rather prefer it to be given in mm as well.
- Figure 9 might be more helpful when ASO observations are included. I also think the plot does not support your interpretation that the drop in total SWE in very high elevations is due to gravitational redistribution (L 387), nor that steeper slopes present less SWE, nor that there is more snow on north facing slopes. Total SWE amounts are presented (in Gt), so this strongly depends on the area covered by this class (i.e. elevation bands). Same is true for slope and aspect bins. You need to scale with the area of a class to allow for these relative comparisons (express SWE in mm). Otherwise, you carrying the information about the catchment topography (e.g. hypsometry). x-axis: Aspect is not in degrees but in cardinal and ordinal directions
- 397 what is meant by “only few examples”? Are there manual SWE observations for other years as well, or is this just a subset of the locations shown in Fig 4b?
- 401 tends to increase
- Figure 12, see fig 9. Since no spatial observations are available in Schnals, I don’t consider this figure very helpful.
- What about an evaluation against the automated snow depth sensor in Schnals?
5 Discussion
- 413 maybe better “…quantitative and qualitative evaluation of the proposed SWE reconstruction over two study areas”
- Figure 16: caption: use the same naming (i.e. “original snow cover map”) as in the legends and captions of the previous Figures (14, 15)
- Section 5.1 only focusses on errors in predicting the accumulation state. Please expound upon the uncertainties associated with predicting the ablation/equilibrium state.
- 480-485 Please clarify/recast: “M” and “potential melting” are used in the same sentence and might confuse readers. Suggestion: “Since potential melting values at the end of the ablation period are high, an erroneous estimation of tSD strongly affects the reconstruction of peak SWE.”
Appendix:
- Figure A1: caption does not state that this is Schnals
- Figure C2: x-axis: Aspect is not in degrees but in cardinal and ordinal directions; As a line plot it would be easier to compare observation and the proposed approach.
Citation: https://doi.org/10.5194/tc-2022-146-RC1 - AC2: 'Reply on RC1', Valentina Premier, 10 Jan 2023
-
RC2: 'Comment on tc-2022-146', Noah Molotch, 28 Oct 2022
My overall impression of the work is that it potentially represents a novel contribution and one that takes the SWE reconstruction approach into a very new direction with potential value. I am excited by this potential and think the authors should be applauded for taking on this work. However, I found some aspects of the paper extremely difficult to follow (identification of catchment state) and in general think the quality of the writing and use of english to be quite problematic. I think the paper is valuable but I would strongly suggest more editorial consideration in the context of sentence structure and grammar as nearly every-other sentence suffers from some type of grammatical error. Most of these errors were quite small and did not interfere with my understanding of the points being made but some errors were more considerable. These errors were far too numerous for me to spend the time to point them all out or to correct them all. My comments are included in the comments margin of the attached PDF. Most of these are broader-context comments that align with my perspective that the paper needs very major revisions. Thank you, Noah Molotch
- AC3: 'Reply on RC2', Valentina Premier, 10 Jan 2023
-
RC3: 'Comment on tc-2022-146', Anonymous Referee #3, 17 Nov 2022
This paper presents a reconstruction method to produce time series of high-resolution (HR) snow water equivalent (SWE) maps from different sources of remotely sensed data, with specific focus on the applicability on mountain areas. The method’s description and results in two selected catchments are discussed, and the potential sources of errors are addressed in the context of general applicability and future improvements.
The topic is relevant for the scientific community in snow-dominated areas due to the lack of HR mapping on a time scale that allows monitoring of quick changes in the snowpack extension and mass change, and the scarcity of direct SWE measurements and methods to provide a dense network of monitoring stations for spatial interpolation approaches. The innovation and relevance of these objectives are sufficiently addressed in the manuscript. However, in its current version, some issues are found that require further assessment before considering further review and potential publication in this journal. Please, find below these major items. I hope that these comments are useful to improve the work and help to further comprehend its context and applicability further than the present results.
- The Introduction section contains good points but requires some structure to get more focused on the specific goals’ context. I would also recommend to present this earlier in the narrative. Lines 100-108 can be easily moved/merged to/with section 2 for the sake of clarity.
- The general objective should be better elaborated in line 99, i.e. not only state what but also what for and some specific scope. For example, the target type of catchment is relevant but it is not declared until lines 130-131 that size is limiting the potential further applicability of the method. Moreover, the order of magnitude of “a not too vast catchment” must be assessed.
- Section 2.1 is determinant in the methodological approach. In the explanation, it is not clear whether the catchment state is identified for each pixel or for the whole catchment area; this needs a revision to be clear throughout the text. Moreover, the spatial definition of the “total delta-SWE” is missing, which is required, and additionally the use of this variable should be uniform for the three states (i.e. is also total in line 149?). In line 149, I am not sure about the meaning of “no changes WITHIN the catchment”, do you mean really that or rather no change when considered as a whole?
- I have doubts on the simplification done on the potential combinations of positive/zero/negative values of delta-SWE and delta-SCA in this section. First, it seems that both variables have different spatial definitions, since pixel changes in terms of SCA are assessed. Additionally, some situations are discarded, for example, accumulation is not allowed to happen with negative delta-SCA values, but this is not infrequent in mountain areas in some regions in the world. Other situations are not included in the three potential states. In general, the assumptions are difficult to be validated in semiarid regions with snow relevance or during patchy snow periods in steep slopes, especially if the catchment state is defined uniformly in space. These issues should have been assessed and their discarding justified or at least clarified in terms of the applicability of the method.
- In section 2.3, two issues require further assessment. First, the use of day-degree modelling for melting rates’ estimation is not the best choice if accurate HR maps are the goal, in my opinion. At least, some justification of this apparent lack of coherence should be included, together with the comparison of the error of SWE estimation associated to the use of such methods and the error from low resolution satellite products. Secondly, the adoption of the temperature threshold is one of the major sources of error in the SWE estimation in mountain areas, as many works have already shown; so, the selected value needs some justification. Thirdly, and more relevant, lines 280-283 involve that melting is the only process in the ablation of the snowpack, which means that sublimation is neglected (but nothing is said on this); this may result in non-negligible loss of mass in the closure of the balance equation, and it is a constraint for the applicability of the method in some regions or during some periods/under some atmospheric conditions. This must be addressed in the description of the methodological assumptions and their validity. Finally, some comments on the scale effects from the subdaily evolution, not operating in the method, should be included.
- In section 4, the results are shown as selected points/transect/ periods in the study catchments, and detailed datasets are included as appendixes. The selection must be justified in all cases. The associated figures and tables’ captions must include the catchment name in all cases (see figures 5 to 7, and table 1). Some sentences lack a proper justification, for example, lines 389 and 399 contain comments that can’t be rigorously concluded in general from what has been shown. Or line 408, regarding Fig. 13, has a mass balance closure test been done? Figure 9 caption, are these “trends”?
- The discussion in section 5 repeats many facts or comments that have been previously presented or commented. Moreover, the discussion is focused on the sources of error at each step of the proposed method. I miss the discussion on the goodness of the results when compared to other products/methods/data sources that provide less resolution, or other standard or alternative existing methods. This is important as HR SWE mapping is the target goal.
- The error indicators in results cannot be properly valued since little information is included from the study catchment in terms of SWE regime, in section 3.
- The discussion/conclusions should also include more reference to what processes can be tracked from the time series obtained of these SWE maps, and what cannot due to the assumptions, etcetera in the approach. This is very relevant to address the further applicability of the method.
Some additional comments:
- In general, the English usage and edition is good, but some revision is recommended.
- Please, review the use of some wording. For example, line 381, “while the others (seasons) are drier” really means snow-scarce, which can also be due to high temperature; or the use of “bias” in the work to define “difference” or absolute error.
- When some references are included in a list, please, use a constant criteria to order (increasing or decreasing date).
- Reference in line 35 looks not recent enough to be a updated review for remote sensing products, at least, some others could have been included.
- Line 65, please provide some reference, there are works on that (i.e. Pimentel et al., 2015;2017; or others).
- Please, assess the error associated to the ASO product, taken as ground-truth to test the results.
- Beyond the comparison of results and ASO in the appendixes, dispersion graphs are needed to further assess the performance of the method, and some selected cases should be included in the results’ section.
Citation: https://doi.org/10.5194/tc-2022-146-RC3 -
AC4: 'Reply on RC3', Valentina Premier, 10 Jan 2023
The authors thank the Reviewer for his/her constructive feedbacks and comments on the manuscript. We went through each point and took advantage of the comments to improve the quality of the manuscript. Our answers are reported in blue in the attached document.
Status: closed
-
CC1: 'Comment on tc-2022-146', Pau Wiersma, 29 Aug 2022
Dear ms Premier,
Congratulations on this great work, I believe it is an important contribution to the field of SWE modeling. I find your solutions very original and at the same time intuitive and straightforward. The manuscript is very readable and detailed, most of the questions that came up while reading were answered later on. Having said that, I do have a number of comments, which I have gathered in the attached document. I hope they are clear and that they will be useful to you, and my apologies if I interpreted the manuscript incorrectly.
Kind regards,
Pau Wiersma
- AC1: 'Reply on CC1', Valentina Premier, 13 Sep 2022
-
RC1: 'Comment on tc-2022-146', Anonymous Referee #1, 13 Sep 2022
I would like to congratulate the authors to this very interesing work, as they have presented a novel methodology and useful contribution to snow science. The presented approach might open new doors for reconstructing snow water resources also on larger scales, by employing multiple data sources. Especially, the usage of SAR backscattering signals to both, detect snow ablation, and indirectly correct daily high resolution SCA maps is an original methodology. The simple, yet efficient basic concept of the reconstruction approach, together with the parsimonious use of in-situ data is highly appealing. In my opinion, the organization of the work is good and the presentation of the results is rather clear. However, many sentences could be restructured to make for an easier read. I have made various comments on the manuscript that I hope will help to improve the paper.
There are still some important details that remain unclear to me, especially concerting the determination of the catchment state:
- It would be helpful to see a more detailed description about how the ablation state is derived from Sentinel 1 data and how this would translate to the three snowpack phases described in Marin et al. (2020) (i.e. moistening, ripening, runoff). In line 185 only a “relevant drop” in backscattering is mentioned. Does this mean the catchment state ablation already starts when any liquid water is present in the snowpack? Since it seems possible (at least with S1A and S1B) to identify the snowpack runoff phase and a separation to previous moistening and ripening phases, the use of a DD-style melt model would be much more justified - as this is practically eliminating the need to track energy states (cold content) in the snowpack modelling. Please clarify if ablation is based on a drop in backscattering (melting phase) like line 185 suggests, or on the minimum of backscattering (runoff phase), as implied line 174. If the latter is the case, there is more physical grounds to employ a DD-style melt model and this should be brought forward in the text. However, if the ablation state is corresbonding to the moistening phase (i.e. a mere drop in backscattering), this decission should be also explained in more detail.
- If I understood correctly, the catchment states ablation and equilibrium can exist both at the same time-step (different pixels have different states), but in accumulation all pixels have this state. Line 104 and 189 contradict themselves in this regard. Please clarify.
- Although bringing the reconstructed SWE time-series in context to the catchment discharge might provide some insights into the estimated snow cover dynamics, however, the way this information is presented in Figure 13 and interpreted in the text (line 409 -411) is not suitable for this purpose. There are basically two statements emerging from this analysis: i) there was more snow in one year than the other, ii) SWE decreases and subsequently discharge increases at some point. As it can be seen in Figure 13 there are very different discharge responses in the spring freshet between the two years. I do not see this analysis to be much helpful in the current state and would advise either to remove it from the manuscript, or expand the analysis (giving more information about precipitation, hydrological characteristics of the catchment, and changing the units in the figure (e.g. to mm)).
- The authors discuss various sources of uncertainty in the methodology and state that due to a number of preprocessing steps a formal sensitivity analysis is difficult to perform. However, it might be still very valuable for the reader to get a feeling about how possible errors might translate to the SWE reconstruction. Please consider the possibility to provide a simplified version of part of the problem, by e.g. perturbing the values of tSD and tSA (and perhaps keeping the states constant during this time) and showing the consequences in terms of peak SWE. This could also help to underline the statements in line 480f.
Additional Comments
Abstract/1 Introduction:
- 1 reconsider recasting the first sentence.
- 3 how about ablation processes, or are you specifically talking about peak SWE?
- 11 At this point, the reader might not follow what you mean with “time-series regularization from impossible transitions”. Mentioned again at 105, 111, and 127 before finally explained in section 2.2 at 194.
- 17 I’m not particularly fond of using present perfect (throughout the manuscript). But that might be personal preference. Please consider using present tense (or simple past) consistently.
- 21 not only on local hydrology, as many regions of the world rely e.g. on the spring freshet hundreds of km downstream
- 23 „from at least 50%“please check again with Vivironi et al (2003). Although snowmelt is a major contributor of mountain water resources, as far as I know, the numbers given by Vivironi and others include mountain waters in general.
- 27 precipitation variability is affected by orography, interpolation is affects by sampling density among other factors, and observations can be erroneous due to e.g. undercatch
- 65 “accumulation and melt = „accumulation and ablation“ (i.e. in this sense ablation includes erosion and evaporation etc.)
- 66 consider topography vs geomorphology (throughout the text)
- 80 they range from DD to complete energy balance models (as used in Bair et al 2016), and these do not require calibration
- 114 (Italy) like (USA)
2 Proposed approach to HR SWE reconstruction
- 120 first sentence is obsolete iMo.
- 132 please specify „too vast“.
- 136 Maybe this would be clearer: „in detail, the catchment state is characterized by the change in SWE, but is also associated with possible changes in SCA“. Or similar.
- 141 „extension“ = extent
- 143 „…dSWE < 0 due to melt water drainage“.
- 145 snow depth, not height. Anyway, better to talk here in terms of mass/SWE. e.g. „..if the snowpack is melting only partially“.
- 167-168 maybe better placed in the discussion section
- 169 potentiality = potential to detect the presence of a melting snowpack…
- 174 most important in terms of what? Certainly not in terms of melt water production as melt rates increase towards later season. Peak SWE is not necessarily the peak of melt water runoff. Please clarify.
- 181 replace “quotes” with “elevations”, “elevation bands” or similar (throughout the manuscript)
- 189 ablation and equilibrium classes can exist at the same time, but 104 states that the state is assumed homogeneous for all pixels of the catchment. Please clarify!
- 196 contamination = obstruction?
- 276 which variable is used as external drift, elevation?
- 296 contemporary = simultaneously ?
- Figure 3, I don’t see this figure referenced anywhere in the text,
3 Study Areas and Dataset Description
- You acknowledge forest canopy as an important source of uncertainty but do not give any information if, and how much of the area is forested.
- 321 Are manual SWE observation only available for one of the two seasons?
- 336 S1 is not introduced in the text (unlike sentinel-2)
- Figure 4: resolution should be improved; it is very hard to read.
- Figure 4b: Why show all SWE observations if you only use a few of them (e.g. in Fig 10)? Or is the mean performance metrics based on all of them? Please clarify.
4 Results
- Figure 6: image resolution should be increased. “Trend” could be recast as “time-series”
- Figure 7: rather small, also image resolution could be increased, caption does not mention ASO
- Figure 8 / Table 1. Although it can make sense to specify total catchment wide SWE in Gt, I would much rather prefer it to be given in mm as well.
- Figure 9 might be more helpful when ASO observations are included. I also think the plot does not support your interpretation that the drop in total SWE in very high elevations is due to gravitational redistribution (L 387), nor that steeper slopes present less SWE, nor that there is more snow on north facing slopes. Total SWE amounts are presented (in Gt), so this strongly depends on the area covered by this class (i.e. elevation bands). Same is true for slope and aspect bins. You need to scale with the area of a class to allow for these relative comparisons (express SWE in mm). Otherwise, you carrying the information about the catchment topography (e.g. hypsometry). x-axis: Aspect is not in degrees but in cardinal and ordinal directions
- 397 what is meant by “only few examples”? Are there manual SWE observations for other years as well, or is this just a subset of the locations shown in Fig 4b?
- 401 tends to increase
- Figure 12, see fig 9. Since no spatial observations are available in Schnals, I don’t consider this figure very helpful.
- What about an evaluation against the automated snow depth sensor in Schnals?
5 Discussion
- 413 maybe better “…quantitative and qualitative evaluation of the proposed SWE reconstruction over two study areas”
- Figure 16: caption: use the same naming (i.e. “original snow cover map”) as in the legends and captions of the previous Figures (14, 15)
- Section 5.1 only focusses on errors in predicting the accumulation state. Please expound upon the uncertainties associated with predicting the ablation/equilibrium state.
- 480-485 Please clarify/recast: “M” and “potential melting” are used in the same sentence and might confuse readers. Suggestion: “Since potential melting values at the end of the ablation period are high, an erroneous estimation of tSD strongly affects the reconstruction of peak SWE.”
Appendix:
- Figure A1: caption does not state that this is Schnals
- Figure C2: x-axis: Aspect is not in degrees but in cardinal and ordinal directions; As a line plot it would be easier to compare observation and the proposed approach.
Citation: https://doi.org/10.5194/tc-2022-146-RC1 - AC2: 'Reply on RC1', Valentina Premier, 10 Jan 2023
-
RC2: 'Comment on tc-2022-146', Noah Molotch, 28 Oct 2022
My overall impression of the work is that it potentially represents a novel contribution and one that takes the SWE reconstruction approach into a very new direction with potential value. I am excited by this potential and think the authors should be applauded for taking on this work. However, I found some aspects of the paper extremely difficult to follow (identification of catchment state) and in general think the quality of the writing and use of english to be quite problematic. I think the paper is valuable but I would strongly suggest more editorial consideration in the context of sentence structure and grammar as nearly every-other sentence suffers from some type of grammatical error. Most of these errors were quite small and did not interfere with my understanding of the points being made but some errors were more considerable. These errors were far too numerous for me to spend the time to point them all out or to correct them all. My comments are included in the comments margin of the attached PDF. Most of these are broader-context comments that align with my perspective that the paper needs very major revisions. Thank you, Noah Molotch
- AC3: 'Reply on RC2', Valentina Premier, 10 Jan 2023
-
RC3: 'Comment on tc-2022-146', Anonymous Referee #3, 17 Nov 2022
This paper presents a reconstruction method to produce time series of high-resolution (HR) snow water equivalent (SWE) maps from different sources of remotely sensed data, with specific focus on the applicability on mountain areas. The method’s description and results in two selected catchments are discussed, and the potential sources of errors are addressed in the context of general applicability and future improvements.
The topic is relevant for the scientific community in snow-dominated areas due to the lack of HR mapping on a time scale that allows monitoring of quick changes in the snowpack extension and mass change, and the scarcity of direct SWE measurements and methods to provide a dense network of monitoring stations for spatial interpolation approaches. The innovation and relevance of these objectives are sufficiently addressed in the manuscript. However, in its current version, some issues are found that require further assessment before considering further review and potential publication in this journal. Please, find below these major items. I hope that these comments are useful to improve the work and help to further comprehend its context and applicability further than the present results.
- The Introduction section contains good points but requires some structure to get more focused on the specific goals’ context. I would also recommend to present this earlier in the narrative. Lines 100-108 can be easily moved/merged to/with section 2 for the sake of clarity.
- The general objective should be better elaborated in line 99, i.e. not only state what but also what for and some specific scope. For example, the target type of catchment is relevant but it is not declared until lines 130-131 that size is limiting the potential further applicability of the method. Moreover, the order of magnitude of “a not too vast catchment” must be assessed.
- Section 2.1 is determinant in the methodological approach. In the explanation, it is not clear whether the catchment state is identified for each pixel or for the whole catchment area; this needs a revision to be clear throughout the text. Moreover, the spatial definition of the “total delta-SWE” is missing, which is required, and additionally the use of this variable should be uniform for the three states (i.e. is also total in line 149?). In line 149, I am not sure about the meaning of “no changes WITHIN the catchment”, do you mean really that or rather no change when considered as a whole?
- I have doubts on the simplification done on the potential combinations of positive/zero/negative values of delta-SWE and delta-SCA in this section. First, it seems that both variables have different spatial definitions, since pixel changes in terms of SCA are assessed. Additionally, some situations are discarded, for example, accumulation is not allowed to happen with negative delta-SCA values, but this is not infrequent in mountain areas in some regions in the world. Other situations are not included in the three potential states. In general, the assumptions are difficult to be validated in semiarid regions with snow relevance or during patchy snow periods in steep slopes, especially if the catchment state is defined uniformly in space. These issues should have been assessed and their discarding justified or at least clarified in terms of the applicability of the method.
- In section 2.3, two issues require further assessment. First, the use of day-degree modelling for melting rates’ estimation is not the best choice if accurate HR maps are the goal, in my opinion. At least, some justification of this apparent lack of coherence should be included, together with the comparison of the error of SWE estimation associated to the use of such methods and the error from low resolution satellite products. Secondly, the adoption of the temperature threshold is one of the major sources of error in the SWE estimation in mountain areas, as many works have already shown; so, the selected value needs some justification. Thirdly, and more relevant, lines 280-283 involve that melting is the only process in the ablation of the snowpack, which means that sublimation is neglected (but nothing is said on this); this may result in non-negligible loss of mass in the closure of the balance equation, and it is a constraint for the applicability of the method in some regions or during some periods/under some atmospheric conditions. This must be addressed in the description of the methodological assumptions and their validity. Finally, some comments on the scale effects from the subdaily evolution, not operating in the method, should be included.
- In section 4, the results are shown as selected points/transect/ periods in the study catchments, and detailed datasets are included as appendixes. The selection must be justified in all cases. The associated figures and tables’ captions must include the catchment name in all cases (see figures 5 to 7, and table 1). Some sentences lack a proper justification, for example, lines 389 and 399 contain comments that can’t be rigorously concluded in general from what has been shown. Or line 408, regarding Fig. 13, has a mass balance closure test been done? Figure 9 caption, are these “trends”?
- The discussion in section 5 repeats many facts or comments that have been previously presented or commented. Moreover, the discussion is focused on the sources of error at each step of the proposed method. I miss the discussion on the goodness of the results when compared to other products/methods/data sources that provide less resolution, or other standard or alternative existing methods. This is important as HR SWE mapping is the target goal.
- The error indicators in results cannot be properly valued since little information is included from the study catchment in terms of SWE regime, in section 3.
- The discussion/conclusions should also include more reference to what processes can be tracked from the time series obtained of these SWE maps, and what cannot due to the assumptions, etcetera in the approach. This is very relevant to address the further applicability of the method.
Some additional comments:
- In general, the English usage and edition is good, but some revision is recommended.
- Please, review the use of some wording. For example, line 381, “while the others (seasons) are drier” really means snow-scarce, which can also be due to high temperature; or the use of “bias” in the work to define “difference” or absolute error.
- When some references are included in a list, please, use a constant criteria to order (increasing or decreasing date).
- Reference in line 35 looks not recent enough to be a updated review for remote sensing products, at least, some others could have been included.
- Line 65, please provide some reference, there are works on that (i.e. Pimentel et al., 2015;2017; or others).
- Please, assess the error associated to the ASO product, taken as ground-truth to test the results.
- Beyond the comparison of results and ASO in the appendixes, dispersion graphs are needed to further assess the performance of the method, and some selected cases should be included in the results’ section.
Citation: https://doi.org/10.5194/tc-2022-146-RC3 -
AC4: 'Reply on RC3', Valentina Premier, 10 Jan 2023
The authors thank the Reviewer for his/her constructive feedbacks and comments on the manuscript. We went through each point and took advantage of the comments to improve the quality of the manuscript. Our answers are reported in blue in the attached document.
Valentina Premier et al.
Valentina Premier et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
699 | 238 | 22 | 959 | 5 | 4 |
- HTML: 699
- PDF: 238
- XML: 22
- Total: 959
- BibTeX: 5
- EndNote: 4
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1