Drivers of Dust-Enhanced Snowpack Melt-Out and Streamflow Timing

Round 1

Reviewer 1 Report

Major comments

The study by Steven R. Fassnacht et al. titled, “Drivers of Dust-Enhanced Snowpack Melt-out and Streamflow Timing” quantify the annual dust-enhanced energy absorption and used this information to model the snowpack melt-out under observed and clean conditions and they determine the difference in snow cover duration between actual  and simulated ideal snowpack to characterize the shifts in melt timing for each year. This paper also deals with the estimation of runoff volume using the runoff model. The model reproduces the observed data well and describes how the presence of dust on the snow surface accelerates snowmelt and its effect on runoff. The paper is complete and concise, and I believe it is ready for publication with the following minor corrections.

(1) I find the description of albedo unclear. The most important thing in this paper is the introduction of albedo into SNOBAL. If the dust effect appears in the visible albedo and not in the near-infrared albedo, I think the description of the observations of the near-infrared albedo is insufficient. In addition, if the albedo is to be calculated separately for the visible and near-infrared regions, it is necessary to divide the solar radiation into the visible and near-infrared regions, and I feel that there is no diagram or description of the yearly variation of this result. It is important to show the observed results for S(TOTAL)=S(VIS)+S(NIR), S(TOTLA)×albedo(ALL)=S(VIS)×albedo(VIS)+S(NIR)×albedo(NIR). By showing the correlation between albedo(VIS) and the dust content, it will be possible to make the reader understand that the year-to-year variation of the visible albedo depends on the dust. Also, calculating albedo(VIS) for clean snow cover at 0.92, which is fresh snow albedo, and fresh snow albedo for the entire snowmelt season is unrealistic and I don't feel that the authors are doing meaningful model calculations. The albedo of fresh snow will change over time due to changes in grain size and other factors. It is difficult to find meaningful comparisons between unrealistic calculations and reality, so I feel it is acceptable to remove all clean snow calculations, but it is important to explain more clearly the validity of calculations using the visible albedo of fresh snow if necessary.

(2) I feel that the abbreviations are not consistent. Once you have defined an abbreviation, I recommend that you use the abbreviation consistently in subsequent sentences. Also, there are places where abbreviations appear out of the blue. Since there are quite a few abbreviations, please consider putting all the abbreviations together in an appendix so that readers will not be confused. For example, line 75: RF comes out of nowhere and there is no explanation of what it stands for; line 331: Snow-all-gone should be written as SAG since the previous sentence defines the abbreviation as SAG; line 610: YW2105 should be WY2015; line 171 and 175: The SBSP and SASP have already been defined in the previous sentence, so this is the second definition, etc.

(3) Line 35: I recommend to add the following two references: Takeuchi and Li (2007: doi: 10.1657/1523-0430(07-094)[takeuchi]2.0.Co;2) and Suzuki and Ohta (2003: DOI:  https://doi.org/10.1175/1525-7541(2003)004<1181:EOLFDO>2.0.CO;2).

Author Response

(1) I find the description of albedo unclear. The most important thing in this paper is the introduction of albedo into SNOBAL. If the dust effect appears in the visible albedo and not in the near-infrared albedo, I think the description of the observations of the near-infrared albedo is insufficient. In addition, if the albedo is to be calculated separately for the visible and near-infrared regions, it is necessary to divide the solar radiation into the visible and near-infrared regions, and I feel that there is no diagram or description of the yearly variation of this result. It is important to show the observed results for S(TOTAL)=S(VIS)+S(NIR), S(TOTAL)×albedo(ALL)=S(VIS)×albedo(VIS)+S(NIR)×albedo(NIR). By showing the correlation between albedo(VIS) and the dust content, it will be possible to make the reader understand that the year-to-year variation of the visible albedo depends on the dust. Also, calculating albedo(VIS) for clean snow cover at 0.92, which is fresh snow albedo, and fresh snow albedo for the entire snowmelt season is unrealistic and I don't feel that the authors are doing meaningful model calculations. The albedo of fresh snow will change over time due to changes in grain size and other factors. It is difficult to find meaningful comparisons between unrealistic calculations and reality, so I feel it is acceptable to remove all clean snow calculations, but it is important to explain more clearly the validity of calculations using the visible albedo of fresh snow if necessary.

 

• Although there is importance in showing the observed results of S(VIS) and S(NIR), the pyranometers at both sites record S(TOTAL) and S(NIR). Therefore, a better equation to show would be S(TOTAL)-S(NIR)=S(VIS). However, as stated in the paper and in Figure 1, the change in albedo for dust present versus dust free is mostly in the visible part of the spectrum.
• Though it is well documented that dust effects visible albedo more than NIR, modeling this relationship is more complicated than it seems. When a dust event occurs, it often comes with a precipitation event. The NIR albedo would likely show normal patterns of fresh snowfall, and assumingly the VIS albedo would decrease – but in many situations the albedo stays consistently high, and no change in albedo can be observed from this. Even on the days following, the VIS albedo and the NIR albedo do not have clearly defined differences from the dust event. We recognize there would be a strong bias in identifying only the small albedo values with each dust event and onwards.
• Finally, the effect from the dust most often happens during melt when these layers re-emerge as the above layers melt from increased shortwave radiation due to seasonal changes as well as increased absorption due to exposure of these dust layers. We know these dust events affect snowmelt but creating a visual of this relationship would imply the filtering of many values – that could be realistic due to the characteristics of the surrounding environment and snowpack.
• Using an albedo of 0.92 for fresh snow “reset” is realistic based on observations from Painter et al. (2012) and Skiles et al. (2012). Albedo does decay, but this decay rate is dynamic (Reimanis, 2021). Adding a fresh snow decay would increase the uncertainty and possibly unneeded, as the albedo of fresh snowfall in the cold, dry temperatures of high-elevation Southwest Colorado decays slowly due to the maintained integrity of the near-surface snow crystals.
• Furthermore, removing the fresh snow events would highly skew the results as precipitation can be common in this area (Figure 3b), and implying there is no fresh snow, even in the melt-out period would lower the accuracy of the model (Figure 4). Finally, as mentioned in section 3.1 of the paper, observed albedo during snowfall events can be highly erroneous due to the upward looking pyranometer getting covered with snow and recording incorrect values. Even if we estimate incoming shortwave radiation using albedo from the windy site (SBSP), visible albedo values were often still larger than one, and considering snowfall events are incredibly important to snowpack dynamics, we thought it was more worthwhile to estimate an albedo for fresh snow, than to filter the data.

 

(2) I feel that the abbreviations are not consistent. Once you have defined an abbreviation, I recommend that you use the abbreviation consistently in subsequent sentences. Also, there are places where abbreviations appear out of the blue. Since there are quite a few abbreviations, please consider putting all the abbreviations together in an appendix so that readers will not be confused. For example, line 75: RF comes out of nowhere and there is no explanation of what it stands for; line 331: Snow-all-gone should be written as SAG since the previous sentence defines the abbreviation as SAG; line 610: YW2105 should be WY2015; line 171 and 175: The SBSP and SASP have already been defined in the previous sentence, so this is the second definition, etc.

 

• We have reviewed the abbreviations and hopefully they are now more consistent.

 

(3) Line 35: I recommend to add the following two references: Takeuchi and Li (2007: doi: 10.1657/1523-0430(07-094)[takeuchi]2.0.Co;2) and Suzuki and Ohta (2003: DOI:  https://doi.org/10.1175/1525-7541(2003)004<1181:EOLFDO>2.0.CO;2).

 

• These citations have been added and all the subsequent references have been shifted.

Reviewer 2 Report

Dear Authors,

Thank you for the interesting work. There is a lot of number crunching in the study, some giving better relations than others. I have one major and one minor comment on the manuscript. The major one is that the study site sections needs a simple but descriptive map of the basin with its observation locations. It is nearly impossible for the reader to follow and understand the processes unless they have an idea about the observation points. The reader is losing pace in the flow of the article when they have to find all this info from references. The minor comment is that at several paragraphs there are too many references cited, even after each sentence. This is just a comment and need not to be altered at this point, but it does get tiring while reading

There are also several comments/suggestions/questions within the manuscript that need attention.

Overall, the work conducted seems to be attractive and favorable for publication in Hydrology.

Best of luck...

Comments for author File: Comments.pdf

Author Response

Thank you for the interesting work. There is a lot of number crunching in the study, some giving better relations than others. I have one major and one minor comment on the manuscript. The major one is that the study site sections needs a simple but descriptive map of the basin with its observation locations. It is nearly impossible for the reader to follow and understand the processes unless they have an idea about the observation points. The reader is losing pace in the flow of the article when they have to find all this info from references.

• A site map has been added. Initially we thought that another map of this area was not relevant, but two reviewers thought a map should be added, so one was added.

 

The minor comment is that at several paragraphs there are too many references cited, even after each sentence. This is just a comment and need not to be altered at this point, but it does get tiring while reading.

• These have been reviewed. We feel that most of these citations are relevant. The issue is the nature of citations in MDPI (like Nature, etc.), such that it is difficult to know which reference is being cited, based only on the number.

 

There are also several comments/suggestions/questions within the manuscript that need attention.

 

• These have been changed in the paper, as per the suggestions.

 

Other comments on the pdf supplied by the reviewer. We have addressed those, with some of the larger comments being noted in the tracked changes version.

Reviewer 3 Report

1. Line 81: It is better to show a study site map, and point out the basin boundary and the locations of two micrometeorological towers.
2. Line 149: Is it possible to show a figure about the annual streamflow as the Figure 2(b) of annual precipitation?
3. Line 203: It is necessary to explain why used the point-based snow energy balance SNOBAL to model hypothetical snowpack. The SNOBAL requires much data preparation. There are other models which could be also suitable for this study. Why didn’t choose other models?
4. Line 340: Could you show me the statistics of the model dust-present SWE versus the derived and manual SWE observations for Figure 5(m) in 2019 based on the root mean square error (RMSE) and the Nash-Sutcliffe Efficiency (NSE) coefficient? The simulation for Figure 5(m) in 2019 is not well. Why?
5. Line 363: For Figure 7a, removing WY2019 improves the correlation (R2 = 0.55); however, for Figure 7c, excluding that year WY2019 reduces the variance explain by almost half (R2 = 0.23). WY2019 is an outlier in terms of modeled versus observed melt-out date (Figure 5m). Is it suitable to be included in this study? The results are obviously different.
6. Line 496: This study used 1 April to SAG, but Skiles et al. [2012, 2015] used 15 April to SAG. These different date could affect the result of the mean daily DEAE. So, maybe the comparison is not appropriate.

Author Response

1. Line 81: It is better to show a study site map, and point out the basin boundary and the locations of two micrometeorological towers.

• A site map has been added.

 

2. Line 149: Is it possible to show a figure about the annual streamflow as the Figure 2(b) of annual precipitation?

• This has been added to Figure 2, and the subsequent figures have been renumbered.

 

3. Line 203: It is necessary to explain why used the point-based snow energy balance SNOBAL to model hypothetical snowpack. The SNOBAL requires much data preparation. There are other models which could be also suitable for this study. Why didn’t choose other models?

• SNOBAL was used here, since this work follows up from the work by Skiles et al. (2012) who used the model for similar purposes at the same site.
• We would have used another model, but the precedence was set by the previous work at the same site.

 

4. Line 340: Could you show me the statistics of the model dust-present SWE versus the derived and manual SWE observations for Figure 5(m) in 2019 based on the root mean square error (RMSE) and the Nash-Sutcliffe Efficiency (NSE) coefficient? The simulation for Figure 5(m) in 2019 is not well. Why?

• The year 2019 was a big snow year with above average precipitation during snowmelt. Based on the model fit for other years, this year was not simulated well. The model statistics are poor with a NSE < 0 for this year. The NSE was greater than 0.50 for all other years.

 

5. Line 363: For Figure 7a, removing WY2019 improves the correlation (R^2 = 0.55); however, for Figure 7c, excluding that year WY2019 reduces the variance explain by almost half (R^2 = 0.23). WY2019 is an outlier in terms of modeled versus observed melt-out date (Figure 5m). Is it suitable to be included in this study? The results are obviously different.

• No, 2019 should not be included, at least not for ΔSAG, as that is derived from the model results. This is most obvious with WY2019 for DEAE, as stated by the reviewer. A sentence was to the Discussion in this regard: “Since there was a 16 day difference in observed versus modeled SAG for 2019 (Figure 6m), it could be acceptable to remove this year from subsequent analyses (Figure 8a).”

 

6. Line 496: This study used 1 April to SAG, but Skiles et al. [2012, 2015] used 15 April to SAG. These different date could affect the result of the mean daily DEAE. So, maybe the comparison is not appropriate.

• Yes, we agree. We used 1 April rather than 15 April to start since peak SWE occurred prior to 15 April in 4 of the 13 study years and we wanted to start modeling prior to peak SWE (see (new) Figures 5b and 6).