the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Evaluation of high resolution snowpack simulations from global datasets and comparison with Sentinel-1 snow depth retrievals in the Sierra Nevada, USA
Abstract. Spatial distribution of mountain snow water equivalent (SWE) is key information for water management. We implement a tool to simulate snowpack properties at high resolution (100 m) by sourcing only global datasets of climate, land cover and elevation. The meteorological data are obtained from ERA5 which makes the method applicable in near real time (5 day latency). We evaluate the output using 49 SWE maps derived from airborne lidar surveys in the Sierra Nevada. We find a very good agreement at the catchment scale using uncalibrated lapse rates. Larger biases at the model grid scale are especially evident at high elevation but do not alter the catchment-scale snow mass accuracy. We additionally compare the simulated snow depth to Sentinel-1 snow depth retrievals and find a similar accuracy with respect to synchronous airborne lidar surveys. However, Sentinel-1 snow depth products are temporally sparse and often masked during the melt season and do not provide SWE.
- Preprint
(3380 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2024-791', Jeff Dozier, 09 May 2024
The paper shows some nice comparisons between model results (ERA5-Land combined with SnowModel) and measurements from the Airborne Snow Observatory. Given the more discouraging conclusions about such models by Liu et al. (2022), I am surprised, but the analysis here seems robust. As the Conclusion notes, the analysis provides a viable method to estimate the water resources in the snowpack in areas with only an austere information infrastructure.
A few comments to improve the manuscript:
Line “06” (106?, bottom of section 2.1). The reported accuracy of SWE, <0.01 m, is ambitious. In their reports to the water agencies, ASO quotes a density RMS uncertainty of ±20 kg/m3, but verification is based on snow courses and snow pillows, which are all on open, flat terrain and have their own uncertainties of SWE and depth.
Especially, you should note that ASO’s translation of snow depth to SWE depends on local measurements of density, typically snow pillows that have a depth sensor also along with snow courses where both SWE and depth are measured.
Section 2.2.2: How are you getting snow albedo for the EnBal part of SnowModel? The ASO spectrometer can be used to retrieve values, but the combined ERA-Land/SnowModel uses the ASO data for validation, not as a driver. The melt rate and disappearance date of the snowpack are sensitive to albedo and consequent radiative forcing by light-absorbing particles (Painter et al., 2010).
Figure 3: The colors used to identify the lines in the plots are too indistinct. Perhaps combine color with line style to make the differences more obvious?
Figure 6: Label the axes. They appear to be UTM zone 11N coordinates, but the identification of the comparison in rotated text is confusing. At first I thought they had something to do with the y-axis.
Figure 7 Line 47 in caption: recommend data “are” instead of “is”.
Line 01 in the Discussion. The phrase “the ASO program has shown that useful SWE products can be derived from remotely sensed snow depth” needs a caveat, in that the ASO model of snow density is adjusted based on in situ measurements of snow density.
Line 21-22 in the Discussion. Perhaps cite the Liu et al. (2022) analysis here?
I agree with the final paragraph of the Discussion. The combination of ERA5, Snow Model, and Sentinel-1 provides a way to analyze the snowpack in mountains with only an austere infrastructure. There are uncertainties of course, but the methods could provide some information in areas where few data exist.
Support for Open Science: The manuscript should identify the sources of data and code availability used in the analyses. I could do my own searches, but statements like “from the Copernicus Climate Change Service (C3S) and can be queried via their application programming interface” (Line 92) could be phrased more helpfully. Similarly, the citation to “Copernicus Digital Elevation Model, 2023” (Line 96) is not in the bibliography. Some information is missing about the “code availability section” mentioned on Line 45.
References
Liu, Y., Fang, Y., Li, D., and Margulis, S. A.: How well do global snow products characterize snow storage in High Mountain Asia?, Geophysical Research Letters, 49, e2022GL100082, https://doi.org/10.1029/2022GL100082, 2022.
Painter, T. H., Deems, J. S., Belnap, J., Hamlet, A. F., Landry, C. C., and Udall, B.: Response of Colorado River runoff to dust radiative forcing in snow, Proceedings of the National Academy of Sciences, 107, 17125-17130, https://doi.org/10.1073/pnas.0913139107, 2010.
Citation: https://doi.org/10.5194/egusphere-2024-791-RC1 - AC1: 'Reply on RC1', Laura Sourp, 15 Jul 2024
-
RC2: 'Comment on egusphere-2024-791', Anonymous Referee #2, 13 May 2024
General comments
The authors present a comprehensive evaluation of high-resolution snowpack simulations forced with globally available datasets, in particular coarse resolution meteorological data downscaled to the model grid. Thus, the study showcases a generic tool for performing snow cover simulations in any region of the world efficiently and with low effort. The simulations presented in the study, performed for the Tuolumne River catchment (Sierra Nevada, USA), were evaluated against high-resolution snow water equivalent (SWE) data derived from Lidar measurements of snow depth and modelled bulk snow densities. The simulations show promising results with comparable performance as satellite-derived snow characteristics for the study basin. In contrast to the remote sensing observations, the snow model results are always available, which is a significant advantage over the occasional satellite retrievals.
Overall, appropriate methods are used in the study and the results are relevant and promising. However, the presentation and discussion of the results sometimes lacks clarity and depth in my opinion. The description of the results deserves a few more details, whereas the discussion requires stronger links to the results themselves (foremost by including more references to specific figures). Furthermore, the paper should likely also be improved language-wise, preferably by a native English speaker. In spite of the shortcoming listed above, the paper is pleasant to read, contains a wealth of interesting results and is a valuable contribution to the snow modelling community. Detailed comments are listed below.
Specific comments
Page 1, line 13: Consider changing “sourcing” to using and “climate” to “meteorology”.
Page 1, line 18: Change from “snow depth to Sentinel-1 snow depth retrievals” to “snow depth to Sentinel-1 retrievals”.
Page 1, abstract: The concluding sentence of the abstract should be improved. One option would be to add a sentence stating directly that the snow model provides results anywhere at anytime in contrast to satellite retrievals.
Page 2, line 34: Please also cite Lievens et al. (2022) and adapt the sentence accordingly.
Page 2, line 46: Include the missing “have”: “There reanalyses have also…”
Page 2, lines 59-60: The sentence “However, the evaluation of these simulations relied on sparse in situ observations or MODIS snow cover area” seems incomplete. What is the drawback with these observations and why are more studies needed? Is it the coarse resolution of MODIS snow covered area?
Page 3, lines 68-79: Consider adding the spatial resolution of the model simulations already here.
Page 5, lines 00-01: Please mention the physical reason why the satellite retrievals do not provide data during the snowmelt period and add a reference supporting the statement.
Page 5, line 06: Important, the statement “…50 m SWE is less than 0.01 m w.e” needs a reference.
Page 5, line 15: What is “grassland rangeland”?
Page 7, line 40: Consider changing from “Appendix Table A1” to “see Table A1 in appendix”.
Page 7, line 58: Consider changing to “very coarse resolution of approximately 31 and 9 km (Fig. 1 and 2)”.
Page 7, lines 62-63: Consider changing to “…the snow depths given by ASO, Sentinel-1, and ERA-SnowModel were…”.
Page 8, lines 65-66: Please reformulate these two sentences. The second sentence needs to reference the first, otherwise it is not clear for what the performance metrics were computed.
Page 8, line 76-78: Please reformulate the sentence. It is too long and hard to read.
Figure 3: Consider using dashed lines for ERA5 and ERA5-Land.
Page 9, lines 84-85: It is likely not needed to describe the lines here since this information is already provided in the legend of the figure.
Page 9, line 89: The sentence “Considering the entire simulation period, 10% of the cells have an RMSE above 0.5 m w.e.” seems somewhat misplaced and is hard to understand.
Page 10, lines 1-2: Why were these two dates selected for the analysis?
Figure 5, caption: Why is the second date not mentioned in the caption?
Page 11, line 14: Is “mean residuals” the same as bias?
Page 11, line 25: Consider changing to “…resolution using upscaled ASO…”.
Page 11, lines 28-29: What does “these missing values are propagated at 1 km resolution” mean?
Page 11, line 30: Is not the exact area used between the methods or the dates, or both?
Figure 7: Consider merging Table 7 into this figure by including texts with the statistics. For an example of what I propose, see Figure 5 in Fontrodona-Bach et al. (2023). The scatter plots could potentially also be improved by showing the scatter density, just like the left panels in the Figure 5 by Fontrodona-Bach et al. (2023).
Page 13, line 53: What discontinuities in ERA5 SWE? Are these visible in Figure 3?
Page 14, line 58-59: Please improve the language of the sentence “We find an overestimation
of snow accumulation in high elevation however which occurs only above 3000 m asl”.
Page 14, lines 66-67: Avalanches move snow from higher to lower altitudes but does not reduce snow amounts. Please rephrase the sentence.
Page 14, lines 75-77: Please refer to Figure 5. Overall, as mentioned in the general comments, provide more links in the discussion to results by adding appropriate cross-references to figures and tables.
Page 15, lines 91-93: The sentence is formulated awkwardly. What does “carries 68 % of the Tuolumne River catchment” mean?
Page 15, lines 1-2: This statement requires at least one reference.
Page 15, line 6: What is hard to understand about the error patterns of Sentinel-1 compared to the other methods?
Page 15, lines 11-14: What has the first part of the sentence about errors has to do with the second part about model differences? Please split this sentence into two, and improve the language.
Page 16, lines 34-35: Consider providing a short description for each components of this tool since many readers start by reading the conclusions of a paper.
Page 16, line 38: What does the “0.08 m” refer to?
Page 16, lines 34-43: Example of paragraph that likely needs language improvements.
Technical comments
Page 3, line 70: Misplaced white space in 50 m.
Page 7, line 39: Missing whitespace.
Page 7, line 56: Missing comma after additionally.
Page 15, line 92 and 93: Wrong reference format.
References
Fontrodona-Bach, A., Schaefli, B., Woods, R., Teuling, A. J., & Larsen, J. R. (2023). NH-SWE: Northern Hemisphere Snow Water Equivalent dataset based on in situ snow depth time series. Earth Syst. Sci. Data, 15(6), 2577-2599. https://doi.org/10.5194/essd-15-2577-2023
Lievens, H., Brangers, I., Marshall, H. P., Jonas, T., Olefs, M., & De Lannoy, G. (2022). Sentinel-1 snow depth retrieval at sub-kilometer resolution over the European Alps. Cryosphere, 16(1), 159-177. https://doi.org/10.5194/tc-16-159-2022
Citation: https://doi.org/10.5194/egusphere-2024-791-RC2 - AC2: 'Reply on RC2', Laura Sourp, 15 Jul 2024
Status: closed
-
RC1: 'Comment on egusphere-2024-791', Jeff Dozier, 09 May 2024
The paper shows some nice comparisons between model results (ERA5-Land combined with SnowModel) and measurements from the Airborne Snow Observatory. Given the more discouraging conclusions about such models by Liu et al. (2022), I am surprised, but the analysis here seems robust. As the Conclusion notes, the analysis provides a viable method to estimate the water resources in the snowpack in areas with only an austere information infrastructure.
A few comments to improve the manuscript:
Line “06” (106?, bottom of section 2.1). The reported accuracy of SWE, <0.01 m, is ambitious. In their reports to the water agencies, ASO quotes a density RMS uncertainty of ±20 kg/m3, but verification is based on snow courses and snow pillows, which are all on open, flat terrain and have their own uncertainties of SWE and depth.
Especially, you should note that ASO’s translation of snow depth to SWE depends on local measurements of density, typically snow pillows that have a depth sensor also along with snow courses where both SWE and depth are measured.
Section 2.2.2: How are you getting snow albedo for the EnBal part of SnowModel? The ASO spectrometer can be used to retrieve values, but the combined ERA-Land/SnowModel uses the ASO data for validation, not as a driver. The melt rate and disappearance date of the snowpack are sensitive to albedo and consequent radiative forcing by light-absorbing particles (Painter et al., 2010).
Figure 3: The colors used to identify the lines in the plots are too indistinct. Perhaps combine color with line style to make the differences more obvious?
Figure 6: Label the axes. They appear to be UTM zone 11N coordinates, but the identification of the comparison in rotated text is confusing. At first I thought they had something to do with the y-axis.
Figure 7 Line 47 in caption: recommend data “are” instead of “is”.
Line 01 in the Discussion. The phrase “the ASO program has shown that useful SWE products can be derived from remotely sensed snow depth” needs a caveat, in that the ASO model of snow density is adjusted based on in situ measurements of snow density.
Line 21-22 in the Discussion. Perhaps cite the Liu et al. (2022) analysis here?
I agree with the final paragraph of the Discussion. The combination of ERA5, Snow Model, and Sentinel-1 provides a way to analyze the snowpack in mountains with only an austere infrastructure. There are uncertainties of course, but the methods could provide some information in areas where few data exist.
Support for Open Science: The manuscript should identify the sources of data and code availability used in the analyses. I could do my own searches, but statements like “from the Copernicus Climate Change Service (C3S) and can be queried via their application programming interface” (Line 92) could be phrased more helpfully. Similarly, the citation to “Copernicus Digital Elevation Model, 2023” (Line 96) is not in the bibliography. Some information is missing about the “code availability section” mentioned on Line 45.
References
Liu, Y., Fang, Y., Li, D., and Margulis, S. A.: How well do global snow products characterize snow storage in High Mountain Asia?, Geophysical Research Letters, 49, e2022GL100082, https://doi.org/10.1029/2022GL100082, 2022.
Painter, T. H., Deems, J. S., Belnap, J., Hamlet, A. F., Landry, C. C., and Udall, B.: Response of Colorado River runoff to dust radiative forcing in snow, Proceedings of the National Academy of Sciences, 107, 17125-17130, https://doi.org/10.1073/pnas.0913139107, 2010.
Citation: https://doi.org/10.5194/egusphere-2024-791-RC1 - AC1: 'Reply on RC1', Laura Sourp, 15 Jul 2024
-
RC2: 'Comment on egusphere-2024-791', Anonymous Referee #2, 13 May 2024
General comments
The authors present a comprehensive evaluation of high-resolution snowpack simulations forced with globally available datasets, in particular coarse resolution meteorological data downscaled to the model grid. Thus, the study showcases a generic tool for performing snow cover simulations in any region of the world efficiently and with low effort. The simulations presented in the study, performed for the Tuolumne River catchment (Sierra Nevada, USA), were evaluated against high-resolution snow water equivalent (SWE) data derived from Lidar measurements of snow depth and modelled bulk snow densities. The simulations show promising results with comparable performance as satellite-derived snow characteristics for the study basin. In contrast to the remote sensing observations, the snow model results are always available, which is a significant advantage over the occasional satellite retrievals.
Overall, appropriate methods are used in the study and the results are relevant and promising. However, the presentation and discussion of the results sometimes lacks clarity and depth in my opinion. The description of the results deserves a few more details, whereas the discussion requires stronger links to the results themselves (foremost by including more references to specific figures). Furthermore, the paper should likely also be improved language-wise, preferably by a native English speaker. In spite of the shortcoming listed above, the paper is pleasant to read, contains a wealth of interesting results and is a valuable contribution to the snow modelling community. Detailed comments are listed below.
Specific comments
Page 1, line 13: Consider changing “sourcing” to using and “climate” to “meteorology”.
Page 1, line 18: Change from “snow depth to Sentinel-1 snow depth retrievals” to “snow depth to Sentinel-1 retrievals”.
Page 1, abstract: The concluding sentence of the abstract should be improved. One option would be to add a sentence stating directly that the snow model provides results anywhere at anytime in contrast to satellite retrievals.
Page 2, line 34: Please also cite Lievens et al. (2022) and adapt the sentence accordingly.
Page 2, line 46: Include the missing “have”: “There reanalyses have also…”
Page 2, lines 59-60: The sentence “However, the evaluation of these simulations relied on sparse in situ observations or MODIS snow cover area” seems incomplete. What is the drawback with these observations and why are more studies needed? Is it the coarse resolution of MODIS snow covered area?
Page 3, lines 68-79: Consider adding the spatial resolution of the model simulations already here.
Page 5, lines 00-01: Please mention the physical reason why the satellite retrievals do not provide data during the snowmelt period and add a reference supporting the statement.
Page 5, line 06: Important, the statement “…50 m SWE is less than 0.01 m w.e” needs a reference.
Page 5, line 15: What is “grassland rangeland”?
Page 7, line 40: Consider changing from “Appendix Table A1” to “see Table A1 in appendix”.
Page 7, line 58: Consider changing to “very coarse resolution of approximately 31 and 9 km (Fig. 1 and 2)”.
Page 7, lines 62-63: Consider changing to “…the snow depths given by ASO, Sentinel-1, and ERA-SnowModel were…”.
Page 8, lines 65-66: Please reformulate these two sentences. The second sentence needs to reference the first, otherwise it is not clear for what the performance metrics were computed.
Page 8, line 76-78: Please reformulate the sentence. It is too long and hard to read.
Figure 3: Consider using dashed lines for ERA5 and ERA5-Land.
Page 9, lines 84-85: It is likely not needed to describe the lines here since this information is already provided in the legend of the figure.
Page 9, line 89: The sentence “Considering the entire simulation period, 10% of the cells have an RMSE above 0.5 m w.e.” seems somewhat misplaced and is hard to understand.
Page 10, lines 1-2: Why were these two dates selected for the analysis?
Figure 5, caption: Why is the second date not mentioned in the caption?
Page 11, line 14: Is “mean residuals” the same as bias?
Page 11, line 25: Consider changing to “…resolution using upscaled ASO…”.
Page 11, lines 28-29: What does “these missing values are propagated at 1 km resolution” mean?
Page 11, line 30: Is not the exact area used between the methods or the dates, or both?
Figure 7: Consider merging Table 7 into this figure by including texts with the statistics. For an example of what I propose, see Figure 5 in Fontrodona-Bach et al. (2023). The scatter plots could potentially also be improved by showing the scatter density, just like the left panels in the Figure 5 by Fontrodona-Bach et al. (2023).
Page 13, line 53: What discontinuities in ERA5 SWE? Are these visible in Figure 3?
Page 14, line 58-59: Please improve the language of the sentence “We find an overestimation
of snow accumulation in high elevation however which occurs only above 3000 m asl”.
Page 14, lines 66-67: Avalanches move snow from higher to lower altitudes but does not reduce snow amounts. Please rephrase the sentence.
Page 14, lines 75-77: Please refer to Figure 5. Overall, as mentioned in the general comments, provide more links in the discussion to results by adding appropriate cross-references to figures and tables.
Page 15, lines 91-93: The sentence is formulated awkwardly. What does “carries 68 % of the Tuolumne River catchment” mean?
Page 15, lines 1-2: This statement requires at least one reference.
Page 15, line 6: What is hard to understand about the error patterns of Sentinel-1 compared to the other methods?
Page 15, lines 11-14: What has the first part of the sentence about errors has to do with the second part about model differences? Please split this sentence into two, and improve the language.
Page 16, lines 34-35: Consider providing a short description for each components of this tool since many readers start by reading the conclusions of a paper.
Page 16, line 38: What does the “0.08 m” refer to?
Page 16, lines 34-43: Example of paragraph that likely needs language improvements.
Technical comments
Page 3, line 70: Misplaced white space in 50 m.
Page 7, line 39: Missing whitespace.
Page 7, line 56: Missing comma after additionally.
Page 15, line 92 and 93: Wrong reference format.
References
Fontrodona-Bach, A., Schaefli, B., Woods, R., Teuling, A. J., & Larsen, J. R. (2023). NH-SWE: Northern Hemisphere Snow Water Equivalent dataset based on in situ snow depth time series. Earth Syst. Sci. Data, 15(6), 2577-2599. https://doi.org/10.5194/essd-15-2577-2023
Lievens, H., Brangers, I., Marshall, H. P., Jonas, T., Olefs, M., & De Lannoy, G. (2022). Sentinel-1 snow depth retrieval at sub-kilometer resolution over the European Alps. Cryosphere, 16(1), 159-177. https://doi.org/10.5194/tc-16-159-2022
Citation: https://doi.org/10.5194/egusphere-2024-791-RC2 - AC2: 'Reply on RC2', Laura Sourp, 15 Jul 2024
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
488 | 139 | 39 | 666 | 15 | 14 |
- HTML: 488
- PDF: 139
- XML: 39
- Total: 666
- BibTeX: 15
- EndNote: 14
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1