the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Mapping snow depth over lake ice in Canada’s sub-arctic using ground-penetrating radar
Homa Kheyrollah Pour
Alex Maclean
Abstract. Ice thickness across lake ice is influenced mainly by the presence of snow and its distribution, as it directly impacts the rate of lake ice growth. The spatial distribution of snow depth over lake ice varies and is driven by wind redistribution and snowpack metamorphism, creating variability in the lake ice thickness. The accuracy and consistency of snow depth measurement data on lake ice are challenging and sparse to obtain. However, high spatial resolution lake snow depth observations are necessary for the next generation of thermodynamic lake ice models. Such information is required to improve the knowledge and understanding of snow depth distribution over lake ice. This study maps snow depth distribution over lake ice using ground-penetrating radar (GPR) two-way travel-time (TWT) with ~9 cm spatial resolution along transects totalling ~44 km over four freshwater lakes in Canada’s sub-arctic. The accuracy of the snow depth retrieval is assessed using in situ snow depth observations (n =2,430). On average, the snow depth derived from GPR TWTs for the early winter season is estimated with a root mean square error (RMSE) of 1.58 cm and a mean bias error of -0.01 cm. For the late winter season on a deeper snowpack, the accuracy is estimated with RMSE of 2.86 cm and a mean bias error of 0.41 cm. The GPR-derived snow depths are interpolated to create 1 m spatial resolution snow depth maps. Overall, this study improved lake snow depth retrieval accuracy and introduced a fast and efficient method to obtain high spatial resolution snow depth information, which is essential for the lake ice modelling community.
Alicia F. Pouw et al.
Status: final response (author comments only)
-
RC1: 'Comment on tc-2022-193', Anonymous Referee #1, 15 Oct 2022
Review on “Mapping snow depth over lake ice in Canada’s sub-arctic using ground-penetrating radar” by Pouw et al.
The snow cover on lake ice is of great significance for the growth and decay of lake ice, lake climatology, limnological hydrology, and lake ecology. It is a positive work to develop a new technology based on the ground penetrating radar to quickly obtain the snow depth over large lake-ice areas. Based on the observation system, the authors carried out observation experiments in four lakes in the Canadian sub-Arctic region, proving the applicability and application value of the observation method, especially proving that the observation ability for the shallow snow layer over the ice surface. Thus, it is a method worth popularizing. The obtained data of large-scale snow observation can be further applied to the numerical simulation of lake ice and limnological hydrological processes, to evaluate the impact of snow and lake ice layers on the ecological environment of frozen lakes, and to evaluate the satellite remote sensing products of snow over the lakes. The paper is well written and structured, the method description is appropriate, the data analysis is basically sufficient, and the conclusion is clear, so it is a research work worth publishing in the TC. However, there are still some problems in the current expressions. It is mainly about the physical analysis of some data statistics results, and the impact of destruction of snowmobile track for natural snow surface on the observation data. Therefore, I recommend that the paper can be considered for publication only after a few minor revisions.
General:
- Some statistical results based on observation data lack the analysis of potential physical mechanisms, for example, the difference of snow depth, density, relevant length in various lakes.
- The author said that snowmobile and sled rolling will increase the snow density and reduce the snow depth to a certain extent. The two impactscan offset each other, so their impacts are not significant. My suggestion here is whether you can further analyze the difference of the impact on thick and thin snow layers, on new and old snow layers, as well as on the snow accumulated in early December and the snow accumulated in late winter.
- This study presents observation dataobtained from one winter. Although the data spatial coverage is relatively large, there is still a lack of data representativeness. Therefore, it is suggested to increase the discussion of data representativeness obtained from the observed winter. How does the snow accumulation on land compare with previous years? What is the difference of the atmospheric precipitation, temperature and other parameters in the winter of the observation related to the climatology? etc. Through such comparison, the application value of observation data can be enhanced.
Special comments:
- Line 15 “~9 cm spatial resolution along transects”9-cm is is the sampling resolution, not the data resolution, because you have not considered the footprint of observation. Therefore, it is recommended to further analyze the observation footprint of single observation.
- 1 Introduction: The application of observation data of snow over the lake ice cannot only focus on the developing of lake ice numerical model, but also be applied to lake ice phenology (e.g., Lei et al., 2012), lake ecology and other fields. The description of research background should be more comprehensive in the introduction.
Ref.: Lei R, Leppäranta M, Cheng B, et al. Changes in ice-season characteristics of a European Arctic lake from 1964 to 2008. Climatic change, 2012, 115(3-4): 725-739.
- Line 46 “Daily snow depths are reported across Canada usinginstruments, such as..” As you mentioned later, the SnowHydro Magnaprobe is a common method for snow depth measurement. Therefore, it should be introduced in introduction, and its advantages and disadvantages should be described, such as manual operation, which is not conducive to obtaining a wide range of snow depth observation data.
- Line 83 “It is expected that the wind fetch and shorelinevegetation affect the snow distribution”, However, in the later data analysis, the impact of these two factors on different lakes has not been discussed enough.
- Table 2: Could you explain why the Long Lakehas a relative large snow density compared to other lakes?
- Line 199 “area = 4 ha”ha is not the International Standard Unit.
- Figure 5: In fact, there are multiple intersections in the observation transects for all lakes, which means that there should be two observations at these intersections. In order to explain the stability of the observation and retrieval results, it is necessary to compare the repeated observation results obtained from these measurement intersections.
- Lines 222, 225 “Long Lake showed the lowest agreement”, “with Vee Lake being the most accurate”: Corresponding to such measurement difference, some physical explanations are required.
- relative error = 11.04 %, and other somewhere: For relative errors, it is not necessary to retain two decimal places, because the accuracy of the evaluation cannot reach this level.
- Line 240 “However, the relative error was improved on Landing-M Lake with a deeper snowpack (5.33 %) than that of Landing-DLake (8.06 %). During the later season, the GPR could derive the minimum snow depths seen on Landing Lake, as opposed to that in the early season, where” Some further explanation is needed here, not only to give the data results.
Citation: https://doi.org/10.5194/tc-2022-193-RC1 - AC1: 'Reply on RC1', Alicia Pouw, 19 Jan 2023
-
CC1: 'Review on “Mapping snow depth over lake ice in Canada’s sub-arctic using ground-penetrating radar” by Pouw et al.', Fei Xie, 20 Oct 2022
Snow accumulation on lake ice can have a significant impact on the evolution of the ice cover, particularly as wind-driven forces can cause significant spatial variation in the distribution of snow on the ice cover. This is more pronounced in areas of low snowfall, where there are surface conditions of both bare ice and snow crossings, which are critical to the overall heat content of the lake ice. It is currently difficult to quantify precisely the spatial distribution of snow thickness, and shallow snow cover is also a dominant natural phenomenon in many mid-latitude regions. This technique allows rapid access to snow depths over large areas of lake ice as opposed to traditional manual measurements and fixed-point automated observations. It is a valuable tool for estimating and analysing the thermal balance of the ice surface over the entire lake ice and for gaining a clearer understanding of the physical processes involved in snow redistribution.
Some questions are as follows:
- The rolling of snowmobile and sled compacts the snow, can the reduction in depth and the increase in density be completely offset? This is because in the case of the study where the snow is deeper, the compaction does not act evenly across the snow layer resulting in an uneven increase in overall density. Would it be better if in the future the snowmobiles were to "push" the sleds instead of "pulling" them, or would it be better if they were to be carried by drones?
- The authors obtained snow depth data with a large spatial coverage and also assessed the accuracy of the data. Consideration could be given to discussing this in the context of climatic background and terrain features to improve the potential application of the data. For example, is the variability in snow depth influenced by the wind speed and direction prior to measurement? Is the greater depth of snow on the banks due to the barrier effect of vegetation or bank slopes?
- Line 17-19, “On average, the snow depth derived from GPR TWTs for the early winter season is estimated with a root mean square error (RMSE) of 1.58 cm and a mean bias error of -0.01 cm. For the late winter season on a deeper snowpack, the accuracy is estimated with RMSE of 2.86 cm and a mean bias error of 0.41 cm.” Is the increase in mean bias error in the late winter season due to the effect of increased snow depth or the effect of deterioration?
- Line 34-36, “As warming is occurring in Northern Canada at twice the global rate and is expected to continue to increase (Zhang et al, 2019)…” Has warming had an impact on snowfall? Is there a gradual increase or decrease in the amount of snow in winter?
- Line 75, “(2) validate the snow-depth retrieval algorithm using in situ observations…” Measuring uncompacted or compacted snow layers?
- In addition to the spatial distribution of snow depth, I would like to know if you have also carried out research on the spatial distribution of ice thickness? Or is your technique actually focused on the identification of the snow-ice interface for shallow snow layers and is not actually an optimal technique for the identification of the ice-water interface?
Citation: https://doi.org/10.5194/tc-2022-193-CC1 - AC3: 'Reply on CC1', Alicia Pouw, 19 Jan 2023
-
RC2: 'Review of tc-2022-193', Anonymous Referee #2, 20 Oct 2022
The manuscript "Mapping snow depth over lake ice in Canada's sub-arctic using ground-penetrating radar" presents a study that takes a commonly used method (GPR) and applies it to snow on lake ice. The study is able to cover great distances with high spatial resolution of observations and compare GPR depth estimates to manual depth measurements with a Magnaprobe. The GPR method resulted in an estimated RMSE of 1.58 cm with a mean bias of -0.01 cm during the early season, and RMSE of 2.86 cm and bias of 0.41 cm later in the season.
Overall, the authors produce a very nice dataset that can be used for modeling efforts and potentially remote sensing validation. However, I do not see anything that justifies this study to be at the level of a "Research Article" in The Cryosphere. Again, this is a great dataset but a more robust analysis of data would need to be presented to be a research article, in my opinion. As it is I think it a great "data paper" or potentially a "technical note" type of manuscript. Unfortunately, The Cryosphere does not publish these types of papers so I recommend either submitting to another journal pretty much as is, or providing further quantitative analysis to bump it up to being a full research article. The variograms are a great start, but I think more information on the spatial variability of the snow on lake ice could be good to include. This could include for example: directional variograms to investigate isotropy, variability or depth as a function of distance to shore or distance to islands, does topography of the shore or presence of trees impact anything. I think that the authors started to go down this route with Figure 9 but it needs to continue for more statistical quantifications, in my opinion.
One reason further analysis would be necessary is because the authors did not develop any new tools advance any of the methods to collect the data. Further minor comments are listed below by line number.
15: 9 cm spatial resolution is the spacing between traces, but after you aggregate the data it is a 1 m raster correct? This is the resolution of the data that should be reported and also incorporates the footprint of observations.
115: "was" should be "were"
158: How was the Wong et al. algorithm applied? Matlab? Python? Please specify.
184-190: How much variability occurred over the 6 m. It seems to me that by choosing only values that closely match one would underestimate the magnitude of the error/bias. As it is written, I do not see a justification for this current method and think the authors should use all values within the 6 m range to calculate the comparison metrics.
200: what is meant by "closed-off areas"
236: Given such low density values, I am not sure that teh Kovacs equation is appropriate. Kovacs was developed for much denser firn. Di Paolo et al. (2018) and Webb et al. (2021) could be good references for a more appropriate equation.
references:
Di Paolo, F.; Cosciotti, B.; Lauro, S.E.; Mattei, E.; Pettinelli, E. Dry snow permittivity evaluation from density: A critical review. In Proceedings of the 2018 17th International Conference on Ground Penetrating Radar (GPR), Rapperswil, Switzerland, 18–21 June 2018; pp. 1–5
Webb, R.W.; Marziliano, A.; McGrath, D.; Bonnell, R.; Meehan, T.G.; Vuyovich, C.; Marshall, H.-P. In Situ Determination of Dry and Wet Snow Permittivity: Improving Equations for Low Frequency Radar Applications. Remote Sens. 2021, 13, 4617. https://doi.org/10.3390/rs13224617
These comments are meant to be constructive. I think this is an excellent dataset and good work.
Citation: https://doi.org/10.5194/tc-2022-193-RC2 - AC2: 'Reply on RC2', Alicia Pouw, 19 Jan 2023
Alicia F. Pouw et al.
Alicia F. Pouw et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
411 | 133 | 18 | 562 | 1 | 1 |
- HTML: 411
- PDF: 133
- XML: 18
- Total: 562
- BibTeX: 1
- EndNote: 1
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1