Implementing spatially and temporally varying snow densities into the GlobSnow snow water equivalent retrieval
- 1Finnish Meteorological Institute, PO Box 503, FIN-00101 Helsinki, Finland
- 2Climate Research Division, Environment Climate Change Canada, Toronto, Canada
- 1Finnish Meteorological Institute, PO Box 503, FIN-00101 Helsinki, Finland
- 2Climate Research Division, Environment Climate Change Canada, Toronto, Canada
Abstract. Snow water equivalent (SWE) is a valuable characteristic of snow cover, and it can be estimated using passive spaceborne radiometer measurements. The radiometer based GlobSnow SWE retrieval methodology, which assimilates weather station snow depth observations into the retrieval, has improved reliability and accuracy of SWE retrieval when compared to stand-alone radiometer SWE retrievals. To further improve the GlobSnow SWE retrieval methodology, we investigate implementing spatially and temporally varying snow densities into the retrieval procedure. Thus far, the GlobSnow SWE retrieval has used a constant snow density throughout the retrieval despite differing locations, snow depth or time of winter. This constant snow density is a known source of inaccuracy in the retrieval. Three different versions of spatially and temporally varying snow densities are tested over a 10-year period (2000–2009). These versions use two different spatial interpolation techniques, ordinary Kriging interpolation and inverse distance weighted regressing (IDWR). All versions were found to improve the SWE retrieval compared to the baseline GlobSnow v3.0 product although differences between versions are small. Overall, the best results were obtained by implementing IDWR interpolated densities into the algorithm, which reduced RMSE (Root Mean Square Error) and MAE (Mean Absolute Error) by about 4 mm and 5 mm when compared to the baseline GlobSnow product, respectively. Furthermore, implementing varying snow densities into the SWE retrieval improves the magnitude and seasonal evolution of the Northern Hemisphere snow mass estimate compared to the baseline product and a product post-processed with varying snow densities.
Pinja Venäläinen et al.
Status: closed
-
CC1: 'Comment on tc-2022-227', Alain Royer, 12 Dec 2022
This article presents a new version of Globe Snow v.3.0 (new product) including a retrieval with variable density. The principle of the approach has already been presented and discussed by P. Venäläinen in TC 2021. In this paper, the improvement is described and analyzed over a larger dataset.
Even if the retrieval of SWE including a variable density in the inversion process improves the SWE a bit (reduction of the bias by 5% on average), the lack of sensitivity of the retrieval for large amounts of snow (SWE>150 mm) remains a major problem with this approach (Fig. 8). This should be recalled in conclusion, even if it is known.
But this long database has the merit to exist and has to be kept up to date.
Some aspects to be clarified (specially Fig. 2):
Abstract: reduced RMSE and MAE by about 4 mm and 5 mm : in %tage?
L.94 For GMON too, the snow density was calculated for SWE and snow depth? But snow depth is not systematically measured at the GMON sites?
L. 170 Figure 2. Not clear : The “Dynamic snow density information’ is derived from Tb (a red arrow is missing?) and from Step 2 : the upper arrow should go the other way?
L. 380 The figures show the SWE retrievals. Wording of the caption not clear: may be confused with “snow” density scatter plots ! To reword more clearly?
- AC3: 'Reply on CC1', Pinja Venäläinen, 31 Jan 2023
-
RC1: 'Comment on tc-2022-227', Nicolas Marchand, 27 Dec 2022
The work presented in this article aims to push further what was already described in P. Venäläinen 2021. The larger dataset used to analyze or validate the SWE retrieval value helps discussing the improvement of the method. The insertion of dynamic snow densities in the retrieval process of the SWE algorithm seems to be an interesting way to move forward, but it is not entirely clear through the paper how relevant the improvements are in terms or relative values (percentages). This might help seeing more clearly the contribution to the proposed improvement on the method. The limitations still existing on the globsnow swe retrievals are not discussed enough in the conclusion of this paper.
L45-47 - SWE retrieval limited by high uncertainties… put an exemple of those high uncertainties, seems rather relevant and would avoid to go find them in the literature, even if all necessary sources are there
L80… Snow density and SWE data... How were taken into account the variabilities of the different sources of the large dataset ? Did you take into account the variability and incertitude on the measurments, or on the methods / models used to obtain them ? You could include some basic informations on those uncertainties in your table 1.
L151 - Could go into more details about those snow free areas… radiometers… which frequencies… optical… what do you use… ? How accurate is it ? Might be relevant to have more insight.
L217 - Don’t you need to go more subscale than that for your variograms fittings ? East and west Canada/USA separately, Europe and Asia separately ? Can you justify this choice ? Have “subcontinental” / “regional variograms” have been tested, and how would their results have compared to the IDWR method ?
L265 - Supports previous point to also look into more detailed characterization (variograms to fit) regarding the areas… west versus east north America for exemple…
L300 - Figure 4… increase police size of legend on the left and right of the plots… very difficult to read
L330 - Paragaph… you put some facts out… might be appreciated for them to be backed with a few references
L337 - Difference in grain size… reference
L465 - Figure 10… increase legends left and right
L495 - Dot missing
L522 - It is not clear whether you put out this specific example a positive or negative consequence ?
L531 - How would you deal with the errors of reanalysis depending on the environment, latitude, lacking or overestimation of precipitations, … ?
L540 – 550 - You don’t make it clear what it is you recommend to be used… one of the 3, or multiples at the same time… or different version depending on the geography ?
- AC1: 'Reply on RC1', Pinja Venäläinen, 31 Jan 2023
-
RC2: 'Comment on tc-2022-227', Jennifer Jacobs, 09 Jan 2023
The inclusion of density in the GlobSnow SWE estimates for both snow grain size and final estimate of SWE are a welcome improvement for global estimates of SWE. The key findings, that dynamic density improves SWE estimates, is not surprising. It is interesting that there is little difference among the three methods that are compared. Additional recommendations for the presentation of the comparisons are suggested below. It is also notable that the performance did not improve as much as one might expect over using a constant density value.
The collection of in situ density values was a massive undertaking. The point made late in the manuscript regarding near real time estimates of SWE not being possible with these same in situ value suggests that a decadal field rather than annual will serve the community better and eliminate the annual collection and QA/QC of density measurements. Since IDWR is recommended and decadal seems to be the most viable and flexible solution, Table 3 should include the performance of decadal IDWR. It is recommended that performance for individual years be assessed using the decadal minus one data set (leave one out) to assess the range of possible performance in any given year. Also, consider making the in-situ density dataset openly available. This resource would extend the value beyond GlobSnow users to the snow community members. For example, there is a rapidly expanding capacity to make snow depth measurements using lidar and structure from motion on airborne and drone platforms that would greatly benefit from insights and data in this current effort.
L24 Provide a measure of the average or percent improvement
L112 “Around 19 GHz…” is an awkward phrasing. The point being that SSM/I and SSMIS have slightly different frequencies might be stated in a clearer manner.
L118 to 119 “removing measurements from stations where the mean March SWE exceeds 150 cm in at least 50% of the years that the station has had at least 20 measurements” This criteria is difficult to follow.
L120 to 121 How was snow wetness determined?
L180 and others “significant differences” implies a statistical test was performed. Please rephrase.
L260 and others Results indicate differences in western versus eastern NA, but are not presented. Perhaps present in supplementary material. Similar for data in Russia later in the manuscript
Table 2 and others Add columns for average values of in situ and modeled
L284 Paragraph break needed starting at “Figure 4”
L294 How was the decision made to use a single semivariogram for such large regions, yet a different semivariogram was determined for each day?
L347 “grain”
Figure 6 Excellent figure, shows that performance varies by month. Additional monthly results would be valuable.
Section 5.2.2 While it is fine to present a single year, please provide information about why that year was selected and whether it is representative for most of the study regions.
L373 concentration
Figure 7. Reduce the number of significant digits to 3.
Figure 8 A density scatter plots would be more useful. Scatter plots should use the same scale
in the x and y -axes (x is much longer). This figure would be valuable to be presented on a monthly basis?
Figure 10 caption should describe the middle row as well.
The discussion needs to be expanded. This first paragraph is unnecessary because it largely repeats the introduction and the methods rather than putting the work in context. There are a number of topics that would be valuable to consider in the discussion. For example,
- It appears that performance is not the same globally. One suggestion is to discuss why North America performance is so poor compared to Eurasia. Another is to address the challenges in Russia in greater detail. Also, does performance differ by year – most applications are interested in changes over time or specific years rather than average conditions.
- There are a number of researchers who have used earlier versions of GlobSnow for applications. The impact and value of these modest improvements in previous research and to the applications in the first paragraph in the introduction could be discussed.
- The differences between the global snow density product produced here versus other products (global or otherwise) and how the approaches researched for this paper might provide value.
Please consider these to be potential topics that this paper is uniquely qualified to comment on and a request to consider at least one broader topic in the discussion as opposed to a request to discuss all of the examples listed above.
L535 Is there a final recommendation on which approach will be used? Will there be a revised GlobSnow dataset in the future or will the algorithm change moving forward?
Overall, this manuscript presents a clear next generation approach to providing improved estimates of SWE globally. Well done.
- AC2: 'Reply on RC2', Pinja Venäläinen, 31 Jan 2023
Status: closed
-
CC1: 'Comment on tc-2022-227', Alain Royer, 12 Dec 2022
This article presents a new version of Globe Snow v.3.0 (new product) including a retrieval with variable density. The principle of the approach has already been presented and discussed by P. Venäläinen in TC 2021. In this paper, the improvement is described and analyzed over a larger dataset.
Even if the retrieval of SWE including a variable density in the inversion process improves the SWE a bit (reduction of the bias by 5% on average), the lack of sensitivity of the retrieval for large amounts of snow (SWE>150 mm) remains a major problem with this approach (Fig. 8). This should be recalled in conclusion, even if it is known.
But this long database has the merit to exist and has to be kept up to date.
Some aspects to be clarified (specially Fig. 2):
Abstract: reduced RMSE and MAE by about 4 mm and 5 mm : in %tage?
L.94 For GMON too, the snow density was calculated for SWE and snow depth? But snow depth is not systematically measured at the GMON sites?
L. 170 Figure 2. Not clear : The “Dynamic snow density information’ is derived from Tb (a red arrow is missing?) and from Step 2 : the upper arrow should go the other way?
L. 380 The figures show the SWE retrievals. Wording of the caption not clear: may be confused with “snow” density scatter plots ! To reword more clearly?
- AC3: 'Reply on CC1', Pinja Venäläinen, 31 Jan 2023
-
RC1: 'Comment on tc-2022-227', Nicolas Marchand, 27 Dec 2022
The work presented in this article aims to push further what was already described in P. Venäläinen 2021. The larger dataset used to analyze or validate the SWE retrieval value helps discussing the improvement of the method. The insertion of dynamic snow densities in the retrieval process of the SWE algorithm seems to be an interesting way to move forward, but it is not entirely clear through the paper how relevant the improvements are in terms or relative values (percentages). This might help seeing more clearly the contribution to the proposed improvement on the method. The limitations still existing on the globsnow swe retrievals are not discussed enough in the conclusion of this paper.
L45-47 - SWE retrieval limited by high uncertainties… put an exemple of those high uncertainties, seems rather relevant and would avoid to go find them in the literature, even if all necessary sources are there
L80… Snow density and SWE data... How were taken into account the variabilities of the different sources of the large dataset ? Did you take into account the variability and incertitude on the measurments, or on the methods / models used to obtain them ? You could include some basic informations on those uncertainties in your table 1.
L151 - Could go into more details about those snow free areas… radiometers… which frequencies… optical… what do you use… ? How accurate is it ? Might be relevant to have more insight.
L217 - Don’t you need to go more subscale than that for your variograms fittings ? East and west Canada/USA separately, Europe and Asia separately ? Can you justify this choice ? Have “subcontinental” / “regional variograms” have been tested, and how would their results have compared to the IDWR method ?
L265 - Supports previous point to also look into more detailed characterization (variograms to fit) regarding the areas… west versus east north America for exemple…
L300 - Figure 4… increase police size of legend on the left and right of the plots… very difficult to read
L330 - Paragaph… you put some facts out… might be appreciated for them to be backed with a few references
L337 - Difference in grain size… reference
L465 - Figure 10… increase legends left and right
L495 - Dot missing
L522 - It is not clear whether you put out this specific example a positive or negative consequence ?
L531 - How would you deal with the errors of reanalysis depending on the environment, latitude, lacking or overestimation of precipitations, … ?
L540 – 550 - You don’t make it clear what it is you recommend to be used… one of the 3, or multiples at the same time… or different version depending on the geography ?
- AC1: 'Reply on RC1', Pinja Venäläinen, 31 Jan 2023
-
RC2: 'Comment on tc-2022-227', Jennifer Jacobs, 09 Jan 2023
The inclusion of density in the GlobSnow SWE estimates for both snow grain size and final estimate of SWE are a welcome improvement for global estimates of SWE. The key findings, that dynamic density improves SWE estimates, is not surprising. It is interesting that there is little difference among the three methods that are compared. Additional recommendations for the presentation of the comparisons are suggested below. It is also notable that the performance did not improve as much as one might expect over using a constant density value.
The collection of in situ density values was a massive undertaking. The point made late in the manuscript regarding near real time estimates of SWE not being possible with these same in situ value suggests that a decadal field rather than annual will serve the community better and eliminate the annual collection and QA/QC of density measurements. Since IDWR is recommended and decadal seems to be the most viable and flexible solution, Table 3 should include the performance of decadal IDWR. It is recommended that performance for individual years be assessed using the decadal minus one data set (leave one out) to assess the range of possible performance in any given year. Also, consider making the in-situ density dataset openly available. This resource would extend the value beyond GlobSnow users to the snow community members. For example, there is a rapidly expanding capacity to make snow depth measurements using lidar and structure from motion on airborne and drone platforms that would greatly benefit from insights and data in this current effort.
L24 Provide a measure of the average or percent improvement
L112 “Around 19 GHz…” is an awkward phrasing. The point being that SSM/I and SSMIS have slightly different frequencies might be stated in a clearer manner.
L118 to 119 “removing measurements from stations where the mean March SWE exceeds 150 cm in at least 50% of the years that the station has had at least 20 measurements” This criteria is difficult to follow.
L120 to 121 How was snow wetness determined?
L180 and others “significant differences” implies a statistical test was performed. Please rephrase.
L260 and others Results indicate differences in western versus eastern NA, but are not presented. Perhaps present in supplementary material. Similar for data in Russia later in the manuscript
Table 2 and others Add columns for average values of in situ and modeled
L284 Paragraph break needed starting at “Figure 4”
L294 How was the decision made to use a single semivariogram for such large regions, yet a different semivariogram was determined for each day?
L347 “grain”
Figure 6 Excellent figure, shows that performance varies by month. Additional monthly results would be valuable.
Section 5.2.2 While it is fine to present a single year, please provide information about why that year was selected and whether it is representative for most of the study regions.
L373 concentration
Figure 7. Reduce the number of significant digits to 3.
Figure 8 A density scatter plots would be more useful. Scatter plots should use the same scale
in the x and y -axes (x is much longer). This figure would be valuable to be presented on a monthly basis?
Figure 10 caption should describe the middle row as well.
The discussion needs to be expanded. This first paragraph is unnecessary because it largely repeats the introduction and the methods rather than putting the work in context. There are a number of topics that would be valuable to consider in the discussion. For example,
- It appears that performance is not the same globally. One suggestion is to discuss why North America performance is so poor compared to Eurasia. Another is to address the challenges in Russia in greater detail. Also, does performance differ by year – most applications are interested in changes over time or specific years rather than average conditions.
- There are a number of researchers who have used earlier versions of GlobSnow for applications. The impact and value of these modest improvements in previous research and to the applications in the first paragraph in the introduction could be discussed.
- The differences between the global snow density product produced here versus other products (global or otherwise) and how the approaches researched for this paper might provide value.
Please consider these to be potential topics that this paper is uniquely qualified to comment on and a request to consider at least one broader topic in the discussion as opposed to a request to discuss all of the examples listed above.
L535 Is there a final recommendation on which approach will be used? Will there be a revised GlobSnow dataset in the future or will the algorithm change moving forward?
Overall, this manuscript presents a clear next generation approach to providing improved estimates of SWE globally. Well done.
- AC2: 'Reply on RC2', Pinja Venäläinen, 31 Jan 2023
Pinja Venäläinen et al.
Pinja Venäläinen et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
263 | 72 | 17 | 352 | 2 | 4 |
- HTML: 263
- PDF: 72
- XML: 17
- Total: 352
- BibTeX: 2
- EndNote: 4
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1