Point-to-Surface Upscaling Algorithms for Snow Depth Ground Observations

Round 1

Reviewer 1 Report

The authors wrote the paper titled "Point-to-surface Upscaling Algorithms for Snow Depth Ground Observations", which is a good insight into the snow depth calculations. However, there are a few points that need to improve, which will help the readers.

1. A conclusion of the abstract is missing. Would you please add a sentence addressing the implication/application of this finding or results?
2. It is great that the authors described all the methods in the paper. However, a section "Data analysis" is missing where authors should provide details on which software, program, or language (e.g., R?) they used for analysis. How do the authors prepare the plots? Please mention which package or tool was used to prepare the plots in the section.   
3. A flow diagram from input data to the results with the decision for each step will be great to visualize the paper at once. It is the authors' decision; however, it would be great if the authors consider it.
4. A spatial plot of topography or elevation with figure 1, would be great to understand the results and discussion. Would you please add a plot of elevation?
5. Table 2 and Table 3- please highlight the best result in each section (maybe "bold" would be enough) that will be helpful for readers to take away from the table.
6. Page 7 and line 10: RMSE height is only 1.87 cm, is this median value or average? Why consider median for the validation, where you reported average on page 6 (Table 2 and Table 3)? An explanation would be helpful to the readers.
7. Would you please add a few sentences at the end of the conclusion section that discusses the implication of this paper or any specific limitation of the study or potential future scope of this research?  

Author Response

Response to Reviewer #1

Firstly, on behalf of all authors, we appreciate your great help and suggestions for this manuscript. Based on your comments, we have revised our manuscript and answered your questions on a point-by-point basis. In addition, the whole manuscript was checked by using a professional English editing service of American Journal Experts. Please see the following replies and the revised manuscript.

Specific comments

1. A conclusion of the abstract is missing. Would you please add a sentence addressing the implication/application of this finding or results?

Response: Thank you for the comment. Did as suggested. Please see the revision on Page 1 lines 25-29:

‘The results of the study are expected to provide a reference for developing a point-to-surface upscaling method based on snow depth ground observations and further solve the uncertainties caused by scale effects in snow depth and other land surface parameters inversion and validation by using remote sensing data’

 

2. It is great that the authors described all the methods in the paper. However, a section "Data analysis" is missing where authors should provide details on which software, program, or language (e.g., R?) they used for analysis. How do the authors prepare the plots? Please mention which package or tool was used to prepare the plots in the section.

Response: In the revised manuscript, the details data process and analysis were described for each method, as well as the software used in this study. We trust that readers can repeat this work according to our description. Please see the revised section 3 ‘Method’. Thank you so much for this comment.

 

3. A flow diagram from input data to the results with the decision for each step will be great to visualize the paper at once. It is the authors' decision; however, it would be great if the authors consider it.

Response: Thank you for the suggestion. We tried to make a data processing flow diagram based on your advice, we regret that several versions of the flow diagram were made but no one was satisfactory. One flow diagram is attached, but we thought that there was no practical value if added to the manuscript. Because the overall data process is very simple, but if each method in detail is involved, the data processing steps are too complex to make an ideal flow diagram. Based on your last comments, the detailed data processes are described in section 3. Therefore, we consider not adding the flow diagram to the revised paper this time. Thank you again.

Figure x. The flow diagram of upscaling ground observations to pixel scale. (Please see the attachment.)

 

4. A spatial plot of topography or elevation with figure 1, would be great to understand the results and discussion. Would you please add a plot of elevation?

Response: As you suggested, the satellite image and DEM were added to figure 1. Thank you for your advice.

 

Figure 1. Upscaling results comparison of (c) ordinary Kriging, (d) simple Kriging, (e) random forest, and (f) Bayesian maximum entropy for a 25 km plot, and the satellite image and DEM for the snow plot are shown in (a) and (b), respectively.

 

5. Table 2 and Table 3- please highlight the best result in each section (maybe "bold" would be enough) that will be helpful for readers to take away from the table.

Response: Thank you for your comments. The best results in each table were highlighted in ‘bold’ as suggested. Please see the revised Tables 2 and 3.

 

6. Page 7 and line 10: RMSE height is only 1.87 cm, is this median value or average? Why consider median for the validation, where you reported average on page 6 (Table 2 and Table 3)? An explanation would be helpful to the readers.

Response: Sorry for the unclear presentation. An explanation for using median and box height is revised as you suggested. Please see the revised manuscript on Page 7 and lines 5-12:

‘Considering that the use of the average error (RSME) may affect by extreme values, and cannot reflect the stability and overall level of the various upscaling methods. Therefore, based on the cross-validation results of different snow plots, the RMSE boxplot is drawn to verify further the results of different point-to-surface upscaling methods (Figure 2). The median in the box diagram represents the overall level of each upscaling method, and the box height (the difference between the 25th percentile and 75th percentile) can indicate whether the RMSE is discrete, which corresponds to the stability of the different methods. The black dots in the box line diagram indicate outliers.’

Thank you for the comments.

 

7. Would you please add a few sentences at the end of the conclusion section that discusses the implication of this paper or any specific limitation of the study or potential future scope of this research?

Response: Thank you for your comments. Did as suggested. Please see the revision on Page 12 lines 1-9:

 ‘So far, the passive microwave remote sensing is still the main data source to acquire the snow depth information on a regional and global scale. The traditional development and validation of the passive microwave snow depth inversion model based on site observation have great uncertainty. Because the large errors may occur by the mismatch of observation scale from ground and space, respectively. The point-to-surface upscaling method can obtain the ground observation data at pixel scale and solve the uncertainty caused by the scale effect. This study is expected to provide a reference for developing a point-to-surface snow depth upscaling scheme and further solve the uncertainty problem caused by scale effects in passive microwave snow depth inversion and validation.’

Author Response File: Author Response.pdf

Reviewer 2 Report

Page 1 line 33, “a passive microwave”?

Page 2 line 4-5, “due to the coarse spatial resolution of passive microwave data, the inversion of snow depth has large errors” is not right. The coarse spatial resolution is not the reason leading to large errors.

Page 2 line 15-25, the “author+year” and “numeric” are two different citation styles. Please do not use them at the same time, i.e., you should try to leave out the author names when you use the numeric citation style.

Page 2 line 33-35, this is a wrong structure: “To provide a reference for developing a point-to-surface upscaling scheme for snow depth and further solve the uncertainties caused by the passive microwave snow depth inversion and validation due to scale effect.”

Page 2 line 38-39, the full name of the NCDC is wrong.

Page 2 line 40-41, what is the difference of “snow samples” and “snow observation points”? These two terms are confusing. Maybe the former is “plots”?

Page 2 line 42-43, the air temperature, altitude, latitude, and longitude are not “snow observation parameters”.

Page 3 from line 17 to the following lines, what is “true pixel scale value”? Do you mean the field observation is not true? It is wrong. I suggest to delete “true”.

Page 3 from line 23 to the following lines, “In the formula” should be replaced by “where”.

Page 4 line 39, what is the difference between “bootstrap aggregation” and “bootstrap aggregation algorithm”? I suggest to delete them.

Page 6 line 4, “Accuracy valuation” should be “Accuracy assessment”.

Page 7-8, Table 3 and Figure 2 are repeating information.

Subsection 4.3 is too long, thus I suggest to divide it to two subsections to present the two different sensitivity analysis. And please do not directly place the figure below the subsection caption.

Author Response

Response to Reviewer #2

Firstly, on behalf of all authors, we appreciate your great help and suggestions for this manuscript. Based on your comments, we have revised our manuscript and answered your questions on a point-by-point basis. In addition, the whole manuscript was checked by using a professional English editing service of American Journal Experts. Please see the following responses and the revised manuscript.

Specific comments

1) Page 1 line 33, “a passive microwave”?

Response: Sorry for this wrong expression. We revised this sentence as (Page 1 lines 39-41):

‘There are large errors and uncertainties in the construction and verification of the snow depth inversion model from a passive microwave remote sensing data based on point-based snow depth ground observation data.’

 

2) Page 2 line 4-5, “due to the coarse spatial resolution of passive microwave data, the inversion of snow depth has large errors” is not right. The coarse spatial resolution is not the reason leading to large errors.

Response: We agree with the comments. The sentence was revised as in (Page 2 line 12-14):

‘However, due to the coarse resolution of the passive microwave remote sensing data, the problem of mixed pixels caused by the high spatial heterogeneity of snow cover is the main reason for the great uncertainties of passive microwave snow depth inversion [8].’

Thank you for the comments.

 

3) Page 2 line 18-28, the “author+year” and “numeric” are two different citation styles. Please do not use them at the same time, i.e., you should try to leave out the author names when you use the numeric citation style.

Response: We have modified the citation style and omitted the authors' names as much as possible. Thank you for the comments. Please see the revision in Page 2 line 24-33.

‘Based on the geography of Greenland, the simple averaging method was used to upscale weather station data and compare it with Climate-SAF surface albedo products, and concluded that the result was consistent with the in-situ observation, but there were problems in areas where the terrain and environment changed dramatically [11]. Kriging method, which belongs to a geostatistical method, is the most widely used in point-to-surface upscaling [12-15]. The conclusion indicated that the co-kriging method is usually superior to the ordinary kriging method, the greater the influence of the covariates on the target parameters, the better the co-kriging results [15]. It has also been shown that the reduced major axis (RMA) is the most practical upscaling method for modeling regional net primary productivity by comparing RMA and Kriging methods [16].’

 

4) Page 2 line 33-35, this is a wrong structure: “To provide a reference for developing a point-to-surface upscaling scheme for snow depth and further solve the uncertainties caused by the passive microwave snow depth inversion and validation due to scale effect.”

Response: We revised this sentence as (Page 2 line 41-44):

‘It is expected to provide a reference for developing a point-to-surface upscaling scheme for snow depth ground observation into pixel scale, and further, solve the uncertainties in construction and evaluation of snow depth inversion model based on the passive microwave remote sensing data.’

Thank you for the comments.

 

5) Page 2 line 38-39, the full name of the NCDC is wrong.

Response: Sorry for the wrong name of the NCDDC. We revised the full name of the NCDDC as ‘National Cryosphere Desert Data Center’ in page 3 line 3-4. Thank you for remind.

 

6) Page 2 line 40-41, what is the difference of “snow samples” and “snow observation points”? These two terms are confusing. Maybe the former is “plots”?

Response: We revised ‘snow samples’ into ‘snow plots’ in the manuscript based on your suggestion. The snow plots are snow observed quadrat in two scales, 500m² and 25km², respectively, in this study. And the snow plots contain many snow observation points. We are sorry for the wrong expression. Thank you for the comments.

 

7) Page 2 line 42-43, the air temperature, altitude, latitude, and longitude are not “snow observation parameters”.

Response: We revised this sentence as (Page 3 line 6-8):

‘The observation parameters include snow depth, air temperature, snow pressure, snow density, altitude, latitude, and longitude.’

Thank you for the comments.

 

8) Page 3 from line 17 to the following lines, what is “true pixel scale value”? Do you mean the field observation is not true? It is wrong. I suggest to delete “true”.

Response: We agree with the comment and deleted the “true” in this sentence as you suggested. Thank you for the comments.

 

9) Page 3 from line 23 to the following lines, “In the formula” should be replaced by “where”.

Response: We replaced all "In the formula" with "where" after the formula. Thank you for the comments.

 

10) Page 4 line 39, what is the difference between “bootstrap aggregation” and “bootstrap aggregation algorithm”? I suggest to delete them.

Response: Did as suggested. Thank you for the comment.

 

11) Page 6 line 4, “Accuracy valuation” should be “Accuracy assessment”.

Response: Did as suggested. Thank you for the comment.

 

12) Page 7-8, Table 3 and Figure 2 are repeating information.

Response: Table 3 and Figure 2 show the mean and median results of the RMSE, respectively. The reason why use the median is to further verify our results. Thank you for the comment. Please see the revised sentences on page 7, lines 5-12:

“Considering that the use of the average error (RSME) may affect by extreme values, and cannot reflect the stability and overall level of the various upscaling methods. Therefore, based on the cross-validation results of different snow plots, the RMSE boxplot is drawn to verify further the results of different point-to-surface upscaling methods (Figure 2). The median in the box diagram represents the overall level of each upscaling method, and the box height (the difference between the 25th percentile and 75th percentile) can indicate whether the RMSE is discrete, which corresponds to the stability of the different methods.”

 

13) Subsection 4.3 is too long, thus I suggest to divide it to two subsections to present the two different sensitivity analysis. And please do not directly place the figure below the subsection caption.

Response: Did as suggested. We split subsection 4.3 into two parts, ‘4.3 Error analysis’, and ‘4.4 Sensitivity analysis’. The figure position was changed based on your advice. Thank you for your great comments and kind reminder.

Author Response File: Author Response.pdf

Reviewer 3 Report

 

The article is devoted to a study of methods for scaling ground-based snow depth observations to the scale of satellite data. Although the authors have done considerable work, particularly in the analysis of the scaling results, the paper has a number of shortcomings.

The authors have not sufficiently described the models used, especially on which dataset and how machine learning and the selection of model coefficients were performed. It requires a detailed description of which data set was used for training, model verification, and which functions were used (as well as probability density).

Satellite data and how it was used is not presented at all, which does not give an idea about the adequacy of the used methods. For example, in Figure 1, I recommend that the data from the satellite be given.

Also, when analyzing the results in Figure 3, it may be useful to indicate which points are from which region (respectively from Table 1), for example, Northern Xinjiang points No. 1-10, Northeast 11-14 Tibetan Plateau 15-16.

On page 9 (lines 20-24) replace "depression" with "descending" and replace the sentence "With the depression and uplift of the terrain..." with "With the ascent or descent of the terrain".

 

After correcting the above deficiencies, article can be published.

Author Response

Response to Reviewer #3

Firstly, on behalf of all authors, we appreciate your great help and suggestions for this manuscript. Based on your comments, we have revised our manuscript and answered your questions on a point-by-point basis. In addition, the whole manuscript was checked by using a professional English editing service of American Journal Experts. Please see the following responses and the revised manuscript.

Specific comments

1) The authors have not sufficiently described the models used, especially on which dataset and how machine learning and the selection of model coefficients were performed. It requires a detailed description of which data set was used for training, model verification, and which functions were used (as well as probability density).

Response: In the revised manuscript, the details data process and analysis were described for each method, as well as the software used in this study. We trust that readers can repeat this work according to our description. Please see the revised section 3 ‘Method’. Thank you so much for this comment.

 

2) Satellite data and how it was used is not presented at all, which does not give an idea about the adequacy of the used methods. For example, in Figure 1, I recommend that the data from the satellite be given.

Response: Did as you suggested. The satellite image as well as DEM are added in Figure 1. The implication of our results was also addressed in the Abstract and Conclusion section, to better understand our results and the significance of this study. Thank you for the comments.

Figure 1. Upscaling results comparison of (c) ordinary Kriging, (d) simple Kriging, (e) random forest, and (f) Bayesian maximum entropy for a 25 km plot, and the satellite image and DEM for the snow plot are shown in (a) and (b), respectively. (Please see the attachment)

 

3. Also, when analyzing the results in Figure 3, it may be useful to indicate which points are from which region (respectively from Table 1), for example, Northern Xinjiang points No. 1-10, Northeast 11-14 Tibetan Plateau 15-16.

Response: In captain Figure 3, we described the points from which region based on your suggestion. Thank you for your advice.

‘Figure 3. Cross-validation results for each plot: (a) 25 km² snow plots (Plots No.1 – 12 are from northern Xinjiang, and No.13- 15 are from the northeast, respectively), and (b) 500 m² snow plots (Plots No.1 - 10 are from northern Xinjiang, No.11 – 14 are from the Qinghai Tibet Plateau, and No.15 – 16 are from the Northeast, respectively)’

 

4) On page 9 (lines 20-24) replace "depression" with "descending" and replace the sentence "With the depression and uplift of the terrain..." with "With the ascent or descent of the terrain".

Response: Thanks for the comments. The wrong expression was revised as you suggested.

 

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Thank the authors improve the manuscript according to the reviewers' comments. It can be published as it is.