Towards large-scale daily snow density mapping with spatiotemporally aware model and multi-source data

Wang, Huadong; Zhang, Xueliang; Xiao, Pengfeng; Che, Tao; Zheng, Zhaojun; Dai, Liyun; Luan, Wenbo

doi:https://doi.org/10.5194/tc-17-33-2023

Articles | Volume 17, issue 1

https://doi.org/10.5194/tc-17-33-2023

© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/tc-17-33-2023

© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 17, issue 1

Research article

|

09 Jan 2023

Research article |

| 09 Jan 2023

Towards large-scale daily snow density mapping with spatiotemporally aware model and multi-source data

Huadong Wang, Xueliang Zhang, Pengfeng Xiao, Tao Che, Zhaojun Zheng, Liyun Dai, and Wenbo Luan

Download

Final revised paper (published on 09 Jan 2023)
Preprint (discussion started on 24 Mar 2022)

Interactive discussion

Status: closed

CC1:
'Comment on tc-2022-45', Xiaofeng Li, 09 Apr 2022
Snow density plays a critical role in estimating water resources and predicting natural disasters such as floods, avalanches, and snowstorms. A GTWNN model was constructed for snow density estimation and achieved daily snow density mapping from 2013 to 2020 in China with the support of remote sensing, ground observation, and reanalysis data. This study provides important spatiotemporal parameters for snow cover hydrology and other aspects. The main suggestions and opinions are as follows:
L115, “Based on the SCA data, the snow cover duration (SCD) is calculated to account for the impact of gravity on snow density”, How to understand that snow density is affected by gravity, and what does it have to do with SCD?

"Spatiotemporally Aware Model" in the title is not mentioned in the manuscript and should be explained.

Whether the lack of observation data in 2019-2020 is related to the epidemic, making it impossible to conduct a large number of observations.

The verification result in Fig.4 is that all the data as a whole is added to the training model, or is the training divided by region and month? Is it the 10-fold validation result of the trained model? Please explain further

L200, Does the reason for the lower accuracies in Northeast China-Inner Mongolia consider the effect of different underlying surfaces on snow density? Forests and farmland in the Northeast, and grasslands in Inner Mongolia may have very different effects on snow density.

The reasons for the slightly lower accuracy in the snow melting and accumulation periods are not only the rapid changes in the snow density itself, and insufficient sampling in observation time and space, but also because the snow accumulation in the early stage of snow accumulation is less, and the water content when the snow melts Therefore, the observation is more difficult, and the observation error is relatively large.

The verification result of the snow density of ERA5 is worse than that of the model in this paper, but many parameters of ERA5 are used in the machine learning model of this paper, so the accuracy of these parameters, if there is also a large error, will not affect the final model accuracy?
Citation: https://doi.org/10.5194/tc-2022-45-CC1
- AC1:
  'Response to CC1', Xueliang Zhang, 03 May 2022
  
  We would like to thank the reviewers for the constructive comments, which helped to substantially improve the manuscript. We have addressed their comments in detail in the pdf supplement.
  
  Citation: https://doi.org/10.5194/tc-2022-45-AC1
  - CC2: 'Reply on AC1', Xiaofeng Li, 04 May 2022
    
    The revisions have successfully addressed the raised concerns
    
    Citation: https://doi.org/10.5194/tc-2022-45-CC2
RC1:
'Comment on tc-2022-45', Anonymous Referee #1, 06 May 2022

This work employed geographically and temporally weighted neural network (GTWNN) model to construct the daily snow density grid products in China, with the support of satellite, ground, and reanalysis data, which is useful for estimating water resources and predicting natural disasters. However, some important issues need to be addressed. The details are as follows.

1.In terms of abstract, the content needs to be specified and well organized. For instance, in Line 6 of this section, the detailed results supporting that the GTWNN model can improve the estimation of snow density should be given; and in Line 10, the specific models should be listed.

2.In the introduction, it is not clear why satellite, ground, and reanalysis data are used.

3.In the end of Section 2.1, the snow season is divided into three periods. Considering the climate and environment show great spatial heterogeneity in snow cover areas in China, this division of snow season should be expounded.

4.In terms of Equation 2, not all the variables are explained in detail.

5.The variables in Figure 2 should be explained in or below this picture.

6.How each kinds of data are used specifically is not given in Section 2 or 3.

7.In Section 3.2, the title of this section is not appropriate for the content. In addition, how to evaluate the GTWNN model (such as the metrics to evaluate the performance) should be described.

8.In table 2, the details about each model are deficient.

9.The text and logic in the manuscript needs improved, particularly in the Results Section. For example, what’s the relationship between Section 4.1 (Descriptive Statistics of Ground Observations) and other results? The position of Section 4.2.3(Importance of the Influencing Variables for Snow Density Estimation) needs consideration.

10.Snow density and its CV in different snow cover regions vary apparently, as well as the monthly changes of snow density. While, the explications about these phenomena are limited in the present manuscript.

Citation: https://doi.org/10.5194/tc-2022-45-RC1
- AC2: 'Response to RC1', Xueliang Zhang, 21 Jun 2022
  
  Thanks for your review and constructive comments. We have addressed their comments in detail in the pdf supplement.
  
  Citation: https://doi.org/10.5194/tc-2022-45-AC2
RC2:
'Comment on tc-2022-45', Anonymous Referee #2, 25 May 2022

The comment was uploaded in the form of a supplement: https://tc.copernicus.org/preprints/tc-2022-45/tc-2022-45-RC2-supplement.pdf

Citation: https://doi.org/10.5194/tc-2022-45-RC2
- AC3: 'Response to RC2', Xueliang Zhang, 21 Jun 2022
  
  Thank you for the feedback on this manuscript. We have carefully revised the manscript and addressed these comments in detail in the pdf supplement.
  
  Citation: https://doi.org/10.5194/tc-2022-45-AC3

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Publish subject to minor revisions (review by editor) (28 Nov 2022) by Chris Derksen

AR by Xueliang Zhang on behalf of the Authors (06 Dec 2022) Author's response Author's tracked changes Manuscript

ED: Publish subject to technical corrections (19 Dec 2022) by Chris Derksen

Thank-you for your thorough responses to the reviewer comments, which are clearly reflected in the revised manuscript. Your paper is now accepted for publication in The Cryosphere. I have just some minor technical corrections for your consideration. Thanks very much,
Chris Derksen

Line 20: change to “20 potentially influencing variables.”
Line 21: change to “an R2 of 0.531”
Line 24: change to “…the GWTNN model in capturing…”
Line 57: “For example, snow density was found to be lower at higher elevations, and even decreased by approximately 0.006 g/cm3 with each 100 m increase in elevation (Zhong et al., 2014)” Can you add specification to the geographic region for which this result from the Zhong et al study applies?
Line 61: remove “In addition, the aspect also affects the snow density through radiation,because sunny-facing slopes that experience” and just start the sentence with “Slopes with high radiation input will be more likely…”
Line 99: change ‘various’ to ‘variable’
Line 101: change to “satellite data can provide information on the snow-related…”
Figure 1: remove the inset map. It is not needed given the geographic focus of the analysis.
Line 145: change to “…different influencing variables on snow density.”
Line 151: consider adding more detail to this statement: “In addition, the min-max normalization method is applied to normalize different influencing variables.” Can you point to which variables are normalized in Figure 2?
Line 232: change ‘involved’ to ‘included’
Line 268: change to “…caused by greater forest cover…”
Line 346: change to “…also provides gridded daily snow density data…”
Line 398-399: remove “in different periods and regions.”
Line 469: change to “…based on the observed snow density.”
Line 480: I suggest adding an extra statement to the new text which emphasizes that the individual correlations with the 20 variables were generally weak, as you note on line 238.

Hide

AR by Xueliang Zhang on behalf of the Authors (21 Dec 2022) Author's response Manuscript

Short summary

The geographically and temporally weighted neural network (GTWNN) model is constructed for estimating large-scale daily snow density by integrating satellite, ground, and reanalysis data, which addresses the importance of spatiotemporal heterogeneity and a nonlinear relationship between snow density and impact variables, as well as allows us to understand the spatiotemporal pattern and heterogeneity of snow density in different snow periods and snow cover regions in China from 2013 to 2020.