the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Estimation of Snow Depth from AMSR-2 Based on an AutoML method over the Qinghai-Tibet Plateau
Abstract. Snow depth (SD) is a crucial parameter for describing the spatiotemporal variations of snow cover, and passive microwave SD products (10–25 km) are widely used for monitoring SD changes. However, as one of the three major snow-covered regions in China, the Qinghai-Tibet Plateau (QTP) has complex terrain and rapid change in snow cover with strong spatial heterogeneity, making it difficult for coarse-resolution SD products to accurately describe its spatiotemporal characteristics. This study proposes a high spatial resolution (500 m) SD estimation method based on AMSR-2 brightness temperature (BT) data and an Automated Machine Learning (AutoML). Firstly, using Pearson correlation coefficient, 19 key factors influencing SD, including AMSR-2 BT, slope, and surface roughness, were selected as input data (independent variables) for AutoML. Meanwhile, a passive microwave downscaled SD data and ground-based SD measurements were introduced as dependent variables for AutoML. Then, the AutoML model was trained separately for four different types of snow cover surfaces (forest, grassland, water, and bare land). Finally, through the ten folds cross validation method, the optimal machine learning model for SD estimation under each type of underlying surface coverage was selected, thus sequential SD datasets were obtained for ten-year snow cover periods of the QTP from 2012 to 2021. Results show that the estimated SD values are consistent with ground-based observations (R=0.81), and the accuracy is high with an RMSE of 3.65 cm. Compared with Landsat-8, the estimated SD spatial distribution is consistent with the snow cover extent on optical images, which can provide reliable data for monitoring snow cover changes in mountainous regions.
- Preprint
(2272 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2875', Anonymous Referee #1, 16 Aug 2025
-
AC1: 'Reply on RC1', Xuan Li, 02 Oct 2025
Dear reviewer,Thank you sincerely for your attention to, and review of, this article. We also appreciate your constructive feedback, which has given us valuable insights into how we can improve our research. We have carefully considered each of your comments and provided detailed responses. For the areas that you suggested needed supplementing and improvement, we have also provided relevant explanations and supporting materials. The specific modification instructions and responses are provided in the attached document for your reference.
-
AC1: 'Reply on RC1', Xuan Li, 02 Oct 2025
-
RC2: 'Comment on egusphere-2025-2875', Anonymous Referee #2, 20 Aug 2025
General comments:
This paper proposed a high spatial resolution (500 meters) snow depth estimation method based on AMSR-2 brightness temperature data and automated machine learning (AutoML), significantly improving the accuracy of snow depth monitoring in the complex terrain areas of the Tibetan Plateau.
Through the Pearson correlation coefficient, 19 key influencing factors were selected, including AMSR-2 brightness temperature, slope, and surface roughness. The method comprehensively considers the impact of various geographic and topographic factors on snow depth, enhancing the model's robustness and accuracy. But the authors ignore influence of the snow properties. How to asses the influence?
However, the paper still needs to improve the writing style and enhance the readability and standardization of the charts, and the details need to check carefully. In conclusion, we suggest a major revision.
Specific comments:
1 Line 20 “Compared with Landsat-8, the estimated SD spatial distribution is consistent with the snow cover extent on optical images, which can provide reliable data for monitoring snow cover changes in mountainous regions”. Sd consistent with SCE, and the logic is not proper.
2 Line 27 "Roof of the World,"
Please remove “,”. It would be better use ‘’ instead of “”, which is the Chinese usage. Please check the whole manuscript.
3 Line 62 Support Vector Machine (SVR), SVM would be better than SVR.
4 In the last part of Introduction, the author need the add the objectives of your work and the contribution or potential usage to this field.
5 The art-to state progress of SD estimation is not enough, and the author need to improve it.
6 What is the amount of SD measurement from climate stations and line measurement? Line measurement is not a proper name. The site is mainly distributed in the middle and the east part and rarely distributed in the west part. What is the influence of limited materials on the remote sensing inversion.
7 Lin 109 Table1 application: usage. Please sperate the usage of auxiliary data.
8 Line 114-116 Please sperate the long sentence.
9 How to deal with the different spatial resolution between different data source.
10 The width of flow line is not consistent.
11 This study marks the introduction of the PyCaret automated machine learning framework into snow depth estimation. It automates data processing and model selection, reducing human intervention while enhancing the efficiency of model selection and parameter optimization. Figure 3 is too simple and needs to improve. How to solve the overfitting?
12 Line 235-240: The introduction of SD estimation in the past should move the Introducion.
Figure 4 the color of meteorological stations and downscaled snow depth sample are too close, which are hard to tell from each other.
13 The discussion lack the comparison of your work and others and please cite more reference.
14 The conclusion needs the quantitative data to support the conclusion.
15 Please check the reference to meet the requirements of magazine.
-
AC2: 'Reply on RC2', Xuan Li, 02 Oct 2025
Dear reviewer,Thank you sincerely for your attention to, and review of, this article. We also appreciate your constructive feedback, which has given us valuable insights into how we can improve our research. We have carefully considered each of your comments and provided detailed responses. For the areas that you suggested needed supplementing and improvement, we have also provided relevant explanations and supporting materials. The specific modification instructions and responses are provided in the attached document for your reference.
-
AC2: 'Reply on RC2', Xuan Li, 02 Oct 2025
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
1,061 | 53 | 19 | 1,133 | 49 | 53 |
- HTML: 1,061
- PDF: 53
- XML: 19
- Total: 1,133
- BibTeX: 49
- EndNote: 53
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
The authors present a 500-m snow depth estimation method using an automated machine learning approach derived from AMSR-2 data, demonstrating good accuracy against ground-based observations. While the work shows potential, several significant issues must be addressed before the manuscript can be considered for publication in this journal.
Major Comments:
Minors:
Lines 35–40: Provide references for claims about SD retrieval challenges.
Lines 41–73: Expand the literature review to include key international studies.
Line 83: Remove "in" for grammatical correctness.
Section 3.1.1: Justify the use of the 23-GHz band given its sensitivity to water vapor.
Lines 213–215: Clarify why all AMSR-2 bands and band differences were selected as input features.
Section 4.1.1: The correlation coefficient is not a robust metric for variable selection. Consider alternative methods (e.g., feature importance from ML models).
Figure 9: Revise captions for clarity (subfigures c and d are unclear).
Section 4.3: The SD-temperature relationship is well-known. Emphasize new insights (e.g., regional variability, model sensitivity) rather than restating basics.