Near Real-Time Monitoring of Large Gradient Nonlinear Subsidence in Mining Areas: A Hybrid SBAS-InSAR Method Integrating Robust Sequential Adjustment and Deep Learning

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper employs deep learning methods to obtain the large gradient deformation phase in mining areas, introducing robust estimation and sequential adjustment methods into the SBAS-InSAR processing flow. This approach partially addresses the limitations of InSAR in mining area applications and improves computational efficiency. I believe this research has great potential for future applications. However, I think several aspects need to be improved before publication.

1. PUNet method was used to obtain the phase unwrapping of large gradient subsidence. In mining area scenarios, the magnitude of surface subsidence can range from tens of centimeters to over ten meters. The PUNet unwrapping method certainly has its limits and cannot be universally applicable to all magnitudes. It is necessary to clarify the specific applicable scenarios of the method used in the article.

2.Lines 126-135 require further refinement for the introduction of the InSAR data preprocessing section.

3.It is recommended to add necessary references to Eq.9 to explain the basis of each parameter.

4.The legend font size in Fig. 3 and Fig. 5 is too small. It is suggested to further improve the details of the figures.

5.The article provides insufficient information about the study area, failing to highlight its characteristics. It is recommended to include more information about the study area's geological background, among other details.

Comments on the Quality of English Language

Language check is needed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

In this article, the authors aim to improve the efficiency of updating the SBAS analysis results as the new data becomes available. The methodology of the article is interesting. The authors explained the core of the methodology in detail. I have some major issues with the drafting of this manuscript. The readability and the structure of the manuscript can be improved significantly. The authors should provide specific information on some methods, data used, etc. to make the results reproducible. I have listed some specific issues below.

1.      What is the innovation/novelty of the manuscript? While the methodology is clearly explained, it is hard to interpret which parts of the methodology were introduced by the authors in this manuscript.

2.      The central theme of the manuscript is not clear. The focus shifts from unwrapping methods to sequential adjustments to mining subsidence several times. Clarify the focus of the manuscript and modify the text to make it clear. If the focus is on the technique, it would be helpful to test it on more areas to strengthen the methodology.

3.      It is not clear why the authors chose only mining subsidence for the PUNet model and the entire analysis in general. Please explain the characteristics exclusive to mining subsidence to justify the study area/technique used. While the authors have mentioned the dynamic subsidence in the mining areas as one reason, Figs 9 & 10 show a spatially consistent and linear subsidence pattern for most of the study period followed by stabilization. What are the limitations of using this technique for other subsidence events with similar characteristics like sinkholes, volcanoes, landslides, etc.?

4.      The references are missing in the manuscript. Please cite the sources for all the statements made in the manuscript, especially in the introduction section. Some sentences that need citations are (line numbers point to the start of the sentence) 50, 51, 57, 59, 102, 108, 111, 146, 359, and more. Also, cite some examples when some techniques are claimed as ‘widely used’, or ‘common techniques’ such as in lines 46, and 102.

5.      Line 87: Please rephrase “all images must be reprocessed”. In reality, only the interferograms must be re-inverted in the general SBAS method. While the interferograms that were produced previously need not be reproduced.   

6.      Line 79: The sentence is not clear. The sentence implies that the revisit period of sentinel-1 is continuously shortened, which is untrue.

7.      The preprocessing steps in the methodology section need to be made more specific to your study to make the results reproducible.

a.      Replace the M and N with the specific numbers. Mention the DEM used and cite the source. Provide the study period.

b.      Specify the coherence threshold applied.

c.      Specify what technique was used to remove the turbulent/stratified atmospheric effects.

d.      Mention the names of the conventional solving methods and cite the sources.

8.      Expand the acronyms the first time it is used. (e.g., DnCNN). Also, it is best practice to expand the acronyms on the first use in the body (in addition to the abstract).

9.      Provide more information on the location, extent of the mines, mining methods, geology of the subsurface, mining period, etc. Also, provide a brief description of the mining-related terms used in the manuscript (e.g., caving method and backfilled mine face.)

10.  The improvement between the SA and RSA does not seem significant (~5%) especially considering that RSA consumes approximately 50% more time according to Fig 13.

11.  Section 3.5: The purpose of this section is not clear. It would make more sense if the authors explained the difference between the two mining methods they mentioned. Also, how was the subsidence under caving mining simulated? This section needs to be modified to explain the purpose, method used, results, and inferences clearly. The inference from this simulation on the effectiveness of your suggested methodology needs to be clearly stated.

12.  Section 4.2: What type of processing configurations were used to obtain the calculation times? How many images were used? what was the multi-look factor? How many connections? Readers, especially those familiar with the generic SBAS technique, can use this information to verify the claims made in this section. Without such information, the results are not reproducible.

13.  Section 4.4: How do you justify using the SBAS technique while considering the pixels as “persistent scatterers”? Why not directly use the PS-InSAR technique instead? There are existing works that used the PS-InSAR technique for the mining subsidence such as the ones below:

 

Li, F.; Liu, G.; Gong, H.; Chen, B.; Zhou, C. Assessing Land Subsidence-Inducing Factors in the Shandong Province, China, by Using PS-InSAR Measurements. Remote Sens. 2022, 14, 2875. https://doi.org/10.3390/rs14122875

Karanam, V., Motagh, M., Garg, S., & Jain, K. (2021). Multi-sensor remote sensing analysis of coal fire induced land subsidence in Jharia Coalfields, Jharkhand, India. International Journal of Applied Earth Observation and Geoinformation, 102, 102439., https://doi.org/10.1016/j.jag.2021.102439

Chen, Y.; Dong, X.; Qi, Y.; Huang, P.; Sun, W.; Xu, W.; Tan, W.; Li, X.; Liu, X. Integration of DInSAR-PS-Stacking and SBAS-PS-InSAR Methods to Monitor Mining-Related Surface Subsidence. Remote Sens. 2023, 15, 2691. https://doi.org/10.3390/rs15102691

Please explain how your method is better suited for performing sequential adjustments compared to the PS-InSAR technique as applied in these articles.  

14.  The figures can be significantly improved to make it more presentable and easier to interpret. The figure captions should contain all the information about the figure. Font size in some figures should be increased. The legend should be clear and show all the details. Some suggestions are given below:

a.      Fig 3: The figure and legend have different colors for dashed lines. The font size in the legend needs to be increased. The data acquisition dates need to be provided in place of/in addition to the number of days.

b.      Fig 4: Explain the rectangles in the caption or legend. Also, what do the colors in the legend represent? It looks like it is a DEM with the values in meters. Specify that.

c.      Fig 5: There are different colors in the fig. But they are not explained in the legend or the caption. Also, the text size for the legend should be increased.

d.      Fig 6: Label the figure sections as (a) and (b). Explain why the legend shows two colors for both SBAS and RSA in each figure. Explain this time series for which two points in the study area.

e.      Fig 7: label the two figures as (a) and (b) and explain them with information such as the dates of reference and secondary images used for creating these interferograms in the caption. Also, briefly describe in the caption, what can be inferred from the figure. Add the title for the color bar.

f.        Fig 8: label the sections (a) and (b). Please specify if the time series values are converted to vertical measurements. Otherwise, the comparison with the leveling results does not make sense. Also, specify the source of the leveling data.

g.       Fig 9: Points A and B are not clearly labeled in the figure. Label it with a dot/star. If point B is right below the letter B, the deformation time series does not make sense as the point is outside the deforming regions. Explain what the authors mean by dynamic subsidence. It looks like the subsidence is spatially and temporally consistent before stabilizing after around 200 days. At the bottom right, ‘benchmark’ is written. Please describe the context and what benchmark the authors want to highlight. In the bottom right figure, there are two lines. Please explain what these lines are. Also, increase the font size.

h.      Figure 10: Why are generic SBAS-InSAR technique results not included in the comparison? Explain why the subsidence stopped after around 200 days. Is it because the mining was not taking place anymore?

i.        Increase the font size for the legend in Fig 11

j.        Fig 12: Provide the dates of images used to calculate the differential interferograms.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

see attached file

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have produced a nice draft, and the current version is significantly improved and in good shape for acceptance. I have only a few minor comments:

Phase unwrapping (mentioned in line 48) is introduced abruptly. It would be beneficial to add a line or two explaining how SAR data is recorded and the rationale behind phase unwrapping. Then, you can connect it with technical details about phase unwrapping. A good example of this is shown in Garg et al. 2022 (in the heading 'Phase unwrapping, time series, and velocity estimation.') where they show why it's a crucial parameter and the significant underestimation that may occur if it is not conducted properly.

Garg, S., Motagh, M., Indu, J., & Karanam, V. (2022). Tracking hidden crisis in India’s capital from space: Implications of unsustainable groundwater use. Scientific Reports, 12(1), 1-17. https://doi.org/10.1038/s41598-021-04193-9

There may be other examples where you can strengthen your argument about phase unwrapping. This will help readers understand why this is crucial in InSAR.

In the final paragraph, it would be beneficial to include a line or two about the limitations and future work.

The equations are not centered, but this seems to be more of a formatting error. Also, the references at the end are numbered twice. 

 

 

Author Response

We have implemented the suggested modifications as follows:

1. Supplementary Explanation on the Importance of Phase Unwrapping

We have augmented the discussion on the significance of phase unwrapping and cited pertinent literature, as evidenced in lines 51-54.

2. Inclusion of Limitations and Future Directions

The constraints of our methodology and prospective avenues for future research have been addressed in Section 4.4.

3. Adjustments to Equation Formatting

Conforming to the journal's style guide, we have corrected the formatting of the equations.

Reviewer 3 Report

Comments and Suggestions for Authors

Please  rescale the figures 4 b. c so that they  are immediately comparable; please improve the identifiability of the position of the mine in Fig. 4b.

Please specify how you calculate the mean coherence in fig. 9c and how the simulation of the decorrelation is carried out in Fig. 12 (lines 410 and ff)

Author Response

We have implemented the suggested modifications as follows:

1. Revisions to Figures 4b and 4c

We have made adjustments to the figures in question, enhancing their comparability. See Figure 4 for the updated illustrations.

2. Explanation of Calculation Methods for Figures 9c and 12

Regarding Figure 9c, the mean coherence was computed by initially assessing the coherence of each interferogram individually using standard techniques, followed by averaging the coherence values for each pixel across all interferograms to yield Figure 9c. Details of this process have been incorporated into the text, found in lines 378-380.

For Figure 12, this illustration is derived from real data processed using the pre-processing steps outlined in Section 2 to generate differential interferograms.