The Early Identification and Spatio-Temporal Characteristics of Loess Landslides with SENTINEL-1A Datasets: A Case of Dingbian County, China

Round 1

Reviewer 1 Report

In this work, the authors describe an improved InSAR-based method for mapping large-scale landslides in loess hilly areas. The activity of all detected loess landslides is assessed and classified at the regional scale using the mean deformation rate, maximum deformation rate, and activity index. Spatial kinematic analysis, correlation analysis of deformation with rainfall were carried out for individual landslides. The work is well written and well structured. The method used seems reasonable and advanced. The discussion is meaningful and extensive, and supported by the data. I recommend taking it in remote sensing with some minor modifications, as the work is interesting, original (as far as I know), and transferable to other similar situations.

Minor comments:

1. The title of 3.2.1 in the manuscript is the same as the title of 3.2.2 and needs to be changed.

2. Figure A1 and Table A1 in line 184, Figure A3 and Figure A4 in line 186 and Figure A2 in line 219 have no corresponding charts in the manuscript.

3. Figure 11 C1 is not identified in the figure.

4. There are problems with some landslide boundaries in Figure 8. For example, Figure 8 (2) and Figure 8 (12) have wrong landslide trailing edge boundaries, which span the other side of the mountain and do not conform to the conventional landslide boundary delineation.

Author Response

Please see the attachment as figures enclosed.

Author Response File: Author Response.docx

Reviewer 2 Report

This work is interesting and innovative.

It gives a clear idea of the application of SAR data to detect landslides in loess areas. A good solution to the task of reducing the number of scenes from Sentinel 1 has been implemented.

I did not understand what software was used?

 

Author Response

Response to reviewer #2

General comments and question:

This work is interesting and innovative.

It gives a clear idea of the application of SAR data to detect landslides in loess areas. A good solution to the task of reducing the number of scenes from Sentinel 1 has been implemented.

I did not understand what software was used?

>> Thanks for your comments and question. The SAR data processing is conducted by GAMMA software, where we proposed tropospheric delay correction based on the quadtree segmentation and the optimal interferograms selection based on the relative error boundary.

Reviewer 3 Report

Dear authors,

thank you for an interesting article. The article seems very valuable, but is quite difficult to understand, partially due to complex mathematics, but partially due to the fact that it is not well explained.

I have several comments:

- abstract: please clarify 50 vs. 16 % of landslides active. I understand that non-active landslides cannot be identified by InSAR

- line 70: "maximum" and minimum baselines. Why maximum?

- section 3.2: do LM and LM-DT methods contain "linear model" only in range and azimuth, or also in elevation?

- section 3.2: please explain the process of re-merging the interferograms. If the ifg crops are atmo-corrected, their phase is not expected to be the same - but small (gaussian) errors are expected. If they were unwrapped independently, systematic errors are expected. How do you treat them?

- line 157: could you give an example of "set window size threshold"?

- line 171: there is 13-day difference between the two dates, shouldn't it be 12?

- line 175-176: STD of residual unwrapped phases : on the overlaps or where? residual with regard to what?

- Figure 3: please change the scale to enhance the differences.

- Section 3.2.2 should be renamed to something like "selection of highest-quality interferograms", as I did not find anything regarding the atmospheric delay correction in it. I understand that this process is done based on the already-merged interferograms, but I am not sure. Please explain.

- Equation (1): please name in detail what is the known and what is unknown (it is only clear from the comment of equation (2).

- Please cite the "relative error boundary theory" already on line 194.

- Please explain clearly the paragraph on lines 204-205.

- line 213: please explain the term "deformation norm"

- line 215: average coherence in which regard? Coherence averaged over all pixels in an interferogram?

- section 3.2.3: if you work with an a-priori defined set of landslides, it should be clear. This is not clear even from the abstract

- line 242: "activity index" is not mentioned before

 - section 4.1: I still don't understand if you got a-priori information about the landslide locations, or you identify the locations based on InSAR data. If you identify the landslides based on the InSAR data, and you - as a user - delineate the rectangles, then the "propotion of active pixels for one landslide" is user-dependent?

- section 4.3: activity index is explained as "mean deformation rate and maximum deformation rate", however, these are two numbers and fig. 10 shows just one

- section 4.3: if you consider the deformation rate, do you consider also the LOS/downslope sensitivity? Some landslides can be fast but "slow" in InSAR due to low sensitivity

- section 4.3: it seems to me that activity index, activeness and "deformation rate" (solely on line 350) are intermixed. If not, please clarify the difference between them

- section 4.3: sometimes with negative signs, sometimes "greater than 20 mm/y"

- section 5.1: multimode velocity discussion should be done based also on the size of landslides and SB filtering window size

- figure 11: are the deformation rates in LOS, or downslope direction?

 

 

Minor mistakes / typos / grammar:

- line 23: experienced -> tested?

- line 64: subjected -> subject

- line 102: belongs -> corresponds?

- If referring to figures/tables A3 etc., please note that they are in the appendix.

- line 171: with frame -> of frame ?

- line 217: robust -> robustness

- line 241: identify the active of surface (??)

 - line 246, 366, 419: kinematic -> kinematics ?

- line 324: almost -> most ?

- line 468: index -> indices?

 

 

Author Response

Please see the attachment as figures enclosed.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Dear authors, thank you for the revised article.

Most of the changes I proposed were reflected.

-line 175: "The phases are averaged  to merge the overlapping parts of adjacent segments": I am afraid of unwrapping errors and simple averaging (without check) would affect the phase values significantly. This is just my note, does not need to be reflected in the article.

section 4.3: mentioning the particular velocities, please refer to the fact that they are in line of sight (?).