Sequential DS-ISBAS InSAR Deformation Parameter Dynamic Estimation and Quality Evaluation

Round 1

Reviewer 1 Report

Interesting and useful paper. A new stepwise temporal phase optimization method is proposed to alleviate the decorrelation dynamically. the post-failure Baige landslide is used to demonstrate the validity of the proposed approach. Some comments and suggestions should be addressed.

(1) To obtain more monitoring targets and avoid the selection of interferograms, ISBAS technology is introduced to generate more interferogram by setting the spatiotemporal baseline threshold in Section 2.2. However, the common SBAS technology can also obtain more interferogram by setting the maximum percentage of critical baseline and the maximum time distance, and then the generated interferogram can be viewed to eliminate the interferogram with poor quality. When ISBAS technology is introduced in Section 2.2, the space-time optimization phase can be obtained. How is it optimized?

(2) In Figure 16, ISBAS-InSAR technology is only compared with GNSS, while ISABA is compared with traditional SBAS. If the monitoring results of SBAS are added to the figure for comparison, can the advantages of high monitoring accuracy of ISBAS technology be better reflected?

Author Response

A point-by-point response to the reviewer’s comments have been uploaded as aPDF file in the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

General Comment

The paper proposes an innovative methodology for retrieving displacement information over areas affected by large deformation gradients and/or relatively high displacement rates. The methodology is based on Small Baseline InSAR analysis. The methodology is applied to a large landslide, which underwent a catastrophic reactivation and successfully investigate the post-failure phase. The validation against more traditional InSAR techniques and ground-based measures appears adequate. Results are described in detail but a prospective discussion of the results is lacking. Also, the authors do not discuss possible limits of their approach and possible fields of application.

The paper is written in an acceptable form and is mostly clear. However, it could greatly benefit from a native english review.

Authors make an effort in using the term ‘near real-time’ to describe their work and the methodology proposed. It appears in the title and in the main objectives described at the end of the Intro section. However, I think that their method do not include anything that justify the use of such adjective. The analysis is real-time in that it uses SAR images that are frequently acquired. I suggest to eliminate such term from the title and the text.

The title is rather wordy and contains acronyms that are not readily and uniquely understood. I suggest to simplify.

The abstract concisely summarize the paper and state the main results. Authors demonstrate a good knowledge of the previous work published on the topic.

The introduction is therefore adequate and puts the proposed work in perspective. However, it does not describe the rational behind the work. Also, the objectives of the paper are not clearly stated at the end of the intro section.

Graphics and supporting information are adequate and explanatory, though some improvements can be done (specific comments).

The paper contains rather clear explanations of the methods, though some clarifications are needed (specific comments).

In my opinion, the paper is fully appropriate for the journal. It includes significant elements of novelty that make it interesting to the vast audience researching on remote sensing. In the meantime, results should be rigorously analyzed and discussed in the light of the most recent literature contributions. In few occasions, I found some unclear or incomplete reasoning that will be later described in detail and that can be fixed with further information and/or data.

Overall, I recommend that the paper is accepted for publication subject to the reorganization of the sections describing, validating and discussing results (major revision).

 

Specific comments

lines 85-86. Sentence unclear, please rephrase.

lines 89-90.Iin the remainder of the paper, it seems that large gradient deformation is the main issue. I don't think that evidences presented here demostrate that the method can be useful in case of decorrelation due to dense vegetation. I suggest to eliminate references to the vegetation cover.

line 264. I think author are talking about decorrelation induced by large displacements. They should specify and also give an idea of the displacement rates and of the deformation gradients that cannot be measured (using Sentinel1).

line 268. Explain how the deformation patterns were mapped.

lines 269-270. Sentence unclear, please rephrase and explain in more detail.

line 274. Do you mean that you consider interferograms with 40 days baseline or shorter?

line 275. ‘208 SB if networks’. It would be useful to see the network, please add a figure. What is the minimum number of interferograms that you use to cover any given instant in your analysis period? Please explain.

lines 279-282. Sentence unclear, please rephrase.

line 285 and Figure 5. Figure 5 do not show any network, please rephrase. I suggest to eliminate Figure 5. It does not give any information that cannot be described in few words.

lines 293-297 and Figure 6. This is a little confusing. Authors stated that they used SB interferograms (<40 days). In my opinion this choice is fully understandable and there is no need to show a 120 days example. I suggest to remove this part of the text and Figure 6.

Figure 8. Place Figure 8 after its citation in the text.

Figures 8, 9 and 10. I suggest to compare your results with fewer optimized interferograms. 3 or 4 is enough to illustrate the point and the bigger size allow the reader to better appreciate the differences. The same applies to subsequent figures 9 and 10.

Figure 10. Please place figure 10 after its citation in the text

Figure 11. Find a better way to compare the two series of data (e.g., symbols or separated charts). Now they are indistinguishable.

lines 373-375. This is described in the figure caption. Here you should explain the rational behind the comparison and comment the figure.

lines 391-392. This point is important and should be expanded in the discussion citing relevant literature.

line 396. ‘cumulative’ not ‘accumulative’.

Figure 13 caption. Meaning of the poligon in figure 13D?

Discussion section. Discussion is poor and the results are not discussed and compared with the most recent literature on the subject. The reader might be interested in your answer to questions like the following: i) can you give some guidance about the range of displacement rates that can be solved with the method you propose?; ii) do you think that your method can give some advantages also when processing large areas? Please expand adequately the discussion.

Figure 15 caption. Please explain that Figure 15B reports the STD?

lines 441-446. Authors have not analyzed the landslides and should limit their discussion to the technical InSAR aspects. I suggest to avoid any geological or geomorphological judgement.

Conclusions section. In the conclusions, authors should add details about: i) the need of using short duration interferograms and ii) the possible advantages that their method implies in terms of computation effort.

Author Response

A point-by-point response to the reviewer’s comments have been uploaded as aPDF file in the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors of this manuscript used SLS-based algorithms and their variants for near-real-time surface deformation analysis, adapting them to the concepts of adaptive interferometric networks proposed by [1] and [2]. Moreover, the authors propose a procedure for evaluating the quality of the results obtained.

As much as the research topic addressed is of considerable interest to the scientific community, I find that the proposed article provides results and findings partly already proposed by other authors and therefore already known. And if there is anything really new, it is not enough clear from the paper.

Below I try to be as clear as possible in making my concerns clear:

1) First, following your mathematical treatment, the proposed method for near-real-time surface deformations uses the concept developed in [1] for estimating residual topography and displacement rates, while the method in [2] for calculating adaptive ground displacement time series. These inherent characteristics are not clearly exposed to a hypothetical reader of the work, but it is as if the message of a totally new approach is being given.

2) The selection of the adequate wrapped SB interferograms for the PTA step seems to be based on unwrapped interferograms; therefore, the optimal network is evaluated based on the information obtained from the unwrapped data with all possible issues related to the chosen phase-unwrapping algorithm. 

3) The advantages of using an InSAR network consisting of a reduced set of redundant small-baseline interferograms in the EMCF-SBAS algorithm, but more generally also for other PTA-like approaches, have already been shown, e.g., in [3]. But in this context, authors should be aware that temporal baselines too small could introduce problems of bias on the deformation estimates [4].

4) Figure 10 shows the temporal coherence maps for the PTA filtering steps with reduced InSAR networks, but probably the authors refer to the triangular coherence, which is calculated by exploiting the wrapped data. However, if so, the triangular coherence estimator was proposed to identify DS coherent targets for subsequent phase-unwrapping procedures, and not to assess the reliability of PTA-like filtering. 

5) Also, regarding the results section, the authors stated that 72 SAR images were used, but in this case, it is not clear whether a block of images was used first and then the remaining ones were added. Because if not, practically the SLS-based method for near-real-time computation is not really used.

[1] Sowter, A.; Bateson, L.; Strange, P.; Ambrose, K.; Syafiudin, M. F. DInSAR estimation of land motion using intermittent coherence with application to the South Derbyshire and Leicestershire coalfields. Remote Sensing Letters. 2013, 4(10), 979-987. 518 

[2] Falabella, F.; Serio, C.; Zeni, G.; Pepe, A. On the use of weighted least-squares approaches for differential interferometric SAR analyses: The weighted adaptive variable-length (WAVE) technique. Sensors. 2020, 20(4), 1103. 

[3] Pepe, A.; Yang, Y.; Manzo, M.; Lanari, R. Improved EMCF-SBAS processing chain based on advanced techniques for the noise-filter and selection of small baseline multi-look DInSAR interferograms. IEEE Trans. Geosci. Remote Sens. 2015, 53, 4394–4417. 

[4] H. Ansari, F. De Zan and A. Parizzi, "Study of Systematic Bias in Measuring Surface Deformation With SAR Interferometry," in IEEE Transactions on Geoscience and Remote Sensing, vol. 59, no. 2, pp. 1285-1301, Feb. 2021, doi: 10.1109/TGRS.2020.3003421.

 

In conclusion, I cannot clearly find the added value in the proposed algorithm (beyond the effective SLS-based algorithms, however, which the authors have previously published) that can make it be published in this version. Certainly, the results obtained are good for analyzing a highly decorrelated scenario, partly because different types of data filtering are adopted. But considering all the steps involved in the processing, it is really difficult to evaluate whether one is assessing the reliability of the near-real-time method or of all the intermediate pre-processing involved in the InSAR chain.

Probably, a thorough reorganization of the work, focusing on the new key steps proposed by the authors and what is already in the scientific literature, could be beneficial for a new resubmission of the work.

Author Response

A point-by-point response to the reviewer’s comments have been uploaded as aPDF file in the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I believe that authors made a good job at improving their manuscript. Before acceptance, they should also consider:

1. modify the title of the paper making it shorter.

2. improve the discussion and conclusion section by putting their results in the context of relevant literature.

Kind regards.

 

 

Author Response

We have uploaded the PDF attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The readability of the article has been improved by the authors, who have partially addressed my concerns. 

 

So, from what I understand from the revised version of the paper, the main novelty lies in the use of truncated SB interferograms for stepwise optimization of the SB interference phase. In view of the large availability of SAR data, the latter feature should lead to a substantial reduction in computation time when applying temporal filtering. 

 

I do not agree at all regarding the non-presence of bias in multi-look SB interferograms. First, this existence has been shown and handled by several mutually independent research groups, whose references I leave here:

 

[1] H. Ansari, F. De Zan and A. Parizzi, "Study of Systematic Bias in Measuring Surface Deformation With SAR Interferometry," in IEEE Transactions on Geoscience and Remote Sensing, vol. 59, no. 2, pp. 1285-1301, Feb. 2021, doi: 10.1109/TGRS.2020.3003421.

 

[2] Y. Zheng, H. Fattahi, P. Agram, M. Simons and P. Rosen, "On Closure Phase and Systematic Bias in Multilooked SAR Interferometry," in IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1-11, 2022, Art no. 5226611, doi: 10.1109/TGRS.2022.3167648.

 

[3]  Y. Maghsoudi, A. J. Hooper, T. J. Wright, M. Lazecky, and H. Ansari, “Characterizing and correcting phase biases in short-term, multi- looked interferograms,” Remote Sensing of Environment, vol. 275, Jun. 2022, Art. no. 113022, doi: 10.1016/j.rse.2022.113022. 

 

[4] F. Falabella and A. Pepe, "On the Phase Nonclosure of Multilook SAR Interferogram Triplets," in IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1-17, 2022, Art no. 5120117, doi: 10.1109/TGRS.2022.3216083.

 

I would like to point out to the authors, that in De Luca et al. (2021) reference is made to a particular processing of the P-SBAS algorithm, which as implemented is in a condition where the effect of bias is somewhat negligible. But likewise, it is important to note that the bias cannot be compensated for by the PTA-like methods in the literature.

 

On this aspect, referring to the experiments shown by the authors of this paper, as the maximum temporal baseline is set at 40 days, the effects of bias could be somewhat present in the deformation estimation itself. Furthermore, it is right for the authors to say that, for simplicity's sake, the hypothetical presence of bias on deformation is not considered, but to claim that multi-look SB interferograms are exempt from the phenomenon is rather dangerous and erroneous, and above all contradicts the recent scientific literature without real counter-evidence.

Author Response

We have uploaded the PDF attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 3 Report

The authors have greatly improved the paper.

I have one last important clarification regarding the phase bias issue on SB ML interferograms, after which the work can be accepted for publication.

I refer specifically to that sentence, "In this study, we use block-shaped SB interferograms to correct the bias by the PL algorithm, in which full interferograms are generated for each block SAR data." I would like to point out that the use of interferograms characterized by a not so small temporal baseline helps to mitigate the effect of the bias to some extent, but this phenomenon is not corrected completely, unless longer interferograms are involved or recently developed techniques are adopted (whose references I have already listed). If full interferometric networks are used for each block, and I am referring to "in which full interferograms are generated for each block SAR data," the mitigation of the bias is strictly proportional to the interferograms with the largest temporal baseline in the block-shaped interferometric network itself, but it is not the PTA-based algorithm that is able to correct the bias. I would like to point this out given the confusion that is currently present in the InSAR community on the issue of phase bias in ML SB interferograms. 

 

Probably, for the test area under consideration using a maximum temporal baseline of 40 days may be somewhat sufficient to avoid major inconsistencies related to bias on interferometric products. 

Ultimately, I suggest smoothing the reported sentence taking into account the suggestions made so far.

Author Response

The attachment is a revised manuscripts and responses.

Author Response File: Author Response.pdf