Review Reports - Rigidity Strengthening of Landslide Materials Measured by Seismic Interferometry

Round 1

Reviewer 1 Report

The authors presented a method to measurement of material rigidity using seismic interferometery. Few minor comments are:

1. Line 85: [27] first observed drops in seismic velocity...

Line 472: [46] noted that frequency-dependet...

Such method of referencing is unsual. Please check other manuscript published in Remote Sensing.

2. This method is effective for slow-moving landslide, but how about the rapid-moving? Is it possible to apply the same method? If so, will the same result obtained?

Author Response

Responses to the Reviewer’s suggestions

Reviewer #1

The authors presented a method to measurement of material rigidity using seismic interferometry. Few minor comments are:

(1) Line 92: [27] first observed drops in seismic velocity… Line 431: [46] noted that frequency-dependent… Such method of referencing is unusual. Please check other manuscript published in Remote Sensing.

Thanks for the comment. We have carefully checked and modified in the revised manuscript.

(2) This method is effective for slow-moving landslide, but how about the rapid-moving? Is it possible to apply the same method? If so, will the same result obtained?

Actually, most of monitoring methods used in present study can be successfully applied to rapid-moving landslide, except for downhole monitoring of displacement and ground water level and GPS measurement. Such downhole and geodetic monitoring, is an invasive tool, cam be limited on rapidly moving landslide that are not safe to monitoring approach. Thus, a noninvasive approach is needed for monitoring both slow- and rapid-moving landslides, which would be very helpful for issuing landslide hazard warnings. A possible solution to continuously monitoring landslide is the coda wave interferometry (CWI) adopted in present study.

Major result presented in current manuscript is that we proposed a conceptual model of the landslide recovery process induced by water loading (undrained water). Where all sliding materials were stable (safety factor > 1.0), unconsolidated landslide colluvium and impermeable sliding surfaces trapped the seepage water to form a water tank, provided that compact facilities increases in the cohesion and strength of landslide materials, thereby increasing the landslide materials’ stability. Landslide sites satisfied the aforementioned conditions probably result the recovery process of landslide body. Our study opens an additional perspective for the study of fluid effects on landslide stability.

Reviewer 2 Report

Review of “ Strengthening the Rigidity of Landslide Materials Measured by Seismic Interferometry », by Kang et al

This article proposes to study the possible effect ground water level and local seismic acceleration on slope stability, and suggest to incorporate relative seismic velocity changes in observational and predictive models. This article has a great potential but requires significant improvement before publications. I therefore recommend Mandatory (Major) Revisions.

Majors comments :

Introduction and discussions: the present study has similarities with the one published by Bontemps et al, Nature Comm. 11 780 (2020), where they described the joint role of precipitations and local earthquake to slope destabilization and relative velocity changes. It would be interesting to compare both works, and exhibit the similar features and possible differences.

From fig 3a, we can estimate that the estimated GWL does not fit with the observed GWL, such that it is very hard to pretend that the hydrological model is significantly right. If so, please explain why and how the misfit could be considered negligible, as it seems to omit to predict the GWL during the 2018 monsoon.

Unfortunately, the presented data and figure, especially FIG 3, are impossible to interpret and hardly readable. It is mandatory to draw new graphs, each quantity with a proper color/line style. Preferably, separate the graphs of different quantities, or limit the amount of graphs on the same figure. This in mandatory to warrant the scientific impact of the publication. Then a clear discussion on each graph would help the reader.

The proposed models are not sufficiently supported by experimental observations yet. More attention should be paid in explaining each behavior of the model, changing one parameter at a time (which in nature may take time, or at least will require careful discussions).

Could the author explain how the present models could be adapted to other site studies? Especially the GWL models and the safety factor seems to be site dependent… which might limit application to other sites.

Minor comments :

P3, end of introduction : forcasting models should be confirmed on other sites and at other times for validation. Otherwise, make clear that the present work is a tentative model.

Fig 1b: please redraw properly. Especially, yellow texts are not visible.

Fig 1 – caption: SP1 and SP2 a far from the landslide (several hundreds of meter). Can the author estimate the horizontal sensitivity kernel of the correlograms to estimate how much the correlograms reveal the internal mechanical evolution of the landslide (and not the material out of it). This might be a technical detail.

Fig 3: plit in more figures and redraw. X-axis : 3600 days?

a/y-axis : simulated or observed GWL

a/y-axis: where are plotted dV/V? dV/V hardly visible

p10 section 5.3: the loss of coherence has been demontrated to be well explained by pore filling with water in multicomposite granular materials. See R. Thery, et al, Tracking fluids in multiple scattering and highly porous materials : toward applications in non-destructive testing and seismic monitoring, Ultrasonics, 102, 106019 (2019).

Author Response

Responses to the Reviewer’s suggestions

Reviewer #2

Major Comments:

(1) Introduction and discussions: the present study has similarities with the one published by Bontemps et al. Nature Comm. 11, 780 (2020), where they described the joint role of precipitations and local earthquake to slope destabilization and relative velocity changes. It would be interesting to compare both works, and exhibit the similar features and possible differences.

Following the reviewer’s suggestion, we have carefully discussed our results with Bontemps et al. (2020) and added the above reference in the revised manuscript. See modified text in Lines 404-408. Actually, Bontemps et al. (2020) highlights the cumulated earthquake forcing of frequent small earthquakes (peak ground velocity is less than 1 cm/s) combined with a high water content in landslide composed of debris avalanche and lacustrine deposits that prevents the landslide from recovering its rigidity. Overall, they not only observed a clear coseismic drop in velocity, which can be attributed by the micro- and macro- fractures caused by the moderate-sized earthquake (M ≥ 5.0), but also found an increasing in relative seismic velocity (dv/v) during the dry season and/or a few weeks after the moderate-sized earthquake.

In our study, the environmental force-related (earthquake, ground water level, and precipitation) dv/v changes can be characterized into three periods: several rainfall episodes without earthquake activity (period 1, P[1]), large effect of earthquake forcing (period 2, P[2]), and prolonged and intense precipitations (period 3, P[3]) (Figure 3a) without earthquake effect, which would be very helpful to investigate the effect of individual forcing in dv/v observations. In P1, an obvious pattern is the typical cycle of reduction and recovery (RAR) in dv/v in response to material changes related to water infiltration and in-filled water running out. During the dry season, as indicated in P2 (Figure 3a), the possible effects of rainfall penetration on the observed dv/v could be neglected as the significant earthquakes occurred in P2. Indeed, a clear scatter pattern in dv/v measurements is evident. We suspect that the frequent strong ground shaking caused dv/v measurements to tend to scatter. Notably, a cycle of RAR in dv/v exhibited a longer cycle duration of approximately 80 days during P3 that coincided with frequent rainfall episodes. We further noticed that the dv/v increased with intense precipitation, but the response of the observed GWL was conspicuously uncorrelated with rainfall data. A large discrepancy in GWL change rates between modeled and observed data was also observed, which implied the occurrence of temporal changes in rainwater infiltration properties (Figure 3a).

(2) From fig 3a, we can estimate that the estimated GWL does not fit with the observed GWL, such that it is very hard to pretend that the hydrological model is significantly right. If so, please explain why and how the misfit could be considered negligible, as it seems to omit to predict the GWL during the 2018 monsoon.

In fact, it is challenging to establish a hydrological model that can predict long-term water level changes, especially in areas with frequent earthquakes or typhoons, where the current situation will change with time or at any event. To effectively quantify the effectiveness of GWL simulations, the mean absolute error (MAE) and the mean relative error (MRE) were evaluated in this study, as shown in equations (1) and (2) in original manuscript, respectively. When the MRE was less than 15% in the present study, parameter calibration was deemed complete. We have presented the values of MAE and MRE in the “Section 5.1 Model Calibration and Groundwater Level Prediction“ of our original manuscript. Please see texts in Lines 321-341.

Overall, results of model simulation exhibit a good reference value before the Julian day of 196 in 2018. However, there is a clear decreasing in model credibility after Julian day of 196 in 2018, even when similar rainfall events occurred afterward. The reduced credibility may imply temporal changes in hydraulic property, resulting changes of preferential paths for fluid infiltration. Several factors can contribute large discrepancies between observed GWL and model result. In this study, we proposed a hypothetical model based on thee water-load compacting force, which is supported by the dv/v observations, geodetic displacement (vertical subsidence), and inclinometer measurements, to interpret the aforementioned discrepancies. Our proposed model has been carefully examined in Sections 6.1 “Factors Influencing Daily Relative Velocity Changes” and 6.2 “Evidence to Support the Water-Load-Based Compacting Model”. Please see texts in Lines 460-523.

(3) Unfortunately, the presented data and figure, especially FIG 3, are impossible to interpret and hardly readable. It is mandatory to draw new graphs, each quantity with a proper color/line style. Preferably, separate the graphs of different quantities, or limit the amount of graphs on the same figure. This mandatory to warrant the scientific impact of the publication. Then a clear discussion on each graph would help the reader.

Modified as suggestion. See Figure 3 in the revised manuscript.

(4) The proposed models are not sufficiently supported by experimental observations yet. More attention should be paid in explaining each behavior of the model, changing one parameter at a time (which in nature may take time, or at least will require careful discussions).

Thanks for the comment. A hypothetical model based on the water-loading compacting force proposed in the present study has been carefully examined in the Section 6.2 “Evidence to Support the Water-Load-Based Compacting Model”, and supported by the dv/v observations, the RTK and inclinometer measurements, the specific conditions of geological model, the laboratory testing of rock strength, the hydrogeological modeling, and the failure scenario. In summary, the formation of a water tank during the rainfall/typhoon seasons requires satisfaction of the following four landslide material preconditions: (1) The impermeable interface (shear zone) acts as an aquitard; (2) preexisting landslide colluvium located (aquifer) at the top of aquitard is required to trap the rainwater; (3) the safety factor for the storm period is > 1.0; and (4) materials underneath the aquitard boundary are compressible. See Lines 481-523.

(5) Could the author explain how the present models could be adapted to other site studies? Especially the GWL models and the safety factor seems to be site dependent… which might limit application to other sites.

Actually, the hydrogeological conceptual model relies very much on a hydrogeological model and in-situ observations of monitoring (e.g., precipitation, groundwater level, and borehole displacement). Indeed, results of safety factors and/or triggering thresholds inferred from the failure scenario are limited to our landslide site. However, the seismic technique, geoengineering-based monitoring, and numerical simulation adopted in the present study can be readily implemented in other places with high landslide activity.

A question which is also raised here, can the monitoring techniques be used for studying the rapid-moving landslide? Most of monitoring methods used in present study can be successfully applied to both slow- and rapid-moving landslides, except for downhole monitoring of displacement and ground water level and GPS measurement. Such downhole and geodetic monitoring, is an invasive tool, cam be limited on rapidly moving landslide that are not safe to monitoring approach. Thus, a noninvasive approach is needed for monitoring both slow- and rapid-moving landslides, which would be very helpful for issuing landslide hazard warnings. A possible solution to continuously monitoring landslide is the coda wave interferometry (CWI) adopted in present study.

Minor Comments:

(6) End of introduction: forecasting models should be confirmed on other sites and at other times for validation. Otherwise, make clear that the present work is a tentative model.

Modified as suggested. See Lines 120-122.

(7) Fig 1b: please redraw properly. Especially, yellow texts are not visible.

See Figure 1 in the revised manuscript.

(8) Fig 1- caption: SP1 and SP2 a far from the landslide (several hundreds of meter). Can the author estimate the horizontal sensitivity kernel of the correlograms to estimate how much the correlograms reveal the internal mechanical evolution of the landslide (and not the material out of it). This might be a technical detail.

Taking the reviewer’s comments into account, we have added the text to present the horizontal sensitivity kernel of the correlograms. See Lines 306-318.

(9) plot in more figures and redraw. X-axis: 3600 days? Y-axis: simulated ot observed GWL. Y-axis: where are plotted dv/v? dv/v hardly visible

See revised Figure 3.

(10) Section 5.3: the loss of coherence has been demonstrated to be well explained by pore filling with water in multicomposite granular materials. See R. Thery et al, Tracking fluids in multiple scattering and highly porous materials: toward application in non-destructive testing and seismic monitoring, Ultrasonics, 102, 106019 (2019)

We have added texts to clarify above statement. See text in Line 401-413.

Reviewer 3 Report

Review comments on

“Strengthening the rigidity of Landslide materials Measured by Seismic Interferometry” by Ken-Hao Kang et al.

In this manuscript the authors selected the Chashan landslide site which has been intensively studied from geological, geophysical, geodetic, geotechnical, hydrological, and seismological perspectives.

This study is relevant with the scopes of the journal, and for this reason my recommendation is to be published after considerable corrections and clarifications which are the following.

The title refers to seismic interferometry measurements although other techniques have been also performed. The study is based in the implementation of several geoengineering techniques (103-112) which are not included in the title. Thus, my recommendation is to change the title appropriately.

In the first four chapters the authors describe in very detail the methods and techniques that they used for this study. In some cases (particularly in the introduction) it is not clear if a method is applied here, or it is mentioned as a general practice.

In more detail, my comments are the following:

289 Please clarify what do you mean with all station pairs. The stations installed here are two.

Fig. 3a. I cannot see red diamonds, only grey and blue. Also, I cannot see the green square.

Fig. 3a. E2 and E4 are rather black than grey.

Fig 3b. The simulated rates after day 160 of 2018 seem to be very different than the observed. Please comment it.

365-366 How and where the PGA was measured?

What is the effect in the landslide triggering of PGA values less than 80 gal, when the safety factor drops below 1.0 in PGA values larger than 530 gal? (397).

432-441: What are the parameters which can cause the variation in seismic velocity due to the earthquake activity? Have you measured or calculate any of these parameters in the test area? It seems that the strong earthquakes are rather far from the test site.

Figure S6 includes all the earthquakes of magnitude larger than M4.1 in the area?

What are the geodetic measurements which conducted in the present study?

483 The frequencies in fig.2 are in the range 2-20 Hz.

Author Response

Responses to the Reviewer’s suggestions

Reviewer #3

This study is relevant with the scopes of the journal, and for this reason my recommendation is to be published after considerable corrections and clarifications which are the following.

(1) The title refers to seismic interferometry measurements although other techniques have been also performed. The study is based in the implementation of several geoengineering techniques (103-112) which are not included in the title. Thus my recommendation is to change the title appropriately.

Thanks for the comment. We investigated the changes in landslide’s rigidity, based on the measurements of relative seismic velocity change (dv/v) over time thanks to seismic interferometry technique. We think that current title not only convey the main topic of the study but also highlight the importance of the research.

(2) In the first four chapters the authors describe in very detail the methods and techniques that they used for this study. In some cases (particularly in the introduction) it is not clear if a method is applied here, or it is mentioned as a general practice.

Most of the monitoring techniques mentioned in the introduction were applied in the present study, except for the GPS measurements.

In more detail, my comments are the following:

(3) Line 281: Please clarify what do you mean with all station pairs. The stations installed here are two.

Modified as suggestion. Lines 284.

(4) Fig 3a. I cannot see red diamonds, only grey and blue. Also, I cannot see the green square. Fig 3a. E2 and E4 are rather black than grey.

We have carefully modified the Figure 3 to make it more easily readable. See Figure 3 in the revised manuscript.

(5) Fig 3b. The simulated rates after day 160 of 2018 seem to be very different than the observed. Please comment it.

A large discrepancy in GWL change rates between modeled and observed data was presented by Figure 3b, which implied the occurrence of temporal changes in rainwater infiltration properties. Actually, a hypothetical model based on the water-load compacting force was proposed in the present study. We have carefully investigated the possible causes and examined our proposed model in Sections 6.1 “Factors Influencing Daily Relative Velocity Changes” and 6.2 “Evidence to Support the Water-Load-Based Compacting Model”. See Lines 460-523.

(6) Lines 358-363: How and where the PGA was measured? What is the effect in the landslide triggering of PGA values less than 80 gal, when the safety factor drops below 1.0 in PGA values larger than 530 gal?

We have conducted a series of failure scenario based on the different tried PGA values to discuss a possible threshold of earthquake forcing for the landslide site. We found that the safety factor (SF) of the B3 potential sliding mass (see Figure S5) dropped to 1.0, indicating the B3 mass reaches the failure condition under a PGA value of 530 Gal. In practical, the PGA of 530 can be used to issue the landslide warning. All potential sliding mass is stable when the PGA values less than 80 Gal.

(7) Lines 432-441: What are the parameters which can cause the variation in seismic velocity due to the earthquake activity? Have you measured or calculate any of these parameters in the test area? It seems that the strong earthquakes are rather far from the test site.

(8) Figure S6 includes all the earthquakes of magnitude larger than M4.1 in the area?

Earthquakes produced the seismic intensities larger than II at the landslide site are used to discuss the earthquake forcing affecting on the dv/v measurements. Thus, the local magnitude of earthquakes ranges from 4.1 to 6.2. We have modified text to clarify above statement. See revised texts in Lines 428-435.

(9) What are the geodetic measurements which conducted in the present study?

We have replaced the word of “Geodetic” by “RTK”. See Line 511.

(10) Line 483: The frequencies in fig.2 are in the range 2-20 Hz.

Result of spectrogram analysis shown in Figure 2 exhibits the strong spectral energy of coda wave distribution within a frequency band of 2-10 Hz. Thus, 2-10 Hz band-pass filtering was applied for the following analysis of the dv/v measurements.

Round 2

Reviewer 2 Report

I think not all comments have been addressed, I still find the figures are very hard to read and understand.

Reviewer 3 Report

after the replies of the authors, this manuscript could be published to the journal