Response to Review Comments Title : The assessment of earthquake-triggered landslides susceptibility with considering coseismic ground deformation

Responses to the Comments Raised by Reviewer #1 1. In the manuscript “The assessment of earthquake-triggered landslides susceptibility with considering coseismic ground deformation”, the authors try to improve earthquake-triggered landslide susceptibility maps by introducing a new parameter that they call “coseismic ground deformation” Assuming as a study case the October 23, 2004, Mw 6.8 Niigata earthquake, the authors provide several landslide susceptibility maps at two different scales, using three different statistical methods, namely, the logical regression (LR), the Support Vector Machine (SVM), and the Artificial Neural Network (ANN). The authors conclude their study, saying that the “coseismic ground deformation” parameter is an “important” factor to evaluate the susceptibility of landslides. In my opinion, the small increase of the area under the receiver operating characteristic (ROC) curve, i.e., from approximately 0.72 to 0.77, obtained by introducing the “coseismic ground deformation” parameter, does not support the conclusion stated by the authors. Such a small improvement is primarily affected by the generally low resolution of the other parameters introduced in the analysis.


different statistical methods (logistic regression, Artificial Neural Network and Support
Vector Machine) are applied to landslide inventory at two different scales: regional and near field. In this last case, coseismic ground deformation is considered as an influencing factor in the susceptibility analysis. From the analyses, the ANN method gives the best results. The objective of the paper is to analyze the importance of the coseismic ground deformation to explain landslide distribution and the benefits of using it when preparing susceptibility maps.
The paper is properly organized and most of figures and tables are of interest.

Authors' reply:
Many thanks for your positive comments and valuable time to improve the manuscript.

2.
Regarding the main objective of the paper, I miss a reflection by the authors about the true usefulness of the parameter in question in the preparation of susceptibility maps.
As the authors point out in the Introduction, these maps constitute the main tool that our society has to establish the areas prone to suffer seismic-induced landslides, and thus define an appropriate use (or restrict their occupation) of the territory. However, the parameter that constitutes the center of the article, the coseismic ground deformation, is a parameter that can only be evaluated afterwards, that is, once the earthquake has occurred. So what real use does it have? Personally, I see this parameter, as well as the distance to the surface of rupture, useful for subsequent studies, to explain why instabilities have occurred in certain contexts or areas, but not to predict their occurrence. In fact, the difference in AUC when considering/not considering this parameter is less than 5%.

Authors' reply:
Many thanks for your useful comments. The landslide susceptibility is defined as the In addition, the coseismic ground deformation will help to reveal the hidden subsurface damage. It should be noted that not all deformation will direct lead the landslides.
However, the area with large coseismic surface deformation often indicates that the movement of the rock mass may be further developed and the integrity of rock mass is reduced, which renders slopes prone to landslip in future earthquakes again. Zhao et al., (2012) explored the localized coseismic deformation in Kizawa (a small village), Japan after the earthquake. The results showed the calculated coseismic deformation in Kizawa is relatively larger, but the landslides are sparse. However, after a detail investigation, it found that the underground structures such as tunnels and wells were severely damaged. The road alignment of the Kizawa tunnel, which was buried 30 m beneath the ground surface, was shifted sideways 1-1.5 m to east-to-southeast direction.
Furthermore, two irrigation well were dislocated at 30 m and 20 m, beneath the ground, respectively. Therefore, it is highly possible that the ground underwent some subsurface damage at locations where the large coseismic deformation. Although the deformation did not form the landslides at these locations in the 2004 Mid-Niigata earthquake, as there were accumulated deformation within the rock and soil, the landslide will easily occur in the next earthquake event. According to the comments above, the coseismic ground deformation should be regarded as a useful influencing factor in the assessment of landslides susceptibility.
Then, in order to evaluate the effects of the coseismic ground deformation on the assessment of landslides susceptibility, the Analysis of Variance method (ANOVA) has been utilized to evaluate the predictive capability of used conditional factors. The factors with higher variance values indicate a higher contribution to landslide models and vice versa. The predictive capability of eight landslide affecting factors was shown in Table 1.  The results showed that the AUC values for the models varied from 0.7253 to 0.801.
So, comparing the increment of AUC values in this study with above mentioned similar studies, it may be concluded that the increasing of AUC is relatively significant.