GIS-Based Landslide Susceptibility Mapping for Land Use Planning and Risk Assessment

Round 1

Reviewer 1 Report

In the manuscript “GIS-based landslide susceptibility mapping for land use planning and risk assessment” the author present the methodology used to derive a shallow landslides susceptibility map of the Portofino Promontory base on a semi-quantitative analytical hierarchy process method. The manuscript is well organized and clearly written but I would like to address in the following general comments some methodological/conceptual issues that, on my opinion, should be explained by the authors.

General comments:

1. In section 2.1 (Study area) it is recalled the presence of several tectonic elements in the area. The authors should point out the reason for which “proximity from tectonic elements” has not been considered as a conditioning factor in section 3. (Landslide conditioning factors).
2. The resulting susceptibility map addresses the generic “shallow landslides” category, while the landslide catalogue used distinguish between more detailed categories. The authors should clarify why decided to generate a comprehensive susceptibility map instead of individual susceptibility maps according with each type of landslide present in the area.
3. In the discussion section I was expecting a quantitative assessment (i.e. accuracy and/or ROC curve evaluation) of the resulting susceptibility map including, for instance, a comparison with respect to the already existing one. In my opinion this analysis is required when proposing this kind of study.

Author Response

Dear Reviewer,

We wish to thank you for reviewing our manuscript and for the suggestions and requests that helped us improving it. 

Please find hereafter our point-by-point response.

1. In section 2.1 (Study area) it is recalled the presence of several tectonic elements in the area. The authors should point out the reason for which“proximity from tectonic elements” has not been considered as a conditioning factor in section 3. (Landslide conditioning factors).

Thank you for your thoughtful comment. Faults and fractures constrain the morphology of the Promontory and the hydrographic network pattern indeed, while the bedding plane does not seem to matter, at least in this specific case. We have selected several conditioning factors for the landslide susceptibility assessment based on the most important and significant ones at the land planning scale, although other elements could certainly be introduced, especially linked to the geological history of the area.

The planimetric definition of a tectonic element is subject to inevitable uncertainties, and in this case we have preferred to manage parameters that can be defined more objectively.

However, we are working on a subsequent research based on a detailed original field survey at 1:2.000 scale in the pilot basins of San Fruttuoso and Paraggi, also with geotechnical in situ tests, which will allow us, among other things, to take more account of geological, tectonic and geomorphological factors.

2. The resulting susceptibility map addresses the generic “shallow landslides” category, while the landslide catalogue used distinguish between more detailed categories. The authors should clarify why decided to generate a comprehensive susceptibility map instead of individual susceptibility maps according with each type of landslide present in the area.

With reference to the classification of landslide based on the type of geomaterials and the velocity of the process proposed by Hungr et al. (2014) as generic “shallow landslides” we mean all the rapid and very rapid mass movements that usually involve the shallow stratigraphy and small volumes of geomaterials, so including each type of landslide that we considered in the manuscript, i.e., shallow landslides, mud flows, debris flows and rock falls. We have clarified the meaning of the expression “shallow landslides” referred to the susceptibility map and then we adopted a more appropriate term in the text.

3. In the discussion section I was expecting a quantitative assessment (i.e. accuracy and/or ROC curve evaluation) of the resulting susceptibility map including, for instance, a comparison with respect to the already existing one. In my opinion this analysis is required when proposing this kind of study.

We implemented results of both qualitative and quantitative assessment of the produced susceptibility map performed using the landslide test set, i.d., a part of the landslide inventory (20%) based on a random selection. For the validation of the results, we performed:

- spatial distribution of landslides within the five LSM classes.

- estimation of the predictive accuracy and efficacy of the proposed model using the ROC analysis.

Furthermore, we compared in quantitative terms the proposed AHP susceptibility map with one already existing using the Cohen's Kappa coefficient.

Sincerely

Reviewer 2 Report

The authors used an analytical hierarchy process (AHP) to identify the areas prone to landslide in the Portofino region of Italy. They used nine controlling factors in landslide occurrence to develop the methodology. They validated the results using an available landslide inventory.

The article is well-written and easy to follow. The AHP process was conducted correctly and the conditioning factors were suitable for the study. However, I have a few comments on the novelty of the work and validation of the results.

1. The AHP has been widely used in landslide susceptibility mapping. The authors should explain the novelty of this study. It was briefly touched in line 100. However, it should be elaborated in the revised manuscript.
2. The authors mentioned that they used a portion of the landslide inventory for validation (line 386). However, no additional information regarding the validation of the results was provided. They could use, for example, the area under the ROC curve (AUC) to validate the results.

Specific comments

Line 273. Please clarify how the slope aspect can accelerate or decelerate landslide occurrence. Use relevant references for the explanations.

Line 470. It would be helpful if you could briefly explain the approach that was used in providing the previous landslide susceptibility map in the study area to compare with the methodology presented in this manuscript.

Line 480. Where are the validation results?

Line 40. Please correct: they are correlated

Line 44. Please correct: potentially unstable

Line 61. Please correct: at a regional

Line 211. Please correct: frequently occur

Line 221 to 226, please restructure this sentence.

Line 245. Remove the duplicated “as the”

Line 245. Please correct: Rainfall is

Line 349 Please correct: we also considered

Line 479 . “training test”? Do you mean training set?

Line 595. Please correct: relies

Line 604. Please correct: landslide zones

Line 629 Please correct: often induce

Good luck!

1/3/2021

Author Response

Dear Reviewer,

We wish to thank you for reviewing our manuscript and for the suggestions and requests that helped us improving it. 

Please find hereafter our point-by-point response.

Comments and Suggestions for Authors

The authors used an analytical hierarchy process (AHP) to identify the areas prone to landslide in the Portofino region of Italy. They used nine controlling factors in landslide occurrence to develop the methodology. They validated the results using an available landslide inventory.

The article is well-written and easy to follow. The AHP process was conducted correctly and the conditioning factors were suitable for the study. However, I have a few comments on the novelty of the work and validation of the results.

1. The AHP has been widely used in landslide susceptibility mapping. The authors should explain the novelty of this study. It was briefly touched in line 100. However, it should be elaborated in the revised manuscript.

The text has been integrated according to the reviewer’s request:

“Further, including anthropogenic factors in the conditioning ones correspond to add the effects of human activity on the Earth’s surface, that is considering man as a morphogenetic factor. The effects of human modification, that may assume a very high importance locally, may result even in landslide generation. In this sense the diffuse presence of abandoned man-made terraces may result in a possible source of shallow landslides.”

2. The authors mentioned that they used a portion of the landslide inventory for validation (line 386). However, no additional information regarding the validation of the results was provided. They could use, for example, the area under the ROC curve (AUC) to validate the results.

We implemented results either qualitative or quantitative assessment of the produced susceptibility map performed using the landslide test set, i.d., a part of the landslide inventory (20%) based on a random selection. For the validation of the results, we performed:

- spatial distribution of landslides within the five LSM classes.

- estimation of the predictive accuracy and efficacy of the proposed model using the ROC analysis.

Specific comments

Line 273. Please clarify how the slope aspect can accelerate or decelerate landslide occurrence. Use relevant references for the explanations.

Section 2.3.2 has been implemented: the role of slope aspect on landslide occurrence has been more detailed and references added.

Line 470. It would be helpful if you could briefly explain the approach that was used in providing the previous landslide susceptibility map in the study area to compare with the methodology presented in this manuscript.

The methodology, which derives from a specific technical regulation, includes some conditioning factors at the catchment’s scale: lithology, slope steepness, land use, hydrological effectiveness, debris cover and related granulometry and the presence of active or inactive landslides. Some other worsening factors are then included, e.g., lithological contact, fault, erosional talweg channel, edge of fluvial erosion scarp, terrace edge, slope angle discontinuity. Weights within every factor are assigned subjectively apart for lithology where a statistical approach is used, considering the proportion of in landslide lithotype. Maps are summed up and after a normalization, susceptibility classes are assigned. Then the methodology is highly affected by subjectivity and strongly relay on the presence of active and inactive landslides. Conditioning factors at local scale, including man-made landforms, are not included in the assessment of landslide susceptibility. Furthermore, areas where active, dormant or inactive/stabilized landslide, deep seated gravitational slope deformation, widespread active or dormant instability phenomena and shallow landslides are detected, are automatically classified as high or very high susceptibility. zones. The obtained map is classed in 5 increasing levels of susceptibility, named from Pg0 to Pg4 (Pg=Pericolosità Geomorfologica, mean Geomorphological Hazard).

Line 480. Where are the validation results?

Line 40. Please correct: they are correlated

Line 44. Please correct: potentially unstable

Line 61. Please correct: at a regional

Line 211. Please correct: frequently occur

Line 221 to 226, please restructure this sentence.

Line 245. Remove the duplicated “as the”

Line 245. Please correct: Rainfall is

Line 349 Please correct: we also considered

Line 479 . “training test”? Do you mean training set?

Line 595. Please correct: relies

Line 604. Please correct: landslide zones

Line 629 Please correct: often induce

We revised the paper and introduced the suggested punctual corrections.

Sincerely,

The authors

Reviewer 3 Report

The authors proposed a landslide susceptibility analysis then comparing it with the existing susceptibility map.

The manuscript present a good scientific report, but 29 pages is too long. While the size of the figures and tables affect the amount of pages, this manuscript can be improved. The sentences can be more concise and only important information or significant findings should be highlighted. Additional or less important sentences should be avoided.

Author Response

Dear Reviewer,

We wish to thank you for reviewing our manuscript and for the suggestions and requests that helped us improving it.

Please find hereafter the point-by-point responses.

Comments and Suggestions for Authors

The authors proposed a landslide susceptibility analysis then comparing it with the existing susceptibility map.

The manuscript present a good scientific report, but 29 pages is too long. While the size of the figures and tables affect the amount of pages, this manuscript can be improved. The sentences can be more concise and only important information or significant findings should be highlighted. Additional or less important sentences should be avoided.

We have revised the text and figures as requested. Due to other reviewers's request we had to add the proper responses and some text and tables.

Sincerely,

The authors

Reviewer 4 Report

Authors assessed the landslide susceptibility of a study area in Italy by using the semi-quantitative analytic hierarchy process (AHP) approach and compare the output map with the one currently used by local stakeholders in land use planning.

The manuscript represents an application of the well-known AHP approach to a case-study. The paper is well written and figures are good, but there are some issues that I feel the authors need to consider.

The main points on which I believe the authors need to supply more information or consider are detailed in the following. Other suggestions are annotated on the pdf.

1. L.38-39. This is the definition of hazard, not of risk. In fact, apart from hazard, the risk assessment requires information on the vulnerability and amount of elements-at-risk (van-Westen et al. 2006). [van-Westen C. J., Van-asch T. W., Soeters R. 2006. Landslide hazard and risk zonation -Why is it still so difficult? Bulletin of Engineering Geology and the Environment, vol.68, 3(4), 167-184. 10.1007/s10064-005-0023-0].

2. The Authors decided to represent landslides as punctual features used as input data in the susceptibility model (L230). What does a point represent? The landslide scarp/crown centroid? The highest point of the scarp/crown? The landslide body centroid? Additional details on the adopted sample criteria should be provided given that the selection of landslide pixel/point samples can affect the resulting landslide susceptibility map.

3. The Authors state that the susceptibility map has been reclassified based on the histogram (L529-530). This sentence without any additional detail has not much meaning giving that different methods based on histogram can be used to classify a map: e.g. quantile, natural breaks, standard deviation, geometric interval etc. Which type of reclassification have you used? You should also explain the rationality behind your choice giving that it strongly affects the extent of each susceptibility class. In the discussion session, some considerations about the influence of the reclassification method on the resulting map are necessary.

4. In order to verify the reliability of the AHP susceptibility model, data concerning the predictive capability of the model based on the validation set require to be provided. e.g. receiver operating characteristics (ROC) analysis, TP (true positives) and TN (true negatives), FN (false negatives) and FP (false positives), prediction rate performance etc…

5. Given the aim of the manuscript to compare the proposed AHP susceptibility map with one already existing (L100), the two maps require to be quantitatively compared. In order to measure the agreement between the two classified maps, the authors can choose among the commonly used indices (e.g. Cohen's Kappa coefficient, Jaccard index, goodness of fit by Hargrove etc...) which can be easily obtained through geoprocessing tools.

Moreover, I suggest to add a map showing where the main differences are located in order to highlight the spatial disagreement between the two susceptibility maps. 

[Cohen J. 1960. A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, pp. 37-46.

Hargrove W.W., Hoffman F.M., Hessburg P.F. 2006. Mapcurves: a quantitative method for comparing categorical maps. J. Geogr. Syst. 8, 187–208.DOI 10.1007/s10109-006-0025-x.

Jaccard, P. 1901, “Etude comparative de la distribution florale dans une portion des Alpes et du Jura”, Bulletin de la Soci ́et ́e Vaudoise des Sciences Naturelles 37(1), 547–579].

Comments for author File: Comments.pdf

Author Response

Dear reviewer,

We wish to thank you for reviewing our manuscript and for suggestions and requests that helped us improving it.

Please find hereafter the point-by-point responses.

Comments and Suggestions for Authors

 Authors assessed the landslide susceptibility of a study area in Italy by using the semi-quantitative analytic hierarchy process (AHP) approach and compare the output map with the one currently used by local stakeholders in land use planning.

The manuscript represents an application of the well-known AHP approach to a case-study. The paper is well written and figures are good, but there are some issues that I feel the authors need to consider.

The main points on which I believe the authors need to supply more information or consider are detailed in the following. Other suggestions are annotated on the pdf.

1. L.38-39. This is the definition of hazard, not of risk. In fact, apart from hazard, the risk assessment requires information on the vulnerability and amount of elements-at-risk (van-Westen et al. 2006). [van-Westen C. J., Van-asch T. W., Soeters R. 2006. Landslide hazard and risk zonation -Why is it still so difficult? Bulletin of Engineering Geology and the Environment, vol.68, 3(4), 167-184. 10.1007/s10064-005-0023-0].

According to the suggested definition, we revised the sentence using correctly the definition of risk and hazard.

2. The Authors decided to represent landslides as punctual features used as input data in the susceptibility model (L230). What does a point represent? The landslide scarp/crown centroid? The highest point of the scarp/crown? The landslide body centroid? Additional details on the adopted sample criteria should be provided given that the selection of landslide pixel/point samples can affect the resulting landslide susceptibility map.

We represented landslides as punctual features due to the incomplete and not homogenous information about area and/or size of the landslides. Point represents the highest point of the scarp.

3. The Authors state that the susceptibility map has been reclassified based on the histogram (L529-530). This sentence without any additional detail has not much meaning giving that different methods based on histogram can be used to classify a map: e.g. quantile, natural breaks, standard deviation, geometric interval etc. Which type of reclassification have you used? You should also explain the rationality behind your choice giving that it strongly affects the extent of each susceptibility class. In the discussion session, some considerations about the influence of the reclassification method on the resulting map are necessary.

To determine the class intervals in the landslide susceptibility map we used three classification systems commonly applied in landslide susceptibility zonation: natural breaks, quantiles and equal intervals and compared the different results. In this study, we selected the natural breaks to divide the LSM values into five classes, from very low to very high according to the histogram values, which is multimodal and shows empty class intervals. Other two methods cannot be used in this study. The quantile classification is not appropriate because data are not linearly distributed; whereas using equal intervals the lowest susceptibility class is too emphasized relative to others.

4. In order to verify the reliability of the AHP susceptibility model, data concerning the predictive capability of the model based on the validation set require to be provided. e.g. receiver operating characteristics (ROC) analysis, TP (true positives) and TN (true negatives), FN (false negatives) and FP (false positives), prediction rate performance etc…

We implemented results of both qualitative and quantitative assessment of the produced susceptibility map performed using the landslide test set, i.d., a part of the landslide inventory (20%) based on a random selection. For the validation of the results, we performed:

- spatial distribution of landslides within the five LSM classes.

- estimation of the predictive accuracy and efficacy of the proposed model using the ROC analysis.

In particular, in order to evaluate the predictive accuracy and efficiency of the map, we plotted the true positive rate (TPR), i.e., the correctly predicted events, opposite to the false positive rate (FPR), i.e., the falsely predicted events and calculated the correlated AUC value.

5. Given the aim of the manuscript to compare the proposed AHP susceptibility map with one already existing (L100), the two maps require to be quantitatively compared. In order to measure the agreement between the two classified maps, the authors can choose among the commonly used indices (e.g. Cohen's Kappa coefficient, Jaccard index, goodness of fit by Hargrove etc...) which can be easily obtained through geoprocessing tools.

We compared in quantitative terms the proposed AHP susceptibility map with the existing regional susceptibility zonation using the Cohen’s Kappa calculation. Outcomes show a general slight concordance between the two maps: medium and high classes display the lowest and no concordance respectively, while the very low one presents the relatively higher degree of concordance confirming the general result even if differentiating between classes. The lowest susceptibility class corresponds substantially with the lower slope gradient areas, that is the lower values of the more critical conditioning factor. On the other hand, the weak concordance in the three higher susceptibility classes highlights the strong differentiation between the two maps where landslide triggering probability is higher

6. Moreover, I suggest to add a map showing where the main differences are located in order to highlight the spatial disagreement between the two susceptibility maps.

We thought that map comparison in Figure 8 shows clearly the spatial differences and disagreements between the two susceptibility models. Moreover, the paper includes already a great number of figures and tables and additional map may further affect the number of pages.

Sincerely,

The authors

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Dear Authors,

after this round of revision I suggest to accept the manuscript in present form in order to be published in the Land journal.

Reviewer 2 Report

All my previous comments have been addressed by the authors. I have no further comments for this manuscript.

Good luck to the authors.

1/23/2021

Reviewer 4 Report

Dear Authors,

thank you for accepting my suggestions and reviewing the manuscript accordingly.

Good luck