Landslide susceptibility assessment in rocky coast subsystem of Essaouira coastal area – Morocco
- 1Higher School of Technology Essaouira, Laboratory of Applied Sciences for the Environment and Sustainable Development (SAEDD), Cadi Ayyad University, Marrakech, Morocco
- 2Universidade Aberta, Lisbon, Portugal
- 3Centre of Geographical Studies, Institute of Geography and Spatial Planning, Portugal
- 4Associated Laboratory Terra, Portugal
- 5Polydisciplinary Faculty of Safi, Safi, Morocco, Department of Earth Sciences, Cadi Ayyad University, Marrakech
- 6Faculty of sciences El Jadida, Geosciences and Environmental Techniques Laboratory, Chouaïb Doukkali University, El Jadida, Morocco
- 1Higher School of Technology Essaouira, Laboratory of Applied Sciences for the Environment and Sustainable Development (SAEDD), Cadi Ayyad University, Marrakech, Morocco
- 2Universidade Aberta, Lisbon, Portugal
- 3Centre of Geographical Studies, Institute of Geography and Spatial Planning, Portugal
- 4Associated Laboratory Terra, Portugal
- 5Polydisciplinary Faculty of Safi, Safi, Morocco, Department of Earth Sciences, Cadi Ayyad University, Marrakech
- 6Faculty of sciences El Jadida, Geosciences and Environmental Techniques Laboratory, Chouaïb Doukkali University, El Jadida, Morocco
Abstract. During the last few decades, many researchers have produced landslide susceptibility maps using different techniques and models including the information value method, which qualified as a wide applied statistical model in several coastal environments. The aim of this study was to evaluate the susceptibility for the occurrence of landslides in Essaouira coastal area using this bivariate statistical method. In this coastal area were identified, inventoried and mapped 588 landslides, of different types, mostly from the observation and interpretation of different data sources, namely high-resolution satellite images, aerial photographs, topographic maps, and extensive field surveys. The rocky coastal system of Essaouira is located in the middle part of Morocco Atlantic coastal area. The study area was split into 1534 cliff terrain units of 50 m width. For training and validation purposes the landslide inventory was divided into two independent groups: training (70 %) and for validating (30 %). Twenty-two layers of landslide-conditioning factors were prepared, including: elevation, slope angle, slope aspect, plan curvature, profile curvature, cliff height, topographic wetness index, topographic position index, slope over area ratio, solar radiation, presence of faulting, lithological units, toe lithology, presence and type of cliff toe protection, layer tilt, rainfall, streams, land-use patterns, NDVI, lithological material grain size, and presence of springs. The statistical relationship between the conditioning factors and the different types of landslides were calculated using the bivariate information value method, in a pixel and in elementary terrain units base model. Validation of the coastal landside susceptibility maps was done using the landslide training group partitions. The ROC curves and Area Under the Curve were used to assess the accuracy and prediction capacity of the different coastal landslide susceptibility models. Two methodologies were adopted to evaluate coastal landslide susceptibility, one considering a pixel base approach and another one using coastal terrain units. The resulted coastal landslide susceptibility maps allowed classifying 38 % of the rocky coast subsystem with high susceptibility to landslides, being the majority of these high susceptible areas located in the southern part of the Essaouira coastal area. Those susceptibility maps would be useful for general planned development activities in the future as well as for environmental protection.
- Preprint
(2538 KB) -
Supplement
(324 KB) - BibTeX
- EndNote
Abdellah Khouz et al.
Status: final response (author comments only)
-
RC1: 'Comment on nhess-2022-76', Anonymous Referee #1, 15 Apr 2022
Dear Editor,
The paper entitled “Landslide susceptibility assessment in rocky coast subsystem of Essaouira coastal area – Morocco” by Abdellah Khouz and co-authors is a timely study that quantitatively assesses landslide susceptibility on a coastal area, through a bivariate statistical method - Information Value. The landslide susceptibility was evaluated in Essaouira coastal area, considering both a pixel based-, and an elementary terrain units’ approach. The authors used a set of 22 conditioning factors and a landslide inventory record with 588 events, of which 70% were used for training and 30% for validation purposes. The accuracy and prediction capacity of the models were assessed by ROC curves and the Area Under the Curve.
General comments:
The paper presents a good database, based on diverse sources, and a considerable amount of work has been done, including field work, which is valuable. Although the method used in this work (Information Value) is not novel, the work is innovative in the area, combining a wide range of data.
The strength of the paper is the analysis of the landslide susceptibility on a coastal area, integrating an interesting number of conditioning factors and a large landslides database, allowing to perform a comparative analysis based on elementary terrain units and pixel-based approaches. The work is original and valuable. The approach consists of a good statistical analysis, which may be of interest for the scientific community.
However, some aspects deserve to be more/better discussed in the text. Additionally, the paper needs a revision of the English. You should be more careful with phrase construction. Sometimes sentences start with lowercase. Sometimes, the sentences end with a comma and starts a new sentence after the comma. A deep revision by the senior co-authors certainly could help to improve. Some figures are not of good quality, i.e., are not readable and this has to be considered.Thus, the paper deserves publication in Natural Hazards and Earth System Sciences, after considering the following issues.
Specific comments:a) Minor comment: in line 43 you refer to “pressure”. What kind of pressure? Urban pressure?
b) Regarding the landslide inventory and the training and validation groups. How were selected the two groups? Randomly? Selected according to any specific criteria? Please state in the methods.
c) Why did you use 70% of the inventory and 30% for the validation? Why not 50% for each? You should state in the methods section why did you use these percentages?
d) The most frequent phenomena are rockfall (149 events). However, translational and rotational slides occupy 85% of the unstable area, mainly occurring in the southern section, where they have higher weight on landslide susceptibility. Is there a higher landslide susceptibility in the southern section because of a higher number of these landslide events or is it because of the area of each landslide, thus performing higher susceptibility?
e) L. 284-286: You state that slope angle does not have the same importance for all types of landslides in your study area. (This would be better stated and discussed in results and discussion section).
Can you state why? Is it only because different types of landslides require different factors and different weight of each factor? Or is it because in your study area, are there other important factors also contributing for slope instability?f) You also mention that slope angle is one of the most influent factors (lines 481-482). However, table S1 shows that some types of landslides do not fit in this assumption.
What does contribute for the low IV score for the highest slope classes (> 35º) for models 10-13, and 15? In the case of rock topple, slopes >15º have negative scores. This should be discussed.g) In table 5 you have the same percentage of landslide susceptibility for translational and shallow translational landslides. What is the explanation? Is it an error or are you assuming all translational landslides as shallow translational?
h) You state that the precipitation is not a “decisive conditioning factor” (L. 588). From a pure statistical point of view, it is true. The reason why you don’t see great differences may be because you are using annual average values of precipitation. However, in drier areas, rainfall intensity may be more important than the annual average amount. Since precipitation is an important triggering factor, it would be expected an increase of landslide events during the rainy season. Didn’t you find any variation? Considering precipitation is not a permanent factor as the others, is it proper to treat it as a conditioning factor based on its (low) annual average?
i) In L. 549-550, you found that eliminating precipitation and TWI of your analysis you get better results (Fig. 11). This is statistically valid. However, considering that this is a dry climate, the effect of humidity and precipitation, when they occur, may be very important for slope instability, but your analysis cannot identify it. It would be important to discuss the limitations of this statistical analysis.
j) Given your results and considering the two approaches (Pixel-based and ETU) used in this work, which is the most suitable one for representing the landslide susceptibility in the area?
Since ETU are defined based on the morphometry of the area, there is a more “guided” analysis in this approach, comparing with pixel-based that is more “random”, some differences between both modelling should be expected.
However, in L. 594, you conclude that both ETU and pixel analyses have similar behaviour. What is causing or contributing to this similarity? The reasons for these similarities and the differences between both approaches should be better discussed in the text.k) One limitation of this bivariate statistical method is that it does not consider possible correlations between variables. This limitation and its impact on possible high scores should be discussed.
l) Another, and very important, drawback of this method is that it uses a part of the landslide inventory to model the susceptibility. Considering this, the validation is not done with the whole inventory, and the landslide dimensions may bias the IV scores. It would be important to discuss this in the text. How do these drawbacks may influence the final results?
FIGURES:
Figure 3: You jumped from C to E and forgot D.Figure 6: This figure is very low. Please make the fond size readable. The legend and the vertical scale are not readable.
Technical issues:
Some issues were found, especially in phrase construction, and the connection between sentences is not always clear. Sometimes it is difficult for the reader to understand your ideas. Please revise your writing.
A brief list of issues below:L. 125-129: Big paragraph, with several sentences separated by semi-colon. Consider rephrasing in shorter and clear sentences.
L. 130-135: you have two sentences starting after a comma, instead a full stop:
- L. 131: “(…) Dufaud et al. 1966, Its existence…”
- L. 132: “(…) Weisrock 1980), It consists (...)” - substitute "," by "."L. 148: you could delete "of the replay"
L. 195-199: “According to the rainfall data, which were made available…”. You stopped this sentence without finishing your idea. Then in L. 196 you end a sentence with a comma and then start a new sentence.
Be very careful with this. You have many examples like this.
This becomes confusing.L. 202-203: Please, show the maximum and minimum values (mm) of precipitation
L. 223: In the end of the line “… (Mennani, 2001), It…” – again you end a sentence with a comma.
L. 225-231: A big sentence that could be divided in two, starting in line 228 “For this reason…”.
L. 284: You start again a sentence after a comma. “… (Epifânio, et al. 2013), Slope angle…”.
L. 292: “… nouthern part…”. Do you mean “northern section”, “northern area”?
You often use in the text the terms northern and southern part. Consider using “section” or “area”…instead of “part”. It is more correct from a geographical point of view.L. 342: Consider substituting "than" by "then".
L. 412-413: Like it is written, it does not make much sense.
Do you mean "... Calcareous crusting and Essaouira sandstone-calcarenite are the two lithological formations most found in the majority of ETU...”?L. 435: that’s – please, avoid word contractions.
L. 452-453: Please revise the sentence. As it is does not make much sense.
Do you mean this? - "These considerably affect the mechanical processes that lead to slope failure and to the subsequent post-failure movements, especially where there are marls or clays."L. 496-497: Or "as they usually happened next..." or "as they are usually next..."
L. 540-545: There is a huge number of commas.
"Tab. 6 shows..." should start as a new sentence.
"...AUC values. Model 1..." Model 1 should also start as a new sentence.L. 575-578: Please revise the text.
L. 579: Substitute “it’s” by “its”.
-
RC2: 'Comment on nhess-2022-76', Anonymous Referee #2, 19 Apr 2022
Dear editor
The paper by Abdellah Khouz and co-authors, “Landslide susceptibility assessment in rocky coast subsystem of Essaouira coastal area – Morocco” focuses on the quantitative assessment and mapping of landslide susceptibility in a coastal area, using the bivariate Information Value statistical method. Were used pixel based and elementary terrain unit approaches, a 588 landslides inventory, a set of 22 landslide conditioning factors. The inventory was partitioned in a 70% training set and a 30% validation set, and modeling results were validated with ROC curves and the corresponding AUCs.
In spite of several issues, the paper contains a certain degree of innovation and deserves to be published after a major revision process.
General comments:
The theme of sea cliff and coastal areas landslide susceptibility and hazard assessment has received very little attention by the research community, despite its relevance in terms of hazard prevention and sustainable land use planning and management. In consequence, the submission of the paper is very welcomed and contains some innovation and interesting results.
The study made included a considerable amount of work to gather the required information, namely for landslide inventory and acquisition of field-based data, on the large number of landslide conditioning factors considered, and also on the extensive modeling made.
However, the paper suffers from various issues which, without detracting its interest and merits, will require substantial revision by the authors.
The manuscript needs a careful English revision for spelling and phrase construction.
The manuscript references suffer from a heavy bias toward in land landslide studies (more than 40 references) while those on sea cliffs are only 13. This problem must be solved because many very important sea cliff studies are missing and are relevant for the topic of the paper. Some suggestions are made along the comments.
There is a lack of clarity on the study area: it is referred in several parts of the paper that study focus on the coastal area, but in Line 107 is stated that the focus is on landslides at the sea cliffs. Later, in lines 113 and 114 and Fig. 1, the coastal subsystems include sandy coast, rocky coast, and anthropic coast. The rocky coast corresponds to sea cliffs or includes sections of low height rocky coast, with no well-defined cliff. It is important to clarify and to use uniform designations along the paper.
The rainfall data would be better expressed with the inclusion of graphs instead of descriptive and incomplete data.
The aerial photographs and satellite images area coverage for the landslides inventory construction should be included in table 1. This is important because any inventory is incomplete by its own nature and depends heavily on the database available. It is also important to clarify that the inventory is of the historical type, with no past date of occurrence limits, and it is also useful to point out its limitations.
In the modeling, 70% of the inventory were used as training set and the other 30% as validation set – explain why those values were used. The validation process is also a matter of debate in the discussion part of the paper.
In the various model results classification why were used the IV values instead of a classification based on the ROC curve results, with limits of unstable areas of, for example 50%, 65%, 80%, 95% of the correctely predicted unstable terrain units.
Although involves some additional work, it would be useful to have the AUC of the ROC curve of each individual factor, at least for all types of movements, to enable the assessment of the more important susceptibility predisposing factors, which could be improved in further studies, in order to obtain better models.
In the paper is missing a discussion of the results obtained and a comparison with other studies of the same type carried out in other coastal cliffs.
One other aspect to address is the validation method: using one part of the inventory to build the model and the other part for validation is a statistically sound method of validation, but it only indicates that the landslide inventory is robust enough and that the inventory partitions are representative samples of the total inventory and have similar relations with the landslides predisposing factors. However, as showed in Queiroz and Marques (2019) a temporal partition of a cliff failure inventory (1947-1980 and 1980-2012) led to very high success ROC AUC values, but to poor prediction rates, which raises fundamental doubts for the true prediction of future evolution behavior of sea cliffs based on its past evolution (as in Guilham et al., 2018). It is the reviewer opinion that this matter should also be subject of discussion and a subject for future work.
Detail comments:
L 42 – Classical references as Sunamura (1992) and Trenhaile (1987) are much more meaningful in this context. Other suggestion: Hampton & Griggs (2004).
L 44 – In the reference it is suggested to ad “e.g. Marques, 2009” but also other relevant references as Teixeira (2006, 2014), Moore and Davis (2015), Gilham et al., (2018) among others.
L 60-62 – The landslide predisposing factors which have been used in published studies are listed along specific cliff factors as the cliff toe protections. This requires some separation, due to the specific context of sea cliffs and also because it was found that the factor is relevant in these studies (Marques et al,. 2011, 2013; Marques, 2018, Guilham et al., 2018, Letortu et al., 2019, Queiroz and Marques, 2019).
L 71 – For sea cliff susceptibility, the terrain unit discussion and one solution were presented in Marques et al. (2011, 2013), which were published before Epifâneo et al. (2014).
L 95 – The phrase seems out of context.
L 121-129 – Rewrite and clarify the setting of the study area and be more specific on the geological structure et relations with geomorphology.
L 130- 166 – The text chaotic and requires clarification, a deep reformulation, and the use of shorter periods.
L 143 – extensional instead of distensional; NNE-SW ??? correct.
L 144 – What is the second direction – only one was indicated above.
L 216 – 231 – The hydrogeological information is relevant for the sea cliffs evolution?
L 246 - 247 – What was the threshold percentage of unstable area in each terrain unit to be considered unstble.
L 295 – 296 – phrase seems incomplete.
L 312 – The references deserve improvement: Lee and Pradhan (2007), Shahabi et al. (2014), Wang et al. (2016) studied sea cliffs? Proper references include Marques et al, 2011, 2013; Epifâneo et al., 2013, 2014.
Table 2 – Replace “limstone” by limestone.
Figure 5 – Replace the pie plots by bar or column plots.
L 417 – What is “limestone barre”? Clarify.
L 528 – Clarify “the respective average of the unstable area, are located more to the souths of study area.”
Suggested References:
Gerivani, H.; Stephenson, W.; Afarin, M. (2020). Sea cliff instability hazard assessment for coastal management in Chabahar, Iran. Journal of Coastal Conservation, 24, Article number 5, 17 p.
Gilham, J., Barlow, J., Moore,R. (2018). Marine control over negative power law scaling of mass wasting events in chalk sea cliffs with implications for future recession under the UKCP09 medium emission scenario. Earth Surf. Process. Landforms 43, pp. 2136-2146 (DOI: 10.1002/esp.4379).
Hampton, M.A., Griggs, G.B. (2004). Formation, Evolution, and Stability of Coastal Cliffs— Status and Trends. U.S. Geological Survey Professional Paper 1693.
Letortu, P.; Costa, S.; Maquaire, O.; Davidson, R. (2019). Marine and subaerial controls of coastal chalk cliff erosion in Normandy (France) based on a 7-year laser scanner monitoring. Geomorphology, Vol. 335, pp. 76-91.
Marques, F. (2018). Regional scale sea cliff hazard assessment at Sintra and Cascais counties, western coast of Portugal. Geosciences 8, 3: 80 (doi:10.3390/geosciences8030080).
Marques, F.; Matildes, R.; Redweik, P. (2013). Sea cliff instability susceptibility at regional scale: a statistically based assessment in the southern Algarve, Portugal. Natural Hazards Earth System Science, 13, pp. 3185-3203.
Moore, R.; Davis, G. (2015). Cliff instability and erosion management in England and Wales. Journal of Coastal Conservation, Vol. 19, No. 6, Special Issue: Conservation, management and restoration of coastal cliffs, pp. 771-784.
Queiroz, S., Marques, F.M.S.F. (2019). Sea cliff instability susceptibility considering nearby human occupation and predictive capacity assessment. Engineering Geology, 253: p.75–93 (doi:10.1016/j.enggeo.2019.03.009).
Sunamura, T. (1992). Geomorphology of Rocky Coasts. Wiley, New York, 302 p.
Teixeira, S. (2014). Coastal hazards from slope mass movements - Analysis and management approach on the Barlavento coast, Algarve, Portugal. Ocean & Coastal Management, Vol. 102, Part A, pp. 285-293.
Trenhaile, A.S. (1987). Geomorphology of Rock Coasts. Clarendon Press: Oxford, UK, 384 p.
Abdellah Khouz et al.
Abdellah Khouz et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
220 | 55 | 12 | 287 | 24 | 6 | 5 |
- HTML: 220
- PDF: 55
- XML: 12
- Total: 287
- Supplement: 24
- BibTeX: 6
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1