the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Jun 2025
Flood and landslide risk mapping based on a multi-criteria analysis (MCA) in Greater Abidjan (Côte d’Ivoire)
Abstract. This study presents a multi-hazard risk assessment of flood and landslide hazards in the Greater Abidjan metropolitan area of Côte d’Ivoire, aimed at enhancing disaster risk reduction strategies in data-poor contexts. Using a semi-quantitative approach within a multi-criteria decision-making framework, specifically the Analytic Hierarchy Process (AHP), we assess both hazard and vulnerability factors contributing to flood and landslide risks, incorporating climatic, environmental, and social aspects. An innovative validation method is introduced, leveraging a multi-source database of past disaster events in the region, combining information from well-established disaster loss databases and results from field surveys, thereby enhancing the robustness and reliability of the results. The findings identify risk-prone areas within Greater Abidjan and provide actionable insights for improving disaster risk management. This approach, which builds on previous research by incorporating both flood and landslide risks, advances a multi-hazard perspective and contributes to a deeper understanding of hazard dynamics in Greater Abidjan. It also offers a model that can be applied to similar urban settings across sub-Saharan Africa.
- Preprint
(2194 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2925', Anonymous Referee #1, 05 Nov 2025
-
AC1: 'Reply on RC1', Habal Kassoum Traore, 20 Jan 2026
We sincerely thank the reviewer for their thorough and constructive comments, which significantly contributed to improving the clarity and robustness of the manuscript. We have carefully addressed all the remarks and revised the manuscript accordingly, as detailed in our point-by-point response below.
Line 110: The paper structure is clearly outlined, though the "Sections" should be labelled according to standard scientific paper structure (e.g., Introduction, Methodology, Results, Discussion)
The article has undergone a structural reorganization. We have adopted the recommended heading hierarchy and, in particular, decoupled the « Results » and « Discussion » into separate sections.
Line 165: There is a mismatch between the locations (4 municipalities listed) and the densities (5 values provided). Please correct this discrepancy.
We have adopted the recommended heading hierarchy and, in particular, decoupled the « Results » and « Discussion » into separate sections. On line 165, population densities are derived from the official statistics provided by the National Statistics Office (INS-CI, 2021). In contrast, the densities presented in Table 5 represent continuous numerical values produced by Meta as part of the « Data for Good » program. These values provide an estimate of the number of people per pixel (≈30 m), generated through deep learning models that integrate satellite imagery, building detection, and census data. In this study, utilizing this continuous variable was essential to meet the specific modeling requirements of the Analytic Hierarchy Process (AHP).
Tables 1, 3, 5: It is unclear on which grounds the classification bins are established. Please add explanation of how thresholds were determined (e.g., natural breaks, quantile, literature-based).
The approach used to define classes in Tables 1, 3 and 5 differs according to the nature of the indicators and is now explicitly described in the revised manuscript.
For Land Use / Land Cover, classes were not derived from statistical thresholding. Instead, land-cover information was extracted from the CORINE Land Cover (CLC) database at Level 1, which consists of five predefined thematic classes (Artificial surfaces, Agricultural areas, Forest and semi-natural areas, Wetlands, and Water bodies). These classes are categorical by definition and therefore do not rely on numerical thresholds.
In the present study, CLC Level 1 classes were used as a conceptual proxy for surface permeability, anthropogenic disturbance and land-cover control on hydrological response and slope stability. Each class was subsequently assigned a relative susceptibility score within the AHP framework, based on published evidence and expert judgement regarding its expected influence on runoff generation, infiltration capacity and slope instability in urban tropical environments.
For continuous environmental indicators (e.g. slope, topographic wetness index and distance-based variables), class boundaries were defined using natural breaks (Jenks) to capture the inherent structure of the data distribution and spatial heterogeneity at the metropolitan scale. For population density (Table 5), a quantile-based classification was applied to ensure a balanced representation of exposure levels across the study area.
These classification choices and their rationale are now clearly documented in the Methodology section, and the captions and legends of Tables 1, 3 and 5 have been revised accordingly to improve transparency and reproducibility.Lines 175-180: The AHP weight determination process needs more detail. Please include: expert selection criteria, number of experts consulted, and the consensus-building process.
The relative weights were derived exclusively from a review of recent peer-reviewed studies applying the Analytic Hierarchy Process (AHP) to similar hazards (floods and landslides) in comparable geographical and urban contexts in West Africa and other tropical environments (e.g., Waseem et al., (2023); Shah and Shah, (2023); Beven and Kirkby, (1979); Montgomery and Dietrich, (1994); Dewan, (2013); Ouattara et al., (2021). The consensus was therefore not established through expert deliberation, but through the convergence of weighting schemes repeatedly reported and scientifically validated in the literature. The most recurrent weight values were extracted in order to ensure objectivity, methodological transparency, and reproducibility of the model.
As specified above, institutional experts from the National Office of Civil Protection (ONPC) and the Ministry of Hydraulics and Sanitation (MINHAS) did not participate in the weight attribution process, in order to avoid subjective bias potentially linked to their operational responsibilities. The experts were selected on the basis of their professional experience and detailed knowledge of risk-prone areas, with one expert representing each administrative department covered by the study. Their contribution was mobilized at a clearly distinct and critical stage of the analysis, namely the identification and spatial localization of historical flood and landslide events based on institutional records and field-based knowledge.
This strict separation between literature-based weight determination and expert-based event mapping allows expert knowledge to be used as an independent validation source. It strengthens the credibility of the results by ensuring that the model outputs, derived from reproducible scientific evidence, are consistent with observed hazard occurrences on the ground.Lines 235 & 258-259: The CR values appear identical (1.3%) for both flood and landslide hazards. Please double-check these calculations and explain if they are indeed identical.
After re-checking the AHP calculations, we identified a reporting error in the landslide CR value. The correct CR values are 1.3% for floods, 0.6% for landslides, and 2.1% for social vulnerability, all well below the 10% threshold.
The manuscript has been corrected accordingly.Detail of verification calculations:
- Flood hazard (Table 2)
Number of criteria (n): 5.
Random Index (RI) for n=5: 1.12.
Maximum eigenvalue (λmax) calculated from the matrix: ≈5.058.
Coherence Index (CI): CI=(λmax−n)/(n−1)=(5.058−5)/4=0.0145.
Coherence Ratio (CR): CR=CI/RI=0.0145/1.12=1.29% (rounded to 1.3% in the text).
Status: Compliant. - Landslide hazard (Table 4) - Correction
Number of criteria: (n): 3
Random index: (RI) for n=3: 0.58
Maximum eigenvalue: (λmax) calculated from the matrix in Table 4: ≈3.007.
Coherence Index (CI): CI= (3.007−3)/(3−1) = 0.0035.
Coherence Ratio (CR): CR=0.0035/0.58=0.60%
Status: To be corrected. The value of 1.3% on line 262 should be replaced with 0.6%. - Social vulnerability (Table 6)
Number of criteria (n) : 3.
Random Index (RI) for n=3 : 0,58.
Maximum eigenvalue (λmax) : ≈3,0246.
Consistency Index (CI) : CI=(3,0246−3)/2=0,0123.
Consistency Ratio (CR) : CR=0,0123/0,58=2,12% (consistent with the 2% announced on line 390).
Status: Compliant.
Line 339: Add a paragraph acknowledging missing vulnerability indicators (income, housing quality, infrastructure, health access).
A new paragraph will be added to explicitly acknowledge the absence of certain key indicators of vulnerability. These shortcomings will be discussed as constraints mainly related to the availability and accessibility of data. This discussion will be based in particular on the work of Dewan (2013), who presents, on page 141, a non-exhaustive list of other vulnerability factors that can be mobilized in the context of the PAA.
Lines 452-510: The results section severely lacks quantitative visualization. Maps alone are insufficient for comprehensive risk communication. Please add: (1) statistical plots showing risk distributions and population exposure, (2) municipal comparison charts, (3) quantitative summary table showing area (km²) and population at each risk level for both hazards.
Using the existing datasets of this study, including the flood and landslide risk maps, municipal boundaries, and population density data, the Results section will be quantitatively strengthened.
Specifically, zonal statistics will be computed by crossing risk classes with population density and administrative units in order to:
(1) produce statistical graphs describing the spatial distribution of risk levels and the associated population exposure;
(2) generate comparative graphs between municipalities based on surface area and exposed population;
(3) compile a summary table reporting, for each hazard and each risk level, the corresponding surface area (km²) and estimated population.
Line 482: "landslides (illustrated by red dots)" - the red dots are not visible in Figure 8. Please make them clearly visible or remove this reference.
This is indeed an error in the text. In Figure 8, landslide occurrences are not represented by red dots but by blue dots. This inconsistency has been corrected in the revised version of the manuscriptLines 515-520: Validation discussion is too brief. Please add ROC curves and confusion matrices to discuss agreement between data and model.
The validation section was strengthened through the integration of ROC curves and confusion matrices, allowing a more accurate assessment of the level of agreement between the susceptibility models and the available observational data, including field surveys as well as the EM-DAT and CATNAT databases.
Validation was based on independent observations of documented events, used as (true positives), while absence points (true negatives) were randomly generated within the study area, excluding zones affected by the observed events. The same methodological approach was applied to ground movement hazards in order to ensure consistency and comparability across the different hazards considered. The ARCSDM tool, integrated within ArcGIS Pro, was used to calculate performance indicators and generate validation graphs, thereby enhancing the robustness and transparency of the results.Lines 524-566: Please group the limitations by themes (data limitations, methodological constraints, scope restrictions) and prioritize by impact on results.
This suggestion has been fully integrated. Limits are now organized into three main categories:
- Data limitations,
- Methodological constraints,
- Boundaries related to the scope and conceptual framework.
They are discussed in terms of their estimated impact on the results.Lines 544-546: Expand discussion on how the identified limitations specifically affect your results and their reliability.
The discussion was further developed to clarify how the identified limitations directly influence the results obtained and their reliability. In particular, the limited number of criteria incorporated into the AHP approach, compared with some recent studies, may constrain the model’s ability to capture the full range of physical and anthropogenic processes controlling flood susceptibility.
For example, although precipitation is a major trigger of flooding, this criterion was not included in the analysis. This choice is explained by the mismatch between the spatial resolution of the available rainfall data and the pixel size used in this study, which could have introduced scale inconsistencies and potentially biased the AHP weighting process. The use of regular-grid satellite products (such as CHIRPS, TRMM, or GPM) was considered, but their coarser spatial resolution could have altered the relative hierarchy of criteria and affected the robustness of the resulting susceptibility map. In addition, as the objective of this study was to characterize the intrinsic susceptibility of the territory independently of event-based rainfall conditions, the integration of precipitation would have required a dynamic approach that lay beyond the methodological framework adopted.
Similarly, several key vulnerability indicators frequently used in the literature (such as characteristics of the built environment, housing quality, income levels, or access to infrastructure and essential services) could not be integrated. This limitation is mainly related to constraints in the availability and spatial compatibility of the data derived from the Meta database used in this study, which does not allow a fine and homogeneous representation of these dimensions at the selected scale of analysis. The absence of these indicators may contribute to an under-representation of certain forms of social and economic vulnerability and therefore represents an avenue for improvement in future work based on more detailed databases or complementary field surveys.All Figures: The figures' quality, size, and font need to be updated for better readability. Ensure minimum 300 DPI resolution and legible text (≥10pt font).
After taking into account the reviewers’ recommendations, we plan to rewrite the entire manuscript using the journal’s official LaTeX template, which will ensure greater layout homogeneity and guarantee a minimum resolution of 300 DPI for all figures, particularly the maps. This will also facilitate better control of font size and legibility, with text consistently set at 10 pt or larger, in order to improve the clarity and readability of the figures in the revised version of the article.
Références :
Beven, K. J., & Kirkby, M. J. (1979). A physically based, variable contributing area model of basin hydrology. Hydrological Sciences Bulletin, 24(1), 43–69. https://doi.org/10.1080/02626667909491834
Montgomery, D. R., & Dietrich, W. E. (1994). A physically based model for the topographic control on shallow landsliding. Water Resources Research, 30(4), 1153–1171. https://doi.org/10.1029/93WR02979
Dewan, A. M. (2013). Floods in a Megacity: Geospatial Techniques in Assessing Hazards, Risk and Vulnerability. Springer, Dordrecht. https://doi.org/10.1007/978-94-007-5875-9
Ouattara, K., Kouamé, K. F., Jourda, J. P., Biémi, J., & Saley, M. B. (2021).
Flood hazard mapping in urban areas using multi-criteria analysis: A case study of Abidjan (Côte d’Ivoire). Natural Hazards, 105, 2537–2563. https://doi.org/10.1007/s11069-020-04428-4
Shah, A. A., & Shah, S. A. R. (2023). GIS-based landslide susceptibility mapping using AHP and machine learning techniques: A comparative study. Environmental Earth Sciences, 82, 154. https://doi.org/10.1007/s12665-023-10812-3
Waseem, M., Khan, A. N., Rahman, G., & Ahmad, S. (2023). Flood susceptibility mapping using the Analytic Hierarchy Process (AHP) and GIS techniques. Arabian Journal of Geosciences, 16, 287. https://doi.org/10.1007/s12517-023-11456-1
Citation: https://doi.org/10.5194/egusphere-2025-2925-AC1 - Flood hazard (Table 2)
-
AC1: 'Reply on RC1', Habal Kassoum Traore, 20 Jan 2026
-
RC2: 'Comment on egusphere-2025-2925', Anonymous Referee #2, 07 Nov 2025
The research by Habal Kassoum Traore and co-authors assesses the risk associated with landslide and flood occurrences in the metropolitan area of Abidjan, Ivory Coast. To achieve this, the authors use a semi-quantitative approach that considers both susceptibility and vulnerability components. The data and information used in the models come from available datasets. The analysis is completed with field surveys. The author highlight that their findings offer insights relevant to disaster risk reduction, especially in data-poor context.
Overall, the research fits well within the scope of NHESS. It is encouraging to see such work carried out in an under-researched type of environment – emphasizing both the societal importance of such a work and also the challenges posed by limited data availability. In this regard, , I fully agree with the authors that using simple, reproducible, and data-efficient approaches represents a relevant and practical strategy.
Nevertheless, there are some important issues in this study that must be stressed.
- The main one to me: the lack of insight into process understanding. What exactly are we studying?
There is too little information with respect to what is being studied/discussed:
What type(s) of flood is considered? Pluvial, fluvial, both? Are we looking at pluvial flooding due to intense (but highlight localised) rainfall events of short duration (like thunderstorm), or are we looking at fluvial flooding associated with few day rainfalls? Different types of floods = different types of drivers and/or weighting in the AHP. What about seasonality? What would be the role of tides in these floods?
In line 311, reference is made about a study on shallow landslides. This is the only part of the manuscript where reference is made about landslide characteristics (although not for this specific study). Overall we clearly miss information about the types of processes that are being studied. Are they recent; old; of natural origin, human induced, triggered by rainfall, favoured by weathering, etc?
Considering the variety of the landscape conditions, and also the climate triggers, it would be normal to have different types of processes, or at least similar processes, but of different ages. In addition to these assumptions that we could have made a few months ago, there is a publication on landslides in Abidjan that has just been released. Although Gnagne et al. (2025)’s work comes with caveats, it still shows some interesting points: different types of landslides (slides,, avalanches, shallow and deep-seated slides), landslides of different ages (different land use/covers than today’s?).
These points with the different types of landslides are that land uses (and dynamics) will not have the same impacts on hazard, whether it is, for example, a shallow or a deep-seated slope failure (Sidle and Bogaard, 2016). Some examples of the role of land use (and land use changes – road construction; deforestation) and urbanization on landslide incidence have been published for similar tropical Africa environments (e.g. Dille et al., 2022; Maki Mateso et al., 2023); these could help to authors to better design their research.
Indeed, beyond the problem of landslide types, the selection of predictor variable, as well as the decisions regarding their categorization and the assignment of weights to these classes; often raises questions. For example, the slope gradient classes do not seems to be based on a geomorphologic consideration (see threshold hillslope concept; e.g. Bennett et al., 2016; Depicker et al., 2021). Roads are known to have an important impact on landsliding (e.g. Tanyas et al., 2022); however this variable, while being considered is the flood analysis, is not considered in the landslide one.
In addition to landslides, as illustrated by Gnagne et al. (2025), gully erosion (large gullies) is also present in the city. These gullies are certainly, in their large majority, induced by urbanisation (Ilombe Mawe et al., 2025). Gullies have influence on landslides and flood (connectivity). Such point would deserve being mentioned (and maybe discussed?)
- The overall modelling approach is questionable.
To study hazard risk, we expect susceptibility and hazard assessment, exposure analysis and vulnerability analysis. In practice we know that all these components of risk are not always easy to assess, especially for one specific region in data-scarce context. In this study, only susceptibility and vulnerability assessments are made. The temporal aspects (hazard) are not considered, nor the exposure one. This is therefore a problematic aspect in the manner the study is being presented and sold.
Another point is that, even for the susceptibility assessments, we do not know what is actually done For example, for landsliding, are the source of landslides being considered or their runout? Depending on those, they may be huge differences in exposure (e.g. Schmitt et al., 2025). In addition, considering the type of speed of the processes (rapid or less rapid landslides, flash floods, etc.), the impacts and hence the overall risk is not the same.
The vulnerability seems to be assumed to be the same for landslide and flood hazards. This cannot be the case considering the different impacts and also the frequency of the processes.
The data used as predictor variables in both flood and landside assessments, as well as for the vulnerability, are not questioned much about their reliability.
EM-DAT data are being used for validation of the models. However, these data come with significant caveats as they are highly biased towards impactful events and also towards regions provided with better communication means and greater wealth. The same hazard with the same intensity and magnitude is likely to receive less attention if it occurs in a remote location or a low-income neighbourhood. Further discussion of these reporting biases and dataset limitations can be found in Stein et al. (2024) and Delforge et al. (2025).
Fiefd survey is not clearly explained.
Hazard zonation; the meaning of the classes?
Landslide assessment provides results for flat areas.
- Muti-hazard analysis
Line 49: “...and lay the groundwork for effective multi-hazard disaster risk management and mitigation strategies.” Here, and in other places in the text, the authors put a focus on multi-hazard assessment. If I look at the definition of UNDRR (https://www.undrr.org/terminology/hazard) about multi-hazard: “Multi-hazard means (1) the selection of multiple major hazards that the country faces, and (2) the specific contexts where hazardous events may occur simultaneously, cascadingly or cumulatively over time, and taking into account the potential interrelated effects”. In this work, I do not see the real contribution with respect to the point (2). The focus of the study must be better defined with respect to what it contributes to the multi-hazard literature.
- State of the art and lack of discussion with respect to that
The state of the art is rather limited in many aspects, including urban contexts of landslide and flood risk, land transformation, population exposure, and multi-hazard interactions. For instance, when considering landslides and floods jointly, only a few studies have addressed this dual perspective. I recommend that the authors consult works such as Ferrer et al. (2024) and Idukunda et al. (2025), and discuss how the context and findings of these studies relate to the dual nature of both exposure types.
Other comments
- Introduction: The focus is strongly put in the case study of Abidjan, while the state of the art about mulit-hazard and risk assessment is not that much developed. One of the main justifications for the work is its aim to complement existing hazard studies conducted in the city. Although this is a valid rationale—particularly given the importance of improving knowledge in such a context—the study lacks a clear anchor point for a broader, international audience. In other words, why should readers unfamiliar with the study area find this work relevant or compelling?
- Line40: “frequency and severity of flooding and landslides in the region have escalated in recent years, highlighting an urgent need to develop more effective multi-hazard risk management strategies”. We would welcome the inclusion of references to support these statements. Studies capable of disentangling such trends are relatively rare, particularly in data-scarce environments. If such studies exist, they should be cited. Moreover, their comprehensive datasets would be highly valuable for both the design and validation of the present research.
- Line 90: “our study offers a more realistic andpolicy-relevant understanding of hazard exposure and vulnerability across the metropolitan continuum of Abidjan”. As stressed here in the introduction, the relevance of the work in DRR policy is emphasized. beyond the production of maps within a basic zonation framework, there appears to be limited consideration of how the results directly contribute to DRR applications. In this sense, the connection to the DRR context seems somewhat overstated.
- Line 95: “To overcome this gap, our approach introduces an innovative validation component: a geo-referenced database of observed past events, compiled from multiple sources including national reports, humanitarian data platforms, and remote sensing-based event detection”. In fact, many studies are based on similar inventories (for instance, in another data-scarce African context—see Nasabimana et al., 2023). From a methodological perspective, there is no real innovation here. Regarding model validation, comparing model outputs with real-world data is almost a compulsory step.
- Too general in many sections (for example section 2.2), bringing not relevant and accurate information for the study
In conclusion, while the topic holds significant potential, the study does not yet appear sufficiently mature. I hope that my comments will assist the authors in further developing and strengthening their work.
References:
Bennett, G.L., Miller, S.R., Roering, J.J. and Schmidt, D.A., 2016. Landslides, threshold slopes, and the survival of relict terrain in the wake of the Mendocino Triple Junction. Geology, 44(5), pp.363-366.
Delforge, D., Wathelet, V., Below, R., Sofia, C.L., Tonnelier, M., van Loenhout, J.A. and Speybroeck, N., 2025. EM-DAT: the emergency events database. International Journal of Disaster Risk Reduction, p.105509.
Depicker, A., Govers, G., Jacobs, L., Campforts, B., Uwihirwe, J. and Dewitte, O., 2021. Interactions between deforestation, landscape rejuvenation, and shallow landslides in the North Tanganyika–Kivu rift region, Africa. Earth Surface Dynamics, 9(3), pp.445-462.
Dille, A., Dewitte, O., Handwerger, A.L., d’Oreye, N., Derauw, D., Ganza Bamulezi, G., Ilombe Mawe, G., Michellier, C., Moeyersons, J., Monsieurs, E. and Mugaruka Bibentyo, T., 2022. Acceleration of a large deep-seated tropical landslide due to urbanization feedbacks. Nature Geoscience, 15(12), pp.1048-1055.
Ferrer, J.V., Samprogna Mohor, G., Dewitte, O., Pánek, T., Reyes‐Carmona, C., Handwerger, A.L., Hürlimann, M., Köhler, L., Teshebaeva, K., Thieken, A.H. and Tsou, C.Y., 2024. Human settlement pressure drives slow‐moving landslide exposure. Earth's Future, 12(9), p.e2024EF004830.
Gnagne, F.L., Schmitz, S., Kouadio, H.B., Hubert-Ferrari, A., Biémi, J. and Demoulin, A., 2025. Predicting Landslide Susceptibility Using Cost Function in Low-Relief Areas: A Case Study of the Urban Municipality of Attecoube (Abidjan, Ivory Coast). Earth, 6(3), p.84.
Idukunda, C., Michellier, C., De Longueville, F., Twarabamenye, E. and Henry, S., 2025. Assessing community vulnerability to landslide and flood in northwestern Rwanda. International Journal of Disaster Risk Reduction, 123, p.105329.
Mawe, G.I., Landu, E.L., Dujardin, E., Imwangana, F.M., Bielders, C., Hubert, A., Michellier, C., Nzolang, C., Poesen, J., Dewitte, O. and Vanmaercke, M., 2025. Mapping urban gullies in the Democratic Republic of the Congo. Nature, 644(8078), pp.952-959.
Nsabimana, J., Henry, S., Ndayisenga, A., Kubwimana, D., Dewitte, O., Kervyn, F. and Michellier, C., 2023. Geo-hydrological hazard impacts, vulnerability and perception in Bujumbura (Burundi): a high-resolution field-based assessment in a sprawling city. Land, 12(10), p.1876.
Ozturk, U., Bozzolan, E., Holcombe, E.A., Shukla, R., Pianosi, F. and Wagener, T., 2022. How climate change and unplanned urban sprawl bring more landslides. Nature, 608(7922), pp.262-265.
Schmitt, R.J.P., Bhandari, S., Vogl, A. and Marc, O., 2025. Leveraging hillslope connectivity for improved large-scale assessments of landslide risk. EGUsphere, 2025, pp.1-34.
Sidle, R.C. and Bogaard, T.A., 2016. Dynamic earth system and ecological controls of rainfall-initiated landslides. Earth-science reviews, 159, pp.275-291.
Stein, L., Mukkavilli, S.K., Pfitzmann, B.M., Staar, P.W., Ozturk, U., Berrospi, C., Brunschwiler, T. and Wagener, T., 2024. Wealth over Woe: Global biases in hydro‐hazard research. Earth's Future, 12(10), p.e2024EF004590.
Tanyaş, H., Görüm, T., Kirschbaum, D. and Lombardo, L., 2022. Could road constructions be more hazardous than an earthquake in terms of mass movement?. Natural hazards, 112(1), pp.639-663.
Citation: https://doi.org/10.5194/egusphere-2025-2925-RC2 -
AC2: 'Reply on RC2', Habal Kassoum Traore, 20 Jan 2026
We sincerely thank the Referee for their insightful, constructive, and meticulous review of the manuscript. The comments provided, particularly regarding the specific research context of the study area, were highly valuable in refining the analysis and strengthening the interpretation of regional dynamics. We also appreciate the extensive bibliographic references suggested, which substantially contributed to reinforcing the conceptual framework of the manuscript. Overall, this thorough review significantly improved the clarity, depth, and scientific quality of the research.
- The main one to me: the lack of insight into process understanding. What exactly are we studying ?
There is too little information with respect to what is being studied/discussed:
What type(s) of flood is considered? Pluvial, fluvial, both? Are we looking at pluvial flooding due to intense (but highlight localised) rainfall events of short duration (like thunderstorm), or are we looking at fluvial flooding associated with few day rainfalls? Different types of floods = different types of drivers and/or weighting in the AHP. What about seasonality? What would be the role of tides in these floods?
The flood risk considered in this study primarily corresponds to urban pluvial flooding, which represents the dominant flooding process affecting the Greater Abidjan metropolitan area. This focus is consistent with both the selected indicators (e.g., topographic wetness index, slope, land use, distance to roads and drainage networks) and the observed flooding patterns, which mainly result from intense seasonal rainfall combined with rapid urbanisation, surface sealing, and insufficient drainage infrastructure.
In the study area, flooding occurs almost exclusively during the long rainy season (April to July), with a marked peak in May–June, as evidenced by recent rainfall analyses and historical flood records. The observed increase in flood impacts is therefore primarily driven by land-use changes rather than by a significant intensification of extreme daily rainfall events.
River flooding and tidal influences may locally affect water levels, particularly in areas close to the Ébrié Lagoon. However, explicitly integrating these processes would require detailed hydrodynamic modelling and high-resolution temporal data that are not consistently available at the metropolitan scale. In a context of data scarcity and within the framework of a semi-quantitative AHP-based approach, we therefore deliberately focused on mapping flood susceptibility associated with recurrent urban pluvial processes, rather than attempting to distinguish multiple flood typologies without sufficient data support.
In line 311, reference is made about a study on shallow landslides. This is the only part of the manuscript where reference is made about landslide characteristics (although not for this specific study). Overall we clearly miss information about the types of processes that are being studied. Are they recent; old; of natural origin, human induced, triggered by rainfall, favoured by weathering, etc?
Landslides are mainly recent events triggered by intense rainfall during the peak of the rainy season, particularly between June and October. They predominantly correspond to shallow landslides and debris flows, which are typical of humid tropical urban environments. While the geological setting, characterised by Mio-Pliocene Continental Terminal formations, favours weathering processes and slope instability, human activities constitute a major aggravating factor. Rapid and often unplanned urbanisation has led to extensive slope modification, vegetation removal, and disruption of natural drainage networks, thereby significantly increasing terrain susceptibility.
These characteristics are consistent with observations reported in tropical urban contexts, where rainfall-induced shallow landslides are strongly controlled by the interaction between lithological conditions, surface weathering, and anthropogenic disturbances (e.g., Dewan, 2013; Guzzetti et al., 2007).
Following the reviewer’s suggestion, a dedicated paragraph has been added to Section 2 (Study Area) to explicitly describe the dominant landslide processes in terms of origin, triggering factors, and typology.
Considering the variety of the landscape conditions, and also the climate triggers, it would be normal to have different types of processes, or at least similar processes, but of different ages. In addition to these assumptions that we could have made a few months ago, there is a publication on landslides in Abidjan that has just been released. Although Gnagne et al., (2025)’s work comes with caveats, it still shows some interesting points: different types of landslides (slides,, avalanches, shallow and deep-seated slides), landslides of different ages (different land use/covers than today’s?).
These points with the different types of landslides are that land uses (and dynamics) will not have the same impacts on hazard, whether it is, for example, a shallow or a deep-seated slope failure Sidle and Bogaard, (2016). Some examples of the role of land use (and land use changes – road construction; deforestation) and urbanization on landslide incidence have been published for similar tropical Africa environments Dille et al., (2022); Maki Mateso et al., (2023); these could help to authors to better design their research.
We agree that landslide processes are not characterized in detail in this study. The paper by Gnagne and al. (2025), recommended by the reviewer, is particularly relevant in the context of Greater Abidjan and has helped us to refine and clarify this point. We fully acknowledge the contribution of this work, as well as of studies highlighting the differentiated influence of land use on shallow versus deep landslides and the role of urbanization dynamics in comparable tropical African contexts. These key references are now more explicitly integrated and discussed in the Discussion section.
Nevertheless, within the scope of the present research, an explicit distinction between landslide types, depths, or ages was not performed. This limitation mainly results from (1) the lack of a spatially homogeneous and exhaustive landslide inventory allowing a reliable classification at the metropolitan scale, and (2) the difficulty of distinguishing between ancient and recent landslides in a densely anthropized urban environment such as Greater Abidjan. This limitation is now explicitly acknowledged and discussed in the section dedicated to limitations and perspectives.
Beyond the issue of landslide typology, we recognize that the selection of predictor variables, as well as decisions related to their categorization and weight assignment, often raises methodological questions. In particular, slope gradient classes are not strictly based on a process-based geomorphological approach relying on threshold hillslope concepts (e.g., Bennett et al., 2016; Depicker et al., 2021). In this study, slope classes were defined following an operational and comparative rationale, consistent with numerous AHP-based susceptibility studies aiming to identify regional-scale spatial patterns rather than to explicitly model failure mechanisms. This limitation is now clearly stated and discussed, emphasizing that the choice of slope classes may locally influence the estimated susceptibility intensity.
Regarding the consideration of roads, we agree that they play an important role in triggering and controlling landslides (e.g., Tanyas et al., 2022). In our approach, roads were explicitly included in the flood analysis due to their direct influence on surface sealing and urban runoff dynamics. However, they were not integrated as an independent variable in the landslide analysis in order to limit redundancy with land-use information. This methodological choice is now clarified and discussed as a potential limitation of the analysis.
Finally, we fully agree with the reviewer concerning the presence and role of erosion gullies in the urban environment of Greater Abidjan, as illustrated by Gnagne et al. (2025) and Ilombe Mawe et al. (2025). Although gully erosion was not explicitly included as a predictive variable in the model, it is now mentioned and discussed in the Discussion section as a cross-cutting process that can locally enhance both landslide and flood hazards through increased surface connectivity. This aspect is also identified as an important perspective for future research.
- The overall modelling approach is questionable.
To study hazard risk, we expect susceptibility and hazard assessment, exposure analysis and vulnerability analysis. In practice we know that all these components of risk are not always easy to assess, especially for one specific region in data-scarce context. In this study, only susceptibility and vulnerability assessments are made. The temporal aspects (hazard) are not considered, nor the exposure one. This is therefore a problematic aspect in the manner the study is being presented and sold.
Another point is that, even for the susceptibility assessments, we do not know what is actually done For example, for landsliding, are the source of landslides being considered or their runout? Depending on those, they may be huge differences in exposure e.g. Schmitt et al., (2025). In addition, considering the type of speed of the processes (rapid or less rapid landslides, flash floods, etc.), the impacts and hence the overall risk is not the same.This study does not constitute a comprehensive risk assessment in the strict sense, as neither the temporal dynamics of hazards nor quantitative exposure are explicitly modelled. The manuscript has therefore been reframed as a semi-quantitative mapping of susceptibility and vulnerability, intended to provide a strategic diagnostic framework. This approach allows the identification of areas where risk-related factors converge, supporting urban planning and resource prioritisation without claiming probabilistic hazard modelling or real-time impact assessment.
Furthermore, the landslide analysis focuses on potential source areas of slope instability rather than on runout or deposition zones. While we recognise that process velocity, particularly for flash floods and shallow landslides, strongly controls impacts in the Greater Abidjan area, these aspects remain beyond the scope of the present study. The results should therefore be interpreted as an assessment of the overall spatial propensity to slope instability and pluvial flooding, providing a consistent metropolitan-scale knowledge base in the absence of detailed hydrodynamic and geotechnical data.
The vulnerability seems to be assumed to be the same for landslide and flood hazards. This cannot be the case considering the different impacts and also the frequency of the processes.
To address the concerns raised, the revised manuscript now specifies that this index should be interpreted as a measure of global socio-environmental vulnerability rather than a process-specific structural vulnerability analysis. We have clarified in the discussion that, although occurrence frequencies differ, integrating these vulnerability factors into a single AHP framework allows for the identification of metropolitan risk « hotspots » where adaptive capacities are most limited in the face of rapid hydro-climatic hazards.The data used as predictor variables in both flood and landside assessments, as well as for the vulnerability, are not questioned much about their reliability.
EM-DAT data are being used for validation of the models. However, these data come with significant caveats as they are highly biased towards impactful events and also towards regions provided with better communication means and greater wealth. The same hazard with the same intensity and magnitude is likely to receive less attention if it occurs in a remote location or a low-income neighbourhood. Further discussion of these reporting biases and dataset limitations can be found in Stein et al. (2024) and Delforge et al. (2025).
In our study, the EM-DAT data were not used as an exhaustive inventory, but as a complementary validation dataset, cross-referenced with CATNAT data and field observations. These limitations and associated biases are now explicitly discussed in order to avoid overinterpretation of the validation results.
Fiefd survey is not clearly explained.
Field surveys were conducted over a three-year period, between 2021 and 2023, with targeted observation campaigns during the rainy seasons. The most intensive fieldwork took place in June 2022, following major flood and landslide events, notably in communes such as Bingerville and Attécoubé, where post-event observations were carried out shortly after the occurrences.
In parallel, a structured socio-environmental survey was conducted across 16 municipalities of Greater Abidjan. A total of 450 individuals were interviewed, based on a statistically defined sampling strategy ensuring metropolitan-scale representativeness while over-sampling high-risk areas. The surveys were administered by 29 trained field investigators, with a minimum of 15 respondents per municipality and higher sample sizes in the most exposed communes (e.g., Abobo, Yopougon, Cocody).
Field observations included the identification of flood marks, damaged dwellings and infrastructure, surface runoff paths, erosion features, slope cuts, and visible signs of shallow slope instability, complemented by photographic documentation and qualitative site descriptions. These field data were used to confirm the spatial occurrence of reported events, support the interpretation of susceptibility patterns, and contribute to the validation of the model in combination with existing databases. The manuscript now explicitly states that these surveys did not constitute a systematic landslide or flood inventory, but rather an independent source of spatial and qualitative validation.
Hazard zonation; the meaning of the classes?
Concerning the zonation of hazards, we now specify that the susceptibility classes represent neither absolute probabilities nor physical thresholds of triggering. They reflect relative levels of susceptibility, resulting from the weighted combination of criteria in the framework of the PAH, and must be interpreted as ordinal categories allowing for spatial comparison of territories, and not as deterministic predictions. This clarification has been added to avoid ambiguity about the actual meaning of the classes.
Landslide assessment provides results for flat areas.
In a multi-criteria AHP framework, susceptibility to landslides is not only determined by the slope. The moderate susceptibility levels observed in flat areas reflect the combined influence of dense urbanization, soil sealing and proximity to road infrastructure. The integration of the topographic humidity index (TWI) makes it possible to identify areas prone to water accumulation, where soil saturation can promote local instabilities. In urban areas, these processes can affect anthropogenic micro-reliefs (e.g., embankments, road trenches, drainage ditches) that are not captured by DTMs at the regional scale, but are indirectly captured by land use indicators. We therefore specify that the vulnerability map represents an overall spatial propensity for ground instability, highlighting areas where anthropogenic pressures increase the fragility of the terrain, even in the absence of steep natural slopes. This perspective is particularly relevant for urban planning in a rapidly expanding metropolis like Abidjan, where human activities are continually reshaping the local topography.
- Muti-hazard analysis
Line 49: “...and lay the groundwork for effective multi-hazard disaster risk management and mitigation strategies.” Here, and in other places in the text, the authors put a focus on multi-hazard assessment. If I look at the definition of UNDRR (https://www.undrr.org/terminology/hazard) about multi-hazard: “Multi-hazard means (1) the selection of multiple major hazards that the country faces, and (2) the specific contexts where hazardous events may occur simultaneously, cascadingly or cumulatively over time, and taking into account the potential interrelated effects”. In this work, I do not see the real contribution with respect to the point (2). The focus of the study must be better defined with respect to what it contributes to the multi-hazard literature.
We acknowledge that, according to the UNDRR definition, a fully developed multi-hazard approach implies the analysis of interactions and cascading effects between hazards. In this study, the term multi-hazard is used in an operational sense, referring to the joint assessment of multiple major hazards within a consistent methodological framework, without explicitly modelling their dynamic or causal interactions (van Westen et al., 2014; Koks et al., 2019; de Ruiter et al., 2020). The manuscript has been revised to clarify this positioning and to explicitly distinguish our approach from fully dynamic, interaction-based multi-hazard frameworks.
- State of the art and lack of discussion with respect to that
The state of the art is rather limited in many aspects, including urban contexts of landslide and flood risk, land transformation, population exposure, and multi-hazard interactions. For instance, when considering landslides and floods jointly, only a few studies have addressed this dual perspective. I recommend that the authors consult works such as Ferrer et al., ( 2024) and Idukunda et al., (2025) and discuss how the context and findings of these studies relate to the dual nature of both exposure types.
Following the reviewer’s suggestion, recent studies by Ferrer et al. (2024) and Idukunda et al. (2025) were consulted and integrated into the state of the art. These works highlight the shared urban drivers of floods and landslides, particularly land-use change and population exposure, and provide relevant insights into the combined exposure patterns addressed in this study
Other comments
Introduction: The focus is strongly put in the case study of Abidjan, while the state of the art about mulit-hazard and risk assessment is not that much developed. One of the main justifications for the work is its aim to complement existing hazard studies conducted in the city. Although this is a valid rationale—particularly given the importance of improving knowledge in such a context—the study lacks a clear anchor point for a broader, international audience. In other words, why should readers unfamiliar with the study area find this work relevant or compelling?
We recognize that the introduction gives an important place to the case study of Greater Abidjan, which reflects both the scarcity of existing work in this context and the need to document tropical urban environments that are still largely understudied. However, this initial choice was not intended to limit the scope of the study to a strictly local interest, but to illustrate more general methodological and conceptual issues, widely shared by many cities in the Global South.
More specifically, this study is part of a broader framework of reflection on how to assess multi-hazard risk in a data-scarce urban context, marked by rapid urbanization, high population exposure and significant constraints in terms of environmental, social and hydrological data. These characteristics are not specific to Abidjan, but concern many tropical metropolises in Africa, Southeast Asia and Latin America.The interest of the work for an international audience thus lies in the proposed methodological approach, based on an integrative semi-quantitative approach (CAM/AHP), reproducible and adaptable to other urban contexts where fine process data are absent or incomplete. The study shows how to articulate heterogeneous indicators of hazards, vulnerability and exposure in order to produce operational risk maps, useful for disaster risk reduction and decision support, beyond the sole case study.
We nevertheless agree that this more general anchoring deserves to be further explained. In response to this remark, the introduction has been revised to reinforce the state of the art on multi-hazard approaches and to emphasize more clearly the transferable scope of methodological choices, as well as the contribution of the study to international debates on risk assessment in tropical urban settings and in the context of data scarcity.
Line40: “frequency and severity of flooding and landslides in the region have escalated in recent years, highlighting an urgent need to develop more effective multi-hazard risk management strategies”. We would welcome the inclusion of references to support these statements. Studies capable of disentangling such trends are relatively rare, particularly in data-scarce environments. If such studies exist, they should be cited. Moreover, their comprehensive datasets would be highly valuable for both the design and validation of the present research.
We have identified several authoritative references that support and appropriately nuance the initial statement. In particular, the IPCC Sixth Assessment Report (AR6, WGII) highlights that observed increases in flood- and landslide-related losses are primarily driven by growing exposure and vulnerability in urban areas, rather than by clearly established trends in hazard frequency or intensity IPCC, (2022) Chapters 4 and 6.
This interpretation is further supported by the Global Assessment Report on Disaster Risk Reduction published by UNDRR, (2019), which emphasises that the rise in disaster losses over recent decades largely reflects increased exposure of populations and assets, especially in rapidly urbanising regions.
At the regional scale, Douglas et al., (2008) document a marked increase in the impacts of urban flooding in African cities, mainly associated with rapid urbanisation, inadequate drainage infrastructure and the settlement of flood-prone areas.
Finally, Stein et al.,( 2024) Delforge et al.,(2025) underline the limitations and reporting biases inherent in global disaster databases such as EM-DAT, stressing that increases in reported disaster events do not necessarily correspond to actual increases in hazard occurrence.
Line 90: “our study offers a more realistic andpolicy-relevant understanding of hazard exposure and vulnerability across the metropolitan continuum of Abidjan”. As stressed here in the introduction, the relevance of the work in DRR policy is emphasized. beyond the production of maps within a basic zonation framework, there appears to be limited consideration of how the results directly contribute to DRR applications. In this sense, the connection to the DRR context seems somewhat overstated.
While we have tempered the language regarding immediate operational tools, the revised manuscript clarifies that this study provides an essential spatially consistent diagnostic base. This work serves as a strategic decision-support framework to guide risk prioritization and resource allocation, bridging the gap between raw data and actionable disaster risk reduction (DRR) policies.
Line 95: “To overcome this gap, our approach introduces an innovative validation component: a geo-referenced database of observed past events, compiled from multiple sources including national reports, humanitarian data platforms, and remote sensing-based event detection”. In fact, many studies are based on similar inventories (for instance, in another data-scarce African context—see Nsabimana et al., (2023). From a methodological perspective, there is no real innovation here. Regarding model validation, comparing model outputs with real-world data is almost a compulsory step.
The manuscript has been revised to remove any reference to « innovative validation » and to specify that the contribution of this step lies mainly in its systematic application at the scale of the entire metropolitan area of Greater Abidjan, as well as in the consolidation of a multi-source inventory (international databases, local reports and field surveys) in the absence of an exhaustive official inventory. This clarification allows us to better position our work in relation to the existing literature.
Too general in many sections (for example section 2.2), bringing not relevant and accurate information for the study.
We have summarized section 2.2 in order to eliminate overly general descriptions. The content has been refocused exclusively on the biophysical and anthropogenic characteristics of Greater Abidjan that directly influence the parameters of the AHP model. Each piece of information retained now serves as a direct justification for the choice of risk and vulnerability indicators presented in the methodology.
Bibliography:
Delforge, D., Wathelet, V., Below, R., Sofia, C. L., Tonnelier, M., van Loenhout, J. A. F., and Speybroeck, N.: EM-DAT: the Emergency Events Database, International Journal of Disaster Risk Reduction, 124, 105509, https://doi.org/10.1016/j.ijdrr.2025.105509, 2025.
Dewan, A. M. (2013). Floods in a Megacity: Geospatial Techniques in Assessing Hazards, Risk and Vulnerability. Springer. https://doi.org/10.1007/978-94-007-5875-9
Dille, A., Dewitte, O., Handwerger, A. L., d’Oreye, N., Derauw, D., Ganza Bamulezi, G., Ilombe Mawe, G., Michellier, C., Moeyersons, J., Monsieurs, E., Mugaruka Bibentyo, T., Samsonov, S., Smets, B., Kervyn, M., and Kervyn, F.: Acceleration of a large deep-seated tropical landslide due to urbanization feedbacks, Nat. Geosci., 15, 1048–1055, https://doi.org/10.1038/s41561-022-01073-3, 2022.
Douglas, I., Alam, K., Maghenda, M., Mcdonnell, Y., Mclean, L., and Campbell, J.: Unjust waters: climate change, flooding and the urban poor in Africa, Environment and Urbanization, 20, 187–205, https://doi.org/10.1177/0956247808089156, 2008.
Ferrer, J. V., Samprogna Mohor, G., Dewitte, O., Pánek, T., Reyes‐Carmona, C., Handwerger, A. L., Hürlimann, M., Köhler, L., Teshebaeva, K., Thieken, A. H., Tsou, C., Urgilez Vinueza, A., Demurtas, V., Zhang, Y., Zhao, C., Marwan, N., Kurths, J., and Korup, O.: Human Settlement Pressure Drives Slow‐Moving Landslide Exposure, Earth’s Future, 12, e2024EF004830, https://doi.org/10.1029/2024EF004830, 2024.
Guzzetti, F., Peruccacci, S., Rossi, M., & Stark, C. P. (2007). Rainfall thresholds for the initiation of landslides in central and southern Europe. Meteorology and Atmospheric Physics, 98, 239–267. https://doi.org/10.1007/s00703-007-0262-7
Gnagne, F. L., Schmitz, S., Kouadio, H. B., Hubert-Ferrari, A., Biémi, J., and Demoulin, A.: Predicting Landslide Susceptibility Using Cost Function in Low-Relief Areas: A Case Study of the Urban Municipality of Attecoube (Abidjan, Ivory Coast), Earth, 6, https://doi.org/10.3390/earth6030084, 2025.
Idukunda, C., Michellier, C., De Longueville, F., Twarabamenye, E., and Henry, S.: Assessing community vulnerability to landslide and flood in northwestern Rwanda, International Journal of Disaster Risk Reduction, 123, 105329, https://doi.org/10.1016/j.ijdrr.2025.105329, 2025.
Ipcc: Global Warming of 1.5°C: IPCC Special Report on Impacts of Global Warming of 1.5°C above Pre-industrial Levels in Context of Strengthening Response to Climate Change, Sustainable Development, and Efforts to Eradicate Poverty, 1st ed., Cambridge University Press, https://doi.org/10.1017/9781009157940, 2022.
Nsabimana, J., Henry, S., Ndayisenga, A., Kubwimana, D., Dewitte, O., Kervyn, F., Michellier, C., Nsabimana, J., Henry, S., Ndayisenga, A., Kubwimana, D., Dewitte, O., Kervyn, F., and Michellier, C.: Geo-Hydrological Hazard Impacts, Vulnerability and Perception in Bujumbura (Burundi): A High-Resolution Field-Based Assessment in a Sprawling City, Land, 12, https://doi.org/10.3390/land12101876, 2023.
Sidle, R. C. and Bogaard, T. A.: Dynamic earth system and ecological controls of rainfall-initiated landslides, Earth-Science Reviews, 159, 275–291, https://doi.org/10.1016/j.earscirev.2016.05.013, 2016.
Stein, L., Mukkavilli, S. K., Pfitzmann, B. M., Staar, P. W. J., Ozturk, U., Berrospi, C., Brunschwiler, T., and Wagener, T.: Wealth Over Woe: Global Biases in Hydro-Hazard Research, Earth’s Future, 12, e2024EF004590, https://doi.org/10.1029/2024EF004590, 2024.
UNDRR: Global assessment report on disaster risk reduction, 2019. https://www.undrr.org/publication/global-assessment-report-disaster-risk-reduction-2019
Citation: https://doi.org/10.5194/egusphere-2025-2925-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,034 | 439 | 40 | 1,513 | 33 | 44 |
- HTML: 1,034
- PDF: 439
- XML: 40
- Total: 1,513
- BibTeX: 33
- EndNote: 44
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors present a significant study on multi-hazard risk assessment for the Greater Abidjan area, which is highly relevant for regional disaster planning. By using the Analytic Hierarchy Process (AHP) within a Multi-Criteria Analysis framework, the manuscript aims to integrate diverse factors (climatic, environmental, and social) to produce valuable risk maps. The conceptual framework is sound, but the manuscript currently needs to strengthen the Methdology and Results section for clarity and transparency as detailed below.
Line 110: The paper structure is clearly outlined, though the "Sections" should be labelled according to standard scientific paper structure (e.g., Introduction, Methodology, Results, Discussion)
Line 165: There is a mismatch between the locations (4 municipalities listed) and the densities (5 values provided). Please correct this discrepancy.
Tables 1, 3, 5: It is unclear on which grounds the classification bins are established. Please add explanation of how thresholds were determined (e.g., natural breaks, quantile, literature-based).
Lines 175-180: The AHP weight determination process needs more detail. Please include: expert selection criteria, number of experts consulted, and the consensus-building process.
Lines 235 & 258-259: The CR values appear identical (1.3%) for both flood and landslide hazards. Please double-check these calculations and explain if they are indeed identical.
Line 339: Add a paragraph acknowledging missing vulnerability indicators (income, housing quality, infrastructure, health access).
Lines 452-510: The results section severely lacks quantitative visualization. Maps alone are insufficient for comprehensive risk communication. Please add: (1) statistical plots showing risk distributions and population exposure, (2) municipal comparison charts, (3) quantitative summary table showing area (km²) and population at each risk level for both hazards.
Line 482: "landslides (illustrated by red dots)" - the red dots are not visible in Figure 8. Please make them clearly visible or remove this reference.
Lines 515-520: Validation discussion is too brief. Please add ROC curves and confusion matrices to discuss agreement between data and model.
Lines 524-566: Please group the limitations by themes (data limitations, methodological constraints, scope restrictions) and prioritize by impact on results.
Lines 544-546: Expand discussion on how the identified limitations specifically affect your results and their reliability.
All Figures: The figures' quality, size, and font need to be updated for better readability. Ensure minimum 300 DPI resolution and legible text (≥10pt font).