the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Jun 2026
Slope stability modelling in Karongi district, Western Rwanda
Abstract. The Karongi District in Western Rwanda is frequently subject to landslides. To date, however, physics-based slope stability assessments remain pending. In this study, we apply a three-dimensional Limit Equilibrium Method (LEM) using Scoops3D software to compute Factor of Safety (FOS) distributions in Karongi District. The model evaluates the effects of the pore-pressure ratio (ru) and horizontal pseudo-static seismic coefficient (keq) on slope stability. Results identify critical thresholds at ru ∼ 0.10 and keq = 0.075, beyond which unstable areas expand rapidly. When combined to pore pressure and at low pore pressure ru ≤ 0.10, seismic loading can significantly amplify slope instability. Model validation using historical landslide inventories shows 80 % spatial agreement between simulated unstable areas (FOS < 1) and observed landslides in two scenarios: (1) ru = 0.18 and keq = 0.10; and (2) ru = 0.35 and keq = 0.03. Although applied to the Karongi district, the methodology presented in this study can be used to assess the relative importance of pore pressure and seismic forcing in slope stability in a seismically active region prone to landslides.
- Preprint
(32243 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- CC1: 'Comment on egusphere-2026-3426', Olivier Dewitte, 06 Aug 2026
-
RC1: 'Comment on egusphere-2026-3426', Anonymous Referee #1, 15 Aug 2026
The manuscript presents a useful first physics-based assessment of slope instability in Karongi District using Scoops3D to investigate the individual and combined effects of pore-water pressure and seismic loading, and the attempt to validate the simulations against historical landslide inventories is valuable. However, the current model relies on several strong simplifications—including homogeneous soil properties over the entire district, uniform ru and seismic forcing, a coarse 100-m computational DEM, and failure volumes substantially larger than the predominantly shallow landslides documented in the region—which limit the physical interpretation and generalizability of the reported threshold values. In particular, representing rainfall solely through a prescribed pore-pressure ratio does not reproduce rainfall infiltration, transient suction loss, groundwater evolution, or the unsaturated hydraulic response that is central to rainfall-induced slope instability. The reported 80% agreement with the inventory is encouraging, but validation should be strengthened quantitatively because two very different combinations of ru and keq produce essentially the same agreement, indicating substantial non-uniqueness in model calibration. Therefore, I recommend major revision, with particular attention to hydro-mechanical representation, spatial variability, scale consistency, uncertainty/sensitivity analysis, quantitative validation, and a substantially deeper comparison with recent literature on unsaturated slope behavior, rainfall infiltration, soil hydraulic characterization, and regional-scale slope stability.
1. How can the authors justify representing rainfall-induced slope instability solely through a spatially and temporally uniform pore-pressure ratio ru? The manuscript acknowledges that this approach does not simulate rainfall infiltration or spatial/temporal rainfall variability and cannot account for negative pore-water pressures. Since many tropical residual slopes initially exist under unsaturated conditions, the authors should more thoroughly discuss the role of matric suction, SWCC, permeability functions, and wetting-induced suction loss. The discussion could be substantially strengthened using the practical unsaturated-slope framework developed by Adiguna et al. (2023), which specifically emphasizes the importance of incorporating soil suction into slope-stability assessment. (doi.org/10.3390/app13031811)
2. Can the critical value ru≈0.10−0.15 genuinely be interpreted as a physically meaningful threshold when ru is prescribed rather than derived from rainfall infiltration or groundwater observations? The manuscript identifies an apparent nonlinear transition beyond approximately ru=0.15, but without linking ru to actual rainfall intensity, duration, antecedent moisture, or measured pore-water pressures, its practical significance remains unclear. The authors should consider linking these results to transient infiltration and evapotranspiration processes; Hamdany et al. (2023), for example, numerically examined how infiltration and evapotranspiration alter soil suction and residual-slope stability and would provide highly relevant context. (doi.org/10.3390/su15118653)
3. Why is the entire Karongi District represented by a single homogeneous soil layer with uniform c, ϕ, and unit weight derived from measurements at a nearby site rather than from the district itself? The adopted values are based on measurements from Bigugu Village in Rutsiro District, whereas the model domain covers approximately 993 km². The authors should quantify how uncertainty and spatial variability in these properties affect the calculated FS and the proposed ru and keq thresholds. Regional soil-property databases and spatial parameterization strategies should also be discussed; the recent database approach of Li et al. (2025) demonstrates how spatially variable soil properties, including biologically influenced properties, can be assembled for regional slope-stabilisation analysis. (doi.org/10.1038/s41598-025-85250-5)
4. There appears to be a major scale inconsistency between the observed landslides and the simulated failure volumes. How does this affect the validity of the conclusions? Observed landslides in Karongi are reported to be shallow, approximately 0.5–2 m deep, with estimated volumes from about 25 to 1.36×105 m³, whereas Scoops3D searches for failure volumes of 107−108 m³. This two-to-several-orders-of-magnitude mismatch requires much stronger justification, and preferably additional higher-resolution simulations representative of the actual shallow landslide dimensions.
5. Is reducing the DEM resolution from 10 m to 100 m appropriate for identifying shallow landslide susceptibility in such steep and dissected terrain? Although the appendix indicates broadly similar vulnerable zones, the percentage of unstable area changes from approximately 0.40% to 1.75% depending on resolution/search settings, which is not negligible when quantitative thresholds are being inferred. A formal grid-resolution/convergence assessment should therefore be provided, particularly because slope gradients and small-scale topographic controls are strongly resolution dependent.
6. How robust are the proposed power-law relationships between unstable area and ru or keq? The manuscript fits power laws only over selected intermediate ranges and then notes departures from those relationships at larger values. The authors should report goodness-of-fit statistics, parameter confidence intervals, sensitivity to the fitted interval, and ideally a physical explanation for the exponents rather than treating empirical curve fitting as evidence of a fundamental threshold.
7. The seismic analysis relies on pseudo-static loading, but how representative are the selected keq values of actual site-specific seismic demand? The manuscript combines event-based estimates with a broad keq=0.03−0.35 range and acknowledges that pseudo-static loading neglects dynamic amplification and shaking duration. The authors should conduct a sensitivity assessment and clearly distinguish scenario analysis from actual seismic hazard prediction; ideally, the adopted coefficient range should be linked more rigorously to PGA, return period, local site effects, and uncertainty.
8. Is an “80% match” sufficient to validate the model when there is no corresponding assessment of false positives or correctly predicted stable areas? Two markedly different parameter combinations—ru=0.18, keq=0.10 and ru=0.35, keq=0.03—both yield approximately 80% agreement, demonstrating non-uniqueness. The authors should calculate more rigorous performance measures such as ROC-AUC, precision, recall/sensitivity, specificity, success/prediction rate, or confusion matrices and demonstrate whether the proposed model performs significantly better than a terrain-only baseline.
9. The manuscript should provide a much deeper discussion of unsaturated hydraulic properties and potential approaches for obtaining them rather than simply acknowledging the absence of pore-pressure data. SWCC and hydraulic conductivity functions provide the physical relationship between water content, matric suction, and infiltration response. For this reason, the authors are encouraged to discuss modern methods for establishing these properties, including the high-suction SWCC measurement approach reported by Prakoso et al. (2024/2025), which demonstrates the importance and feasibility of obtaining suction-dependent hydraulic information over an extended range. Such discussion would also clarify how future versions of the Karongi model could replace prescribed ru with physically calculated transient pore-water pressures. (doi.org/10.3390/su17010218)
10. The discussion of mitigation and practical applicability should be expanded beyond identifying unstable zones. What engineering interventions would follow from the modelling results? For rainfall-dominated slopes, surface protection, drainage, vegetation, and modifications to infiltration can substantially change the hydraulic boundary condition rather than merely changing static strength parameters. The authors should therefore discuss relevant rainfall-protection strategies and recent examples such as the recycled-concrete-aggregate capillary-barrier study by Rahayu et al. (2024), which evaluates how slope-cover systems alter pore-water pressure and stability during rainfall. (doi.org/10.1016/j.rineng.2024.103244) Together with the wetting-induced slope framework of Adiguna et al. (2023), transient infiltration–evapotranspiration modelling of Hamdany et al. (2023), regional soil-property database of Li et al. (2025), and SWCC measurement study of Prakoso et al., these references would help the authors place the present simplified ru-based framework within the broader state of the art and identify a credible pathway toward a more physically based model.
Citation: https://doi.org/10.5194/egusphere-2026-3426-RC1 -
RC2: 'Comment on egusphere-2026-3426', Anonymous Referee #2, 14 Sep 2026
Dear Editor,
I read with interest the manuscript entitled “Slope stability modelling in Karongi district, Western Rwanda”. The manuscript addresses the relevant problem of rainfall and seismic triggered landslides, with an application to a poorly investigated region that is susceptible to the natural hazard.
Although the importance of the topic, however, I believe that in this present form the manuscript is not sufficiently robust to be published in NHESS.
To my opinion, there is inconsistency between the processes that the paper aims to interpret and those actually represented by the model and the application. Please read in the following the specific comments:
- Rainfall-triggered hydrological dynamics are not simulated. A spatially and temporally uniform pore-pressure ratio is instead prescribed as a proxy for hydrological forcing. Consequently, the proposed “critical pore-pressure thresholds” cannot be interpreted as rainfall or hydrology-based thresholds for Karongi. The main limit of this approach is that pore pressure is assumed as hydrostatic (this is not stated in the work) and they are not associated to the soil moisture dynamics determined by the hydrological processes, like in the most common coupled hydrology-stability models, i.e. TRIGRS (Baum et al., 2008) , tRIBS-Landslide (Arnone et al., 2011, Lepore et al., 2013), GEO-Top (Simoni et al, 2008), HIRESS (Rossi et al., 2013) and so on.
- Arnone, E., Noto, L. V., Lepore, C., and Bras, R. L.: Physically-based and distributed approach to analyze rainfall-triggered landslides at watershed scale, Geomorphology, 133, 121–131, https://doi.org/10.1016/j.geomorph.2011.03.019, 2011.
- Baum, R. L., Savage, W. Z., and Godt, J. W.: TRIGRS - A Fortran Program for Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability Analysis, Version 2.0, https://doi.org/10.3133/ofr20081159, 2008.
- Lepore, C., Arnone, E., Noto, L. V., Sivandran, G., and Bras, R. L.: Physically based modeling of rainfall-triggered landslides: a case study in the Luquillo forest, Puerto Rico, Hydrol. Earth Syst. Sci., 17, 3371–3387, https://doi.org/10.5194/hess-17-3371-2013, 2013.
- Rossi, G., Catani, F., Leoni, L., Segoni, S., and Tofani, V.: HIRESSS: a physically based slope stability simulator for HPC applications, Natural Hazards and Earth System Sciences, 13, 151–166, https://doi.org/10.5194/nhess-13-151-2013, 2013.
- Simoni, S., Zanotti, F., Bertoldi, G., and Rigon, R.: Modelling the probability of occurrence of shallow landslides and channelized debris flows using GEOtop‐FS, Hydrol. Process., 22, 532–545, https://doi.org/10.1002/hyp.6886, 2008.
Therefore, there is no actual investigation of the interaction between hydrological processes and landslide triggering.
- There are a few mistakes in the theoretical presentation of the factor-of-safety formulation.
-
- Equations (1)-(3) mix stresses and forces and are dimensionally inconsistent as currently written. Eq. (2) expresses a Mohr–Coulomb shear strength in stress units, whereas Eq. (3) expresses a force; in fact, this is confirmed by Eq. (5) where W/A_h provides force over unit area, i.e. quantity consistent with pressure.
- The term W is described as the normal force acting on each column (line 75), although this should be the weight of soil so that Wsin(alpha) represents, as usual, the downslope component of the weight, as a result of the equilibrium of forces against T (Wsin(alpha) is the mobilizing force due to the weight of soil). See classical forces- equilibrium formulation in standard slope stability analysis.
- Eq. (4) defines pore pressure from the pressure head, but the manuscript does not explain how (h) is determined. In the actual simulations, pore-pressure conditions are represented through a prescribed (r_u), independently of rainfall infiltration and transient hydrological processes. This important limitation should be made explicit.
- The limits of integration of Eq. 5 are not specified, therefore it is not clear to which soil thickness or volume the weight refers.
- It is not clear why ru=0.5 should correspond to the saturated conditions. Although Scoops3D itself presumably implements the complete LEM formulation correctly, the equations presented in the manuscript do not consistently describe that formulation.
- In addition, no final expression for the FOS is provided, and it is not clearly stated for which type of failure mechanism the adopted formulation is valid.
- At this regard, the manuscript acknowledges that the observed landslides are mainly shallow translational slides/debris flows, whereas Scoops3D assumes spherical rotational failure surfaces.
- There is no investigation regarding the validity of homogamous soil. The model assumes homogeneous geotechnical properties over the entire district and analyses rotational failures of 107–10810^7–10^8 m³ using a 100 m DEM, whereas the observed inventory is dominated by shallow landslides with reported volumes mostly below 10510^5 m³. This scale and process mismatch substantially limits the physical interpretation of the results.
- The reported 80% agreement with the historical inventory does not constitute a sufficiently rigorous validation. Two combinations of ru and keq are selected that provide similar agreement, but there is no quantitative assessment of false positives, specificity, predictive skill, or parameter identifiability. Most importantly, landslide dates are largely unavailable; therefore, the relative roles of rainfall-induced pore pressure and seismic forcing cannot be established from the inventory.
- Ultimately, some statements of the introduction may be too strong. For example, at line 39-40 “Although the effects of rainfall on landslides are known, physical models describing the relationship between rainfall and pore-pressure remain to be developed (Ngaboyigihugu et al., 2025)”. This statement is clearly not correct in general, given the large body of physically based rainfall-infiltration/slope-stability modelling already available in the literature. Do the authors probably refer to the only specific study area?
For all these reasons, I recommend rejection for this journal at this step of the work.
Citation: https://doi.org/10.5194/egusphere-2026-3426-RC2 -
RC3: 'Comment on egusphere-2026-3426', Anonymous Referee #3, 18 Sep 2026
This manuscript presents a physics-based slope stability assessment considering simultaneous variations in pore water pressure and seismic conditions. It represents an important contribution to a region with significant exposure to geohazards, but I have notable concerns about the modeling approach and other components of the work:
- The manuscript states that this is the first application of a physics-based slope stability assessment in the region, but it does not adequately describe previous work to assess landslide susceptibility and what gap this particular study aims to fill. I encourage the authors to better explain the previous work in the region, what those works have learned, and what additional insight is expected from applying a physics-based approach.
- The choice to use SCOOPS3D seems misaligned with the landslide scale and mechanisms described in the manuscript. SCOOPS is a 3D model aimed at evaluating spherical slip surfaces. This is better utilized for rotational, and often deeper landslides, rather than the shallow, translational slides described in the manuscript. Further, the raster resolution of 100-m seems far too course to evaluate shallow, colluvial failures, which are strongly governed by finer-scale geomorphic features.
- Related to the previous comment, I would encourage the authors to consider use of a more appropriate slope stability model that is better suited to shallow, colluvial failures and that can evaluate transient hydromechanical processes in unsaturated soils (e.g., TRIGRS). Further, more discussion of the nature of landslides in the region should be included to better justify modeling choices. For example, do landslides occur under unsaturated or saturated conditions? Are there any coseismic landslides in the current inventory? How do your modeling choices reflect these observations?
- The validation technique appears to be inappropriate. Given that large swaths of the landscape appear to be classified as unstable, a validation scheme that represents false positives would be more appropriate here. Visually, it appears that the current model may significantly overpredict instability.
Overall, I think the study addresses an important problem and that a physics-based assessment of landslide susceptibility in Rwanda could provide a valuable contribution in a region with significant exposure to geohazards and a potentially complex combination of landslide triggers. However, the issues with the current modeling framework, representation of landslide processes, and validation are substantial enough that they cannot be adequately addressed through minor revision. Therefore, I recommend rejection in its current form, while encouraging the authors to substantially reconsider the modeling approach and resubmit a revised study.
Citation: https://doi.org/10.5194/egusphere-2026-3426-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 111 | 69 | 16 | 196 | 15 | 10 |
- HTML: 111
- PDF: 69
- XML: 16
- Total: 196
- BibTeX: 15
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear Sylvain Barayagwiza and co-authors,
I read your study on slope stability modelling in Rwanda with great interest. To further enrich your analysis, I would like to draw your attention on several points that, I believe, are relevant.
I hope you find these references helpful.
Kind regards,
Olivier Dewitte
References
Deijns, A.A., Dewitte, O., Thiery, W., d'Oreye, N., Malet, J.P. and Kervyn, F., 2022. Timing landslide and flash flood events from SAR satellite: a regionally applicable methodology illustrated in African cloud-covered tropical environments. Natural Hazards and Earth System Sciences, 22(11), pp.3679-3700.
Deijns, A.A., Michéa, D., Déprez, A., Malet, J.P., Kervyn, F., Thiery, W. and Dewitte, O., 2024. A semi-supervised multi-temporal landslide and flash flood event detection methodology for unexplored regions using massive satellite image time series. ISPRS Journal of Photogrammetry and Remote Sensing, 215, pp.400-418.
Depicker, A., Jacobs, L., Delvaux, D., Havenith, H.B., Mateso, J.C.M., Govers, G. and Dewitte, O., 2020. The added value of a regional landslide susceptibility assessment: The western branch of the East African Rift. Geomorphology, 353, p.106886.
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.
Depicker, A., Govers, G., Jacobs, L., Vanmaercke, M., Uwihirwe, J., Campforts, B., Kubwimana, D., Mateso, J.C.M., Bibentyo, T.M., Nahimana, L. and Smets, B., 2024. Mobilization rates of landslides in a changing tropical environment: 60-year record over a large region of the East African Rift. Geomorphology, 454, p.109156.
Dewitte, O., Dille, A., Depicker, A., Kubwimana, D., Maki Mateso, J.C., Mugaruka Bibentyo, T., Uwihirwe, J. and Monsieurs, E., 2021. Constraining landslide timing in a data-scarce context: from recent to very old processes in the tropical environment of the North Tanganyika-Kivu Rift region. Landslides, 18(1), pp.161-177.
Monsieurs, E., Kirschbaum, D.B., Tan, J., Maki Mateso, J.C., Jacobs, L., Plisnier, P.D., Thiery, W., Umutoni, A., Musoni, D., Bibentyo, T.M. and Ganza, G.B., 2018b. Evaluating TMPA rainfall over the sparsely gauged East African Rift. Journal of Hydrometeorology, 19(9), pp.1507-1528.
Monsieurs, E., Jacobs, L., Michellier, C., Basimike Tchangaboba, J., Ganza, G.B., Kervyn, F., Maki Mateso, J.C., Mugaruka Bibentyo, T., Kalikone Buzera, C., Nahimana, L. and Ndayisenga, A., 2018a. Landslide inventory for hazard assessment in a data-poor context: a regional-scale approach in a tropical African environment. Landslides, 15(11), pp.2195-2209.
Monsieurs, E., Dewitte, O. and Demoulin, A., 2019. A susceptibility-based rainfall threshold approach for landslide occurrence. Natural Hazards and Earth System Sciences, 19(4), pp.775-789.
Sibomana, P., Vanmaercke, M., Depicker, A., Tychon, B., Hubert, A. and Dewitte, O., 2025. Effects of agricultural terraces on landslide occurrence: Insights from a tropical mountainous region (Rwanda, Africa). Catena, 253, p.108898.
Uwihirwe, J., Hrachowitz, M. and Bogaard, T.A., 2020. Landslide precipitation thresholds in Rwanda. Landslides, 17(10), pp.2469-2481.
Uwihirwe, J., Hrachowitz, M., and Bogaard, T., 2022a. Integration of observed and model-derived groundwater levels in landslide threshold models in Rwanda, Nat. Hazards Earth Syst. Sci., 22, 1723–1742, https://doi.org/10.5194/nhess-22-1723-2022.
Uwihirwe, J., Riveros, A., Wanjala, H., Schellekens, J., Sperna Weiland, F., Hrachowitz, M. and Bogaard, T.A., 2022b. Potential of satellite-derived hydro-meteorological information for landslide initiation thresholds in Rwanda. Natural Hazards and Earth System Sciences, 22(11), pp.3641-3661.