Engineering Geology, vol.323, 2023 (SCI-Expanded)
Physically-based models are reliable techniques for landslide susceptibility assessment, therefore, several models have been proposed in the literature. In this study, two different hydrological models (FSLAM and TRIGRS) and two slope stability calculation methods (infinite slope and limit equilibrium method with spherical failure surface in SCOOPS3D) were compared. The rainfall-triggered shallow landslides that occurred in Kaptanpaşa, Rize (Turkey) in 2017 were selected as a case study. The main focus is investigating the impact of landslide morphology on the success of the slope stability models. For this purpose, landslides in the inventory were grouped according to their morphological characteristics, such as the landslide depth, length, width and their ratios, and their compatibility with model assumptions (e.g. the infinite slope assumption). Then, the model performances in different landslide types were evaluated separately. To calibrate the model parameters a new time-saving, semi-automated calibration strategy, based on a sensitivity analysis, is proposed. Concretely, stability parameters c′∅′Z were calibrated via a MATLAB code by considering the area under the curve (AUC) and extreme conditions: saturated and initial conditions. Overall, SCOOPS3D is deduced to have better performance than infinite slope analysis. However, the accuracy of SCOOPS3D is revealed to decrease for landslides with a high ratio of length/depth and width/depth. From the comparison of hydrological models, FSLAM yielded more plausible results than TRIGRS with the available data. It is demonstrated that the morphological features of the landslides significantly influence the performance of the physically-based stability models. Furthermore, it is highlighted that the resolution of the Digital Elevation Model in comparison to the width and length of the landslides in an inventory can be influential in the performance of the physical model. 10 m resolution of DEM is deemed to be suitable for the study area judging the predictive performance in small landslides.