Rock mass characterization is crucial for evaluating slope stability and recommending effective prevention mechanisms. This study presents a comparative analysis of three approaches for discontinuity analysis: (1) conventional field survey, (2) digital manual measurement on 3D models generated with UAV-based photogrammetry, and (3) semi-automatic analysis based on clustering algorithms (K-NN) for point cloud segmentation. All three methods were applied to the same slope, allowing their performance to be evaluated in terms of accuracy, efficiency, and replicability. The results showed that the semi-automatic method achieved the highest coverage (81%) and identified 586 discontinuities, with RMSE values of 2.58° for orientation, 0.087 m for spacing, and 2.05 m for persistence, using the conventional method as a reference. The digital manual method, with 19% coverage, yielded very low error (RMSE of 3.27° for orientation, 0.012 m for spacing, and 0.063 m for persistence), validating it as a complementary and reliable alternative. In contrast, the conventional method required the longest execution time (10 h) and achieved only 19% coverage, being the least replicable due to its dependence on expert judgment. Overall, the comparison highlights the advantages of digital methods, especially the semi-automatic approach, in improving efficiency, safety, and replicability, while providing robust information to recommend prevention strategies for rock slope stability.

Traditional rock mass evaluation methods have three main limitations: (1) their application depends largely on specialist judgment; (2) their discrete classification approach, such as RMR (Rock Mass Rating) and SMR (Slope Mass Rating), leads to abrupt transitions between categories; and (3) the interaction among geomechancial parameters is limited; these aspects reduce their ability to represent slope behavior in a gradual manner. The main objective of this research was to develop a model capable of representing gradual transitions between geomechanical conditions and the interaction among parameters related to susceptibility to instability. The model uses a hierarchical modular framework based on the Mamdani fuzzy inference mechanism, allowing the incorporation of expert knowledge through linguistic rules. It is implemented in a graphical environment that allows users to directly use geomechanical parameters obtained through conventional characterization or from three-dimensional digital models derived from UAV (Unmanned Aerial Vehicle) photogrammetry. Model consistency was evaluated through a sensitivity analysis, which verified the model’s response coherence across variations in input parameters. The graphical evaluation tool was then applied to three real case studies with different geomechanical configurations, and the results were compared with those from traditional methods (RMR and SMR). The results showed differences between traditional and fuzzy approaches, as our proposal links recommendations to specific geomechanical conditions across different evaluation levels, identifying conditions for potential intervention measures. In addition, the model enables the zonification of instability susceptibility, facilitating its use in future risk analyses. Our model is intended for application under normal slope conditions, without accounting for extreme events or external dynamic loads, such as seismic activity, groundwater level variations, infiltration, or high-mountain conditions.