Abstract. Landslides triggered by rainfall are very common phenomena in complex tropical environments such as the Colombian Andes, one of the regions of South America most affected by landslides every year. Currently in Colombia, physically based methods for landslide hazard mapping are mandatory for land use planning in urban areas. In this work, we perform probabilistic analyses with r.slope.stability, a spatially distributed, physically based model for landslide susceptibility analysis, available as an open-source tool coupled to GRASS GIS. This model considers alternatively the infinite slope stability model or the 2.5-D geometry of shallow planar and deep-seated landslides with ellipsoidal or truncated failure surfaces. We test the model in the La Arenosa catchment, northern Colombian Andes. The results are compared to those yielded with the corresponding deterministic analyses and with other physically based models applied in the same catchment. Finally, the model results are evaluated against a landslide inventory using a confusion matrix and receiver operating characteristic (ROC) analysis. The model performs reasonably well, the infinite slope stability model showing a better performance. The outcomes are, however, rather conservative, pointing to possible challenges with regard to the geotechnical and geo-hydraulic parameterization. The results also highlight the importance to perform probabilistic instead of – or in addition to – deterministic slope stability analyses.

Los estudios básicos de susceptibilidad y amenaza por la ocurrencia de movimientos en masa son un elemento fundamental para la actualización de los planes de ordenamiento de los municipios del territorio colombiano. Dado lo anterior, la Ley 1523 de 2012 establece la política marco, y el Decreto 1807 de 2014, compilado en el 1077 de 2015, establece los lineamientos técnicos que tales estudios deben seguir y las condiciones mínimas que se deben cumplir. Por este motivo, se realizó una selección de algunas metodologías reconocidas en la literatura, que, al ser adecuadas y validadas según las condiciones propias de cada municipalidad, pueden ser utilizadas para la realización de tales estudios, sean tanto para el área rural y para suelo urbano y de expansión, como para cada uno de los factores que pueden detonar estos eventos.

<p>Colombia is located in a tropical environment with mostly warm and humid climatic conditions and complex mountainous terrain, where landslides triggered by intense rainfall are very common. Therefore, determining the occurrence and propagation of these events is of great interest in risk management and territorial planning programs.</p><p>Landslide propagation is difficult to predict due to uncertainties of rheological properties as well as initiation and dynamics of rock or sediment mobilization.  In Colombia, methodologies and models for landslide propagation have been less addressed than those corresponding to the occurrence, even though the consequences on people and infrastructure are generally are strongly related to travel distances and impact areas. Most propagation models are based on empirical methods to establish the travel distance of the sliding material, employing geometric approximations or geomorphological interpretation. In the last decades, physically based dynamic landslide propagation models have been proposed. These models use digital elevation models in combination with flow parameters. Their application represents a complex task because of the difficulty in constraining – depending on the model – sometimes the large number of relevant flow parameters.</p><p>In this study, we apply two open source models that work as extensions to the GRASS GIS software: (i) the <em>r.slope.stability</em> model for slope stability assessment using a limit equilibrium model for different sliding surface geometries, together with a probabilistic analysis applied to a range of geotechnical parameters (cohesion, internal friction); and (ii) <em>r.avaflow </em>for landslide propagation, which employs a multi-phase model considering solids and fluids. The models are implemented in the catchment area known as La Arenosa (9.9 km<sup>2</sup>), located in the municipality of San Carlos (Antioquia, Colombia). On September 21, 1990, an event of rainfall of short duration and high intensity precipitated on La Arenosa catchment. approx. 200 mm of precipitation fell within the study area in less than 3 hours, triggering approx. 700 landslides many of which have converted into hillslope debris flows. The zones categorized with a high probability of failure through <em>r.slope.stability</em> are defined as source areas and propagated down with <em>r.avaflow.</em> The results are evaluated against the landslide inventory in order to evaluate the potential of the proposed model combination for predictive simulations.</p>

Landslides triggered by rainfall are very common phenomena in complex tropical environments such as 15 the Colombian Andes, one of the regions most affected by landslides every year. Currently in Colombia, physically based methods for landslide hazard mapping are mandatory for land use planning in urban areas. In this work, we perform probabilistic analyses with r.slope.stability, a spatially distributed, physically based model for landslide susceptibility analysis, available as an open-source tool coupled to GRASS GIS. This model considers alternatively the infinite slope stability model or the 2.5D geometry of shallow planar and deep-seated landslides with ellipsoidal 20 or truncated failure surfaces. We test the model in the La Arenosa catchment, northern Colombian Andes. The results are compared to those yielded with the corresponding deterministic analyses and with other physically based models applied in the same catchment. Finally, the model results are evaluated against a landslide inventory using a confusion matrix and Receiver Operating Characteristic (ROC) analysis. The model performs reasonably well, the infinite slope stability model showing a better performance. The outcomes are, however, rather conservative, 25 pointing to possible challenges with regard to the geotechnical and geo-hydraulic parameterization. The results also highlight the importance to perform probabilistic instead ofor in addition todeterministic slope stability analyses.

No abstract