Debris flow generation in volcanic zones in the southern Andes has not been widely studied, despite the enormous economic and infrastructure damage that these events can generate. The present work contributes to the understanding of these dynamics based on a study of the 2017 Petrohué debris flow event from two complementary points of view. First, a comprehensive field survey allowed us to determine that a rockfall initiated the debris flow due to an intense rainfall event. The rockfall lithology corresponds to lava blocks and autobrecciated lavas, predominantly over 1500 m a.s.l. Second, the process was numerically modelled and constrained by in situ data collection and geomorphological mapping. The event was studied by back analysis using the height of flow measured on Route CH-255 with errors of 5 %. Debris flow volume has a high sensitivity with the initial water content in the block fall zone, ranging from 4.7×10⁵ up to 5.5×10⁵ m³, depending on the digital elevation model (DEM) used. Therefore, debris flow showed that the zone is controlled by the initial water content available previous to the block fall. Moreover, our field data suggest that future debris flows events can take place, removing material from the volcanic edifice. We conclude that similar events could occur in the future and that it is necessary to increase the mapping of zones with autobrecciated lava close to the volcano summit. The study contributes to the understanding of debris flows in the southern Andes since the Osorno volcano shares similar features with other stratovolcanoes in the region.

The interaction of geological processes and climate changes has resulted in growing landslide activity that has impacted communities and ecosystems in northern Chilean Patagonia. On 17 December 2017, a catastrophic flood of 7 × 10⁶ m³ almost destroyed Villa Santa Lucía and approximately 3 km of the southern highway (Route 7), the only land route in Chilean Patagonia that connects this vast region from north to south, exposing the vulnerability of the population and critical infrastructure to these natural hazards. The 2017 flood produced a paradigm shift on the analysis scale to understand the danger to which communities and their infrastructure are exposed. Thus, in this study, we sought to evaluate the susceptibility of landslides in the Yelcho and Rio Frio basins, whose intersection represents the origin of this great flood. For this, we used two approaches, (1) geospatial data in combination with machine learning methods using different training configurations and (2) a qualitative analysis of the landscape considering the geological and geomorphological conditions through fieldwork. For statistical modeling, we used an inventory of landslides that occurred between 2008 and 2017 and a total of 17 predictive variables, which are geoenvironmental, climatological and environmental triggers derived from volcanic and seismic activity. Our results indicate that soil moisture significantly impacted spatial susceptibility, followed by lithology, drainage density and seismic activity. Additionally, we observed that the inclusion of climatic predictors and environmental triggers increased the average performance score of the models by up to 3–5%. Based on our results, we believe that the wide distribution of volcanic–sedimentary rocks hydrothermally altered with zeolites in the western mountains of the Yelcho and Rio Frio basin are highly susceptible to generating large-scale landslides. Therefore, the town of Villa Santa Lucia and the “Carretera Austral” (Route 7) are susceptible to new landslides coming mainly from the western slope. This requires the timely implementation of measures to mitigate the impact on the population and critical infrastructure.

Landslide inventories are crucial to studying the dynamics, associated risks, and effects of these geomorphological processes on the evolution of mountainous landscapes. The production of landslide maps is mainly based on manual visual interpretation methods of aerial and satellite images combined with field surveys. In recent times, advances in machine learning methods have made it possible to explore new semi-automated landslide detection methodologies using remotely detected images. In this sense, developing new artificial intelligence models based on Deep Learning (DL) opens up an excellent opportunity to automate this arduous process. Although the Andes mountain range is one of the most geomorphologically active areas on the planet, the few investigations that use DL mainly focus on mountain ranges in Europe and Asia. One of the main reasons is the low density of landslide data available in the Andean areas, making it difficult to experiment with DL models requiring large data volumes. In this work, we seek to narrow the existing gap in the availability of landslide inventories in the area of the Patagonian Andes. In addition, the feasibility and efficiency of DL techniques are studied to develop landslide detection models in the Andes from the generated datasets. To achieve this goal, we generated in a manual process a datasets of 10,000 landslides for northern Chilean Patagonia (42–45°S), being the largest freely accessible landslide datasets in this region. We implement a machine learning model, through DL, to detect landslides in optical images of the Sentinel-2 constellation using a model based on the DeepLabv3+ architecture, a state-of-the-art deep learning network for semantic segmentation. Our results indicate that the algorithm detects landslides with an accuracy of 0.75 at the object level. For its part, the segmentation reaches a precision of 0.86, a recall of 0.74, and an F1-score of 0.79. The correlation of the segmentation measured through the Matthews correlation coefficient shows a value of 0.59, and the geometric similarity of the correctly detected landslides measured through the Jaccard score reaches 0.70. Although the model shows a good response in the testing area, errors are generated that can be explained by geometric and spectral relationships, which should be solved through new training approaches and data sets.