Effect of geomorphology and land cover on landslide distribution at a local spatial scale: An example of the central Andes, Colombia

Tropical mountains are particularly vulnerable to landslides due to their susceptibility to climate warming combined with changes in land use driven by development and social transformations. Therefore, landslides in these regions pose a serious challenge to local land management, infrastructure development, and the conservation of soil and water resources. The Colombian Andes is a region where landslides are widespread, which, combined with the dense population, makes it prone to geohazards.

Numerous studies focused on national and regional inventories of landslides; however, small landslides (<1 km²) are often neglected despite having strong consequences for local communities. A detailed inventory of past landslides is essential for analysing the geomorphological processes related to landslide initiation and for calibrating and validating landslide susceptibility models. This study focuses on the impact of land cover and geomorphology on the distribution of small landslides (less than 1 km²). To minimise the influence of other factors, we concentrated on a single catchment area characterised by relatively uniform geology and precipitation. The main objectives of the study were (1) to document and analyse the spatial distribution of landslides; (2) to investigate factors potentially responsible for their development, specifically examining the differences in the frequency of landslides between forested and non-forested areas in a local spatial scale.

Landslides were identified using high-resolution satellite imagery from Ikonos, WorldView, and Pleiades from 2000/2003, 2013/2014, and 2019/2020. Landslides were visually interpreted from the images based on factors such as image tone, texture, vegetation cover, and visible disturbances of the surface. The identified landslides were vectorised as polygons, and a point representing the centre of the headscarp was also added for each landslide. The mapping results were verified during fieldwork in 2017, 2018, and 2019. Basic morphometric and descriptive parameters were attributed to each landslide, including area, type of landslide, and land cover. In the final step, frequency ratio modelling was employed to investigate the relationship between topographical and land cover factors and the distribution of landslides.

We mapped more than 900 small landslides ranging in size from 10² m² to 10⁴ m². Most of these landslides were found in cultivated areas, such as pastures, farms, and plantations, or along local roads. Our findings revealed four potential scenarios for landslide activity: (1) an intensification of landslide processes and an increase in the overall landslide area; (2) active landslides that remain stable in terms of size; (3) the activation of new landslides; and (4) deactivation of existing landslides accompanied by vegetation succession. The activation of new landslides and the intensification of existing ones were primarily linked to direct human modifications of the terrain, mainly through constructing new roads or repairing existing ones. The results indicate that the spatial distribution of landslides at a local scale is marked by significant clustering, with the zone of pastures characterised by the biggest concentration of landslides.

The research was funded by the Polish National Science Centre, Poland (Project number 2015/19/D/ST10/00251)