The role of land use and land cover on landslide susceptibility assessment

• Autor(a) principal: Melo, Raquel
• Data de Publicação: 2023
• Outros Autores: Zêzere, José, Oliveira, Sérgio, Garcia, Ricardo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10451/59325
• topic: Land use and land cover; Landslide susceptibility; Urban areas

The role of land use and land cover (LULC) as a conditioning factor for landslide susceptibility has been the subject of numerous scientific studies worldwide. Regarding this topic, Quevedo et al. (2023) presented an extensive literature review of 536 scientific papers, published from 2001 to 2020, where the LULC was investigated in relation to landslide susceptibility. In general, the aim of these studies is to understand how different LULC patterns and changes can influence the likelihood of landslide occurrence. Other recent study (Rohan et al., 2023) highlights that the susceptibility assessment must be undertaken separately for non-urbanized and urbanized areas, in order to consider the influence of the latter on landslide activity. Moreover, the authors suggest that the susceptibility models developed with landslide inventories from non-urbanized areas may wrongly evaluate the susceptibility in urbanized areas. In the present work we argue the relevance of LULC as a conditioning factor for landslide occurrence, in areas of dominant shallow slides and rapid urban growth. Instead, we hypothesize that this independent variable leads to a decrease in landslide susceptibility in urban areas, creating a false sense of security and leading to a potential increase in human exposure. Our hypothesis is tested in Rio Grande da Pipa basin, a 110 km2 area located in northern Lisbon that is prone to landsliding. Based on a landslide inventory with 272 occurrences triggered in 2010 (70% for training and 30 % for validation) and seven independent variables (slope, aspect, plan curvature, topographic position index, slope over area ratio, soil thickness, lithology), four landslide susceptibility models were developed using a bivariate statistical method (the information value). Three models integrated the variable LULC from 1995, 2010 and 2018; and the fourth was computed only with the seven independent variables above mentioned, which means no LULC was included. The four models were compared by computing the ROC curves and estimating the Area Under the Curve (AUC). According to the AUC results, the model with the worst predictive capacity was the one that included the LULC from 2010, however, it was expected that this model would return the best AUC since the landslides were triggered in 2010. Moreover, the model with the best predictive capacity was the one that integrated the LULC from 1995. Regarding the information value scores, all models that included the LULC variable classified urban areas with a negative score, which means that, in the overall model, this variable is forcing a decrease in landslide susceptibility in urban areas. Given our extensive knowledge about the Rio Grande da Pipa basin, we believe this assumption is unwise and can be linked to the incompleteness of landslides inventories, especially in recently urbanized areas, due to changes in the original topography. Lastly, by comparing the landslide susceptibility models without LULC and with the LULC from 2010, the urban area in the high and very high susceptibility classes is about 4 % in the former and 0.1 % in the latter. These results support the hypothesis that using the LULC variable produces a decrease in landslide susceptibility in urban areas.