Unveiling the drivers of rainfall-triggered landslide spatial distribution: Insights from event-based and historical landslide inventories in Lisbon region, Portugal

This study examines the factors that influence the spatial distribution of rainfall-triggered landslides and evaluates the impact of using different types of inventories in assessing landslide susceptibility in Lisbon region, Portugal. Two questions guided the analysis: (1) How does the spatial distribution of landslides during a specific event relate to predisposing and triggering factors? and (2) Can event-based inventories capture the patterns of historical inventories and vice versa? The results show that predisposing factors such as slope, aspect, and elevation are key elements to identify regions prone to landslide occurrence. Although their ranking of importance differs between data-driven landslide susceptibility models using event-based and historical landslide inventories, the partial dependence plots (PDP) show no substantial changes in how the predisposing factors influence the outcome. Model comparisons and cross-inventory testing highlighted the individual strengths of event-based and historical inventories, and their spatial differences in susceptibility map outputs that can be attributed to highly localized rainfall. The variations observed in the susceptibility maps highlight the importance of integrating both types of datasets for a comprehensive analysis. Also, the consideration for using event-based inventories for general susceptibility mapping should be cautious, since an event-based inventory could represent a highly localized intense rainfall and the absence of occurrences in certain areas may simply be due to the lack of rainfall at those locations, even if the conditioning factors are the same or potentially more favorable for landsliding. This same bias can result in underestimated prediction rates when event-based landslide inventories are used to validate susceptibility models that were developed using historical landslide data.