Asturias is one of the most landslide prone areas in the north of Spain. Most landslides are linked to intense and continue rainfall events, especially between October and May. This fact points out precipitation as the main triggering factor in the study area. Thirteen rainfall episodes that caused 1064 landslides between 2008 and 2016 have been selected for its study. Landslide records come from the Principality of Asturias Landslide Database (BAPA) and meteorological data from the Spanish Meteorological Agency (AEMET). Meteorological conditions which took place during each period have been characterized by using NCEP/NCAR Reanalysis data. Four main landslide-triggering meteorological patterns have been identified for the Asturian territory: Strong Atlantic Anticyclone pattern (SAA), Atlantic Depression pattern (AD), Anticyclonic ridge pattern (AR) and Cut-off Low pattern (CL).

Similar to many mountainous regions of the world, landslides are a recurrent geological hazard in the Gipuzkoa province (northern Spain) that commonly cause damage to communication infrastructure, such as roads and railways. This geomorphological process also threatens buildings and human beings, albeit to a lesser degree. Over time, different institutions and academic research groups have individually collected crucial information on historic and ancient landslides in this region, resulting in various landslide inventories. However, these inventories have not been collectively assessed, and their suitability for landslide susceptibility modelling projects has often been assumed without comprehensive evaluation. In this study, we propose a simplified method to explore, describe, and compare the various landslide inventories in a specific study area to assess their suitability for landslide susceptibility modelling. Additionally, we present the results of an illustrative experiment that demonstrates the direct effect of using different inventories in landslide susceptibility modelling through a data-driven approach. We found that out of the five digitally available inventories in the study area, only three provide sufficient guarantees to be used as input data for susceptibility modelling. Furthermore, we observed that each individual inventory exhibited inherent biases, which directly influenced the resulting susceptibility map. We believe that our proposed methods can be easily replicated in other study areas where multiple landslide inventory sources exist, and that our work will induce other researchers to conduct preliminary assessments of their inventories as a critical step prior to any landslide susceptibility modelling project.