Influence of sample and terrain unit on landslide susceptibility assessment at La Pobla de Lillet,Eastern Pyrenees,Spain

The reliability of susceptibility maps depends largely on the quality of the information used for its evaluation. This study seeks to analyze the influence of sample size and type on the results of discriminant analysis applied to shallow landslide susceptibility assessment. The study also assesses the role of the terrain unit in discriminant analysis. To this end, two databases based on fieldwork (slope unit) and GIS with 15- and 45-m grid cells (grid cell-based unit), were compared in the same zone at La Pobla de Lillet, Spanish Eastern Pyrenees. The results show that although there is no significant influence of the type of sample, it is necessary to use at least half of the individuals of the sample in order to obtain good results from discriminant analysis. It is the terrain unit that exerts the biggest influence on the result of susceptibility. Some morphometric parameters related to landslides were compared in the databases. The slope unit of the fieldwork database better reflects the land characteristics than the regular grid used by GIS. The values of the variables obtained by GIS procedures are smooth, obtaining mean errors for the slope angle variable of 19.5 and 33.5% for the grids of 15 and 45 m, respectively, in the study area. One-way and T tests demonstrate that the smoothness of the values exerts a decisive influence on the discriminant results. Kappa’s analysis shows that there is no significant equivalence between some of the categorical variables used in both databases. The use of these variables demand the application of clearly defined criteria. The cell size should match the dimensions of the phenomenon analyzed given the unsuitability of the grid of 45 m in this study.