Brenning, A. (2008): Statistical geocomputing combining R and SAGA: The example of landslide susceptibility analysis with generalized additive models. In: J. Böhner, T. Blaschke & L. Montanarella (eds.), SAGA – Seconds Out (= Hamburger Beiträge zur Physischen Geographie und Landschaftsökologie, vol. 19), p. 23-32. Statistical geocomputing combining R and SAGA: The example of landslide susceptibility analysis with generalized additive models Alexander Brenning Department of Geography, University of Waterloo, Ontario, Canada Abstract The integration of statistical software with geographical information systems is required to be able to efficiently combine the most powerful tools and techniques available in both environments. The RSAGA package, which provides access to SAGA GIS geoprocessing functions from within the R statistical data analysis environment, is a recent contribution to this endeavor. The present work gives an overview of the structure of the RSAGA package, and demonstrates its usefulness in the context of landslide susceptibility modeling with terrain attributes and generalized additive models (GAMs). The GAM is an extension of the generalized linear model (e.g. linear and logistic regression). It is able to model nonlinear relationships, but retains an interpretable additive structure. In the case study on landslide distribution in the Ecuadorian Andes, several local as well as catchment‐related morphometric attributes are important, mostly nonlinear predictors of landslide occurrence. Other applications that can benefit from an integration of modern statistical computing techniques and GIS‐based digital terrain analysis include pedometrics, Precision Agriculture, and species habitat studies. Introduction Spatial analysis using Geographical Information System (GIS) techniques on one side and the statistical analysis of environmental data on the other provide two different views of spatial data analysis problems that are too often separated by disciplinary and software‐related barriers. However, both perspectives have much to contribute to spatial data analysis: GIS software provides a rich set of tools for spatial data manipulation, queries, and visualization, and statistical data analysis software can offer spatial and non‐spatial techniques required for understanding data or applying predictive models. These models may range from traditional statistical ones such as linear regression to complex black‐box models such as the support vector machine, a powerful, emerging machine‐learning technique. Given the fast progress in geographical information science and computational statistics, efficient ways of coupling GIS and data analysis software are required. This integration has been an important topic in recent years in the developer community of R (see e.g. Bivand, 2000; Brenning & van den Boogaart, 2001; Bivand et al., 2008), an open‐source data analysis environment that is widely used in statistical