The poster will present the results of a recent study which has investigated different approaches to upscaling plot data from forest soil monitoring, geomatching and regression combined with spatial data analysis (regression kriging). Geomatching builds on the representativity of inventory plots in mapping units (e.g. soils), while regression kriging develops regionally adjusted and optimized (sub)models, of which the regional uncertainties can be assessed from kriging residuals and the remaining model error. Different plot densities from several monitoring systems were compared and the uncertainties assessed (16x16 km, 8x8 km, 4x4 km). Regression kriging has been introduced in order to allow refined analysis of uncertainties. Regression analysis is particularly suitable in forest landscapes, because continuous (e.g. topography) and discrete effects (e.g. liming, bedrock, soil type, forest stand effects) may be modelled together. After performing the multiple linear regressions, the residuals were analyzed for spatial neighbourhood effects. The so-called autocorrelation had to be taken into account in most cases using geostatistics. Because several data sets from different countries and regions were available, a systematic testing of the method could be performed. A method was introduced which requires careful exploratory data analysis in order to develop meaningful (regional) submodels which greatly improve model fit. Regionalization (or upscaling) is a typical problem of multivariate statistics, because the researcher has to deal not only with one relevant key process, but rather with a set of them which is characterizing the regional peculiarity of a landscape. Compared to other regionalization studies, the new features of this method are the combination of predictor variables in forested landscapes, the interaction with forest management options and the development of scenarios for visualizing and evaluating forest management impacts. Differences in model results could be related to bedrock and forest stand composition, climatic factors, and mesoscale soil moisture regime. A high intensity of the evaluation was focused on using landscape morphology as a predictor for the soil chemical status. Besides topographic variables (e.g. elevation, transformations of aspect, slope, curvature, topographic wetness index and many others), classification of the parent material, forest site survey indices and stand characteristics from forest inventory data were the auxiliary variables that provided indirect information about the soil chemical status. To discuss and evaluate landscape related influencing factors and their interactions, constellations of hypothetical events (model scenarios) could be modelled on the basis of the regression equations.