Regionalizing Indicator Values for Soil Reaction in the Bavarian Alps from Averages to Multivariate Spectra

摘要: We present an approach to produce maps of Ellenberg values for soil reaction (R-value) in the Bavarian Alps. Eleven meaningful environmental predictors covering GIS-derived information on climatic, topographic and conditions were used predict R-values. As dependent variables, indicator queried from plot records vegetation database WINALPecobase. additive georegression model, which combines complex prediction models increased accuracy a boosting algorithm. In addition we included spatial effects into model account autocorrelation. particularly interested usefulness averaged R-values prediction, applied two different models: (1) geo-additive regression that estimates mean (2) proportional odds predicting probability distribution over 1 9. found dependencies between R-value our predictors. Both produced same pattern predictions. Spatial had impact only first model. The main drawback is oversimplification reaction, entailed by averaging values. Therefore, regionalized average provide limited site-ecological characteristics. Model failed range shapes original spectra precisely. contrast, second provided more sophisticated picture reaction. To make multivariate output 2 comparable 1, propose three-dimensional color-space. addition, comparison both based multiple linear resulted R2 0.93. promising also other regions as well ordinal-scaled ecological parameters.