The objective of the present study was to map dominant plant communities of an alpine area in the northeastern Alps (Austria), based on computer modelling. We employed gradient analysis by means of Canonical Correspondence Analysis (CCA) as a prediction tool and image segmentation as a filter for reducing the number of incorrect predictions. Topographical variables reflecting relief properties at different scales were used as surrogates for environmental conditions in combination with spectral band values from infrared orthophotographs. Coupling topographic correlation using CCA and image analysis proved practicable to map the distribution of alpine plant communities. Although plant communities often showed similar spectral response, they were mapped according to their specific topographical niches. Generally, topographic variables, indicative of environmental gradients controlling plant distribution, provided this information in most cases. The importance of spectral vs topographic variables varied among plant communities. Whereas the correlation between topography and plant species distribution was particularly significant for mapping alpine grasslands, spectral texture measures proved to be of major importance in discriminating between pioneer communities. Post-processing by image segmentation improved overall accuracy by 12%. A total of 17 plant communities and their mosaics were mapped, with an overall accuracy of 69.4% and a κ value of 0.64. Inaccuracy resulted from insufficient resolution of the available digital elevation model and confounding effects of additional controls like land use history, which could not be accounted for by topographic descriptors.
Abbreviations: CCA = Canonical Correspondence Analysis; DCA = Detrended Correspondence Analysis; DEM = Digital Elevation Model; GIS = Geographic Information System; NDVI = Normalized Difference Vegetation Index.
Nomenclature: Adler et al. (1994).