Modelling species richness from HRSC and QuickBird data in the Turtmann valley, Switzerland

摘要

Species richness is a fundamental biodiversity parameter in Landscape Ecology. At a local scale, we analysed the relationship between species richness and topographic, spectral and texture parameters derived by remote sensing data in an alpine environment. Landform parameters were calculated from a 1m digital surface model (DSM). Spectral and texture information were derived by QuickBird 2 data. Vegetation was mapped during field campaigns 2005 and 2006. For predicting species richness the nonlinear multivariate regression method of the partial least squares (PLS) was used as a statistical technique. The best model with 74% explained variance of the species richness of the test plots was created using spectral and texture information. Reducing all remote sensing derived parameters to the most important ones, still 72% of the variance could be explained. We conclude that for modelling species richness reflectance and NDVI as well as elevation information and homogeneity measures are the most important parameters. Consequently, biodiversity could successfully be predicted with remote sensing derived parameters.