THE CHARACTERIZATION OF NATURAL VEGETATION USING 1ST-ORDER AND TEXTURE MEASUREMENTS IN DIGITIZED, COLOR-INFRARED PHOTOGRAPHY

A video-digitized, aerial colour-infrared photograph of a species-rich mire site was analysed to explore its information content with respect to ecologically important site characteristics, viz., species richness and structural diversity. First-order (mean and standard deviation) and texture measures (semivariance, angular second moment and inverse difference moment) were calculated for image windows representing 2 by 2m vegetation quadrats. For each quadrat the number of herbaceous species (SPPH), the cumulative cover of herbaceous species (COVH) and moss (COVM) and a structure-diversity index (DIVC) were derived from the complete recordings of species composition and species cover per quadrat. The diversity index is made up of structure classes based on species morphology and maximum height. Stepwise regression was used for the explanation of the variance in these statistics observed between the quadrats. The additional value of the texture measures was assessed relative to a base model containing red and near-infrared only. CO VH and DIVC appeared to be sufficient as descriptors of species richness and structural diversity; 30 and 60 per cent of the variance respectively is explained by the base model. The addition of semivariance resulted in an improved model only for SPPH and COVH. Biological and physical arguments are provided for explaining the behaviour of each statistic.