An integrated predictive vegetation modelling approach: combining remote sensing, GIS, and biological theory

Abstract

Many diverse approaches have developed in order to characterize the spatial distribution of landscape features (particularly vegetation pattern). While technically variable, the data used in landscape characterization can be assigned into one of two categories - those that are spectrally derived from remotely sensed data and those that are ecologically relevant resource gradients. The remote sensing approach uses empirical relationships between land cover classes and spectral reflectance to classify a landscape, whereas the biological approach uses physiological and ecological relationships in defining geographic suitability for different species. Remote sensing approaches are limited in their ability to distinguish vegetation classes at the species scale, while biological approaches are limited in their ability to identify nonvegetated features on the landscape. In this thesis, an integrated method that combines these complimentary approaches is developed and compared to the individual spectral and biological approaches. The result is an accurate land cover pattern distribution that is able to resolve tree cover at the species level, while retaining the ability to define disturbed and non-vegetated areas.