Automated identification of alpine bogs using remote sensing and Bayesian classifiers
Sphagnum bogs play a pivotal role in the hydrology of Australian alpine catchments. Identifying such bogs from multispectral and hyperspectral data sources using traditional classification methods has proven suprisingly difficult, since they are invariably mixed with a variety of other psecies, and the spectral signatures of alpine vegetation have not enabled these vegetation types to be separated. Manual methods of identifying and classifying such bogs have been used in limited areas, but are unsuitable for use at the landscape scale.
This paper describes a methods of incorporating a range of landscape features, such as topography, hydrological position, valley bottom flatness, spectral signatures, texture and calibrated wetness indices as inputs to a Bayesian expert system which achieves 95% accuracy in identifying alpine wetlands in the Victorian alps.