SPECIES COMPOSITION RELATED TO SPECTRAL CLASSIFICATION IN AN AUSTRALIAN SPINIFEX HUMMOCK GRASSLAND

This paper demonstrates a methodology for relating objective vegetatio n classifications to spectral classifications in order to map variatio n in species composition within natural vegetation. Landsat MSS data w as used to map spinifex-dominated vegetation units for an island conse rvation reserve and oil production field on the north-western shelf of Western Australia. A significant relationship was established between an agglomerative hierarchical classification of ground samples, chara cterized by percentage cover of plant species and physical cover compo nents, and a similar classification of spectral means for sample pixel s. Assignment of spectral means to mapping classes was guided by both ground and spectral sample clustering. The strong relationship between the spectral classification and vegetation groups meant that cover cl asses mapped on the basis of spectral properties could be characterize d by quantitative ground data meaningful to vegetation ecology. The re sultant groups were differentiated largely on the basis of percentage cover of the three major spinifex species and the proportion of plant litter and exposed soil and surface rock. The study confirms the utili ty of ground cover as a quantitative variable for developing relations hips with spectral classifications, and demonstrates a methodology whi ch may have a wider application for mapping natural vegetation communi ties.