Summary
Subantarctic islands contain some of the most remote and least disturbed wilderness regions of the world. Despite best efforts at conservation, the unique and highly specialised plant communities of these islands already show the effects of human impacts; most notably the impacts of introduced species and global climate change. The vegetation responds by changes in local abundance or cover of certain species at the expense and possibly extinction of others, changes in the extent and boundaries of certain communities, or migrations involving whole plant communities. Remote sensing provides a rapid and non-intrusive means of mapping and monitoring vegetation in sensitive environments. The synoptic overview provides advantages that field sampling cannot, particularly in such remote areas where access is very limited. However, there are limits to the level of information that remote sensing can supply. An inappropriate choice of sensor resolution or unrealistic expectations of the features that will be discriminated in the imagery will lead to disappointing results. This study assessed the potential for vegetation mapping and monitoring on Macquarie and Heard Islands. Ground spectrometry was used to produce a spectral library for classification, to assess whether the spectral reflectance of individual plant species and vegetation communities were likely to be discriminated in remotely sensed imagery and to determine the most appropriate spectral resolution and bandset for this task. The capabilities of three satellite sensors for detailed mapping of the vegetation were compared to assess the level of information that each could provide; i.e. a high spectral, low spatial resolution sensor (EO-1 Hyperion), a low spectral, high spatial resolution sensor (Ikonos-2) and a traditional broadband sensor (Spot XS). The accuracy of each classification was assessed using quadrat data of plant species and percent cover collected from Macquarie and Heard Islands, as well as recent vegetation maps produced by ANARE.