Mapping salt-marsh vegetation using high-spatial and hyper-spectral satellite data to assist wetland inventory

Abstract

There is a lack of objective information on wetland condition or importance that could be used to inform management decisions. Salt marshes are complex ecosystems that are not well mapped and understood. This research was conducted to assess the potential of high-spatial and high-spectral resolution satellite data to map and monitor salt marsh vegetation communities. In particular, the study aimed to verify whether different salt marsh vegetation species could be differentiated using high-spectral and high-spatial resolution imagery and whether these could be linked to wetland condition. The results were then to be used to develop a GIS-based method of assessing the condition of wetlands and prioritising wetlands for rehabilitation. The main salt marsh vegetation species studied included 'Sarcocornia' and 'Sporobolus'. Two types of satellite imagery were used for this project. To compare sensor capabilities in discriminating salt marsh vegetation, it was decided to use high-spatial data sets from Quickbird and high-spectral data sets from Hyperion. Quickbird images have 0.7m pixel resolution in the panchromatic mode and 2.4m resolution in the multi-spectral mode. The multi-spectral mode consists of four broad bands. Hyperion images have 242 narrow bands and a pixel resolution of 30m. A hybrid unsupervised and supervised classification procedure was used to assess the wetland mapping potential of the Quickbird and Hyperion data. The supervised classification results were much more interesting and showed greater promise. Most of the vegetation species were better mapped. Water bodies were exceptionally well delineated by both types of imagery. One area of concern was the misclassification of 'Sporobolus' into grasses, mainly while using Quickbird imagery. This was mainly the case where the 'Sporobolus' was tall and dry; looking very similar to the tall reedy grass.