Mapping a Cyano-Bacterial Bloom (Lyngbya majuscula) in Moreton Bay, Australia : A Preliminary Comparison of Hyperspectral and Multispectral Approaches

摘要

The aim of this paper is to present the first stage in a project comparing the accuracy and efficiency of multispectral and hyperspectral images for mapping a toxic algal bloom in Moreton Bay, eastern Australia. Lyngbya majuscula is a toxic cyanobacteria of major concern in coastal areas of Eastern Australia due to its widespread coverage of critical seagrass and coral environments. An essential component of managing these blooms is developing accurate and cost-efficient approach to map and monitor changes to their extent and density. Previous work has described the successful development of a Landsat 7 ETM+ based L. majuscula mapping program. A larger project was initiated to determine the improvements possible in L. majuscula mapping by using hyperspectral data. For the first stage in this project, Landsat 7 ETM+ and EO-1 Hyperion image data were collected near simultaneously on March 21, 2001, within one day of an extensive field survey to map the extent of L. majuscula blooms. The Landsat and Hyperion data were both processed using a supervised classification approach (Spectral Angle Mapper) to derive a map of seagrass and L. majuscula cover. Both the Landsat and Hyperion maps were compared against field survey data and expert field knowledge of the area to determine their relative mapping accuracy. Consistent patterns in seagrass and L. majuscula distribution were observed between the two output maps. The Hyperion map appeared to map L. majuscula cover down to lower density levels that the Landsat 7 ETM. Further comparison of the cost, time, expertise and software used to derive each map product indicates two different approaches could be taken. One is more suited to relatively low-cost, government monitoring, and the other to providing a scientific basis for future monitoring programs.