A multitemporal multiple density slice method for wetland mapping across the state of Queensland, Australia

• Published in: International journal of remote sensing 2009-01, Vol.30 (13), p.3365-3392
• Main Authors: Knight, A. W., Tindall, D. R., Wilson, B. A.
• Format: Article
• Language: English
• Subjects: Animal, plant and microbial ecology; Applied geophysics; Assessments; Biological and medical sciences; Classifications; Density; Earth sciences; Earth, ocean, space; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; General aspects. Techniques; Internal geophysics; Mapping; Spectra; Statistics; Teledetection and vegetation maps; Vegetation; Wetlands
• ISSN: 0143-1161; 1366-5901
• DOI: 10.1080/01431160802562180

The Australian and Queensland Governments are developing comprehensive wetland maps at a scale of 1:100 000 for the state of Queensland, Australia. Spectral classifications for water features were developed using Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM+) imagery acquired over a 16-year period. A multiple density slice/supervised classification method, the Standing Water Body (SWB) method, was developed to separate the main spectral and land cover elements of wetlands (vegetation, water and shadow cast by vegetation and topographic relief) and used rules to combine spectral classes to provide multitemporal (MT) information on wetland extent and water inundation regimes for features of at least 0.25 ha. Accuracy assessment in four trial areas compared the SWB method to the Normalized Difference Water Index (NDWI). The assessments of classified features were scale adjusted to maximum class-area proportions to enable statistical comparison and to account for the large area of non-wetland in the four trial areas. The average overall accuracy for wetland classification was 95.9% for the SWB method and 95.3% for the NDWI. The average unadjusted KHAT statistic for the wetland classification was 0.84 and 0.90 for the SWB and NDWI, respectively. The scale-adjusted KHAT statistic was much lower for both methods, averaging 0.45 for the SWB and 0.39 for the NDWI, mainly due to large omission errors. A method for the implementation of the SWB method for systematic and repeatable mapping of wetland areas is presented. The study recommends enhancement of the SWB classification through the inclusion of the NDWI classification and ancillary data such as vegetation mapping and drainage networks.