RECENT DEVELOPMENTS IN LANDSAT-BASED CONTINENTAL SCALE LAND COVER CHANGE MONITORING IN AUSTRALIA

摘要

Land use changes associated with agriculture and forestry are a significant component in Australia's carbon budget. In response to the need to quantify the resulting greenhouse gas emissions, the capability for continental monitoring of land cover changes using Landsat data has been developed through collaboration between the Australian Department of Climate Change (DCC), the CSIRO, and other partners. The project, called the National Carbon Accounting System -Land Cover Change Project (NCAS-LCCP) uses some 5000 Landsat MSS, TM and ETM+images to monitor land cover attributes including the presence/absence of perennial vegetation, plantation types and urban extent, at 25m resolution for fifteen time periods since 1972. The derived information is used to inform carbon accounting, environmental reporting and management. The original demands for the spatial and temporal resolution used arose from the development of the reporting rules (Marrakech Accord) guiding the implementation of accounting procedures for the Kyoto Protocol. However the remote sensing program evolved to more broadly cover interests in monitoring land use change generally. To meet Australia's current definition of forest, a land cover must have 20% tree crown cover and the potential to reach greater than 2m in height. However, there are large areas of Australia where vegetation communities do not meet this canopy density threshold, such as the drier rangeland regions or within agricultural regions. Increasingly there is a requirement to map and monitor current and historical changes in such areas, at regional and national scales, to provide information for conservation and natural resource management questions. Landsat Thematic Mapper data from 1989 to the present is being used. Sparse perennial vegetation cannot be reliably discriminated from other ground cover types based solely on the spectral information in this data; however by including information from image texture the results can be significantly improved. The timing and extent of land cover change (to or from forest) is important for accounting of carbon fluxes. In areas where new forest growth has been identified, the carbon modelling can be improved by using growth data specific to the new forest type. Given new forest, forest type (hardwood and softwood plantations, environmental planting and native regrowth) is of interest. Research was undertaken to examine whether an operationally feasible methodology for labelling the type of new plantings could be developed using the existing time series of Landsat imagery. An operational methodology for mapping new post-1990 plantation type was developed using Landsat TM imagery, but the method could not be applied to pre-1990 forest types due to the reduced discrimination using Landsat MSS imagery. An alternative strategy was developed using the spectral growth patterns of different plantation types through the time series to best separate them from native forest. This paper will present the results from the recent pre-1990 plantation mapping and sparse cover monitoring activities which are now being implemented at national scale in the NCAS-LCCP program.