Measuring soil salinity using airborne multi-frequency polarimetric synthetic aperture radar and its application to the north Australian tropical coastal wetlands

  • Darren Bell

    Student thesis: Doctor of Philosophy (PhD) - CDU


    Soil salinisation is a major problem in coastal and inland Australia. This thesis focuses on the use of airborne multi frequency quadpolarised synthetic aperture radar data for mapping soil salinity. Four major contributions have been made. The first is the application of known soil moisture inversion algorithms to the tropical wetlands in northern Australia. This required the development of an appropriate filter for selecting data, from a highly heterogeneous environment, that satisfy the known validity region of the models. The second is a new methodology developed for isolating the imaginary part of the dielectric constant, which is correlated to the electrical conductivity of the target. The third is a non-linear model for vegetation correction of the airborne radar data for mapping soil salinity. The fourth is a highly accurate methodology for mapping bio-indicators of saltwater intrusion using a data fusion process.

    The first contribution provides an analysis of the application of existing inversion algorithms to the tropical wetlands. This is a significant step from their common application, primarily homogeneous agricultural environments. This lead to the development of the data selection filter which also allowed assessment of the degree of vegetation cover and density beneath which soil parameters could be assessed. These contributions demonstrate the transportability of the algorithms to the natural tropical wetlands of northern Australia. However, these algorithms are not able to determine soil salinity.

    Soil salinity information is known to affect the radar backscatter data. The second contribution outlines the development of the Combined Model (CM), providing a methodology to extract the soil salinity information from these data. However, vegetation cover proved to be a problem for this model. The third contribution deals with vegetation cover problems experienced in the development of the Combined Model. Vegetation cover is known to attenuate the radar backscatter resulting in an underestimation of the dielectric constant. The development of a non-linear regression model for vegetation correction significantly improved the results of the Combined Model inversion for all cover types and densities that were accepted by the data selection filter.

    The fourth contribution further improves the salinity mapping procedure by analysis of recent saltwater intrusion. Recent Saltwater intrusion is often characterised by dead trees, which are excluded from assessment using the vegetation corrected combined model by the data selection filter. Melaleuca spp. are known to be excellent bio-indicators of wetland health and are particularly good indicators of recent saltwater intrusion. The developed classification methodology of the fused Landsat Thematic Mapper data with Cloude decomposed radar data achieved highly accurate results. The location and spatial extent of dead Melaleuca, assessed using this methodology, is extrapolated as a measure of recent saltwater intrusion.

    A synergism of the vegetation corrected combined model results, with the results obtained from the data fusion process, provides a map of classified saline soils and clearly identifies recent saltwater intrusion.
    Date of Award2002
    Original languageEnglish
    SupervisorWaqar Ahmad (Supervisor)

    Cite this