A multistaged approach to mapping soil salinity in a tropical coastal environment using airborne SAR and Landsat TM data

This paper provides a methodology to map soil salinity and a recent saline intrusion in the coastal fringe of Kakadu National Park, Australia. A model inversion from AirSAR data to map soil salinity on the bare and partially vegetated floodplain was used. This algorithm has been developed by combining the small perturbation model and the Dubois model to extract the imaginary part of the dielectric constant as an improved indicator of salinity. A vegetation correction to the combined model has been developed to allow salinity extraction from floodplain areas with moderate vegetation cover. Recent saline intrusion areas are often characterised by stands of dead Melalecua (paper bark trees). These are unsuitable for the inversion algorithms due to pronounced corner reflector effects on the radar. These areas were identified with high accuracy using a supervised classification on a fused AirSAR/TM image. The resulting map provides salinity levels for most of the floodplain and clearly identifies recent saltwater intrusion.