Predictive mapping of intertidal and shallow subtidal habitats using ALOS imagery and LiDAR bathymetry

The Corner Inlet-Nooramunga marine embayment and barrier islandcomplex in eastern Victoria includes extensive sand and mud flats with a dendritic network of tidal channels in the west in Corner Inlet and a more complex arrangement of channels and islands to the east in Nooramunga. Maps of the predicted distribution of shallow subtidal and intertidal habitats in two separate areas of the region covering a total of 34,000 ha) were developed from satellite spectral and high resolution bathymetry data via an automated classification tree (Quick Unbiased Efficient Statistical Tree) approach. Spectral data from ALOS (Advanced Land Observing Satellite) imagery, corrected for atmospheric effects using ground based spectral samples and CSIROs AtCor process were combined with bathymetric data from airborne LiDAR (Light Detection and Ranging) surveys at a 10m resolution. The main habitats identified in both areas were seagrassbeds of either Zosteraceae species or Posidonia australis, bare sand and mud and areas dominated by the filter feeder Pyura stolonifera. Small areas of fringing reefwere also identified in Corner Inlet. A total of 791 stratified ground-truthing points were used to train and assess the automated classification which resulted in overall classification accuracies of 73% for Corner Inlet and 85% for Nooramunga study areas. Advantages of this approach, compared to previous habitat mapping exercises, include the ability to accurately and independently assess errors in habitat mapping, improved atmospheric correctionof imagery and repeatability for use in future assessments of change. Comparisons with prior mapping of these areas were limited by differences in methodologies and were sensitive to the level of resolution in habitat categorization. Effects of and approaches to dealing with this unquantifiable uncertainty in long-term monitoring and management of shallow water habitats are explored using examples from this project.