Regional Characterisation and Mapping of Australian Forest Structural Types from ALOS PALSAR data
Richard Lucas(1), John Armston(2), John Carreiras(3) and Bunting Peter(1)
(1) Aberystwyth University, Llandinam Tower, Aberystwyth, United Kingdom
(2) Department of Natural Resources, 80 Meiers Road, 4068 Brisbane, Australia
(3) Tropical Research Institute, Travessa do Conde da Ribeira, 1300-142 Lisbon, United Kingdom
Forests occurring in the tropics and sub-tropical regions of Australia are diverse ranging from rainforest on the coastal margin to semi-arid scrub in the interior. In Queensland, one of the largest states in Australia, mapping of pre-clearing vegetation communities has been undertaken by the Queensland Herbarium through reference to historical stereo aerial photography and various forms of ecological and historical knowledge. Changes to these vegetation communities through deforestation are mapped and monitored at a state-wide level using time-series analysis of Landsat sensor data and derived canopy attributes (namely Foliage Projected Cover or FPC) by the Queensland Remote Sensing Centre. The floristic composition and structural type of these vegetation communities (“Regional Ecosystems”) are continuously updated by extensive field sampling and ground survey by the Queensland Herbarium.
This paper provides an overview of the use of Japan’s Advanced Land Observing Satellite (ALOS) Phased Arrayed L-band Synthetic Aperture Radar (PALSAR), singularly and in combination with Landsat data, for advancing the mapping of quantitative descriptions of forest structure and detecting change in these descriptions. For the study, mosaics of Kyoto and Carbon (K&C) Initiative fine beam dual polarization (HH+HV) strip products were used. To maximise co-registration accuracy with FPC and Regional Ecosystem data, the PALSAR data were orthorectified using Shuttle Radar Topographic Mission (SRTM) Digital Elevation Model (DEM) data and nominal 12.5 m spatial resolution Landsat panchromatic mosaics.
Comparison of mapped Regional Ecosystems with the ALOS PALSAR strip product and also historical Japanese Earth Resources Satellite (JERS-1) SAR indicated a close correspondence of the backscatter data with the spatial distribution of different ecosystems. Significant within-class variability was also observed and reference to limited available ground-truth data indicated that such variability was attributed to differences in height, density and crown cover (which relate to biomass). The study also highlighted areas where ALOS PALSAR data, in combination with Landsat-derived FPC, could be used to better map the extent of regenerating forest and dead standing timber on previously deforested and disturbed areas.