The aim of this study was to refine the predictions of forest structure and biomass of two structurally different plant communities using remote sensing techniques in topographically complex landscapes. Small footprint Light Detection and Ranging (LiDAR) and Landsat5 Thematic Mapper multispectral satellite data along with a statistical modelling approach were employed in this study. Two forested areas, eucalypt dominated open-canopy vegetation of the Richmond Range National Park (RRNP) and the tall closed-forest community of the Border Ranges National Park (BRNP) were selected in north-eastern New South Wales, Australia.
The key findings of the study were; (1) the physically based radiometric correction method improved the performance of all topographic correction methods, and the Processing Scheme for Standardised Surface Reflectance model performed significantly better in terms of yielding a marked improvement of Foliage Projective Cover (FPC) prediction from Landsat5 TM compared to other correction models; (2) the magnitude of error for predicting structural parameters of vegetation using LiDAR was much higher in subtropical rainforest than those documented in other studies; (3) fusing LiDAR with Landsat5 TM derived variables increased overall performance for the RRNP and combined sites data by describing extra variation of field estimated plot scale AGB except subtropical rainforest; (4) employing digital elevation model based estimated insolation, topographic wetness index (TWI), and topographic variables enabled a characterisation of the plot scale structural attributes of forests in relation to the variation of topography. The key finding of this investigation was that adopting this methodology to characterise vegetation structure in relation to environmental and topographic variations in eucalypt dominated open canopy is possible. The results indicated that physical and chemical properties of soil, and geology possibly have significant influences on potential vegetation structure more than the TWI, and insolation in subtropical forest.
Overall, this study presented different investigations for understanding the potential for remotely sensed data to refine the prediction of fundamental forest structure and biomass of two plant communities. This finding is significant because as far as we are aware, this is one of few studies to demonstrate the application of remotely sensed data to investigate forest structure of hilly plant communities in Australia. Thus, this information strengthens the understanding of methods of prediction of Australian woody plant communities over large areas of landscape. Furthermore, despite some limitations, these results add to the existing knowledge base and contribute to a better understanding of how to predict spatial configuration of vegetation structure of plant communities in topographically complex landscapes.