Spot-5 Multispectral Image for Pine Plantation Structure Mapping

The huge extents of the softwood plantations in Australia’s national forests (about 1,020,000 hectare in 2009) require continuous silviculture operations. Consequently a reliable and continuous quantification of the plantation structure is needed to sustainably manage them. Remote sensing data is considered as a cost-effective and efficient tool for this task and several remotely sensed datasets have been examined for this purpose, including SPOT-5 data. In this study, different derivatives of SPOT-5 multispectral images including spectral and textural indices were examined for estimating inventory parameters of a Pinus radiata plantation in New South Wales, Australia. The spectral derivatives included individual bands, band ratios, principal components (PCs), and 19 vegetation indices extracted for 61 plots collected randomly over the study area. Grey level co-occurrence matrix (GLCM) indices were also calculated for individual bands, band ratios and PCs, for different window sizes and orientations. Stepwise multiple-linear regression was used to examine the relationship between each spectral and textural attributes and structural parameters including mean height, mean diameter at height breast (DBH), stand volume, basal area, and stocking. The results showed textural indices perform better than spectral attributes for estimating the structural parameters of the plantation. Moreover, adding spectral derivatives to the textural indices did not improve the results derived from the models achieved from textural indices. Among different structural parameters, mean height and mean DBH were estimated with errors of 13% and 17.1% which are better than the acceptable error of forest sampling, while the errors of estimation for other structural parameters are more than 20%.