Vol. 8, No. 2, 93-106, 2009

Abstract
Forest structure parameters such as diameter at breast height (DBH), tree height (H), stem density (STDENS), basal area (BA) and crown extension (CROWEX) are major variables of a forest inventory, which can also be investigated using satellite data. Remote sensing is often the only practical means of acquiring information on these variables over large areas, particularly in tropical regions with inaccessible terrain. This paper investigates the suitability of LANDSAT ETM+ imagery for remote estimation of forest structure parameters in a tropical rainforest, namely the Lore-Lindu National Park in Central Sulawesi, Indonesia. All LANDSAT ETM+ bands (except the 6^th band) and four multi-spectral vegetation indices derived from these bands were examined for their performance to express the forest structure parameters DBH, H, STDENS, BA, and CROWEX. The field data used for the correlation and regression analysis were collected from 60 plots throughout the area. The study showed that middle-infrared wavelengths of LANDSAT (bands 5 and 7) were best correlated with the forest structural parameters and could be used for prediction of DBH, H, BA and CROWEX. The tree height was predicted most reliably, while predictions of diameter at breast height, basal area and crown extension were less accurate. Stem density could not be predicted with sufficient accuracy from canopy reflectance only. The vegetation indices derived from red and near-infrared wavelengths (bands 3 and 4) such as the normalised difference vegetation index and the simple ratio were weakly correlated with the selected forest parameters. The vegetation indices incorporating band 5 showed much stronger correlations with forest parameters. The results revealed a great potential for the estimation of forest structure parameters with satellite optical data in tropical regions, where an extensive field survey is not feasible.

ISSN 1729-3782