Estimating relative abundance of tree species in tropical rainforest using remotely sensed data

Abstract

Mixed pixel occurrence in remote sensing imagery is a main source of problems in classifying ground features, especially when dealing with complex ecosystems such as tropical rainforest areas due to its high diversity of tree species. Pure pixel composed of a single species is very rare in most remote sensing imagery even in some advent ultrafine spatial resolution. In order to achieve an optimum output in classification of tree species in the forest, mixed pixel must be spectrally unmixed using sub-pixel approaches. This study was carried out in order to estimates the composition of tree species in Pasoh Forest Reserve by estimating the relative abundance of the tree species. The estimation of relative abundance was carried out using two types of spectral unmixing approaches which are Mixture Tuned Matched Filtering (MTMF) and modified Canopy Fractional Cover (mCFC). MTMF and mCFC were employed to Hyperion EO-1 satellite image with 30 meters spatial resolution. The relative abundance of Chengal trees was firstly estimated at a plot of 50 hectare. The correlation coefficients between the relative abundance obtained from MTMF and mCFC with the relative abundance of ground data in 50 hectare plot was 0.46 and 0.67, respectively. Therefore, mCFC was selected as it gives more encourage result in order to estimate relative abundance of Chengal trees at wider area such as compartment level. The model obtained from this study would be useful in forest monitoring and management