Discrimination of peat swamp forest types with hyperspectral data

In tropical peat swamp forest, forest fire and illegal logging are major problems, which cause forest succession from grass just after disturbance to completely recovered forest. In order to distinguish each recovering stage, discrimination of forest types, such as primary forest and secondary forest, is very important. In general cases, a pixel-based classification is one of the most attractive choices for forest monitoring. However, since difference between primary and secondary forest comes in distribution ratio between the number of small-diameter trees and the number of large-diameter trees, only the pixel-based approach for the classification is not sufficient. In this paper, we use both spectral and spatial information from hyperspectral data to develop a high accurate biomass prediction model. Moreover, forest type classification scheme considering spatial distribution of biomass is proposed.