LiDAR patch metrics for object-based clustering of forest types in a tropical rainforest.

Abstract

Tropical rainforests support a large proportion of the Earth’s plant and animal species within a restricted global distribution, and play an important role in regulating the Earth’s climate. However, the existing knowledge of forest types or habitats is relatively poor and there are large uncertainties in the quantification of carbon stock in these forests. Airborne Laser Scanning, using LiDAR, has advantages over other remote sensing techniques for describing the three-dimensional structure of forests. With respect to the habitat requirements of different species, forest structure can be defined by canopy height, canopy cover and vertical arrangement of biomass. In this study, forest patches were identified based on classification and hierarchical merging of a LiDAR-derived Canopy Height Model in a tropical rainforest in Sumatra, Indonesia. Attributes of the identified patches were used as inputs for k-medoids clustering. The clusters were then analysed by comparing them with identified forest types in the field. There was a significant association between the clusters and the forest types identified in the field, to which arang forests and mixed agro-forests contributed the most. The topographic attributes of the clusters were analysed to determine whether the structural classes, and potentially forest types, were related to topography. The tallest clusters occurred at significantly higher elevations (> 850 m) and steeper slopes (> 26°) than the other clusters. These are likely to be remnants of undisturbed primary forests and are important for conservation and habitat studies and for carbon stock estimation. This study showed that LiDAR data can be used to map tropical forest types based on structure, but that structural similarities between patches of different floristic composition or human use histories can limit habitat separability as determined in the field.