Estimation of above ground biomass of mangrove forest plot using terrestrial laser scanner

Abstract

Above-Ground Biomass (AGB) is an important parameter in the conservation of mangrove ecosystem owing to their ecological and economic benefits. LiDAR technologies in forest studies have become popular, due to its highly accurate 3D spatial data acquisition. In this study, we propose an end-to-end framework for estimating AGB of mangroves from Terrestrial Laser Scanner (TLS) point clouds. The framework includes pre-processing of data, segmenting the wood and foliage at tree level using Weighted Random Forest (WRF) classifier and constructing Quantitative Structure Model (QSM) of the wooden components to estimate its biomass. The flow was extended to AGB estimation of 33 x 33 m plot by integrating tree level framework. The study also finds a unique solution to estimate the contribution of pneumatophores in the AGB. Segmentation of wood/foliage of tree point cloud using WRF yielded better results with an increment of 15.27 % in Balanced accuracy, 0.2 of Cohen’s Kappa coefficient, and 7.45 % in F1score than RF classifier. AGB estimation of mangroves using our approach using TLS data is 47.54 T/ha which has a mean bias of 0.0044 T/ha and RMS variation of 0.026 T/ ha when compared with the allometric methods. A Breadth-first graph-search segmentation approach was used to count the pneumatophores, aerial roots seen in few mangrove species (R2 = 0.94 with manual counting) and estimate its contribution to AGB of mangroves which is first of its kind using TLS point cloud. This outcome could also aid future studies in modeling the underlying root network and estimating the below-ground biomass.

Keywords

• Above-ground biomass
• Mangroves
• Pneumatophores
• Terrestrial LiDAR
• Machine learning
• Random forest