Forest biomass estimation using machine learning with lidar data

About

Accurate estimation of forest biomass is important for calculating carbon storage in trees which is required for carbon accounting to inform climate change mitigation policies. Direct measurement of biomass by collecting samples in situ is time-consuming and only feasible for small areas. In contrast, lidar remote sensing provides 3D data over large areas offering the opportunity to estimate tree biomass at forest or farm scale. This project aims to develop a machine-learning methodology for accurate above-ground biomass estimation from lidar point clouds. Our preliminary results show discrepancies of about 10% with field measurement of above-ground biomass, which is significantly lower than existing segmentation-based methods and FullCAM which yield discrepancies as large as 70%.