Methods for LiDAR-based estimation of extensive grassland biomass

Abstract

Abstract Biomass estimation derived from Terrestrial Laser Scanning (TLS) is already an established technique in forestry, whereas TLS measurements are less well investigated for use in grassland ecosystems. Detailed information provided by survey systems can enhance management strategies and support timely measures. Field measurements were made in the “UNESCO biosphere reserve Rhon” in Central Germany with a TLS station (Leica P30). Four methods for estimating biomass from 3d point clouds have been applied to the data, which were Canopy Surface Height (CSH), Sum of Voxel, Mean of 3d-grid Heights, and Convex-Hull. The optimum set of model specific parameters to increase model stability and performance was identified. The methods were compared in terms of model performance and calculation speed. For each method the effect of the number of scans used for each point cloud was assessed. The best fit for fresh biomass determination was achieved with a mean CSH value derived from the top 5% of all CSH values (adj. R2 0.72). In all cases, models for dry biomass estimation had less explanatory power than those for fresh biomass. CSH models based on point clouds, which were merged from two opposite scans, achieved the highest average accuracy both for fresh and dry biomass (adj. R2 0.73 and 0.58 respectively).