研究目的
To test if terrestrial laser scanning, as a non-destructive method, can be used to estimate extensive grassland biomass, specifically comparing four analytical approaches for calculating biomass from TLS point cloud data.
研究成果
Canopy Surface Height was the most accurate method for estimating both fresh and dry biomass, with two scans from opposite directions providing the best performance. TLS shows promise for non-destructive biomass estimation in grasslands, but further research is needed for practical implementation, including combining height and spectral information.
研究不足
Models for dry biomass had less explanatory power than fresh biomass, with up to 50% variation unexplained. The study was limited to small plot scales and heterogeneous grassland vegetation, which may affect generalizability. Issues with DEM creation led to exclusion of some samples. The methods may not capture enough information for accurate dry biomass estimation, and sensor fusion with spectral data is suggested for improvement.
1：Experimental Design and Method Selection:
The study used TLS to collect 3D point cloud data in grassland ecosystems, applying four methods (Canopy Surface Height, Sum of Voxel, Mean of 3d-grid Heights, and Convex-Hull) for biomass estimation. Linear regression models were developed to correlate point cloud-derived variables with destructive biomass measurements.
2：Sample Selection and Data Sources:
Data was collected from two sites in the UNESCO biosphere reserve Rh?n, Germany, with three plots each of 8x8 meters, sampled on three dates in 2016. Sub-plots of 1 m2 were used for destructive biomass sampling.
3：Sub-plots of 1 m2 were used for destructive biomass sampling.
List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: TLS station (Leica P30), tripod, highly reflective targets, scissors for cutting biomass, drying oven for dry biomass determination, software (Cyclone 9.1, R-Studio with specific packages).
4：1, R-Studio with specific packages).
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: TLS scans were performed from four sides of each plot, merged into point clouds, and processed. Biomass was destructively sampled by cutting at 5 cm height, dried, and weighed. Point clouds were analyzed using the four methods with varying parameters.
5：Data Analysis Methods:
Linear regression models were developed in R, with adjusted R2 and rRMSE used for evaluation. Calculation times were measured using the microbenchmark function.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
