Evaluating forest aboveground biomass estimation model using simulated ALS point cloud from an individual-based...

Area of Science:

• Forestry and Remote Sensing
• Ecological Modeling
• Geospatial Analysis

Background:

• Area-based approaches (ABA) using airborne laser scanning (ALS) are common for estimating forest aboveground biomass (AGB).
• Model generalization and variable selection for ABA-AGB estimation face scalability challenges across diverse forest types and regions.
• Accurate reference AGB data is often lacking, hindering the assessment of variable selection's influence.

Purpose of the Study:

• To evaluate the transferability of ABA-based AGB estimation models using computer simulations.
• To assess the generalization of ALS-derived variables for AGB estimation across different forest types.
• To identify crucial ALS-derived variables for robust AGB estimation.

Main Methods:

• Coupled FORMIND (individual-based forest growth model) and LESS (3D radiative transfer model) for generating simulated ALS data.
• Developed AGB estimation models using the random forest algorithm on virtual and real-world forest data.
• Assessed model transferability via cross-comparison and validated with field plots and ALS data.

Main Results:

• Canopy surface area (α-shape) and Weibull distribution parameters are identified as vital variables for ALS-based AGB estimation.
• Incorporating these variables significantly improved AGB estimation accuracy across various forest types and regions.
• The optimized model achieved high performance: R² = 0.945, RMSE = 34.22 t/ha, MAE = 20.53 t/ha.

Conclusions:

• Computer simulations offer a valuable tool for exploring challenges in ABA-AGB estimation and refining variable selection.
• The study deepens understanding of the relationship between forest vertical structure metrics and AGB.
• Identified key variables enhance the scalability and accuracy of ALS-based AGB estimation models.