Timber cruising data-driven forest biomass estimation model for airborne LiDAR point cloud

Airborne laser scanning (ALS) is a dependable and precise source for aboveground biomass (AGB) mapping. The area-based approach (ABA) utilizes statistical relationships between field-measured AGB and LiDAR-derived parameters to generate geospatial AGB representations at grid-cell scales. However, collecting field-measured data is both labor-intensive and costly, significantly increasing the expenses associated with ALS-based AGB mapping. Timber cruising data records AGB information for treatment sub-compartments before and after management, demonstrating potential as an effective substitute for field plot data. Thus, we introduce a sub-compartment-based approach (SCA), utilizing timber cruising data as an alternative calibration data to field-measured data for AGB estimation models. This approach estimates and maps AGB at the treatment sub-compartment scale, delineated manually and automatically, rather than at the grid-cell scale. To accommodate variations in the collection time and area of timber cruising data, an automated machine learning (AutoML) model was implemented, which supports automated model selection and hyperparameter tuning. Additionally, we performed a comparative analysis with the linear model with ridge regularization (LMR) model. Validation experiments were conducted at multiple scales, the results demonstrate that the SCA, particularly the model optimized with AutoML, consistently delivered high accuracy across at the sub-compartment scale (optimal model: R2 of 0.81, rRMSE of 0.21) and field plot scale (optimal model: R2 of 0.88, rRMSE of 0.26). Furthermore, at the forest farm level, the AGB estimated by SCA for both larch and mixed species closely aligned with those estimated by ABA. However, for forest farm managers, the advantage of adopting the SCA lies in avoiding the additional costs associated with new field plot measurements compared to the ABA model. This research illustrates the feasibility of substituting traditional field plot data with timber cruising data, reducing costs while maintaining precision and efficiency in AGB estimation.