Modelling Forest Inventory and Biophysical Variables for an Uneven-Aged Forest Using Multi-Source Remotely-Sensed Data
Forest resource inventory (FRI) information is critical to sustainable forest management. Airborne Laser Scanning (ALS) offers a cost-effective option for modelling forest inventory, biophysical and ecological variables over large areas. Given that traditional ALS-based FRIs rely primarily on height data, the objective of this research was to examine the potential of ALS intensity data, multi-seasonal multispectral imagery, and digital aerial photogrammetry (DAP) for enhancing traditional ALS-based FRIs using a combination of non-parametric and parametric modelling techniques. For size class distribution estimation, the results of k-nearest neighbor imputation and random forest regression demonstrated that the combination of ALS height- and intensity-based metrics improved accuracy compared to models based on either type of metric alone. Using a hierarchical variable clustering technique, ALS intensity data were found to carry unique information complementary to passive near-infrared data, despite their similarity in wavelengths. Compared to ALS data alone, the addition of multi-seasonal imagery contributed to more accurate models of basal area and species mixture. In contrast, ALS height- and intensity-based metrics exhibited unparalleled utility for modelling stem density compared to optical imagery. Among the three multispectral sensors examined (i.e., Landsat-5 TM, Sentinel-2A and WorldView-2), Sentinel-2A proved to be the most cost-effective for enhancing ALS-based FRI, owing to its sufficient spatial resolution and inclusion of key spectral bands (i.e., red-edge and shortwave infrared). Compared to ALS, similar functional groups of metrics were found in DAP data, but DAP metrics lacked the capacity for characterizing canopy permeability. Due to the lack of penetrating echoes, gap fraction information was not well represented by DAP, resulting in suboptimal accuracy for LAI estimation compared to ALS. However, a comparison of functional groups between DAP and ALS identified tasks for which DAP is suitable (e.g., volume, forest successional stages, and species mix). Overall, this research demonstrates that ALS-based FRIs can be enhanced by additional sources of input, such as ALS intensity data and multispectral imagery; thereby demonstrating greater potential for more advanced FRIs for Canadian forests.
URI for this recordhttp://hdl.handle.net/1974/24295
Request an alternative formatIf you require this document in an alternate, accessible format, please contact the Queen's Adaptive Technology Centre
The following license files are associated with this item: