Assessing terrain modelling and forest inventory capabilities of digital aerial photogrammetry from autonomous aerial systems

Autonomous aerial systems-based digital aerial photogrammetry (AAS-DAP) is an emerging technology that has the capacity to generate dense three-dimensional (3D) point clouds similar to airborne laser scanning (ALS). Over forested stands, these point-clouds can be used to model forest attributes using an area-based approach however, model accuracy is dependent on digital elevation model quality used to gather vegetation heights above ground. It is known that canopy occlusion contributes to larger gaps in terrain registration from AAS-DAP compared to ALS point clouds. Due to the recent emergence of AAS-DAP as a cost-effective remote sensing platform, few studies have investigated the terrain modelling and forest inventory capacity of AAS-DAP over complex conifer forests. In Chapter 3, through the use of a sensitivity analysis, I established a set of optimal ground points from AAS-DAP by using commercially provided ALS ground points as reference. This optimal set of ground points was then used to test common terrain surface interpolation routines in Chapter 4. Interpolation routines include inverse-distance weighted, natural neighbour, triangulated irregular network, and spline with tension. Using field-measured tree height and stem diameter, allometric relationships were established for dependent variables: mean tree height (Hmean), Lorey’s height (HLorey) and stem volume per hectare (Vstem). Models were then fit among dependent variables and metrics calculated from the vertical distribution of the AAS-DAP point cloud normalized by the different AAS-DAP terrain surfaces in addition to a reference surface generated from commercially provided ALS ground points. A Kruskal-Wallis with Dunn’s posthoc test found no significant difference between predictions derived from different terrain surfaces for all three dependent variables; however, the inverse-distance-weighted method produced a distribution of predictions most similar to those from the ALS-DEM. The best performing forest attributes models for Hmean, HLorey and Vstem yielded mean root-mean-square errors (RMSE) of 1.19 m (7.29%), 0.92 m (5.04%) and 54.55 m³·ha⁻¹ (26.66%) respectively across the four AAS-DAP terrain surfaces generated. Model performance was higher yet comparable when using the ALS-DEM for point cloud height normalization with RMSE of 0.73 m (4.43%), 0.59 m (3.24%) and 37.31 m³·ha⁻¹ (18.24%).Forestry, Faculty ofGraduat