Assessing and Correcting Topographic Effects on Forest Canopy Height Retrieval Using Airborne LiDAR Data

Abstract

Topography affects forest canopy height retrieval based on airborne Light Detection and Ranging (LiDAR) data a lot. This paper proposes a method for correcting deviations caused by topography based on individual tree crown segmentation. The point cloud of an individual tree was extracted according to crown boundaries of isolated individual trees from digital orthophoto maps (DOMs). Normalized canopy height was calculated by subtracting the elevation of centres of gravity from the elevation of point cloud. First, individual tree crown boundaries are obtained by carrying out segmentation on the DOM. Second, point clouds of the individual trees are extracted based on the boundaries. Third, precise DEM is derived from the point cloud which is classified by a multi-scale curvature classification algorithm. Finally, a height weighted correction method is applied to correct the topological effects. The method is applied to LiDAR data acquired in South China, and its effectiveness is tested using 41 field survey plots. The results show that the terrain impacts the canopy height of individual trees in that the downslope side of the tree trunk is elevated and the upslope side is depressed. This further affects the extraction of the location and crown of individual trees. A strong correlation was detected between the slope gradient and the proportions of returns with height differences more than 0.3, 0.5 and 0.8 m in the total returns, with coefficient of determination R2 of 0.83, 0.76, and 0.60 (n = 41), respectively.

Determinants of Aboveground Biomass across an Afromontane Landscape Mosaic in Kenya

“…Additionally, the slope can bias ALS height percentiles [66], which were used in the final AGB model. This could increase AGB predictions in the steep slopes with an effect on the BRT analysis (Section 4.2), and future studies should consider applying a slope correction [67] before computing ALS height metrics. Nonetheless, there was no sign of systematic overestimation in the steep slopes when assessing the model residuals against the slope.…”

Airborne LiDAR Remote Sensing for Individual Tree Forest Inventory Using Trunk Detection-Aided Mean Shift Clustering Techniques

“…In addition to the indicators of "recall" (r) and "precision" (p), the "F-score" (F = 2 × (r × p)/(r + p)) is used in the comparison to evaluate the overall detection accuracy. Except the adaptive mean shift-based clustering algorithm presented in [26], the approaches chosen for this accuracy comparison include: a region growing approach presented in [2], a k-means clustering approach presented in [19], a mean shift-based approach described in [15] and a CHM-based approach adopted in [40]. For more detailed descriptions about these individual tree detection approaches, please refer to [26].…”

Adaptive Mean Shift-Based Identification of Individual Trees Using Airborne LiDAR Data

“…The overall detection performance (namely the indicators of "recall" and "precision") of our proposed approach is compared to other point cloud-based approaches, as shown in Table 5. The CHM-based method (multi-resolution segmentation using watersheds) adopted in [32] is also implemented in this comparative experiment to provide a reference. From the data listed in Tables 4 and 5, it is visible that the proposed adaptive mean shift-based method can produce greater values for the overall detection indicators ("recall" and "precision") than other methods, which indicates that its detection performance outperforms the CHM-based and point-based approaches reported in representative academic literature.…”