Performance, effectiveness and computational efficiency of powerline extraction methods for quantifying ecosystem structure from light detection and ranging

National and regional data products of the ecosystem structure derived from airborne laser scanning (ALS) surveys with Light Detection And Ranging (LiDAR) technology are essential for ecology, biodiversity, monitoring. However, noises like powerlines often remain, hindering accurate measurement 3D structures LiDAR. Currently, there is a lack studies assessing powerline noise removal in context generating ALS point clouds. Here, we assessed (1) performance accuracy, (2) effectiveness, (3) time efficiency execution three extraction methods (i.e. two point-based based on deep learning eigenvalue decomposition, respectively, one hybrid method) removing when quantifying landscapes varying canopy heights vegetation openness. Twenty-five LiDAR metrics representing key dimensions height, cover, vertical variability) across 10 study areas Netherlands were used our assessment. The method had best showed highest accuracy various landscape types (average F1 score = 96%), closely followed by 95%). In contrast, decomposition was lower 82%) depended composition (e.g. decreased 96% to 63% average height increased landscapes). Powerline effectiveness closest those manually labeled ground truth data) capturing cover low- high-vegetation layers. Time (processed points per second) method, yet reduced > 50% compared (ranging 20% 89% different composition). Based findings, recommend upscaling multi-terabyte datasets or national extent because its high computational efficiency. Remaining misclassifications could be further minimized improving training dataset models including shapes transmission towers datasets). Our findings provide novel insights into methods, how their varies mapping LiDAR, extent.