Evaluating the Potential of Digital Aerial Photogrammetry (DAP) versus Airborne Laser Scanning (ALS) for Individual Tree Detection and Segmentation in Subtropical Chir Pine ( Pinus Roxburghii ) Forests

Individual tree detection (ITD) algorithms have often relied on airborne laser scanning (ALS) data for delineating trees. Digital Aerial Photogrammetry (DAP) has emerged as a viable alternative to ALS, leveraging sophisticated image-matching algorithms for 3D point cloud generation and subsequent ITD. However, so far, few studies have compared ITD results between ALS and DAP 3D data. We present a detailed comparison of five ITD algorithms using both ALS and DAP data in a subtropical Chir Pine forests, Pakistan. Our analysis, which included 284 field-measured trees, assessed two categories of ITD algorithms: those applied to raster-based Canopy Height Models (CHMs) and those which are directly applied to the point clouds. We evaluated work-flows using fixed window size (FWS) and variable window size (VWS) as well as, unsmoothed and smoothed CHMs generated from ALS and DAP data. Among window sizes, 3 × 3 FWS and 2 × 2 FWS performed best, yielding F Scores of 0.66 and 0.63 using unsmoothed CHMs from ALS and DAP data, respectively. Among point cloud methods, mean shift algorithms consistently outperformed others, achieving F Scores of 0.67 and 0.61 with ALS and DAP data, respectively. The Dalponte2016 algorithm exhibited superior performance in crown segmentation, consistently producing crown radii within 0.5 m of the reference field measured crowns, for ALS data and under 0.6 m for DAP data. Overall, both ALS and DAP achieved comparable results; however, ALS data yielded slightly higher F scores in tree matching and exhibited a stronger correlation with field data compared to DAP. Our findings suggest that DAP-derived point clouds, when normalized by precise DTMs such as those obtained from ALS data, can be effectively utilized for ITD in Chir Pine forests, offering compatibility comparable to ALS data.