Validation of UAV-LiDAR–Derived Effective Leaf Area Index (eLAI) in a Conifer Forest: Comparison with High-Density Litter-Trap Measurements

Leaf area index (LAI) is a key variable in environmental and ecological research and is widely used in many models. Although destructive sampling can provide more direct LAI estimates, it is extremely labor-intensive and typically limits observations to the individual-tree scale. Litter-trap measurements enable stand-scale LAI estimation, yet applications remain rare in conifer forests, where evergreen species dominate. Recent advances in UAV-LiDAR have enabled the estimation of effective LAI (eLAI), which does not account for leaf clumping, over broad areas at high spatial resolution. In conifer forests, comparisons of eLAI between instruments (e.g., UAV-LiDAR versus LAI-2200) have been reported; however, studies validating UAV-LiDAR-derived eLAI against ground-measured LAI are still very limited. Consequently, it remains unclear to what extent UAV-LiDAR eLAI represents true LAI at the stand scale.

In this study, we conducted intensive litter-trap sampling in a deciduous conifer plantation of Dahurian larch (Larix gmelinii) and compared ground-based LAI with UAV-LiDAR-derived eLAI. The study was carried out in Mikasa, Hokkaido, Japan. We deployed 100 litter traps (1 m × 1 m) in a grid to collect needle litter within a 10 m × 10 m plot, thereby deriving a ground-reference LAI and its spatial variability. Concurrently, we conducted a UAV-LiDAR survey to validate LiDAR-based eLAI and to assess the importance of key parameters and processing settings used in the gap-fraction approach.

UAV-LiDAR point clouds were processed using the R package lidR. eLAI (without clumping correction) was computed from the Beer–Lambert relationship based on gap fraction. To evaluate parameter sensitivity, we systematically varied the extinction coefficient (k), the minimum gap-fraction threshold (P_gap), scan-angle correction, and a minimum height threshold for including first returns. These settings were altered stepwise to generate 144 parameter combinations, and the resulting eLAI estimates were compared with litter-trap-based LAI. The relationship between eLAI and LAI was most strongly affected by k and P_gap, whereas the other settings had minor effects within the parameter ranges evaluated for this plot. Overall, agreement between UAV-LiDAR-derived eLAI and ground reference LAI was low, with correlation coefficients ranging from 0.02 to 0.21 across all parameter combinations. The mean measured LAI in the plot was 2.04 m² m⁻², whereas LiDAR-based eLAI was substantially higher (6.12–14.08 m² m⁻²).

These results indicate that UAV-LiDAR-derived eLAI can markedly overestimate LAI unless woody contributions are removed and clumping is explicitly corrected. In particular, k and P_gap critically influence estimation accuracy, highlighting the need for careful calibration and species-/site-specific parameter selection. Our findings caution that using UAV-LiDAR eLAI directly as LAI in conifer-forest studies may lead to substantial bias and should be avoided without appropriate corrections.