Modeling approaches to estimate effective leaf area index from aerial discrete-return LIDAR

Leaf area index (LAI) has traditionally been difficult to estimate accurately at the landscape scale, especially in heterogeneous vegetation with a range in LAI, but remains an important parameter for many ecological models. Several different methods have recently been proposed to estimate LAI using aerial light detection and ranging (LIDAR), but few systematic approaches have been attempted to assess the performance of these methods using a large, independent dataset with a wide range of LAI in a heterogeneous, mixed forest. In this study, four modeling approaches to estimate LAI using aerial discrete-return LIDAR have been compared to 98 separate hemispherical photograph LAI estimates from a heterogeneous mixed forest with a wide range of LAI. Among the four approaches tested, the model based on the Beer–Lambert law with a single parameter (k: extinction coefficient) exhibited highest accuracy (r2 = 0.665) compared with the other models based on allometric relationships. It is shown that the theoretical k value (=0.5) assuming a spherical leaf angle distribution and the zenith angle of vertical beams (=0°) may be adequate to estimate effective LAI of vegetation using LIDAR data. This model was then applied to six 30 m × 30 m plots at differing spatial extents to investigate the relationship between plot size and model accuracy, observing that model accuracy increased with increasing spatial extent, with a maximum r2 of 0.78 at an area of 900 m2. Findings of the present study can provide useful information for selection and application of LIDAR derived LAI models at landscape or other spatial scales of ecological importance.