Scaling effects are a central problem in ecology. Individuals, species, communities, landscapes, and biomes are affected by environmental parameters operating at different scales. As landscape-level ecological studies rely on remote sensing, data resolution determines what scientists see and how they interpret the signal. One remote sensing technique that has become popular with ecologists is LiDAR (light detection and ranging) which maps 3-dimensional vegetation structure at fine scales. This study explores how aggregating LiDAR data at different pixel sizes (5x5, 10x10, 25x25, 50x50, 100x100 m, etc.) impacts derived structural metrics of an architecturally complex forest.
In 2013, 1x1 m-resolution LiDAR point cloud data were collected in the Chiquibul Forest, a contiguous, a 107,000 ha minimally-disturbed Belizean neotropical forest with varying topographic relief. 30 1x1 km study blocks were segregated into areas of high and low topographic relief based on standard deviation of elevation within each block. LiDAR-derived forest metrics were quantified at different pixel resolutions using the USDA LiDAR processing software FUSION. Average values were calculated for maximum and mean canopy height, rugosity, vertical diversity, and canopy closure over the different pixel resolutions within each 1x1 km block. Linear mixed models were used to assess statistically significant differences due to scaling.
Results/Conclusions
Preliminary results show large differences in maximum canopy height in both high and low relief regions. At the 5x5 m pixel size, average maximum canopy height was 24.3 m and 22.6 m at high and low relief sites respectively. Average maximum canopy height increased with pixel size, and at the 100x100 m pixel was 35.7 m and 33.1 m at high and low relief respectively. There were large differences in average vertical diversity (using the Jost index) depending on topography (5.6 high relief, 6.1 low relief at 5x5 m pixel vs. 8.8 high relief, 8.9 low relief at 100x100 m pixel). Mean canopy height, average rugosity, and average canopy closure measures showed less variation based on pixel size.
Results from this study will aid ecologists who use LiDAR to account for scaling effects. Larger pixels are better able to capture emergent canopy trees within neotropical forests. Since these tall trees exist in low densities, their signal is often drowned out at smaller pixel sizes which are dominated by shorter canopy trees. Larger pixels also captured higher vertical diversity within these forests. Pixel size does not seem to affect mean canopy height, average rugosity, or average canopy closure over a landscape.