SPATIAL ANALYSIS OF AIRBORNE LASER SCANNING POINT CLOUDS FOR PREDICTING FOREST STRUCTURE

The arrangement of trees with respect to each other plays a role in various forestry decisions. In this study, the arrangement of trees was summarized by three different structure indices. Their values were determined from field measurements and predicted with the well-known k-nn estimation method using data obtained by airborne laser scanning (ALS). ALS-derived predictions are often assisted by vertical summaries of the pulse returns. Our goal was to identify spatial summaries of the ALS point cloud that can improve predictions based on commonly used feature metrics. We explored the horizontal distribution of the pulse returns through canopy height models thresholded at different height levels. We introduce completely new metrics based on 1) the Euler number, which is the number of patches of vegetation minus the number of gaps, and 2) the empty-space function, which is a spatial summary function of the gap space. Data from a study site in Central Finland was available with circular field plots with 9 m radius. We find that small sample plots can be challenging. Still, we present evidence that the use of spatial feature metrics improves the prediction of forest structure indices and has potential for improvements for other forest variables related to gap structures.