Terrestrial lidar reveals new information about habitats provided by large old trees

Large old trees have been described as keystone habitats for several species. However, current research does not fully explain why these species show a preference for such trees. In this study, we combined field observations of birds with terrestrial lidar scans and computational feature-recognition to describe habitats provided by trees at an unprecedented level of detail. We conducted field observations of birds at 62 trees and used parameters including branch angle, branch diameter, branch state (living or dead), and trunk diameter at breast height (DBH) to develop a generalised linear mixed model (GLMM) that could predict which types of branch birds are more likely to visit. We then quantified angles, diameters, and states of 78,006 branch objects representing the complete canopies of 16 trees. By combining these two models we predicted that large trees (>80 cm DBH) contained, on average, 383 m of branches that were highly suitable for birds (i.e., the predicted probability of observing a bird was >= 0.5), which was more than seven times the average length of highly suitable branches provided by medium trees (51-80 cm DBH). Only one of the sampled medium trees contained highly suitable branches. Small trees (<50 cm DBH) contained none. Our analysis provides new knowledge about characteristics that make large old trees disproportionately attractive to birds and presents a novel method of assessment that can apply to other complex habitat structures.