Remotely sensed forest structural complexity predicts multi species occurrence at the landscape scale

Abstract

Along with plant species composition forest structural complexity is an important determinant of forest biodiversity, but difficult to predict in space from field data. We analyzed forest structural complexity based on a comprehensive set of variables derived from nationally available, area-wide remote sensing, particularly LiDAR data. We generated variables related to vertical and horizontal structural heterogeneity, as well as site factors potentially indicating the abundance of weakened trees or snags. We used them to predict the occurrence of four bird species with narrow and complementary structural habitat requirements, together being indicative of structurally diverse forests. Presence/absence data of Capercaillie (Tetrao urogallus), Hazel Grouse (Bonasa bonasia), Three-toed Woodpecker (Picoides tridactylus) and Pygmy Owl (Glaucidium passerinum) from three biogeographic mountain regions in Switzerland were used to calibrate species distribution models (boosted regression trees BRT) for each species individually, as well as for the sympatric occurrence of at least three of the four target species. The predictive deviances explained (D2) and the AUC values obtained from cross-validation ranged from 15.5% to 63.1% and 0.77% to 0.97% respectively. Sympatric species occurrence reflecting overall forest structural complexity was predicted best, with an outstanding accuracy. To support management and monitoring schemes we identified variable threshold effects based on partial dependence plots. Variables related to vertical foliage distributions were most important, followed by horizontal structural attributes such as canopy height variations, forest edges and gaps. Site factors such as topographic position improved all models and were most important for the species depending on weakened trees and dead wood. We conclude that recent advances in remote sensing allow for large-scale determination of forest structural characteristics suitable for developing species and habitat distribution models of considerable generality, while keeping an unprecedented level of detail. Our approach allows forest managers to amend regional and countrywide management plans with reliable maps depicting areas of high forest structural complexity and habitat quality, which will facilitate the integration of conservation-relevant information into multifunctional forestry.