Assessing object-oriented LiDAR metrics for characterizing bird habitat in a management perspective

Abstract

Light Detection and Ranging (LiDAR) provides detailed information on the three dimensional structure of the environment, and is increasingly used in habitat modeling for a wide variety of species including birds. LiDAR has been shown to improve predictive performance of species distribution models. It is recommended that explanatory variables in habitat models should be meaningful from the species point of view in order to best explain species distribution within a landscape [2]. However, is good predictive performance of a habitat suitability model sufficient to impact local conservation actions? In order to take appropriate and more efficient management decisions, we hypothesize that the metrics explaining the species distribution need to be also meaningful for managers. Some LiDAR point clouds metrics such as the standard-deviation of penetration ratio between 0.5-10m [1] are not easy to interpret. However, metrics extracted using object-oriented methods may fill this gap by giving metrics based on existing landscape components. Instead of calculating metrics over a surface unit (the pixel), an object-based classification group together neighboring points because they belong to the same overall structure which define an object type (tree, road, building, gap). The aim of this study is to improve forest management planning by using LiDAR predictors meaningful for both the species and managers. We are here focusing on the case of the Capercaillie (Tetrao urogallus), an avian species of conservation concern occurring in the French Jura Mountains. Capercaillies favor old mixed forest with a mosaic of structurally different habitats (gap openings, moderate canopy cover area, isolated resting trees, presence of shelters) and the species is threatened by habitat loss and degradation. Habitat restoration planning is thus a fundamental aspect of species conservation actions. We extracted a range of object-oriented metrics from LiDAR datasets, defined with the support of conservation experts and forest managers. We then compare habitat suitability models based on inhomogeneous point process models, such as Maxent, fitted with either commonly used “points cloud” or new “object-oriented” LiDAR metrics. Preliminary results indicate that both categories of metrics yield similarly accurate predictions of Capercaillie habitat suitability. Thus, we hope that the use of object-oriented metrics, with their likely improved interpretability, will allow for more practical recommendations supporting forest management planning in favor of Capercaillie conservation.

[1] Bae, Soyeon, Bjoern Reineking, Michael Ewald, and Joerg Mueller. 2014. “Comparison of Airborne Lidar, Aerial Photography, and Field Surveys to Model the Habitat Suitability of a Cryptic Forest Species–the Hazel Grouse.” International Journal of Remote Sensing 35 (17): 6469–89. [2] Johnson, Chris J, and Michael P Gillingham. 2005.