Presentation Abstract

Title: Modeling marbled murrelet habitat using LiDAR-derived canopy data.
Session Title: Conservation and Management of Birds
Session Number: 67
Session Time: Wednesday, Oct 17, 2012, 8:30 AM -12:20 PM
Presentation Time: Wednesday, Oct 17, 2012, 10:40 AM -11:00 AM
Author(s): Joan C. Hagar1, Bianca N. I. Eskelson2, Patti Haggerty1, S. Kim Nelson2, David G. Vesely3, 1USGS, Corvallis, OR, 2Oregon State University, Corvallis, OR, 3Oregon Wildlife Institute, Corvallis, OR, Contact: joan_hagar@usgs.gov
Abstract Body: LiDAR can substantially improve the accuracy and precision of characterizations of habitat relationships for wildlife species known to be responsive to forest structure by providing fine-scale vegetation data that are inaccessible by both ground plot and other remote sensing methods. We used LiDAR data to determine occupancy probability for the federally threatened marbled murrelet in the Oregon Coast Range. Our goal was to provide a predictive tool for developing better occupancy maps for threatened and endangered species and to guide habitat management. Our objective was to identify LiDAR-derived variables that can be used as reliable predictors of murrelet occupancy at the stand-level spatial scale, and nesting habitat at finer spatial resolution. We used murrelet occupancy and nest location data collected following Pacific Seabird Group protocol, and canopy metrics calculated from discrete return airborne LiDAR data using the FUSION software package, to fit a logistic regression model predicting the probability of occupancy. Our final model for stand-level occupancy included distance to coast, stand size, and four LiDAR-derived variables (SD of the 99th percentile of returns, the mean of the average elevation of returns, the mean of the kurtosis of the elevation of returns, and the maximum canopy surface ratio derived from the LiDAR highest hits raster). With an area under the curve value (AUC) of 0.82 this model had excellent discrimination, but only moderate agreement when evaluated with Cohen’s κ (κ = 0.52). Our model provided better discrimination between occupied and unoccupied stands than a model using variables derived from Gradient Nearest Neighbor maps that were previously reported as important predictors of murrelet occupancy (AUC = 0.72, κ = 0.33). Our model can be used to identify sites that have high probability of occupancy by murrelets based on the fine-scale canopy characteristics to which this species is known to respond.



Technical Support Help Desk
Phone: 217-398-1792
Email: Help Desk