OOS 4-3
Estimating marine bird abundance in offshore wind development areas: Integrating uncertain species identification in transect surveys

Monday, August 10, 2015: 2:10 PM
316, Baltimore Convention Center
Nathan J. Hostetter, Department of Forestry and Environmental Resources, North Carolina State University, Raleigh, NC
Beth Gardner, Department of Forestry and Environmental Resources, North Carolina State University, Raleigh, NC
Holly F. Goyert, Department of Forestry and Environmental Resources, North Carolina State University, Raleigh, NC
Andrew T. Gilbert, Biodiversity Research Institute, Gorham, ME
Kathryn A. Williams, Biodiversity Research Institute, Portland, ME
Emily E. Connelly, Biodiversity Research Institute, Gorham, ME
Melissa Duron, Biodiversity Research Institute, Gorham, ME
Richard R. Veit, Division of Humanities and Social Sciences, College of Staten Island, Staten Island, NY

Understanding marine bird abundance and distribution in offshore waters is important for marine spatial planning and management of target populations. Line and strip transect surveys are common methods for estimating marine bird abundance and are widely used in many ecological studies. Individuals observed during transect surveys, however, are not always identified to the species level. These partial observations (e.g., individuals identified to genus, but not species) present a major challenge when estimating species specific abundance. We present an approach to predict species specific abundance when species identity is not perfectly observed. We treat species identity as a latent variable, and use spatial covariates and data collected during boat surveys (line transect) to inform abundance estimates in aerial surveys (strip transect). We demonstrate this approach using marine bird data collected during shipboard and aerial transect surveys off the coast of Delaware, Maryland, and Virginia, U.S.A. during 2012-2014. Aerial surveys used high resolution digital video that allowed for large-scale surveys of the region, while shipboard surveys were conducted in a subset of the study area and obtained high rates of species identification.


Species identification rates varied by survey and species, but were generally lower in digital video aerial surveys (<45% of individuals identified to species) than during shipboard surveys (>70% of individuals identified to species). Spatial covariate relationships (distance to shore, sea surface temperature, chlorophyll-a, and salinity) varied across species and survey, but often assisted in predicting species identity. For instance, shipboard surveys identified 96% of loon observations to species (Common Loon or Red-throated Loon), while digital video aerial surveys identified 15% of loon observations to species. Covariate relationships indicated that Common Loons were more likely to be farther offshore relative to Red-throated Loons, which informed species predictions and abundance estimates. The ability of boat survey data to predict species identity was validated by comparing predicted and observed species identity. Overall, integrating spatial covariates often improved species predictions, but success varied by survey. Our results indicate that integrating multiple sources of information can improve species specific abundance estimates in studies where species identity is imperfectly observed.