Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Predictive models of marine bird distribution as a spatial planning tool: a case study in the U.S. Mid-Atlantic

Kinlan, B.P., Rankin, R.W. and Caldow, C. (2013) Predictive models of marine bird distribution as a spatial planning tool: a case study in the U.S. Mid-Atlantic. In: Pacific Seabird Group 40th Annual Meeting, 20 - 24 February, Portland, Oregan


Marine birds are highly mobile organisms that range widely and respond to dynamic features in their physical and biological environment at time scales from minutes to decades. Marine spatial planning can help minimize risks to marine birds from new offshore ocean uses such as renewable energy facilities, but requires high-quality spatial information. Maps are needed that characterize the persistent spatial features of at-sea marine bird occurrence probability and relative abundance, as well as uncertainties arising from incomplete sampling and inherent variability. Yet, moving from scattered, heterogeneous at-sea survey transect data to gap-free, high-resolution distribution maps at the relatively fine spatial scales (~1km horizontal resolution or better) often needed for marine spatial planning is a formidable statistical challenge. We present a case-study from the U.S. Mid-Atlantic in which we combine >30 years of at-sea survey data from 32 datasets with a large database of high-resolution oceanographic and environmental predictor variables to predict marine bird occurrence and abundance. An ensemble machine-learning technique known as component-wise boosting was adapted to account for complex interactions and non-linearities, spatial correlation, temporal effects and zero-inflation. We use this model to develop predictions of individual species and functional group occurrence and abundance, as well as community metrics such as species richness and diversity. We show how model outputs can be tailored to maximize their use to planners, decision-makers, and stakeholders in marine spatial planning processes. Finally, we give real-world examples of how modeling products have been used, and are anticipated to be used, to reduce conflicts between marine bird habitat and ocean energy facilities in the U.S. Atlantic.

Item Type: Conference Paper
Item Control Page Item Control Page