Spatiotemporal modelling of sea duck abundance: implications for marine spatial planning

TL;DR

This paper presents a flexible, model-based approach for estimating the distribution and abundance of sea ducks, aiding marine spatial planning and impact assessment for offshore wind energy developments in North America.

Contribution

The study introduces a comprehensive modeling framework that combines occupancy and abundance models with covariate effects, enhancing spatial planning for vulnerable marine bird populations.

Findings

Applied the model to sea ducks in Nantucket Sound, Massachusetts.

Demonstrated the approach's ability to incorporate multiple covariates and spatiotemporal trends.

Provided estimates to inform offshore wind energy impact assessments.

Abstract

Effective marine spatial plans require information on the distribution and abundance of biological resources that are potentially vulnerable to anthropogenic change. In North America, spatially-explicit abundance estimates of marine birds are necessary to assess potential impacts from planned offshore wind energy developments (OWED). Sea ducks are particularly relevant in this context as populations of most North American species are below historic levels and European studies suggest OWEDs. We modelled species occupancy using a generalized additive model and conditional abundance with generalized additive models for location, scale, and shape; the models were subsequently combined to estimate unconditional abundance. We demonstrate this flexible, model-based approach using sea ducks (Common Eider [Somateria mollissima], Black Scoter [Melanitta americana], Surf Scoter [M. perspicillata], White-winged Scoter [M. deglandi], and Long-tailed Duck [Clangula hyemalis]) in Nantucket Sound, Massachusetts, USA, which supports some of the largest concentrations of wintering sea ducks in eastern North America and where a 454-MW OWED is proposed. Our approach to species distribution and abundance modelling offers several useful features including (1) the ability to model all conditional distribution parameters as a function of covariates, (2) integrated variable reduction and selection among many covariates, (3) integrated model selection, and (4) the efficient incorporation of smooth effects to capture spatiotemporal trends poorly explained by other covariates. This modelling approach should prove useful for marine spatial planners in siting OWEDs while considering key habitats and areas potentially vulnerable to anthropogenic stressors. Moreover, the approach is equally suitable for terrestrial or aquatic systems.