Integrating habitat and partial survey data to estimate the regional population of a globally declining seabird species, the sooty shearwater

Tyler J Clark,Jason Matthiopoulos,Anne-Sophie Bonnet-Lebrun,Letizia Campioni,Paulo Catry,Ilaria Marengo,Sally Poncet,Ewan Wakefield

doi:10.1016/j.gecco.2019.e00554

Abstract

Many animal populations are thought to be in flux due to anthropogenic impacts. However, censusing organisms to understand such changes is often impractical. For example, while it is thought that over half of pelagic seabird populations are declining, most breed in burrows or on cliffs, in large, remote colonies, making them difficult to count. Burrow-nesting sooty shearwaters (Ardenna grisea) are abundant but declining in their core (South Pacific) breeding range, potentially due to introduced rodents and habitat loss. In contrast, Kidney Island, their largest colony in the Falkland Islands (Southwest Atlantic), purportedly grew by several orders of magnitude since the mid-1900s. This island is rodent-free, and native tussac grass (Poa flabellata) has increased following cessation of historical exploitation. To estimate the sooty shearwater population in the Falkland Islands, and its relationship with breeding habitat availability, we sampled burrow density and occupancy on Kidney Island and modeled these as functions of habitat. Both indices responded positively to a proxy for historical increases in tussac cover. We estimate that breeding sooty shearwaters occupy ∼140,000 (95% CI: 90,000–210,000) burrows on Kidney Island. Moreover, using additional survey data and Generalized Functional Response models to account for intra-island variation in habitat availability, we estimate that 25,000 (95% CI: 20,100 - 30,500) burrows could be occupied on nearby islands from which non-native rodents have been recently eradicated. Our study shows that habitat selection functions, generalized where necessary, not only improve population estimates but provide biological insights needed to reverse declines in seabirds and other species.

Highlights

Many animal populations are thought to be in decline due to human activities (Halpern et al, 2008) while others are growing following the alleviation of historical impacts (Buxton et al, 2014)
Model I suggests a positive relationship between burrow density and slope, elevation, tussac cover, and change in tussac density; and a negative relationship with distance from the coast (Table 2)
Tussac cover and tussac height interacted synergistically to result in a decreased burrow density (Table 2; Fig. A1)

Summary

Introduction

Many animal populations are thought to be in decline due to human activities (Halpern et al, 2008) while others are growing following the alleviation of historical impacts (Buxton et al, 2014). Simple “design-based” estimation assumes that population size can be estimated by multiplying the density in a subset of randomly placed sample plots by the area occupied. The latter is often defined based on assumptions about what constitutes suitable habitat. By modeling dependencies of population size on resource or habitat availability, for example using habitat selection functions (HSFs), more accurate “model-based” estimates can be made (e.g., Scott et al, 2009). Generalized Functional Response models (GFRs), which account for differences in animals’ responses to changes in habitat availability, can be used to overcome this (Matthiopoulos et al, 2011) but GFRs have not yet been applied widely to the problem of population size estimation

Objectives

Methods

Results

Conclusion