Skip to main content

    SEASONAL ESTIMATION OF ABUNDANCE BY BOOTSTRAPPING INEXACT RESEARCH DATA (SEABIRD): A METHOD FOR ASSESSING ABUNDANCE AND UNCERTAINTY FROM HISTORICAL COUNT DATA USING ADELIE PENGUINS AS A CASE STUDY

    Request Meeting Document
    Document Number:
    WG-EMM-PSW-08/11
    Author(s):
    J.P. McKinlay and C.J. Southwell (Australia)
    Abstract

    In addition to a review of published studies on Adelie penguin abundance at breeding sites in Antarctica, Southwell (2004) proposed a general abundance estimator appropriate for a species of that kind. The present study attempts to implement this estimator in the form of a parametric bootstrap model, utilising as input data published counts of Adelie penguins and estimates of their uncertainty at breeding sites in Antarctica. To achieve this task, a menu-driven suite of routines titled SEABIRD (Seasonal Estimation of Abundance by Bootstrapping Inexact Research Data) has been developed in the R language for statistical computing (R Development Core Team 2008). Software is reliant on data being presented in a specified format congruent with CCAMLR databases designed to store historical survey data relating to penguin abundance. Usual sampling methodology considerations reported in work of this kind are accommodated, such as availability and perception bias, as well as many of the vagaries associated with combining diverse data collected in a variety of ways over many decades. Of particular concern was to ensure that different types of counts (eg. nests, chicks or adults), perhaps made at different time points in a breeding season, might usefully be combined in order to obtain regional-scale estimates of abundance. This was achieved by using independent estimates of availability throughout a breeding season collected at a few, frequently sampled sites, in order to standardise historical estimates to a common reference point of breeding chronology. Equally important was the idea that estimates of uncertainty associated with historical counts be faithfully incorporated and preserved, and methods have been developed to allow these to be combined or interpreted in several different ways. In order to help understand how these and other elements of the procedure contribute toward estimates of uncertainty when combining data, as far as practical different components of the estimation process can be switched on or off to assess their effect. Confidence intervals for final estimates at different scales of spatial aggregation are determined by examining the bootstrap distribution of population estimates at selected percentile intervals. While tailored for Adelie penguins, the method and implementation is sufficiently general to potentially be adapted for other Antarctic species showing seasonal variation in availability to sampling methodology. At the time of writing, SEABIRD is well developed but still very much an evolving work. It is anticipated that use of the software and discussion of the estimation issues involved will identify possibilities for improvement.