Project topics and materials
Get Complete Project Material File(s) Now! »
Analysis
For analysis, only the last resight of multiple resightings of each individual within any given year was used. If an animal was sighted only once after an absence of four years and then never again, the entry was judged to be erroneous and was removed from analyses. Single sightings of any individuals during the course of a year were carefully inspected alongside previous and subsequent (year) records and the validity of such single resights during a year judged against previous haul-out behaviour (i.e. philopatry to habitual haul-out area, social status of the individual at the time of haul-out etc.). However, the resighting schedule (see above) resulted in comparatively few single resights during a year compared with multiple resights, thereby allowing confidence in correct resighting data. All seals were assumed to age on 15 October, which is the peak adult female haul-out date at Marion Island (Kirkman et al. 2004). Capture-history matrices were constructed using the resighting (recapture) data from the 1983 to 2002 cohorts, up to but not including the commencement of the 2008 breeding season. This allowed 24 years of resighting history for the 1983 cohort and 5 years for the 2002 cohort, effectively doubling the number of cohorts and resighting history timespan over that of the Pistorius et al. (1999b) study.
These capture-history matrices (depicting absence or presence of individuals per year as 0 or 1 respectively, over time) were condensed to 40 sex-specific input files for the 20 cohorts (mimicking the analytical design used by Pistorius et al. [1999b]). These capture-history matrices were used as input files for the software package MARK (White and Burnham 1999), an application for the analysis of marked individuals, used to obtain maximum-likelihood estimates of survival and resight
probability. MARK provides parameter estimates under the essential Cormack-Jolly- Seber (CJS) model and under several models that appear as special cases of this model (Lebreton et al. 1992). As it was impossible to distinguish between mortality and permanent emigration, we imply apparent rather than absolute survival. The two fundamental parameters of these models are:
Φi = the apparent survival probability for all animals between the ith and (i +1)th sample (i = 1, …, k – 1), and ρi = the recapture probability for all animals in the ith sample (i = 2, …, k).
The first step in the mark-recapture analyses involves Goodness-of-Fit (GOF) tests for the CJS model, and we used Program RELEASE to validate the model assumptions. Despite some support for no age dependence in apparent survival (hereafter ‘survival’) of primiparous adult female southern elephant seals from this population (Pistorius et al. 2004), our inclusion of both sex and all age categories in analysis resulted in age-dependence being assumed in this study and as such Test 3.Sm was retained (see Lebreton et al. 1992). In this study time-dependent and agedependent survival could not be differentiated due to time and age intervals being equivalent, as a result of the exclusive use of single cohorts as separate input matrices for MARK.
Five models were considered for each sex. Firstly, a candidate set of 4 models with varying constraints on survival and resighting probability, exactly replicating those used by Pistorius et al. (1999b) were considered for each sex to establish the survivorship schedule. Based on indications from McMahon et al. (2003) and Pistorius et al. (2004, 2008a), an additional age-constrained model describing agedependence in survival up to (and including) earliest age at primiparity (3 yrs), followed by constant survival, was added for females. A fifth model, describing constant survival of males after age six was defined, based on the relative cessation of the secondary growth spurt evident in males between the ages of 4 and 6 (Pistorius et al. 2005) and indications of breeding by some males in this younger age bracket at Marion Island (M.N. Bester unpubl. data). The most parsimonious model out of the set of 4 or 5 models per sex per cohort (with and without the addition of the extra models to be comparable with Pistorius et al. [1999b]), was selected using the small sample corrected Akaike Information Criterion (AICc) (Lebreton et al. 1992; Anderson et al. 1994). AICc weighs the deviance (quality of fit) and the precision (via number of estimable parameters) to select a model that best describes the data (Lebreton et al. 1992). Violation of one or more of the CJS model assumptions, as identified in GOF testing, would require correction for extra-binomial variation using a variance inflation factor (ĉ) by adjusting AICc estimates (QAICc) for the CJS and nested models. Despite over-dispersion in their data, ĉ adjustments to AICc estimates were not performed by Pistorius et al. (1999b). Thus, the model sets (with four models per sex per cohort) exactly comparable with the Pistorius et al. (1999b) procedure were not adjusted for overdispersion (AICc), while the model sets (with five models per sex per cohort) were adjusted for overdispersion (QAICc). In so doing, the relative effects of adjustment and non-adjustment of mark-recapture results, corollary to model assumption violations (see also Appendix 1 in de Little et al. 2007), are provided to illustrate potentially erroneous biological interpretation of survivorship data. In acc with Pistorius et al. (1999b), a major aim of this study was to determine which age categories were most closely related with the population state change, and thus estimates from the simplest model (constant survival and capture) were not selected for estimate outputs.
Age-specific survival estimates
First year survival estimates for both sexes have remained relatively constant and high (~ >50%) for the entire study period (1984-2003), although particularly high survival between 1996 and 1998 for both sexes is evident, following a trough in estimates during 1993 by comparison (Fig. 4.1a). Mean second year survival for both sexes showed a slow increase between 1992 (males = 0.733, females = 0.797) and 1997 (males = 0.869, females = 0.892), followed by a considerable trough through 2001 (both sexes ~ 0.710), recovering slightly before a recent decrease in 2004 (Fig. 4.1b). Third year female survival showed a slight overall increase after 1993, with 1998, 2002 and 2005 identified as higher mortality years, concomitant with the trough in second year survival after 1997 to 2002 (Fig. 4.1c). Lower 4th, 5th and 6th year survival in pubescent males was progressively associated with 2003, 2004 and 2005 (Fig. 4.1d, e, f), while lower 4th, 5th, 6th, 7th and 8th year female survival was progressively associated with 2003 through to 2007 (Fig. 4.1d, e, f, g, h). Adult male survival increasingly fluctuated, especially if model selection replicating Pistorius et al. (1999b) were used for estimates, while adult female survival remained stable (~ 0.800) through to the 14th year of life (Fig. 4.1i, j, k, l, m, n). High mortality in 4th year females during 1993, progressed annually to 14th year survival in 2005, based on estimates from the selection of ‘constant-survival-after-age-3’ models (Table 4.2) from the full candidate set of ĉ-adjusted models (Fig. 4.1d, e, f, g, h, i, j, k, l, m, n). The estimates from the models replicating the Pistorius et al. (1999b) criteria offered varied annual descriptors of survival, particularly for adult females (older than 3 yrs), identifying 1993 and 2003 as high mortality years for females in their 4th and 5th years (Fig. 4.1d, e). Similarly, females in their 6th, 7th, 8th and 9th years experienced high mortality during 2006, based on these model outputs (Fig. 4.1 f, g, h, i).
1. GENERAL INTRODUCTION
Introduction
Southern Elephant Seal Biology
Southern Elephant Seal Distribution
Present Worldwide Population Status
Study Area and Marine Surrounds
Aims and Objectives of this study
Thesis Structure
Literature Cited
2. HOW TO WEIGH AN ELEPHANT SEAL WITH ONE FINGER: A SIMPLE THREE DIMENSIONAL PHOTOGRAMMETRIC APPLICATION
Abstract
Introduction
Methods
Study area
Field techniques
Photogrammetric analyses
Results
Discussion
Literature Cited
3. TEMPORARY MARKING OF UNWEANED SOUTHERN ELEPHANT SEAL PUPS
Abstract
Introduction
Methods
Study area
Data collection
Data analyses
Results
Discussion
Literature Cited
4. IMPROVED SURVIVORSHIP, AND IMMIGRATION, DRIVE A POPULATION STATE CHANGE IN SOUTHERN ELEPHANT SEALS AT MARION ISLAND
Abstract
Introduction
Methods
Study site
Tagging and resighting of seals
Analysis
Results
Discussion
Literature Cited
5. FERTILITY, LONGEVITY AND REPRODUCTIVE SENESCENCE IN FEMALE SOUTHERN ELEPHANT SEALS AT MARION ISLAND
Abstract
Introduction
Methods
Study area and mark-recapture experiment
Longevity schedule
Actuarial senescence
Reproductive senescence
Fertility
Results
Discussion
Literature Cited
6. USING COMPLEX ECOLOGICAL MODELLING SOFTWARE REQUIRES CAREFUL THOUGHT, A THOROUGH UNDERSTANDING OF THE SOFTWARE AND METICULOUS EXPERIMENTAL DESIGN
Abstract
Introduction
Southern Elephant Seal Case Study
Methods
Tagging and resighting of seals
Analysis of tag resightability
Results
Discussion
The Argument
Literature Cited
7. GENERAL CONCLUSION
Synthesis
Literature Cited
