UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An ecosystem study of the Prince Edward Archipelago (Southern Ocean) Gurney, Leigh Josephine 2013

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2014_spring_gurney_leigh.pdf [ 22.96MB ]
Metadata
JSON: 24-1.0165739.json
JSON-LD: 24-1.0165739-ld.json
RDF/XML (Pretty): 24-1.0165739-rdf.xml
RDF/JSON: 24-1.0165739-rdf.json
Turtle: 24-1.0165739-turtle.txt
N-Triples: 24-1.0165739-rdf-ntriples.txt
Original Record: 24-1.0165739-source.json
Full Text
24-1.0165739-fulltext.txt
Citation
24-1.0165739.ris

Full Text

    AN ECOSYSTEM STUDY OF THE PRINCE EDWARD ARCHIPELAGO  (SOUTHERN OCEAN) by LEIGH JOSEPHINE GURNEY  BSc, University of Cape Town, 1994 BSc, Honours Rhodes University, 1995 MSc, Rhodes University, 2000    A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Oceanography)    THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) December 2013     ? Leigh Josephine Gurney, 2013   Abstract This study brings together the wealth of data available for the Prince Edward Islands marine ecosystem and consolidates it into a network, mass-balanced model (using Ecopath). Biomass estimates for the land based top predators show penguins dominate the system for all three time periods assessed (1960s, 1980s and 2000s). The islands appear to have a carrying capacity which may be declining. A consumption model shows a change in prey for the land based top predators from one in which both crustaceans and myctophid fish were of equal importance in the 1960s, to one dominated by myctophids for the 2000s period. The contribution of the sources of primary production were assessed through the ecosystem model with open ocean productivity dominating at all but the smallest scale (shelf region), where the macrophyte production was important. The model describes the marine ecosystem for each of the above mentioned time periods at the scale of the Exclusive Economic Zone and, when compared to other subantarctic and Antarctic systems for which there are ecosystem models, the system was most similar to the neighbouring Kerguelen Islands. An investigation into the ecosystem boundary size was conducted, with all constituents able to satisfy their energetic requirements if considered at the scale of the EEZ. Using the dynamic temporal simulation approach (Ecosim), the model was able to successfully hindcast three past events: the fur seal exploitation, Patagonian toothfish fishery, and the effect of cat predation on small flying birds. In each instance the model performed well for the directly impacted groups. Potential ecosystem effects of climate change were explored through simulations of increasing and decreasing productivity. No single scenario was able to replicate observed patterns and a suite of drivers needs to be considered to reproduce observed patterns. The inclusion of energetic density of prey led to improvements in consumption rate estimates for the static models and should be incorporated into estimates to improve ecosystem model parameterization. The work constitutes the first ecosystem model for the PEIs that can be used as a tool for an ecosystem approach to marine resource management.    ii  Preface This work brings together two parts of my earlier professional development by linking field and research experience with the South African National Antarctic Programme, (1995 ? 2000) with experience gained through my work at the Tasmanian Aquaculture and Fisheries Institute (2001 ? 2004) and combining them in the construction of an ecosystem model, developed at the UBC Fisheries Centre. I was responsible for identifying and developing the research idea in consultation with my supervisor, Dr E. A. Pakhomov. Guidance through the program was provided by my supervisory committee. The work itself was completed solely by me, with advice and guidance in analysis and interpretation provided by my committee members in accordance with their area of expertise. Chapter 4 of this dissertation has been published as L.J. Gurney, E.P. Pakhomov and B.P.V. Hunt, (2011). Life-support system of the Prince Edward Archipelago: overview of local and advected resources. In: First symposium on The Kerguelen Plateau: Marine Ecosystem and Fisheries (Duhamel G. and Welsford D., eds), pp. 217-232. Paris: SFI. Initial formulation of the links in the food web were discussed by all three authors. The work itself was completed and written by me, with reviewing and editing contributions by the two co-authors. No other sections of the thesis have yet been published, though 4 sections have been presented at conferences (Chapter 5, Fisheries and Marine Ecosystems, Canada, 2006; Chapter 6, SCAR Symposium Russia, 2008; Chapter 7 Advances in Marine Ecosystem Modelling Symposium (AMEMR), Plymouth 2011; and Chapter 8, SCAR Symposium 2010, Argentina).  This work comprises a series of chapters tackling different aspects of the Prince Edward Islands marine ecosystem. The work did not follow a linear progression and in compiling the thesis, the structure has a number of sections which need to be clearly outlined.  Chapter 1 includes an introduction to the ecosystem approach to marine resources and ecosystem modelling which is followed by an overview of the study area: the Prince Edward Islands (PEIs) (Southern Ocean). Chapter 2 presents a full review of the PEI marine ecosystem and quantifies the system in terms of biomass estimates, for three time periods for which there are data (1960s, 1980s, 2000s), and illustrates the long term trends of seals and seabirds. In Chapter 3, the iii  biomass estimates from Chapter 2 are used in combination with local dietary data and a consumption model assesses the changes in consumption of broad prey categories for the land based top predators over the same three time periods (1960s, 1980s and 2000s). In Chapters 2 and 3, part of the data compilation that was necessary to construct the ecosystem models for the marine component of the PEIs was prepared, and forms the basis of the data behind all subsequent chapters.  The following three chapters (4,5 and 6) are static mass balanced models of the marine PEI ecosystem used to address different ecological questions. Chapter 4 is an assessment of the local and advected resources at the PEIs and was done in preparation for a workshop which was held in France in 2010, and subsequently published in the proceedings of the workshop. The model is unique, in that it was a simplified (collapsed) early version of the full model (has 21 functional groups), representing only one time period (1980s) and was focused on the primary producers at the islands. Data input for this model were slightly different from that used in all subsequent chapters. Although there is overlap with other sections of this thesis, the manuscript has been submitted here unaltered in text. The figures and tables have, however, been renumbered to fit in with the full document?s structure.  The updated version of the PEI ecosystem model at the scale of the Exclusive Economic Zone (EEZ) includes a higher resolution of the model in terms of functional groups (37) as well as improvements in model parameterization across 3 time periods (1960s, 1980s and 2000s). This model is described in Chapter 5, and used in Chapter 6 to address the question of ecosystem boundary size of the land-based top predators based at the islands by investigating the spatial extent of these groups in the system through the resolution of their energetic requirements.  An improvement to the parameterization of the final model version involved important changes to the consumption estimates, and this work is presented as a standalone chapter, Chapter 7. The work focused on incorporating local energetic values of the prey when estimating consumption and helped to resolve model balancing issues that had proved challenging.  The ability to explore management options is often the goal of developing an ecosystem model. Before exploring future policy scenarios, it is useful to test if the model is capable of hind-iv  casting, with any degree of success, past known events for which there are data. It can also provide a useful tool for exploring future potential scenarios when developing risk assessments, associated management options and optimizing outcomes according to policy aims and objectives. With this in mind, Chapter 8 illustrates the models capability of hindcasting three known events (culling of fur seals, the Patagonian toothfish fishery and a demonstration of the deleterious effect of the cat infestation on the local small bird population). Chapter 9 demonstrates scenario testing of future potential climate change events at the islands.  A short summary of conclusions are presented in Chapter 10 to complete the work.   v  Table of Contents Abstract ........................................................................................................................................... ii Preface............................................................................................................................................ iii Table of Contents ........................................................................................................................... vi List of Tables .................................................................................................................................. x List of Figures .............................................................................................................................. xiii Acknowledgements ...................................................................................................................... xix Dedication ................................................................................................................................... xxii Chapter 1 Introduction .................................................................................................................... 1 1.1 Ecosystem approach to management of marine resources ................................................... 1 1.2. Ecosystem models ................................................................................................................ 2 1.3. The Prince Edward Archipelago .......................................................................................... 7 1.3.1. Background and history ................................................................................................ 7 1.3.2. Geology and geography ................................................................................................ 8 1.3.3. Marine research ............................................................................................................. 9 1.4. Aims of the project............................................................................................................. 13 Chapter 2 Review of the Prince Edward Island marine ecosystem: Quantifying the system in terms of biomass and highlighting long term trends ..................................................................... 16 Introduction ............................................................................................................................... 16 2.1. The marine ecosystem: A review and biomass estimates .................................................. 17 2.1.1. Primary producers ....................................................................................................... 17 2.1.2. Zooplankton ................................................................................................................ 21 2.1.3. Benthic ecosystem ...................................................................................................... 22 2.1.4. Cephalopods ................................................................................................................ 24 2.1.5. Fish fauna .................................................................................................................... 25 2.1.6. Avian fauna ................................................................................................................. 31 2.1.7. Mammals..................................................................................................................... 35 2.2. ?Carrying capacity? for the islands? ................................................................................... 39 vi  Chapter 3 Consumption estimates of marine resources by top predators at the Prince Edward Islands ........................................................................................................................................... 52 3.1. Introduction ........................................................................................................................ 52 3.2. Methods.............................................................................................................................. 53 3.3 Results ................................................................................................................................. 55 3.3.1. Assimilation efficiencies ............................................................................................. 56 3.3.2. Energetic content of prey ............................................................................................ 56 3.4. Discussion .......................................................................................................................... 58 3.5. Conclusions ........................................................................................................................ 61 Chapter 4 Quantifying local and advected resources: Relative importance of the producers in the system at different spatial scales ................................................................................................... 66 4.1. Introduction ........................................................................................................................ 66 4.2. Methods.............................................................................................................................. 68 4.3. Results ................................................................................................................................ 75 4.4. Discussion .......................................................................................................................... 77 4.5. Conclusions ........................................................................................................................ 80 Chapter 5 An ecosystem model of the Prince Edward Island Archipelago .................................. 87 5.1. Introduction ........................................................................................................................ 87 5.2. Methods.............................................................................................................................. 89 5.2.1. Study area.................................................................................................................... 89 5.2.2. Modelling approach .................................................................................................... 89 5.2.3. Data ............................................................................................................................. 92 5.3. Results ................................................................................................................................ 98 5.3.1. Balancing the model ................................................................................................... 98 5.3.2. Data quality ............................................................................................................... 100 5.3.3. Model output ............................................................................................................. 102 5.3.4. Ecosystem network analysis ..................................................................................... 104 5.3.5. Summary statistics .................................................................................................... 107 5.4. Discussion ........................................................................................................................ 108 5.5 Conclusions ....................................................................................................................... 110 Chapter 6 Exploring ecosystem boundary size ........................................................................... 126 6.1. Introduction ...................................................................................................................... 126 vii  6.2. Methods............................................................................................................................ 128 6.3. Results .............................................................................................................................. 128 6.4. Discussion ........................................................................................................................ 133 6.5. Conclusions ...................................................................................................................... 136 Chapter 7 The importance of incorporating diet quality into consumption rates for ecosystem modelling studies ........................................................................................................................ 141 7.1. Introduction ...................................................................................................................... 141 7.2. Method ............................................................................................................................. 143 7.3. Results .............................................................................................................................. 144 7.4. Discussion ........................................................................................................................ 146 7.5. Conclusions ...................................................................................................................... 147 Chapter 8 Population dynamics at the Prince Edward Islands:  Hindcasting of three known events..................................................................................................................................................... 154 8.1. Introduction ...................................................................................................................... 154 8.2. Method ............................................................................................................................. 155 8.2.1. Fur seal exploitation .................................................................................................. 156 8.2.1. Patagonian toothfish fishery ..................................................................................... 157 8.2.1. Cat predation on small flying birds ........................................................................... 158 8.3. Results .............................................................................................................................. 160 8.3.1. Fur seal exploitation .................................................................................................. 160 8.3.2. Patagonian toothfish fishery ..................................................................................... 161 8.3.3. Cat predation of small flying birds ........................................................................... 162 8.4. Discussion ........................................................................................................................ 163 8.4.1. Fur seal simulation .................................................................................................... 163 8.4.2. Patagonian toothfish fishery simulation.................................................................... 164 8.4.3. Cat predation on small flying birds ........................................................................... 165 8.5. Conclusions ...................................................................................................................... 166 Chapter 9 Population dynamics at the Prince Edward Islands:  Forecasting of climate change scenarios through forcing of primary producers ......................................................................... 177 9.1. Introduction ...................................................................................................................... 177 9.2. Method ............................................................................................................................. 180 9.3. Results .............................................................................................................................. 180 viii  9.4. Discussion ........................................................................................................................ 185 9.5. Conclusions ...................................................................................................................... 188 Chapter 10 Conclusion ................................................................................................................ 200 Bibliography ............................................................................................................................... 207 Appendix 1.A. Remotely sensed satellite data of chlorophyll-a centered on the Prince Edward Islands. ................................................................................................................... 233 Appendix 2.A. Summary of bird population data and biomass estimates .......................... 246 Appendix 2.B. Summary of Southern Elephant seal population and biomass estimates ... 248 Appendix 3.A. Diets of the top predators at the Prince Edward Islands ............................ 249 Appendix 3.B. Table of energetic content of prey (summary from the literature) ............. 258 Appendix 4.A. Species list for the Life Support System model ......................................... 259 Appendix 5.A. Data preparation for the ecosystem model of the Prince Edward Islands .. 261 Appendix 5.B. Mixed trophic impact (MTI) values for the 1980s model of the Prince Edward Islands .................................................................................................................... 315 Appendix 6.A. Data used to drive the fur seal simulations of the PEI ecosystem model... 317 Appendix 7.A. Data used to drive the Patagonian toothfish simulations of the PEI ecosystem model. ................................................................................................................ 319 Appendix 8.A. Data used to drive the cat predation on small flying seabird simulations of the PEI ecosystem model. ................................................................................................... 320    ix  List of Tables Table 2.1. List of codes used to identify functional groups and the corresponding species names (or Phylum/Class/Order, whichever appropriate) included in this study of the marine ecosystem of the Prince Edward Islands. ....................................................................................................... 41 Table 2.2. Area (in km2) considered to be between the contours between 0 and 1500m, as calculated using the GEBCO_08 Grid Version 20100927 from www.gebco.net ........................ 43 Table 2.3. Total biomass (B, t) of all land based top predators for 3 time periods, including adjusted estimates for time spent away from the islands (BTA, t). .............................................. 44 Table 2.4. Breeding cycles for the seabirds at the Prince Edward Islands (adapted from Cooper and Brown 1990)........................................................................................................................... 45 Table 3.1. Species list (common names), average weight, field metabolic equations used, daily consumption rates for individuals (in terms of wet weight), population estimates for the 1960s, 1980s and 2000s and proportion of time spent at the islands. ...................................................... 62 Table 3.2. Energetic content of six prey categories provided with summarised diet matrix, average energy density of diet and assimilation efficiencies for each species. ............................ 63 Table 3.3. Consumption (in t per year) of each species/group by prey category for A)1960s, B)1980s and C)2000s. ................................................................................................................... 64 Table 4.1. Input parameter estimates for the four Life-Support System Models (P/B, Production to biomass ratio; Q/B, Consumption to biomass ratio; U/Q, Unassimilated consumption). ........ 81 Table 4.2. Diet Matrix for the four Life-Support System models ................................................ 82 Table 5.1. Model parameters used for input for all three time periods. Includes Biomass (t or t.km-2) calculated for the Exclusive Economic Zone (EEZ) of the Prince Edward Islands (PEIs) for 1960s, 1980s and 2000s; Estimates of Production to Biomass (yr-1); Consumption to Biomass (yr-1) and Unassimilated Consumption (UC) is provided. Model outputs of Trophic Level (TL) x  and Production to Consumption rates (P/Q) given. Black text indicates the original estimates (done in t or t.km-2, whichever was appropriate). Blue text is the conversion to relevant units (from the black text figures). * 0.08 for 1960s & 1980s models, 0.13 for 2000s model; **5.11 for 1960s, 5.09 for 1980s, 4.98 for 2000s; ***4.74 for 1960s & 1980s, 4.67 for 2000s; 5.10 for 1960s, 5.15 for 1980s, 5.18 for 2000s. ....................................................................................... 112 Table 5.2. (A) Diet matrix for each functional group except Orcas, Southern Elephant Seals and Giant Petrels with contributions summing to 1 for all consumers for all three time periods (1960s, 1980s, and 2000s); (B) Diet matrix for Orcas, Southern Elephant Seals and Giant Petrels with unique diets provided for each time period (1960s, 1980s, and 2000s) (see text for details)...................................................................................................................................................... 113 Table 5.3. Index of Data Pedigree generated for each functional group based on the data quality of three input parameters (Biomass (B), Production to Biomass (P/B) and Consumption to Biomass (Q/B) estimates). Key for generating the index provided. ........................................... 115 Table 5.4. Transfer efficiencies summarised by Trophic Level (TL) for flows from the producers, the detritus and all flows combined for all three time periods. A summary of the flows for TLs 2 to 4 for each case also provided. ................................................................................................. 116 Table 5.5. Summary statistics of the Prince Edward Island model for each time period (1960s, 1980s and 2000s) as compared to 8 other Southern Ocean/Antarctic Ecopath models. ............. 117 Table 5.6. Qualitative assessment to highlight where scientific research efforts should be focussed in future. ....................................................................................................................... 118 Table 6.1. Description of the model size according to the length of the radius (nm) used, area included in the model (km2) and the proportion of each model that could be considered to represent the shelf region (as a percentage). ............................................................................... 137 Table 6.2. Input parameters for each model. MBM = Mixed Balanced Model. B = Relative biomass (t.km-2). P/B = Rate of production to biomass (yr-1). Q/B = Rate of consumption to biomass (yr-1). UC = Unassimilated consumption (between 0 and 1). Numbers in black = Pelagic xi  associated groups. Numbers in grey = Land based or benthic associated groups. * = value = 0.08 for 1980s model and 0.13 for 2000s model. ............................................................................... 138 Table 6.3. Ecotrophic Efficiency output estimates for all Model sizes 1-4 for the 1990s and 2000s time period, as well as for the Mixed Balanced Model created for the 1980s. Ecotrophic Efficiency values that exceed 1 are printed in red text. .............................................................. 139 Table 7.1. Daily food ingestion rates (g.ind-1.d-1) and annual consumption to biomass rate estimates (yr-1) of the species of birds found breeding at the Prince Edward Islands using Field Metabolic Rates (FMR) as defined by Nagy et al. (1999) (with amendments by Ellis and Gabrielson 2001) calculated using average mass estimates from Ryan and Bester (2008). ....... 149 Table 7.2. Diet matrices of the breeding bird species of the Prince Edward Islands summarised into six prey categories, with energetic density of prey categories given, and average energetic density of diet (kJ.g-1) and assimilation efficiency (as a fraction of the diet) for each bird species provided. ..................................................................................................................................... 150 Table 8.1. Estimates of the maximum and minimum % difference as compared to the starting biomass (t.km-2) for each functional group in the PEI marine ecosystem model for each hindcasting scenario run. ............................................................................................................ 168 Table 9.1. Matrix of climate change scenarios (S) from 1 through 8 with Forcing functions (F) 1 and 2 acting on each of the 4 primary producers at the Prince Edward Islands. ........................ 189 Table 9.2. Estimates of the maximum and minimum % difference as compared to the starting biomass (t.km-2) for each functional group in the PEI marine ecosystem model for each forecasting scenario run (S1-S8). ................................................................................................ 190    xii  List of Figures Figure 1.1. The Antarctic continent, the sub-Antarctic Islands and the approximate position of the Antarctic Polar Front (APF) (Australian Antarctic Division Data Centre, SCAR Map Catalogue, #13137) ....................................................................................................................... 15 Figure 2.1. Satellite derived chorophyll-a estimates from the region surrounding the Prince Edward Archipelago (Southern Ocean). Rectangles illustrate area from which satellite remote sensing data was extracted a) Large rectangle represents the 1km SeaWiFS LAC data extracted from January 1998 to December 2004, b) Medium rectangle represents the 1km MODIS LAC data extracted from January 2005 to December 2008; c) Smallest rectangle represents the 1km area LAC data extracted from SeaWiFS (Jan 1998 ? Dec 2004) and MODIS (Jan 2005 ? December 2008), used to quantify the island associated blooms. ................................................ 46 Figure 2.2. Summary of satellite derived chlorophyll-a data for the open ocean biomass and local area biomass in the vicinity of the Prince Edward Islands. Estimates for 1998 to 2004 were derived from the SeaWiFS Satellite, while estimates for 2005 to 2008 were derived from the MODIS Aqua satellite (see text for details). Error bars represent 1 standard deviation. ............. 47 Figure 2.3. Bathymetry contour lines in the vicinity of the Prince Edward Islands estimated from the GEBCO_08 Grid Version 20100927 (www.gebco.net) bathymetry data. ............................. 48 Figure 2.4.A-L. Figures illustrating bird biomass time series for penguins, albatross, Giant Petrels and Skua species at the Prince Edward Islands from the 1960s to 2010; ......................... 49 Figure 2.5. Land based predator biomass for three time periods (1960, 1980, 2000) for the Prince Edward Islands. ............................................................................................................................. 51 Figure 3.1. Community consumption estimates of the six prey categories for A) Seabirds, B) Mammals, C) All land based top predators .................................................................................. 65 Figure 4.1. Map of the Indian sector of the Southern Ocean showing the position of the Prince Edward Islands and main frontal features (STC, Subtropical Convergence; SAF, Sub-Antarctic Front; APF, Antarctic Polar Front). .............................................................................................. 83 xiii  Figure 4.2. Map of the Prince Edward Islands, indicating the spatial distribution of the three primary producers. ........................................................................................................................ 84 Figure 4.3. Flow diagram with relative importance of each primary producer at 200nm scale. Box size is proportional to the square root of the biomass of the functional group. .................... 85 Figure 4.4. Relative biomass of each primary producer for the four model sizes (model sizes equivalent to circles with given radii). .......................................................................................... 86 Figure 4.5. Ecotrophic efficiencies for functional groups of the first trophic level of the four ?Life-Support System? models. ..................................................................................................... 86 Figure 5.1. Study areas. Map showing the Exclusive Economic Zone (EEZ) of the Sub-Antarctic Prince Edward Islands situated southeast of South Africa. ........................................................ 119 Figure 5.2. Schematic representation of the food web. Each functional group is represented by a circle which is scaled to the square-root of the biomass (t); groups are distributed with increasing Trophic Level (TL) on the y-axis and all trophic linkages indicated by grey lines. ................... 120 Figure 5.3. Results of the sensitivity analysis. The index is the count of estimated parameters of the model affected by at least 30%, given the changes (between -50% and 50%) in the input parameters of each functional group (listed on the y-axis). Effects within groups have been eliminated from the analysis. ...................................................................................................... 121 Figure 5.4. Plot showing the model results of the Ecotrophic Efficiencies (EE) (0-1), an index of how much of the production for each functional group is used in the system, for each of the three time periods (1960s, 1980s and 2000s)....................................................................................... 122 Figure 5.5. Diagram showing the trophic flows as summarised in the Lindeman spine for each of the three time periods for which models were constructed, A) 1960s, B) 1980s  and C) 2000 (TST = Total System Throughput, TE ? Transfer Efficiency). .................................................. 123 Figure 5.6. Bar plot showing the Biomass (B) by trophic level for each of the three time periods (1960s, 1980s, and 2000s). ......................................................................................................... 124 xiv  Figure 5.7. Plot showing the relative trophic impact (RTI) of each functional group for each time period (1960s, 1980s, and 2000s). .............................................................................................. 125 Figure 5.8. Scatter plot of the relative trophic impact (RTI) plotted against biomass (B), providing an indicator of ?keystone? species. ............................................................................. 125 Figure 6.1. Illustration of the theoretical boundaries of each of the four models used in the study. The outermost ellipse represents Model 1 (an area equivalent to a circle with a radius of 200nm centered on the islands, representing the Exclusive Economic Zone of the Prince Edward Islands). The innermost ellipse represents Model 4 (an area equivalent to a circle with a radius of 20nm, representing the shelf region of the islands). Bathymetry data source: The GEBCO_08 Grid, version 20100927, http:www.gebco.net. ........................................................................... 140 Figure 7.1 Illustration of the variation in the consumption to biomass rate estimates (yr-1) as calculated for selected A) mammals, B) large seabirds and C) small seabirds found at the Prince Edward Islands. ........................................................................................................................... 151 Figure 7.2. Graph showing the percentage difference between consumption to biomass rate estimates (Q/B, yr-1) calculated using Nagy et al.?s (1999) field metabolic rates with standard processing versus using local diet and energetic density of prey for all breeding bird species found at the Prince Edward Islands. ........................................................................................... 152 Figure 7.3. Illustration of the A) diets of the penguin species summarised into six prey categories (?fish general?, ?fish ? mesopelagic?, ?cephalopods?, ?crustaceans?, ?vertebrates? and ?other?), energetic density of the B) prey categories and C) penguin diets with and without assimilation efficiencies as compared to standard processing (shown in red). ............................................... 153 Figure 8.1. Reconstruction of fur seal exploitation driven by hunting with model results compared to time series survey biomass data for the PEI EEZ from 1800 to 2010 for Antarctic Fur Seals A) and Subantarctic Fur Seals B)................................................................................ 169 xv  Figure 8.2. Trends in biomass estimates (t.km-2) from 1800 to 2010 for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the fur seal industry. ....................................................................................... 170 Figure 8.3. Reconstruction of the Patagonian toothfish fishery from 1980 to 2006 presented with fishing mortality ( as catch data, t.km-2) used to drive the model (A) and model output showin in (B), with and without cetacean depredation as compared to time series relative abundance data (from catch-per-unit-effort (CPUE)) data provided by Brandao and Butterworth (2009) for the PEI EEZ. ..................................................................................................................................... 171 Figure 8.4. Trends in biomass estimates (t.km-2) from 1980 to 2006 for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the Patagonian toothfish fishery. .................................................................... 172 Figure 8.5. Reconstruction of the cat predation on small flying birds (Prions and Petrels) from 1940 to 1990 with predation estimates (derived from cat population and consumption estimates from Bester et al. 2000) driving the model (with original estimates (CAT S1), 1/7th original estimates (CAT S2) and 1/10th original estimates (CAT S3)). ................................................... 173 Figure 8.6. Trends in biomass estimates (t.km-2) for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the cat predation on the small flying birds (Prions and Petrels) from 1940 to 1990 based on original consumption estimates (CAT S1). .............................................................................................. 174 Figure 8.7. Trends in biomass estimates (t.km-2) for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the cat predation on the small flying birds (Prions and Petrels) from 1940 to 1990 based on estimates 1/10th of original predation estimates (CAT S3). ........................................................................ 175 Figure 8.8. Spawning biomass estimates taken from Brandao and Butterworth?s (2009) Figure A.1. for the Patagonian toothfish at the Prince Edward Islands. Original caption from figure: Spawning biomass estimates (note that recruitment can vary prior to the onset of harvesting). Estimates are given for the Optimistic, Intermdeiate, Less Pessimistic and Pessimistic scenarios xvi  (details of the conditioning of these scenarios to the data are provided in the text; see also the caption to Table A.4). All results shown assume a cetacean depredation factor z=1, i.e. recent losses to cetacean depredation are equal to the landed longline catch. ...................................... 176 Figure 9.1. Plot of satellite derived chlorophyll-a values from 3 sensors (SeaWiFS, MODIS-Aqua, MODIS-Terra) from 1997 to 2010 centered on the Prince Edward Islands (see text)..... 191 Figure 9.2. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a positive effect on the PIA production term.............................................................................................................. 192 Figure 9.3. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a negative effect on the PIA production term.............................................................................................................. 193 Figure 9.4. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a positive effect on the POL production term. ........................................................................................................... 194 Figure 9.5. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a negative effect on the POL production term. ........................................................................................................... 195 Figure 9.6. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a positive effect on the POS production term. ............................................................................................................ 196 Figure 9.7. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a negative effect on the POS production term. ............................................................................................................ 197 Figure 9.8. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a positive effect on the PMA production term. .......................................................................................................... 198 xvii  Figure 9.9. Trends in biomass estimates (t.km-2) from 2000-2099 for all functional groups of the PEI marine ecosystem following a linear forcing function created to produce a negative effect on the PMA production term. .......................................................................................................... 199    xviii  Acknowledgements There are many people who I would like to acknowledge for their support during this project. I would like to thank all scientists from the South African National Antarctic Program (SANAP) who have collected scientific data from the PEIs and the South African National Research Foundation (NRF) for the funding of research projects at the PEIs. Without these scientific contributions over the past 60 years the construction of the present models would not have been possible. Various discussions with researchers in South Africa were invaluable. I would like to thank those who met with me to discuss this work at various stages of its conception and completion including Henry Valentine, Rob Crawford, Marthan Bester, Nico de Bruyn, and Peter Ryan. I would especially like to thank Peter Ryan for finding my resolution of balancing the model by adding a ?dead penguin? group inadequate, which resulted in a lengthy diversion into the consumption rates and ultimately made a valuable contribution to the work. I would like to thank Lucy Bastin for her assistance with GIS, working with the GEBCO bathymetry data and producing the associated maps and data, and for working with me to follow through with the sensitivity testing. I would also like to thank her for her enthusiasm for the project and for her encouragement. I would like to thank Bruce Monger for his excellent teaching and for assistance with providing the facilities for processing the satellite data used in this work. A number of people reviewed the project at various stages through conception to completion and I would like to thank them for their valuable contributions: Lynne Shannon for reviewing the project proposal, Chiara Piroddi for reviewing much of the work and for her assistance with accessing UBC Fisheries Centre summary statistics for the Antarctic systems, and Sarah Richards for her presence throughout the entire duration of the PhD, for her steadfast support and encouragement and her excellent editing skills. Thanks also to Katherine Tattersall for her assistance in the final hour.  To my supervisory committee: It was a privilege to have each of you on my committee and I would like to thank you for your guidance throughout the process and thank you for not giving up on me, even in the middle where it seemed I might never make it through. I would like to thank John Dower for his time and commitment to the project. I would like to thank Susan Allen for always being willing to enter into discussions on all topics broached, for extending my xix  knowledge in physical oceanography and for her ability to get the best out of me in all circumstances. I would like to thank Villy Christensen for being an excellent tutor in my development as a ?would be? ecosystem modeller. Thank you for your guidance and assistance from inception to the very end, always finding the time to help me advance my understanding. To Evgeny Pakhomov, I travelled to Canada to take up this opportunity to study under your supervision and it has been a privilege to do so. Thank you for allowing me the freedom to explore a multitude of avenues which allowed this process to be one where I learned a great deal and became a truly interdisciplinary study. Through your insight and familiarity with biological data in general and with Antarctic systems in particular, along with the effortless manner in which you share your knowledge, your guidance was invaluable. Thank you also for the funding which made this project possible, and for your support. The initial hand in procedure of this dissertation was initiated two years before it finally became a reality and during that time my professional career changed direction as I took up a postdoctoral grantholder position at the Joint Research Centre, EU. I was fortunate to be able to continue various aspects of this work in my new position. I would like to thank Nicolas Hoepffner, for supporting my involvement in the IndiSeas II project and for encouraging me to complete this work. I would like to thank Fr?deric M?lin for reprocessing the satellite data for the islands and providing excellent advice. I would like to thank Marta Coll and Jeroen Steenbeek for collaborations in the development of the project to drive an Ecopath model with satellite derived primary production. I would also like to thank Mineral Services Canada for generously allowing me the use of the residential facility for the duration of my studies.  Finally, I would like to thank the underwater hockey community for brightening my being each time I entered the water to play the game. Thanks to my Australian, Canadian, Italian and South African team mates for the pleasure of your company. I would like to thank my family and friends for their support and encouragement, including: Grant and Angela Bender, Eric and Ingrid Kalnins, Marie-Renee Blanchet, Lisa Cooke, Melanie Johnson, Tim Minto, Dawn Koop, Cynthia Clapera, Karen Engelberts, Catherine Engelberts, Chiara Piroddi, Sarah Richards, Maria Domenici Larysa Pakhomova, Jean Kerby-Feast, James Gurney, Kim Gurney, Dot Gurney and xx  John Gurney. Lastly, and most importantly to David McGaughey, for your patience and belief in me - this would not have been possible without you. Thank you.   xxi  Dedication   For Dorothy Anne Harris Gurney (nee Storie)      ?When one tugs at a single thing in nature, he finds it attached to the rest of the world?  - John Muir, Naturalist (1838 ? 1914)  xxii  Chapter 1 Introduction  1.1 Ecosystem approach to management of marine resources  Following the collapse of many fish stocks the world over, due at least partly to failure of traditional fisheries science management strategies to sustainably manage fisheries, a worldwide resolution to adopt a more holistic approach to management of marine resources has been made. Management strategies are turning to ecosystem studies to move from single species assessment models, to multi-species and ultimately to ecosystem based management. This approach comes with the recognition that resource management of marine ecosystems should incorporate a broader scope of variables and drivers including environmental, biological and socio-economic factors.  International meetings and conventions have played a role in highlighting the need for an ecosystem approach to marine resource management. These include: the 1972 Stockholm Conference on the Human Environment; the 1992 Conference on Environment and Development; the 1992 Convention on Biological Diversity; the 1995 Kyoto Declaration on Sustainable Contribution of Fisheries to Food Security; the 2001 Reykjav?k Declaration; and the 2002 World Summit on Sustainable Development (Garcia et al. 2003; Shin and Shannon 2010). Policy has been changing globally with this new perspective and a push to include ecosystem considerations is being incorporated into operating procedures. This has created a need for the development of models that can be used for such an ecosystem approach, which currently range from individual based models through to ecosystem models (see overview in the following section, 1.2).  Ecosystem models for use in management are still in their early developmental phase but many management policies now require that ecosystem based models are used in assessing marine resources to address management options alongside traditional single stock assessments (Christensen and Walters 2011). Such models also provide the ability to test ?what if? scenarios, which can be useful in exploring ecosystem responses to change.  Ecosystem models are time consuming to construct but provide a useful tool for managers to optimize for fisheries and/ or 1  societal benefits and/ or conservation objectives, while taking into consideration the ecosystem as a whole and not solely the targeted species. It was with this in mind that this ecosystem study was initiated for the Prince Edward Islands (PEIs). The main objective was therefore to create a model of the PEIs that can be used as a tool to explore a number of theoretical and applied ecological issues.  The PEIs are an archipelago situated in the southeast sector of the Southern Ocean and, like other Sub-Antarctic equivalents, are host to millions of seabirds (including penguins) and seals, which use the islands as a breeding ground and a refuge while moulting. Historically, seals were harvested at the islands and there is an on-going fishery for Patagonian toothfish (Dissostichus eleginoides). The islands host approximately 13% of the worldwide population of King penguins (Aptenodytes patagonicus) with the remaining three species of penguins making relatively small contributions to the world populations (Macaroni (Eudyptes chrysolophus) 4%, Southern Rockhopper (Eudyptes chrysocome) 5% (but 17% of the filholi race), and Gentoo (Pygoscelis papua) 0.5%) (Crawford and Cooper 2003; Ryan and Bester 2008).  In terms of the conservation status, according to Birdlife International, the Gentoo penguin is classified as ?Near Threatened?, both the Macaroni?s and Southern Rockhopper?s as ?Vulnerable?, while the Kings are ?Least Concern? (IUCN Red List for birds (Downloaded from http://www.birdlife.org on 27/09/2011). The islands therefore have both fisheries and conservation concerns. They fall under South African jurisdiction and are within the Conservation of Antarctic Marine Living Resources (CAMLR) Convention Area (Areas 58.7, 58.6 and 58.4.4) (CCAMLR 2011).  The Commission of the CAMLR has mandated that the PEIs must be managed using an ecosystem approach. Prior to this work no ecosystem model has existed for the archipelago. 1.2. Ecosystem models  The primary role of models in ecosystem science is to permit controlled exploration of a complex reality. Models in marine ecology have been developed for individuals through to ecosystems (Hollowed et al. 2000). A review of 20 models used to assess ecosystem approaches to fisheries by Plaganyi (2007) provided the framework for this assessment of modelling tools currently available. Individual based models (IBMs) track the fate of individuals through their life span with the assumption that the behaviour of an individual represents the population as a 2  whole. These models are usually single species but may be multi-species or represent an ecosystem, as for example OSMOSE (Shin et al. 2004). Dynamic multi-species models or Minimally Realistic Models (MRM) are restricted to those species most likely to have important interactions with the species of primary interest and are usually system specific, for example the original MRM (Punt and Butterworth 1995), ESAM (Extended Single-species Assessment Models (Hollowed et al. 2000)), MSVPA (Multi-Species Virtual Population Analysis), MULTSPEC (Tjelmeland and Bogstad 1998), BORMICON (Bjoernsson et al. 1998), SEASTAR (Tjelmeland and Lindstr?m 2005), GADGET (Taylor and Stefansson 2004), CCAMLR predator-prey models (e.g., Mori and Butterworth 2006), and Multi-species Statistical Models (Plaganyi 2007). Bioenergetic models, which are based on bioenergetics and allometric equations, have also been used. Examples include those of Yodzis and colleagues (Yodzis and Innes 1992; Yodzis 1998; Koen-Alonso and Yodzis 2005). Finally, whole ecosystem models, which take into consideration all trophic levels in a system, include Atlantis (Fulton et al. 2004a; Fulton et al. 2005) and the Ecopath with Ecosim (EwE) software (Christensen et al. 2008). The intention of this study was to create a whole ecosystem model, and therefore the discussion which follows focuses on these two models.  The Atlantis model was created through a series of amalgamations and developments from earlier models. The European Regional Seas Ecosystem Model (ERSEM) (Baretta et al. 1995) is a generic model, which focused on the lower trophic levels (plankton and regeneration cycle) and could be coupled to a physical model (e.g., Allen et al. 1998; Blackford et al. 2004; Sole et al. 2006; Siddorn et al. 2007). An updated version of this model (ERSEM II) was combined with the Port Phillip Bay integrated model (PPBIM) of Murray and Parslow (1999) to create the Integrated Generic Bay Ecosystem Model (IGBEM) (Fulton et al. 2004a). This model was later combined with Bay Model 2 (Fulton et al. 2004a) to create Atlantis (Fulton et al. 2005).  The EwE software developed from an ecosystem model (the original Ecopath) initially created by Polovina (1984). Christensen and Pauly (1991) developed this into a mass balanced ecosystem model (Ecopath II) which later incorporated developments from the independently developed ecosystem model (also based on Polovina 1984) called NETWRK by R.E. Ulanowicz (Ulanowicz 1986). Subsequent developments of the model include a temporal component 3  capable of simulations (Ecosim) (Walters et al. 1997) and a spatial component (Ecospace) (Walters et al. 1999), which enables exploration of changes in habitat. The model has been recognized as an important tool and was nominated by NOAA as one of the top ten scientific breakthroughs in their 200 year history (see http://celebrating200years.noaa.gov/breakthroughs/ecopath/). The EwE model is rooted in ecosystem theory and has a user friendly Windows-based interface that enables a user with limited mathematical and programming skills to construct a useful model of an ecosystem (Shannon et al. 2004). This ease of access has been criticised because it can result in misuse (or insufficient depth of use) of the program (Plaganyi 2007). Other cautions regarding the use of EwE are that the default settings should not be used without careful consideration and alternative functional relationships should be considered (Plaganyi 2007) as these can produce markedly different model outputs (Koen-Alonso and Yodzis 2005). Ongoing software development for this model has allowed it to be the fore-runner in its field.  EwE is the most widely used ecosystem model and capable of addressing the widest range of topical ecosystem approaches to fisheries research questions (Plaganyi 2007).  EwE and Atlantis both have the capacity to represent all trophic levels and are appropriate for addressing broad ecological questions. Both have the capability of including age structured functional groups and spatially resolved data, important considerations in models (Fulton et al. 2004b). Atlantis has a modular structure, which allows substitution of sub-models, a development now incorporated into EwE version 6 (Christensen et al. 2008), and one which will help with the transparency of the model and accessibility of the coding to allow for additional modules to be written and customised for particular ecosystems. For a summary of the strengths and weaknesses of each of these models, please refer to Plaganyi (2007). Models of food webs aim to represent the links between predators and prey in an ecological community and as such are complex networks (Quince et al. 2005b). Important considerations in modelling include understanding the extent to which the model structure (Fulton and Smith 2004; Plaganyi 2007) and underlying model assumptions predetermine or have implications for the results obtained (Plaganyi and Butterworth 2004). This means that the models themselves need to be closely scrutinized to understand the extent to which these factors influence the model 4  outputs. The model structure (and even its development history) can have significant implications for the potential range of dynamics displayed (Fulton and Smith 2004; Plaganyi 2007). Important considerations for the development of ecosystem models include the effect of specific formulations on model outputs. Plaganyi (2007) gives three examples of this: feeding functional responses (Fulton et al. 2003b); predator-prey relationships (Yodzis 1994); and the structuring of competition (Rice 2000). The treatment of uncertainty and the feasible representation of biodiversity also need to be considered (Plaganyi 2007). There is current agreement that ideally more than one model should be used and outputs from different models should be compared (Shannon et al. 2004; Koen-Alonso 2005; Field et al. 2006; FAO 2008). For instance, in Aydin et al. (2005) three independent modelling assessment models (a NPZD model, Ecopath with Ecosim model and a bioenergetics model) were all used to investigate Pacific salmon in the Alaskan subarctic gyre ecosystem. It is important for models used to incorporate ?top down? as well as ?bottom up? forces in marine ecosystems (Reid et al. 2000; Aydin et al. 2005; Field et al. 2006; Christensen and Walters 2011).  Another issue that is important in developing a model is to determine the model complexity (Raick et al. 2006). From a practical standpoint it is not necessary to consider all the species to be found in a community as not all are equally important in determining the nature of the whole community. The major part of the energy flow involves a comparatively small number of species and we can learn much about communities and populations by concentrating on dominant populations in the major trophic levels (Odum and Odum 1959). Two issues arise when addressing the question of determining the nature of the community as a whole. The first is that with increasing model complexity (to represent biological realism) there is an associated increase in uncertainty and in scientific imprecision in estimates of the associated parameter values (Blackford et al. 2007; Plaganyi 2007). Secondly, it has been found that it is not necessary to include all of reality in terms of linkages as only a relatively small number of variables is often sufficient for effective models because ?key factors? or ?integrative factors? often govern or control a system (Odum 1971; Quince et al. 2005b). The conclusion, for model complexity, is that important groups should be focused on, and groups should only be added when there is good reason to look at those particular aspects in more detail.   Those that only add noise to the system should be omitted (Shannon et al. 2004). Adding complexity for completeness does not 5  contribute anything if the resulting model is of poor quality and model comparisons have shown that intermediate model complexity results in optimal performance (Fulton et al. 2003a).  Typically there is a trade-off between the range in trophic levels considered and the corresponding detail with which each group is represented. Determination of which functional groups should be included in a model needs to be carefully assessed. Two suggested guidelines for aggregation of groups include guarding against aggregation of serially linked groups or combining groups (or age classes) that have rate constants that differ by 2 to 3 fold (Fulton et al. 2003a). Transgressions of these two suggested guidelines have been shown to decrease model performance (Fulton et al. 2003a). A suggestion of exclusion of feeding links that represent less than 10% of consumption, both by and of any species, has been shown to have minimal affect on model predictions, and this can amount to up to 44% of the links being removed as demonstrated by Yodzis (1998). Above this threshold for linkage strength the model predictions start to become unreliable. In a study on weak linkages by Pinnegar et al. (2005),  it was found that, following a perturbation, models in which ?weak links? were aggregated into other groups created a more stable model output than models in which weak links were removed from the model completely. Implications of this suggest that considerations on whether weak links are removed by ?lumping? or ?chopping? may have very different system consequences (Pinnegar et al. 2005). An assessment of reducing model complexity through simplifying by aggregation found that even the simplest NPZD (nutrient, phytoplankton, zooplankton, detritus) models were able to recreate the global ecosystem features, however a greater degree of sophistication was required to simulate the various realistic behaviours (Raick et al. 2006). The findings of this study showed that a 9-compartment model that reduces functional groups but includes the bacterial loop and unbalanced algal growth performed best with respect to producing realistic behaviours (e.g., the phytoplankton competition, the potential carbon or nitrogen limitation of the zooplankton ingestion, the model trophic closure, etc.) (Raick et al. 2006).  The decision to use the EwE software to develop the first quantitative, network based ecosystem model of the Prince Edward Islands was based on a number of considerations, which were assessed at the start of the project: 1. The lack of a physical oceanographic model of the region meant that the capabilities of this component of the Atlantis model would be absent; 2. The lack 6  of accessibility to the Atlantis model code and absence of available documentation (pers comm. with B. Fulton); 3. The accessibility of the EwE software and supporting documentation (free download from www.ecopath.org) and training (graduate course followed at UBC); 4. The ease of access to the EwE software creators and developers, being based on UBC campus and; 5. The time required to develop an Atlantis model exceeds what was available for this study. A brief synopsis of the islands is provided below (Section 1.3), including information on background, history, geography, geology and oceanography.  This is followed by a summary of the data collection for the marine ecosystem available for the construction of the model for which the guidelines outlined above were taken into consideration. 1.3. The Prince Edward Archipelago 1.3.1. Background and history The Prince Edward Islands (PEIs) are a sub-Antarctic archipelago situated in the south-western Indian Ocean and are South African territory. Together with Crozet (France), Kerguelen (France), MacDonald (Australia) and Heard (Australia), these islands form the South Indian Ocean Province, one of three provinces of the Sub-Antarctic region (H?nel and Chown 1998). These islands, along with South Georgia (UK), Auckland (NZ), Campbell (NZ) and Macquarie (Australia), are all sub-Antarctic ecosystems (see Figure 1.1). The PEIs are found in a High Nutrient Low Chlorophyll zone (HNLC), and like other sub-Antarctic islands, are ?hotspots? with higher biological productivity in the vicinity of the islands than in the surrounding water. The islands are also host to millions of birds and seals, which use the islands as a refuge and breeding ground.  The Prince Edward archipelago consists of two islands: Marion (46? 54? S; 37? 45? E) and Prince Edward (46? 38? S; 37? 57? E) (Figure 1.1). Found approximately 1800 km southeast of Port Elizabeth (South Africa), 2300 km north of L?tzow-Holm Bay (the closest point on the Antarctic continent) and 950 km west of the French Crozet Islands (H?nel and Chown 1998), the islands are remote. They were first sighted in 1663 by the Dutch East Indiaman, the Maerseveen, under the command of Barent Barentszoon Ham. No landing was made, but the names Maerseveen (Marion) and Dina (Dena/Denia) (Prince Edward) were given to each of the islands. Over a 7  hundred years later the French naval officer MM Marion du Fresne of the Frigate Le Mascarin passed the islands in 1772 and called them Isle of Hope (Marion) and Isle of Cavern (PE), but renamed them the Frigid Islands on leaving (H?nel and Chown 1998). In 1776 Captain James Cook renamed the islands again, calling both islands the Prince Edward Islands. Marion was later renamed in honour of the French naval officer from the earlier voyage. Initially, the islands were used as a stop off point for whaling ships en route to the Antarctic. In the 1800s sealing (for both elephant and fur seals) began at the islands. The last known visit by sealers was in the 1930s. The islands were annexed by South Africa in 1947/48 and a meteorological station was established. In the 1960s the station was expanded to include scientific research, which continues today. Terrestrial and marine research has been ongoing at the islands for over 50 years, with a resultant rich published literature (H?nel and Chown 1998). 1.3.2. Geology and geography The two islands are the twin peaks of a coalescing shield volcano and are estimated to be less than 1 million years old (the oldest recorded lava on Marion is estimated to be 450 000 years old but most flows date between 15000 and 400 years) (H?nel and Chown 1998). The volcano is still active with the most recent eruptions recorded in 1980 (Verwoerd et al. 1981) and 2004 (pers. obs. by D.W. Hedding in Sumner et al. 2004). Glacial history is evident on Marion, with at least five glaciations having occurred during the Quaternary, along with evidence of a glacial event during the Late Pleistocene (H?nel and Chown 1998). Almost all of Marion Island was covered by ice at the last glacial maximum.  Marion is the larger of the two islands, 293 km2 in size, with approximately 70 km of coastline (Beckley and Branch 1992). The highest peak on Marion is Mascarin, at a height of 1231 m. In the 1950s and 1960s this peak was covered in an ?ice plateau?, which used to extend down to 1000 m, but now no longer exists due to rapid melting (Sumner et al. 2004). Prince Edward Island is smaller at 46 km2, and its highest peak, Van Zinderen Bakker, is considerably lower at 672 m (H?nel and Chown 1998).  The terrain on the islands is wet and marshy, with many lakes and ponds. Slopes are generally fern-covered while hillsides and peaks of lava have no vegetation. The coastline is mostly cliff-8  faced. Of the coves on Marion Island, most are boulder, rocky or pebble beaches, with only two categorized as sandy (Ship?s Cove and a small beach at Goodhope Bay) (H?nel and Chown 1998). Prince Edward island has a more rugged coastline and fewer beaches (H?nel and Chown 1998).  1.3.3. Marine research 1.3.3.1. Oceanography The PEIs lie east of the Southwest Indian ridge and southwest of the shallow Del Cano Rise, directly in the path of the Antarctic Circumpolar Current (ACC) (Deacon 1983; Lutjeharms 1985). Rising up from a depth of 3000m, Marion and PE are approximately 20km apart and separated by an inter-island shelf which varies between 45 and 260m depth (Pakhomov and Chown 2003).  The islands fall in the Antarctic Polar Frontal Zone (APFZ) of the ACC between the Sub-Antarctic Front (SAF) in the north and the Antarctic Polar Front (APF) in the south. The SAF has been found between 45?15? S and 47?25? S (a range of 241 km) and has an average temperature of 7.0?C while the APF has been found between 49?39? S and 50?47? S (a range of 126 km) with an average temperature of 3.4?C (Lutjeharms and Valentine 1984). These fronts are also important biogeographic boundaries (Deacon 1983). Both fronts are consistent and well-defined oceanic thermal fronts (Deacon 1983; Lutjeharms 1985; Nowlin and Klinck 1986), however their position has been shown to be dynamic and variable (Lutjeharms and Valentine 1984; Duncombe Rae 1989a; Lutjeharms et al. 2002).  The hydrodynamics in the vicinity of the islands are affected by the proximity of the SAF to the islands (Pakhomov et al. 2000a; Perissinotto et al. 2000; Ansorge and Lutjeharms 2002). Although the SAF is generally found to the north of the islands it has also been found south of the islands (Lutjeharms 1990), and on one occasion the SAF was found split into two branches, one passing north of the islands and the other to the south (Ansorge and Lutjeharms 2002). It is thought that when the SAF is close to the islands, increased current velocities associated with this front leads to a flow-through system between the islands (Pakhomov and Froneman 1999b; Ansorge and Lutjeharms 2000; Pakhomov et al. 2000a). When the SAF is further north the lower velocities of the inter-frontal zone result in water retention over the inter-island shelf and as a consequence little exchange of inshore/offshore waters over the inter island region occurs (Perissinotto and Duncombe Rae 9  1990). A number of other mechanisms for on shelf water retention have been postulated, including upwelling (Grindley et al. 1985), eddy formation (Allanson et al. 1985) and Taylor column formation (Perissinotto and Duncombe Rae 1990). Under conditions of water retention on the shelf, whatever the mechanism, phytoplankton blooms have been observed to occur between the islands (Allanson et al. 1985; Duncombe Rae 1989a; Perissinotto and Duncombe Rae 1990). The oceanography in the area is complex because of the combination of the frontal features, the bathymetry and the interaction with the islands themselves (Ansorge and Lutjeharms 2002; Ansorge and Lutjeharms 2003; Ansorge and Lutjeharms 2005). Oceanographic surveys downstream of the islands have shown north-south meanders in the predominantly easterly flow which are thought to be the result of a wake downstream of the islands (Perissinotto et al. 2000). Both warm and cold-core mesoscale eddies have been observed upstream and downstream of the islands (Ansorge and Lutjeharms 2002; Ansorge and Lutjeharms 2003; Ansorge et al. 2004).  Water in the vicinity of the islands is generally Sub-Antarctic Surface Water (Deacon 1983) but both Subtropical and Antarctic Surface waters have been observed (Miller et al. 1984; Perissinotto et al. 2000). Nutrient levels in the Sub-Antarctic are relatively high and augmented in the vicinity of the islands by fresh water run-off that carries nutrients from guano and the feathers of moulting oceanic birds (Ismail 1990; Perissinotto and Duncombe Rae 1990). The concentrations of nutrients reported during different studies vary considerably but all lie within the range expected for Sub-Antarctic waters (Miller et al. 1984; Allanson et al. 1985; Ismail 1990; Perissinotto and Duncombe Rae 1990; Thomalla et al. 2011). Nitrogen, in both ammonia and urea forms, shows a concentration gradient spreading outwards and downstream from the islands to approximately 80 km off shore (Perissinotto et al. 2000). 1.3.3.2. Biology  Following annexation of the islands in 1947/48, a research station was established.  Biological observations date back to the early 1950s and 1960s for many of the seal and seabird populations (Rand 1962; Van Zinderen Bakker Sr et al. 1971) with descriptions of the breeding populations of the land-based top predators described for the first time. In the 1970s oceanographic studies 10  on the pelagic system began (El-Sayed 1976) and resulted in the first comprehensive study of the phytoplankton and zooplankton of the area (El-Sayed et al. 1979a; El-Sayed et al. 1979b; Grindley and Lane 1979). These studies were followed by work done from the 1980s through to the early 2000s describing the community structure and variations in abundance related to oceanographic features in the vicinity of the islands (Miller et al. 1985; Boden and Parker 1986; Perissinotto 1989; Perissinotto and Boden 1989; Perissinotto and McQuaid 1992b; Hunt et al. 2001; Hunt et al. 2002; Bernard and Froneman 2003; Hunt and Pakhomov 2003; Hunt et al. 2008). It has been found that the islands do not have an endemic zooplankton community but species of subtropical, sub-Antarctic and Antarctic origin are all found (Pakhomov and Froneman 1999b; Pakhomov et al. 2000a; Pakhomov and Froneman 2000; Hunt et al. 2002) and the abundance and distribution of zooplankton groups varies on both temporal and spatial scales (Hunt et al. 2001; Hunt et al. 2002). Studies focused on particular zooplankton groups were also conducted through the late 1990s and early 2000s, including work on euphausiids (Gurney et al. 2001; Gurney et al. 2002; Bernard and Froneman 2006; Bernard et al. 2007), pteropods (Bernard 2006) and amphipods (Froneman et al. 2000b).  The benthic community was comprehensively studied in the 1980s using both dredging and diving surveys that produced a comprehensive list of species present and a review of the ecology of the community (Blankley 1984; Blankley and Branch 1984; Blankley and Branch 1985; Blankley and Grindley 1985; Arnaud and Branch 1991; Branch et al. 1991a; Branch et al. 1991b; Beckley and Branch 1992; Branch and Williams 1993; Branch 1994). The benthic community comprises approximately 550 species with seven benthic community groups. The benthic decapod Nauticaris marionis (Bate 1888), an endemic species, has the second highest crustacean biomass of the benthic community. It is believed that this species provides a link between the benthic production and the near-shore pelagic predators of the islands, and numerous studies have focused on this decapod because of its perceived key role in the ecosystem (Perissinotto and McQuaid 1990; Kuun et al. 1999; Pakhomov et al. 1999; Pakhomov et al. 2000c; Pakhomov et al. 2004).  Scientific sampling of marine fishes around the PEIs dates back to 1873 when the Challenger expedition visited the region and collected three species as new to science (Gunther 1880, in Gon 11  and Klages 1988). Almost 100 years passed before any further additions were made. Andriashev (1971) identified the few fish specimens brought back from the South African expedition in 1965/66 resulting in two new distributional records (Gon and Klages 1988). French scientists added to the species list considerably in 1976 when several trawls were made with the RV Marion-Dufresne during their Sub-Antarctic inter-island research cruise (Hureau 1979; Duhamel et al. 1983). The list of 33 species compiled by Gon and Klages (1988) was then extended by 36 new records to the area following the only fisheries independent survey in 2001 by the MV Iris (Pakhomov et al. 2006).  The only species to be targeted by a commercial fishery in the Exclusive Economic Zone of the islands, is the Patagonian toothfish Dissostichus eleginoides. Within 5 years of the fishery officially opening, the catch statistics indicated that the stocks were severely depleted (Brandao et al. 2002). A much reduced fishery is still ongoing today. The fishery is a long-line fishery and has in the past had serious impacts of by-catch on the sea bird population (Nel et al. 2002c; Nel et al. 2003), but data show that mitigation efforts have been successful and the by-catch significantly reduced (to zero) starting in the 2005/06 season (CCAMLR 2010).  The mesopelagic community has been neglected in studies to date. Very little data are available for this community in the vicinity of the islands. Studies that consider the abundance and/or diet of myctophid fish are rare (Perissinotto and McQuaid 1992b; Pakhomov et al. 1996) and no sampling has been targeted for studies of the cephalopod community. Data available for this important component of the ecosystem are lacking and currently limited to species lists established from the diet analyses of the seabirds and seals (Berruti and Harcus 1978; Adams et al. 1985; Cooper et al. 1988; Gartshore et al. 1988; Hunter and Klages 1989). Research on birds at the PEIs has been ongoing since the inception of the scientific research at the islands and there is a rich published literature on the birds of the islands, which provides a solid base of data for this important component of the ecosystem (Williams et al. 1975; Siegfried 1978; Williams et al. 1979; Cooper and Brown 1990; Crawford and Cooper 2003). Population dynamics of penguins (Crawford et al. 2003b; Crawford et al. 2003c; Crawford et al. 2003d; Crawford et al. 2009), albatross species (e.g., wandering and grey-headed) and other large birds (e.g., the northern and southern giant petrels)  have been monitored with long term trends observed (Nel et al. 2002a; 12  Crawford et al. 2003d; Ryan et al. 2009). The effect of fisheries on marine birds is a worldwide concern (Tasker et al. 2000; Cury et al. 2011). Recent population declines have been linked to the increase in tuna long lining as well as recent large scale Illegal Unregulated and Unreported (IUU) long-line fishing for Patagonian toothfish (Nel et al. 2002b; Nel et al. 2003) though as already stated, recent mitigation measures have been successful.  Datasets for the mammals began with the earliest observations on populations recorded in Rand (1956) and Van Zinderen Bakker Sr et al. (1971). Following these, research on the breeding populations of the seals began in earnest (e.g., Condy 1977; Condy 1978a; Condy 1981) and has continued to today (Kerley 1983b; Wilkinson and Bester 1990a; Hofmeyr et al. 1997; Hofmeyr et al. 2006; Bester et al. 2009). In addition to these population studies, biological data on diet, foraging ranges and breeding have also been conducted and have resulted in a comprehensive dataset for the islands (Condy and Bester 1975; Klages and Bester 1998; Pistorius et al. 2001a; Pistorius et al. 2001b; Makhado 2002; Kirkman et al. 2003; Pistorius et al. 2008; de Bruyn et al. 2009). 1.4. Aims of the project It is evident from the overview provided here that an ecosystem model of the Prince Edward Islands has the potential to provide a platform to collate the available biological data of the marine system, which can then be used to address ecological questions. With a model in place, there is the potential to use it in the future for practical management applications.  The overall aim of this study was to create the first ecosystem model of the PEIs marine system.  The model will be based on the existing comprehensive biological and ecological dataset, and be used as a tool to explore a number of theoretical and applied ecological questions, specifically: 1. To describe and parameterize the food web of the marine component of the PEI ecosystem in terms of biomass (Chapter 2) and consumption (Chapter 3); 2. To characterise the ecosystem using a mass balanced ecosystem modeling approach (Ecopath) to: 13   - investigate the sources of productivity and their relative importance to the system at various spatial scales (Chapter 4); and  - describe the ecosystem across a range of time periods (1960s, 1980s, 2000s) (Chapter 5);  - determine the ecosystem boundary size of the top predators (Chapter 6) 3. To investigate the population dynamics at the PEIs using a dynamic temporal simulation approach (Ecosim) looking at past changes, which are directly related to human activity (Chapter 8) and potential effects of climate change (Chapter 9) In the process of developing the ecosystem model for the PEIs, an investigation into the consumption rate estimates used in the model was initiated, which resulted in a revision of the available estimates. This work is presented in Chapter 7, and the methodology is included in all models.   14       Figure 1.1. The Antarctic continent, the sub-Antarctic Islands and the approximate position of the Antarctic Polar Front (APF) (Australian Antarctic Division Data Centre, SCAR Map Catalogue, #13137)   15  Chapter 2 Review of the Prince Edward Island marine ecosystem: Quantifying the system in terms of biomass and highlighting long term trends Introduction An extensive published literature exists for the marine ecosystem of the Prince Edward Islands with many of the publications referred to in Chapter 1 of this dissertation. The aim of this section is to summarise the literature to provide the reader with a comprehensive overview of the ecosystem starting with the primary producers through to the top predators. This includes a summary of the pelagic components of the system (phytoplankton, zooplankton, nekton, fish), the benthos and the land-based vertebrates (seals and seabirds). A species list is provided in Table 2.1, which aggregates the marine ecosystem into 37 functional groups. Biomass estimates for the major components of the marine system have been quantified in preparation for construction of the ecosystem model, the spatial extent of which is the Exclusive Economic Zone (EEZ). Data for the islands exist in a variety of forms, dependent on the field of expertise / discipline of the original research. In many instances, raw data in the form of abundance estimates (in either absolute or relative terms) had to be converted into biomass estimates, which involved a number of conversion factors and assumptions. All conversion factors have been described and the final biomass estimates presented in absolute values (metric tonnes) or per unit area (t.km-2/g.m-3), whichever is more appropriate. (In general, biomass estimates for land based fauna are given in absolute terms while aquatic groups are given per unit area). Source data and derivations are explained in the text. Through this data compilation, where available, the long term changes in populations of the fauna of the marine ecosystem are illustrated and, finally, a note on the possibility of the islands having a ?carrying capacity? is discussed.    16  2.1. The marine ecosystem: A review and biomass estimates 2.1.1. Primary producers Open ocean production As described in the oceanography section of Chapter 1 (Section 1.3.3), the islands fall in the Antarctic Polar Frontal Zone of the Antarctic Circumpolar Current between the Sub-Antarctic Front (SAF) in the north and the Antarctic Polar Front (APF) in the south. The Sub-Antarctic water found to the south of the SAF, is cool (less than 10?C) and fresh with high nutrient concentrations but with low chlorophyll concentrations (Hempel 1985), known as HNLC waters. The open ocean phytoplankton is dominated by the nano- and picophytoplankton size fractions with the contribution of microphytoplankton to the whole phytoplankton community usually accounting for less than 20% of the total community (El-Sayed et al. 1979a; Froneman et al. 1998; Read et al. 2000; Thomalla et al. 2011). It has been frequently argued that in the Southern Ocean a lack of stability and deep mixing lead to light limitation and accounts for low chlorophyll biomass in this region (Mitchell-Innes 1967; Nelson and Smith 1991). However, Read et al. (2000) found the shallow layer (less than 25 m) of westward moving surface water had an adequate light environment and determined iron limitation to be the most probable cause for the absence of a diatom bloom in these surface waters. Beneath the surface layer, the observed sub-surface biomass maximum of small nanoplanktonic cells (less than 20 ?m) is largely trapped in the layer 40-80 m between two density maxima, although low cholorophyll concentrations are found to depths of 120 m (Read et al. 2000).  Elevated production is associated with the frontal features (both the SAF and the APF) as they appear to have particularly pronounced vertical stability and are associated with enhanced upwelling of nutrients and leakage of nutrients across the front (Read et al. 2000). This observation for the fronts in the region of the PEIs, is also true of the eddies observed in the vicinity of the islands where shallow mixed layers and enhanced nutrients provide favourable conditions for phytoplankton growth (Lutjeharms et al. 1985; Pakhomov et al. 2000b; Ansorge et al. 2004). Elevated phytoplankton standing stock has also been observed downstream of the islands (approximately double the upstream observations), associated with a downstream wake 17  (Perissinotto et al. 2000). This increased productivity filters up through the food web to the higher trophic levels. For instance, at-sea distribution of Grey-headed albatross and King penguins have been shown to be closely related to frontal features (Bost et al. 1997; Guinet et al. 1997; Nel et al. 2001).  In situ data for the phytoplankton community of the PEIs is almost exclusively limited to the austral autumn because the SA Agulhas (the research and supply vessel used by the South African National Anatarctic Programme) usually visits the islands just once a year to resupply the base. During this period, the marine surveys are conducted. Open ocean in situ measurements of chlorophyll-a (chl-a) in the vicinity of the PEIs range from <0.1 - 0.8 mg.m-3(Froneman and Balarin 1998; Froneman et al. 2000a; Pakhomov et al. 2000b; Bernard and Froneman 2005; Hunt et al. 2008; McQuaid and Froneman 2008).  In an effort to increase both the spatial and temporal coverage of the dataset, remotely sensed ocean colour satellite data were analysed.  The data used was at 1 km resolution over a six by six degree area centered over the islands (44?S to 50?S, and 35?E to 41?E) from SeaWiFS (1998 to 2004) and a two by two degree area (45.8?S to 47.8?S and 36.8?E to 38.8? E) from MODIS-Aqua (2005-2008) (see Figure 2.1). For both instruments, daily Level-2 standard chlorophyll products were downloaded from the Ocean Color Web (Feldman and McClain 2007b) and mapped to a cylindrical coordinate system that retained the original 1-km spatial resolution. SeaWiFS standard chlorophyll (Reprocessing Version 5.1) was derived using the OC4v4 pigment algorithm (O?Reilly et al. 1998; O'Reilly et al. 2000). MODIS-Aqua standard chlorophyll (Reprocessing Version 1.1) was derived using the OC3 pigment algorithm (O'Reilly et al. 2000). Data from both instruments were derived using near real time meteorological and ozone data for atmospheric correction. Standard Level-2 quality flags were used to mask poor quality data (Feldman and McClain 2007a).  The OC3 pigment algorithm is intended to yield MODIS-Aqua chlorophyll estimates that are directly comparable with chlorophyll estimates derived from the SeaWiFS instrument and the OC4v4 pigment algorithm (See for example: Franz 2005; Zhang et al. 2006).  18  Images of monthly composites are included in Appendix 1.A. of this thesis. Average annual chl-a values from the larger area and earlier period (1998 ? 2004) from the SeaWiFS satellite ranged between 0.22 mg.m-3 and 0.26 mg.m-3 (mean 0.24, ? SD 0.04 mg.m-3), while data for the small area (and later time period 2005 - 2008) were lower  with  a range between 0.17 mg.m-3 and 0.21 mg.m-3 (mean 0.19 ? SD 0.01 mg.m-3) (Figure 2.2). An annual average of 0.22 mg chl-a.m-3 was calculated from all years and is in agreement with historic in situ data (e.g. 0.22 mg chl-a.m-3 (El-Sayed et al. 1979a); 0.2 mg chl-a.m-3  (Boden 1988), 0.22 mg chl-a.m-3 (Gurney et al. 2002)).  To convert the chl-a estimates to biomass, a chl-a:Carbon ratio of 32.25 was used (based on the relationship from Hewes et al. (1990): C = 80 chl a0.6, using the value of 0.22 mg chl-a.m-3); a conversion factor of x 10 for Carbon to wet weight (Dalsgaard and Pauly 1997); and a euphotic depth of 120m (Read et al. 2000), resulting in an estimate of 8.51 t.km-2 (or a total of 3667929 t for the EEZ). This was divided into a microphytoplankton component (size >20 ?m) of 1.70 t.km-2 (~20% of the total phytoplankton) and a nano and picophytoplankton component of 6.81 t.km-2 (~80% of the total phytoplankton) based on the relative contributions of these size fractions to the open ocean phytoplankton assemblages.  Island associated blooms Since the earliest oceanographic studies at the islands, phytoplankton blooms in the vicinity of the islands have been observed. The perception of the blooms was that they were a relatively local phenomenon. With the advent of remote sensing of chla a from satellite observations, it has become clear that the blooms extend away from the islands in various directions (see Appendix 1. A of images from SeaWiFS and MODIS-Aqua), and are therefore referred to here as ?island associated blooms?. Results from in situ measurements of chl-a from within the vicinity of the islands range from 0.01 - 2.8 mg chl-a.m-3 (El-Sayed et al. 1979a; Miller et al. 1984; Allanson et al. 1985; Boden 1988; Duncombe Rae 1989a; van Ballegooyen et al. 1989; Perissinotto et al. 1990b; Froneman et al. 2000a; Perissinotto et al. 2000). Phytoplankton blooms are considered to occur when chl-a concentrations exceed 1.5 mg.m-3 (Boden 1988; Duncombe Rae 1989a). The blooms are 19  dominated by diatoms (Boden 1988). They occur seasonally and are usually the result of increased production of the chain-forming Chaetoceros radiacans (Boden et al. 1988), Rhizoselena curvata, and Dictyocha speculum (Perissinotto 1992) or Fragilariopsis spp. (McQuaid and Froneman 2008). They are thought to occur during periods of water retention in the vicinity of the islands, when reduced flow through results in a build-up of fresh water runoff carrying nutrients from the islands, and creating water column stability (Perissinotto and Duncombe Rae 1990; Perissinotto et al. 2000). Water retention needs to be prolonged (at least 14 days duration) to allow for phytoplankton build up (Perissinotto et al. 1990b). Under non-bloom conditions, chl-a concentrations range between 0.05 and 0.45 mg.m-3, with the composition dominated by nano- and/or picophytoplankton (Perissinotto et al. 2000; Bernard and Froneman 2002) similar to the open ocean phytoplankton composition.  The in situ data is limited in its temporal resolution (mostly austral autumn estimates). In an effort to increase the spatial and temporal resolution of the data, remotely sensed ocean colour satellite chl-a data at a 1km resolution from a subarea centred on the islands (46.5?S to 47.1?S and 37.5?E to 38.3?E) from 1998 to 2008 was processed (SeaWiFS data 1998 ? 2004, MODIS data 2005-2008) (see Figure 2.1). Monthly averages peaked during the summer months and values of up to 1.4 mg chl-a.m-3 showed clear seasonal blooms in the vicinity of the islands. The annual averages for these data range between 0.18 and 0.42 mg chl-a.m-3, with the average chl-a for the sub-area 0.27 mg chl-a. m-3 (Figure 2.2). Therefore an increase of 0.05 mg chl-a.m-3 over and above the open ocean value of 0.22 mg chl-a.m-3, was attributed to the elevated production associated with the islands, contributing 20% to the productivity in the area in which it occurs. To convert the chl-a estimate to biomass: the value of 0.05 mg chla. m-3 was divided by 1000 to convert to g (from mg), then multiplied by 32.25 to convert from chl-a to carbon (see above), then multiplied by 10 to convert from carbon to wet weight (Dalsgaard and Pauly 1997) and multiplied by 25 based on the euphotic depth being approximated to be 25m (Knox 2007a). This resulted in an estimate of 0.403t.km-2 for the area in which it occurs (assumed to be an area within a circle of radius 20nm, equivalent in size to the subarea of satellite data used), equating to an estimate of approximately 1737.47 t total (or 0.00403 t.km-2 for the PEI EEZ). 20  Interestingly, zooplankton grazing studies during bloom conditions showed no evidence of grazing on microphytoplankton (>20 ?m) (the size fraction which dominates during such blooms), but grazing occurred in the nano- and pico- size fractions of the phytoplankton (Perissinotto 1992). This lead to the theory that the production from the blooms falls out of the surface waters and provides a direct transfer of primary production from the pelagic to benthic subsystem. This has been confirmed by studies of stable isotope signatures in the pelagic and benthic communities (Kaehler et al. 2000).  Benthic macrophytes Two kelp species, Macrocystis laevis and Durvillaea antarctica dominate the macrophyte biomass and, while the combined production per unit area of these two has been estimated to be greater than that of the phytoplankton production, it was thought to contribute less to the seas around the PEIs because of its limited spatial coverage (Attwood et al. 1991). Macrocystis laevis is endemic to the islands (Hay 1986) and occurs along the lee shore of the islands between 5 and 20 m depth (Attwood et al. 1991). This kelp has been found growing at a depth of 68 m in the open shelf area (Perissinotto and McQuaid 1992a) and the offshore limit is apparently controlled by availability of suitable rock substrata with the inshore limit apparently set by maximum height of storm-induced waves (Perissinotto and McQuaid 1992a). Quantitative estimates of the macrophytes made from photographs and from diving surveys conducted in the 1980s are 63 500 t for M. laevis (Attwood et al. 1991) and 3 300 t for D. antarctica (Haxen and Grindley 1985). Input from macrophyte production into the system by way of particulate or dissolved organic carbon through fragmentation was also thought to be limited as it was suspected that almost all of the production was exported to the open ocean pelagic environment (Attwood et al. 1991). This theory has subsequently been adjusted as evidence from stable isotope signatures has shown that the input of the macrophytes, particularly as particulate carbon, is substantial to the near shore benthic community (Kaehler et al. 2006).  2.1.2. Zooplankton The islands do not have an endemic zooplankton community. Species of subtropical, sub-Antarctic and Antarctic origin are all found (Pakhomov and Froneman 1999b; Pakhomov and 21  Froneman 2000; Hunt et al. 2002). The abundance and distribution of zooplankton groups varies at both temporal and spatial scales (Hunt et al. 2001; Hunt et al. 2002). Zooplankton communities in the vicinity of the islands have been found to be associated with different water masses and water temperature, these factors accounting for as much as 69% of variation in community structure (Hunt et al. 2002). Elevated zooplankton densities are associated with the SAF region and within close proximity to the island plateau (Pakhomov and Froneman 1999a). Results from acoustic surveys suggest that large plankton and micronekton are mostly washed around rather than across the inter-island shelf region (Pakhomov and Froneman 1999a) with the shelf region generally characterised by low average zooplankton size and biomass (Hunt and Pakhomov 2003). Therefore, upstream, inter-island and downstream communities have been shown to have different characteristics with biomass estimates in the inter-island area being the lowest. While the community has been shown to be highly variable, euphausiids (E. vallentini, E. longirostris and Nematoscelis megalops), amphipods, fish, salps (Salpa thompsoni) and chaetognaths (Sagitta gazellae) often dominate the larger zooplankton size fraction (Miller 1982; Pakhomov and Froneman 2000). The mesozooplankton size fraction (and often the biomass on the whole) is dominated by copepods (Grindley and Lane 1979; Hunt et al. 2001; Bernard and Froneman 2002).  A review by McQuaid and Froneman (2008) estimated zooplankton biomass to range between 0.55 and 62.70 mg dry weight.m-3. Pakhomov and Froneman (1999) summarise an intermediary range of between 17 and 45 mg dry weight.m-3 and this estimate was used for the zooplankton biomass in this study.  A dry weight to wet weight conversion of 5 (Cushing et al. 1958) with a depth integration to 200 m, divided by 1000 (mg to g) was used to convert from dry weight to integrated wet weight biomass. Therefore the range in g.m-2 was equivalent to the above estimates quoted in mg dry wt.m-3. An intermediate value of 28 g.m-2 was used and divided between three zooplankton groups with 5 g.m-2 for large crustaceans, 16 g.m-2 for small crustaceans and 7 g.m-2 for other zooplankton (estimates are equivalent to t.km-2). 2.1.3. Benthic ecosystem The benthic community comprises approximately 550 species with seven benthic community groups. There are 200 macrobenthic species, numerically dominated by polychaetes, crustaceans, 22  mollusks, nematodes and echinoderms. Biomass of the benthos has been found to increase with increasing depth (0.12 kg.m-2 at 5 m, 0.34 kg.m-2 at 10 m, 0.46 kg.m-2 at 15 m) (Beckley and Branch 1992). Using an estimate of the shelf area from bathymetry data (Table 2.2, Figure 2.3; source GEBCO_08 Grid version 20100927; http://www.gebco.net), the area around the islands which falls within the 0 ? 500 m is 887.37 km2. Using the value at 15m depth (0.46 kg.m-2), an estimate for the model area (EEZ) would be approximately (408 190 t/ 431014 km-2 =) 0.947045 t. km-2. Data from Perissinotto and McQuaid (1990), however, provide an estimate of approximately 6.4 g.m-2 in the area where the benthic fauna occurs (bryozoa 4.5 g.m-2, asteroidea 1.0 g.m-2; Echinus sp. 0.5 g.m-2, ophiuroidea 0.2 g.m-2, bivalves 0.1 g.m-2, polychaetes 0.1g.m-2). This lower estimate results in a value of 0.013176 t.km-2. For this study a biomass value of 0.5 t.km-2 was used as the average of these two estimates.  Benthic decapod Nauticaris marionis, the benthic decapod, has the second highest crustacean biomass and numerous studies have focused on this species because of its perceived key role in the ecosystem (Perissinotto and McQuaid 1990; Kuun et al. 1999; Pakhomov et al. 1999; Pakhomov et al. 2000c; Pakhomov et al. 2004). The decapod is consumed by some of the top predators on the islands (Brown et al. 1990) and therefore provides a link between benthic production and the higher vertebrates. It is widely distributed around both Marion and Prince Edward islands, but occurs mainly within the 200 m depth contours (Perissinotto and McQuaid 1990a). On the inter-island plateau the decapod occurs in a supra-benthic layer which extends 5-10 m above the bottom. The shallowest depth that it has been recorded at is between 30 ? 33 m, while it has been found at depths as great as 606 m and 775 m (Perissinotto and McQuaid 1990a). Abundance estimates from dredge samples are as high as 25 ind.m-2 but photographs of the bottom show much higher densities with a maximum of about 80 ind.m-2 on the south-east coast of Marion Island and even this is considered to be a significant underestimate. Biomass estimates have been based on data from Branch et al. (1993) where abundance estimates of N. marionis are 20 ind.m-3. If swarms are considered 3.5m deep (or alternatively using a depth integrated estimate of 70 ind.m-2), with a wet weight biomass estimate of an individual equal to 226 mg. The average length of N. marionis is between 25-30mm, so 27mm was used, and the wet weight was 23  estimated at 266mg. Therefore the estimate of biomass was (70 ind.m-2 x 0.266g =) 18.62 g.m-2. Assuming the decapod is found in the area between 0 and 500m (= 887.37 km2; Table 2.2; Figure 2.3; GEBCO_08 Grid version 20100927; http://www.gebco.net), a total of 16 522 t of decapods may be found at the islands. The biomass estimate for the EEZ model is therefore (16 522/431014 t.km-2) 0.0383 t.km-2.  2.1.4. Cephalopods The cephalopods are a key component of the marine ecosystem, yet no directed studies have focused on this important group, and no quantitative data on the cephalopods in the vicinity of the islands exists. Information on species composition comes largely from diet analysis of the top predators (Berruti and Harcus 1978; Cooper et al. 1990; Cooper et al. 1992; Cooper and Klages 1995) which has provided an extensive species list. Ashmole (1968) has suggested that surface feeding seabirds are essentially non selective with respect to the taxonomic affinities of their prey, and therefore suggest that species compositions from such diet analyses may be a true reflection of those in the vicinity of the islands (Berruti and Harcus 1978). Such an assumption however, may bias the species considered to the near surface species. Despite this potential issue, the prey of the two sooty albatrosses (Phoebetria fusca and P. palpebrata) at Marion Island, where 3 295 beaks were analysed (Berruti and Harcus 1978), along with other studies (Cooper and Klages 1995), were used to identify the most important cephalopod families, which were Onychoteuthidae, Histiotethidae, Chiroteuthidae and Cranchiidae. Of the 23 taxa found in the diets locally at the PEIs, Kondakovia longimana was identified as a key species in the diet of albatross species (Wandering and Grey-headed, (Cooper et al. 1992); Dark-mantled Sooty (Cooper and Klages 1995)) and penguins (Kings (Adams and Klages 1987); Macaroni and Rockhoppers, (Cooper et al. 1990)) with Moroteuthis species also important in Grey-headed albatross diets (Cooper et al. 1992) and recorded as the dominant cephalopod in the Patagonian toothfish diets (Pakhomov and Bushula 2003; Pakhomov et al. 2006). Due to the lack of local data, biomass estimates for cephalopods were derived from information from models of other Subantarctic systems. For the Kuergelen Islands model a biomass estimate of 0.355 t.km-2 (Pruvost et al. 2005) was used, and for the Falklands model an estimate of 0.35 t.km-2 (for large cephalopods) (Cheung and Pitcher 2005). Because the PEI system is, in general, 24  less productive than both the Kerguelen and Falklands systems, the estimate has been approximated at one third of these and then partitioned between the small and large groups (approximately 0.1183 t.km-2, divided between large 0.0650 t.km-2 and small 0.0450 t.km-2). These estimates were adequate to supply the system needs and were a best approximation as no data exist.  2.1.5. Fish fauna The two most important families in terms of the fish fauna considered at the islands are the Nototheniids and the Myctophids. At the PEIs the Nototheniids are found in a range of habitats including the inshore small demersal group, large demersals and large pelagics and are the dominant species in these habitats. The Myctophid fish, on the other hand, dominate the small pelagic biomass. The Nototheniids are considered the most successful migrants to the waters of the Sub-Antarctic islands of the Indian Ocean both in terms of diversity and abundance (Gon and Klages 1988). Gobionotothen marionensis and G. acuta are found inshore, the Painted notie Lepidonotohen larseni is found on the continental slope, and the three larger demersal species, the Grey rockcod Lepidonotothen squamifrons, the Black rockcod Notothenia coriiceps, and the Marbled rockcod N. rossii are found in deeper water. In the pelagic domain, the smaller species Paranotothenia magellanica (previously known as Notothenia macrocephalata) is found, along with Dissostichus eleginoides (Patagonian toothfish) which dominates the large pelagic fish community, and for which there is an ongoing fishery. The only other species which has been identified as being of potential commercial value is the Nototheniid, the Grey rockcod (Duhamel et al. 1983; Pakhomov et al. 2006).  Myctophids are the most widely distributed and abundant pelagic fish in the Southern Ocean (Gjosaeter and Kawaguchi 1980; Kozlov 1995; Sabourenkov 1991), and at the PEIs 17 of the 35 species of small pelagic fish belong to this family. In Sub-Antarctic waters the four most abundant species are Electrona carlsbergi, Electrona antarctica, Krefftichthys (Protomyctophum) anderssoni and Gymnoscopelus nicholsi. Seasonality in the occurrence of myctophids has been observed in the diets of top predators at the islands, and this indirect 25  evidence suggests that there is seasonality in the occurrence of these species with different species dominating the diets during different times of the year. However, no pelagic trawling data exist for the islands. The importance of the Nototheniids (particularly the Patagonian toothfish) is evidenced through the fishery. The key role of the myctophids in the system is clear through the diet composition of the vertebrate top predators (e.g., Adams and Klages 1987; Brown et al. 1990), as has also been shown for neighbouring systems (e.g., Crozet Islands (Cherel et al. 1993), Kerguelen Island (Lea et al. 2002)). The myctophids form an important component of the diet of the land based top predators, including the Sub-Antarctic fur seals (> 96% of the diet of Sub-Antarctic fur seals (Makhado 2002), as well as the penguins (e.g., King penguins - making up 86% of diet (Adams and Klages 1987; Brown et al. 1990; Cooper et al. 1990)). They also form an important part of the diet of other nekton (cephalopods) (Lipinski and Linkowski 1988) and fish species (e.g., Grey nototheniid, Lepidonotothen squamifrons) (Pakhomov and Bushula 2003; Pakhomov et al. 2006)).  Studies on Chondrichthys (sharks and rays) have not been conducted at the islands, though rays (Bathyraja tuff, Rajella barnardi, and another unidentified Raja sp.) were collected in the MV Iris survey (Pakhomov and Bushula 2003; Pakhomov et al. 2006). Along with these species, it is presumed that species described from the Kerguelen Islands are also present at the PEIs. Three species are described: the Greenland shark Somiosus microcephalus, Porbeagle Lamna nasus, and Lanternshark Etmopterus granulosus (Cherel and Duhamel 2004). The only quantitative fisheries independent survey for the islands was conducted by the MV Iris in 2001. This trawl survey targeted predominantly demersal species with 54 trawls between 200 m and 1500 m made using a 100 mm mesh net (Pakhomov and Bushula 2003; Pakhomov et al. 2006). The results of the survey showed major shifts in fish community composition occurred at 500-600 m and 800-900 m, which was thought to be a result of physical and biological vertical zonation (Pakhomov et al. 2006). In terms of species, Dissostichus eleginoides was found to dominate at depths less than 500 m, Lepidonotothen squamifrons was found to dominate between 450 and 750 m, Macrourus carinatus at greater than 600 m and Echiodon cryomargarites between 600 and 1000 m (Pakhomov and Bushula 2003). The Commission for the Conservation 26  of Antarctic Marine Living Resources (CCAMLR) fisheries reports on catch data for the Patagonian toothfish fishery, for that portion of catch which falls within the area of CCAMLR, is reported annually in the CCAMLR report series. Estimates of by catch of fish and birds are also included in the reports, as are incidents of by catch of seabirds). Data on catch per unit effort (CPUE) for the Exclusive Economic Zone (EEZ) of the PEIs has been reported and estimates of landed catch as well as initial biomass estimates have been made (Brandao et al. 2002; Brandao and Butterworth 2009). The paucity of data on the fish community at the PEIs made estimating quantitative biomass particularly challenging and highlighted the urgent need for more data on this community. The only quantitative estimates for the fish community are from the abundances recorded for the most common inshore species (Bushula et al. 2005) and data for the Patagonian toothfish, which has been derived from a combination of the fisheries independent survey from 2001 and fisheries catch data for the islands (Brandao et al. 2002; Brandao and Butterworth 2009). With the coefficient of variation around the estimate at 213%, the estimates are considered far from accurate. As no other data exist, relative abundances of the Patagonian toothfish estimates to other community groups from the MV Iris data (Pakhomov et al. 2006) were used to make estimates of this community. In addition, relative abundances of the different fish community groups from the neighbouring Kerguelen system (Duhamel and Hautecoeur 2009) were used to assist in quantifying the biomass of contributions of the respective fish populations using the Patagonian toothfish as the reference quantity. The data from the Kerguelen Islands  in particular proved useful (Duhamel and Hautecoeur 2009), though as noted by Gon and Klages (1988), the fish fauna at the PEIs are inferior to that at the Kerguelen islands and do not support the same biomass, and this was taken into consideration. This has been attributed to the much larger shelf area of the Kerguelen plateau which supports larger populations of more species.  In preparation for constructing an ecosystem model for the PEIs, decisions on which species should be grouped together into functional groups had to be made. These decisions were based on fish size (large considered to be any fish greater than 50cm in length at maturity) and diets, with similar species being grouped together (Table 2.1). Biomass estimates for each of the functional groups for the fish fauna are outlined below. In all instances in this study where the 27  text refers specifically to a functional group and its input or output from the model, it is Capitalized and italicized. Where left in regular text, it refers to information or data that is not specific to the model, but to the species or group in any other capacity.  The large pelagic fish groups include the Sharks and Rays, Large Pelagic Fish and the Patagonian Toothfish. For the Sharks and Rays, catch rates for just one species from this group Bathyraja tuff  (Pakhomov et al. 2006) (in this paper as taaf) show very low occurrence of  0.02 ind.h-1 out of 45.45 ind.h-1 for deep demersals (in assemblage B of their study). The resulting estimate would be very low (at 0.843 t) for the system. A biomass estimate for two species of Bathyraja at the Kerguelen islands found 28431 t, or 12% of the Patagonian toothfish biomass estimate (Duhamel and Hautecoeur 2009). Based on this, the estimate was determined as 0.00036 t.km-2 (from 0.12*0.003; the most recent toothfish survey data from the PEIs). This is in contrast to the estimate of 0.001 t.km-2 made in the Kerguelen model (Pruvost et al. 2005). Currently no estimates for the sharks exist, but the group remains so named in the hope that improvements to the model in future will allow for the inclusion of shark data in this functional group. The biomass estimates for the Large Pelagic Fish group were made following the relative abundance of large pelagic fish in comparison to the local estimates made for the Patagonian toothfish (FPT) (Pakhomov et al. 2006). The value estimated is half the estimated biomass of the FPT stock and was set to 0.042t. km-2, less than half of the estimate for the Kerguelen stock (0.0940 t.km-2, model generated, Pruvost et al. 2005). For the FPT a biomass estimate of 1168 t with a coefficient of variation of 213% was estimated from the voyage of the MV Iris using swept area (Brandao et al. 2002). This amounts to a biomass of (1168/434101.4 =) 0.00271 t.km-2 for the EEZ of the PEI for the 2000s time period. This biomass estimate is presumed to be a small percent of the original biomass that would have been in the system prior to the fishery crash of the mid 1990s. If the current estimate is 5% of original biomass (Brandao et al. 2002), estimated biomass prior to the start of the fishery is assumed to have been between 0.054196  t.km-2 (or 23 359t) and 0.116004 t.km-2 (or 49 999t).  For the Kerguelen Islands, an estimate for the adults of this species was 0.129 t.km-2 (approx 55 600t) (Pruvost et al. 2005), with a more recent estimate of current biomass higher at 0.288 t.km-2 (124000 t), which is approximately half 28  of the estimated fish biomass for the Kerguelen Island system in total (245000 t, 0.57 t.km-2) (Duhamel and Hautecoeur 2009). Estimates for the mesopelagic fish community in the Southern Ocean are 4.5 t.km-2 (Gjosaeter and Kawaguchi 1980), a figure later revisited for many regions but remaining unchanged for this area (Lam and Pauly 2005), which has been used for this study. Estimates for Myctophid fish in the Southern Ocean are generally in this range (e.g., Filin et al. 1991; Kozlov et al. 1991), though these estimates exceed those made by Pakhomov et al. (1996).  Pakhomov et al.?s (1996) data were based on 5 voyages of the South African National Antarctic Program between 1985 and 1995 where dry weight estimates ranged from 0.01 to 1.1 g dry wt.m-2 with an average of 0.138 g dry wt.m-2. Using a dry to wet weight conversion of 5, the average biomass would be 0.69 t.km-2, approximately six and a half times lower than that used in this model. However, the sampling gear used in Pakhomov et al.?s (1996) study was not ideal for sampling myctophids, and may account for the low estimate. There is an urgent need for further research of this particular group and the parameter used in this model should be revised once a comprehensive sampling of the area has been carried out. Large demersal fish species were divided into two groups: a general Large Demersal Fish and a family specific Large Nototheniid Demersal Fish group. For the general Large Demersal Fish group, the data from the fisheries independent survey for all assemblages (Pakhomov et al. 2006) showed the Patagonian toothfish (FPT) Dissostichus eleginoides catch rate was 38.88 ind.hr-1 compared to the rate for the general large demersal fish species (as categorized for this model) which was approximately 55.45 ind.hr-1. Keeping in mind that the trawls were not targeting FPT, the toothfish were probably under represented. A recent trawl survey at Kerguelen showed the large demersal fish group to consist of approximately 50% of the FPT biomass and this estimate was used in combination with the local FPT data. The Large Nototheniid Demersal Fish were considered to be 12% of the FPT population, which leaves a remaining 38% to be classified as Large Demersal Fish. Using this approach with the pre-crash FPT data from the 1980s time period model (0.075t. t.km-2), the approximate value would be 0.0285 t.km-2. This estimate is less than that for the Kerguelen model (0.5 t.km-2) (Pruvost et al. 2005) and Falklands model (>0.4t.km-2) (Cheung and Pitcher 2005). For estimating the biomass of the Large Notothenid 29  Demersal Fish group, the relative proportion of this group, compared to the FPT estimate, was used. Lepidonotothen squamifrons was caught at 4.53 and 1.45 ind.hr-1, compared with 35.55 and 1.93 ind.hr-1 for FPT in community assemblages A and B respectively (which represent different trawl depths) (Pakhomov et al. 2006). The relative contribution of L. squamifrons was therefore between 12 and 75 % of the toothfish biomass for these assemblages. The data for Assemblage A is similar to that found for the Kerguelen population (Duhamel and Hautecoeur 2009) where L. squamifrons and N. rossi were found to have a biomass of approximately 10% of the FPT estimate. Using the lower estimate (12% of the FPT data) with the intermediate estimate of the FPT data (pre-crash) (0.075t.km-2) we get an estimate of  0.00955 t.km-2, or 4119t. The small demersal fish groups were divided into two: the Small Continental Slope Demersal Fish and the Small Inshore Demersal Fish. Of the Small Continental Slope Demersals, Lepidonothen larseni occurs at a density of 0.2 ind.m-2 and the mass varies in wet weight from 0.2 to 8 g (Bushula et al. 2005). If an average fish is considered to weigh 5 g, the biomass estimate would be 1 g.m-2 (equivalent to t.km2). As this fish resides on the slope area there is a need to scale for this area only.  Using bathymetry data (GEBCO_08 Grid version 20100927; http://www.gebco.net) from the vicinity of the islands, an estimate of the shelf area extending from the islands between the depths of 300 m to 1500 m is 1693.11 km2 (see Table 2.2, Figure 2.3). If we presume that this one species is one third of the total for this group, the resulting biomass estimate is (1693.11km2 x 3 t/km2 =) 5079.33 t, or 0.01178 t.km-2. Calculations for the Small Inshore Demersals were based on Gobionotothen marionensis as estimates were available for this species. Using an average weight for G. marionensis of 10 grams (midpoint of that found by Bushula et al. 2005), then the biomass estimate for this species is (0.2 ind x 10 g =) 2.0 g.m-2 (equivalent to 2.0 t.km-2). To scale for the area in which this species occurs, bathymetry data (GEBCO_08 Grid version 20100927; http://www.gebco.net) from the islands was used to estimate the area from 0 m to 300 m depth (= 530.18 km2) (see Table 2.2, Figure 2.3). As G. marionensis is only one of the three species included in this group, the estimate for this group was made using the biomass estimate of this species (2.0 t km-2 x 530.18 km-2) multiplied by 3, equals 3181.08 t which equates to 0.007 t.km-2 for the ecosystem (EEZ of PEIs). 30  2.1.6. Avian fauna In excess of 2.5 million birds from 29 species are recorded as breeding at the Prince Edward Islands, only one of which is restricted to land (the Lesser sheathbill). There are four penguin species (King, Macaroni, Rockhopper and Gentoo) which currently dominate the avian biomass (between 92 and 96%) as well as the overall total land based top predator biomass (between 58% and 78%, depending on time period and method of calculation. See Table 2.3). Five albatross species (including Wandering, Yellow-nosed, Grey-headed, Light and Dark-mantled Sooty) breed at the islands, while numerous others are sighted in the area (Black-browed, Shy, Salvin's, Southern and Northern royal albatross) but are not considered resident. They contribute between 0.35 and 2.12 % of the avian biomass and less than 2% of the total biomass of the land based top predators. In addition there are three large flying birds at the islands, the Northern and Southern Giant petrels and Sub-Antarctic skuas. These are considered apex predators that feed mainly on seabirds and marine mammals during the breeding season (de Bruyn and Cooper 2005; de Bruyn et al. 2007).  Fourteen small flying seabirds are found at the islands. These include the Fairy prion, Salvin?s prion, Blue petrel, Great-winged petrel, Soft-plumaged petrel, Kerguelen petrel, Grey petrel, White-chinned petrel, Grey-backed petrel, Black-bellied storm petrel, South Georgian Diving petrel, Common Diving petrel, the Antarctic and Kerguelen terns. They make up between 2 and 6% of the avian biomass and between 1 and 5% of the total land based predator biomass (depending on time period and adjustments for time spent at the islands) (Table 2.3). Of these, the Salvin?s prions and Blue petrels have by far the highest number of breeding pairs on the islands (population estimates for the two combined at 1 million birds) with hundreds of thousands of pairs of the remaining 12 species (9 petrels, 2 terns and 1 prion). The kelp gull, a predominantly coastal forager and the Crozet shag also breed on the islands.  A review of the known breeding cycles for the islands was completed by Cooper and Brown (1990), and the data from this study is presented in Table 2.4. In summary, King penguins and Wandering albatrosses have the longest breeding cycles, both of which are longer than one year. The Gentoo penguin, the Great-winged and Grey petrels are all winter breeders, while the remaining species are summer breeders. Most of the species are migratory and only spend the breeding portion of the year at the islands.  31  Of the birds found on the islands many are classified globally as ?Near-Threatened? (e.g. Gentoo penguin, Light-mantled Sooty Albatross) or ?Vulnerable? (Macaroni and Rockhopper penguins; Wandering, Grey-headed albatross) as well as ?Endangered? (Indian Yellow-nosed albatross and Dark-mantled Sooty albatross) by BirdLife International (2011) (IUCN Red List for birds. Downloaded from http://www.birdlife.org on 27/09/2011), and are a priority for conservation concerns at the islands. Long term population trends, and corresponding biomass estimates have been summarised for this study (Figures 2.4.A - L) and the data provided (Appendix 2.A.). Average bird mass (summarised in Ryan and Bester 2008) was used with population estimates to construct biomass time series estimates. Descriptions for each group are given below. Penguins The four species of penguin (Kings, Macaronis, Southern Rockhoppers and Gentoos) dominate the seabird community making up approximately 95% of the biomass (Table 2.3). Extensive research on all aspects of their biology has been published on all four species and provides a comprehensive biological data set for this group at the islands (Williams 1982; Adams and Brown 1983; Lacock et al. 1984; Adams et al. 1985; Adams 1987; Adams and Klages 1987; Adams and Wilson 1987; Adams and Brown 1989; Adams and Klages 1989; Brown et al. 1990; Duplessis et al. 1994).  The most recent survey for King penguins sets the breeding population at 442000 individuals, and a corresponding total biomass of approximately 5000 t. This makes the species the single largest contributor to the biomass of the land based top predators at these islands. Population censuses for this species have not been conducted as regularly as those of the other three penguin species, but the data suggest that the population at the PEIs is stable to increasing (Figure 2.4.A, and Crawford et al. 2003d). Declines in populations are evident however for both the Macaroni and the Rockhopper populations. The most recent population estimate for the Macaroni population is approximately 600 000 individuals (Crawford et al. 2009). Between the 1980s and the early 2000s, the Macaronis showed a greater decline in the population on PE (50% decline) (Ryan et al. 2003) than the one on Marion (10-15% decline) (Crawford et al. 2003b), though data 32  from the most recent surveys show declines on both islands to be approximately 30% (Figure 2.4.B, data in Appendix 2.A, and Crawford et al. 2009). This equates to a decline in biomass from approximately 3883 t in the 1980s to 2778 t in 2009. The Southern Rockhoppers have declined even more than the Macaronis, from approximately 180 000 breeding pairs in 1994/5 to 67 000 pairs in 2002/3 (Crawford et al. 2003c), and down to 45 000 pairs in the last census (Crawford et al. 2009), which is one quarter of the 1994/5 population estimate (See Figure 2.4.C). In terms of biomass, decline was from approximately 1 140 t to 449 t over this time period. The Gentoo population is far smaller than those for the other penguins (contribute less than 0.4% of the penguin biomass for all time periods) with current estimates of breeding birds around 3500 individuals and historical fluctuations ranging between 1920 and 4235 individuals. The numbers of Gentoos declined from the mid 1990s, but show recent recovery (Figure 2.4.D, Appendix 2.A.) (Crawford et al. 2003a; Crawford et al. 2009), and have the lowest biomass of the four penguin species at between 15 t and 21 t. It has been suggested that the relative success of the different species of penguins may be linked to the differences in their foraging ranges and hence their dependence on pelagic or near-island resources. King penguins are able to travel the furthest (Adams 1987) and this allows them to forage at the productive frontal systems, which may be some distance from the islands. The smaller penguins are, however, restricted in their foraging range and are therefore more dependent on the island ecosystem. Of the smaller penguins, the Macaroni travels the furthest (59 to 303km) (Brown 1987), followed by the Southern Rockhopper (4 to 157km) (Williams and Siegfried 1980; Brown 1987) and lastly the Gentoo (80% of all trips <40km, (Adams and Wilson 1987)). As both the Macaronis and the Rockhopper populations have been  in decline, the reason may also be related to the food source of each of these species, as both have a predominantly crustacean diet, while the Kings and the Gentoos have a high contribution of fish to their diets (see section to follow on diets). The reduced breeding success that has been documented over the last decade for the Rockhopper penguins has pointed to the birds returning to the islands to breed in sub-optimal condition with lower recorded weights on arrival and sub-optimal conditions for feeding at the over-wintering grounds has been implicated in the breeding failure for these species (Crawford et al. 2008). 33  Breeding cycles of the four penguins species found at PE islands differ with both the Kings and Gentoos present year round while the Macaronis and Rockhoppers are summer breeders (see Table 2.4 (Rand 1955; Cooper and Brown 1990). Gentoos nest in winter, rearing chicks for a few months, while Kings lay eggs in mid-summer and rear chicks for 10 months. Rockhoppers and Macaronis arrive towards the end of October to breed then moult, leaving by the end of April, having spent between six and seven months at the islands (Rand 1955; Williams et al. 1975; Cooper and Brown 1990).  Large flying birds  As already mentioned, this group includes the five albatross species that breed at the islands (Wandering, Yellow-nosed, Grey-headed, and Light- and Dark-mantled Sooty albatross) as well as the Giant petrels (BGPs) (Northern and Southern) and the skuas. Together the albatross populations account for less than 1.65% of the total biomass of land based predators, and less than 2.2% of the total avian biomass. Population dynamics of albatross species (e.g. Wandering and Grey-headed) and other large birds (e.g., the Northern and Southern Giant petrels) have been monitored with long term trends observed (Nel et al. 2002a) and summarized here (Figures 2. 4. E-I, Appendix 2.A). Recent population declines have been linked to the increase in tuna long lining as well as large scale Illegal Unregulated and Unreported (IUU) long-line fishing for Patagonian toothfish (Nel et al. 2002b; Nel et al. 2002c; Nel et al. 2003), though following extensive mitigation measures the most recent estimates for by-catch of birds on the fishing vessels has been significantly reduced (CCAMLR report 2011). Biomass patterns, summarized in this study, show variability for the albatross species over the past decade, though all species show stable (to increasing) estimates. Feeding of these species is not limited to the islands themselves as their foraging ranges are extensive (Nel et al. 2001; Bost et al. 2009). Southern Giant petrels outnumber the Northern Giant petrels approximately 5:1 and species show variation in population sizes through the late 1990s with a general trend to higher populations from the first surveys (1970s) to the present (Williams et al. 1979; Cooper and Brown 1990; Ryan et al. 2009, Figures 2.4.J and K). In contrast the Subantarctic Skua numbers have shown a gradual decline over the same time period (Figure 2.4.L).  34  In terms of breeding patterns, the Wandering albatross are present year round, while all the remaining species may be considered summer breeders (summarized in Table 2.4, Cooper and Brown 1990), spending between 7 (Grey-headed) and 9.5 (Dark-mantled Sooty) months at the islands. Data on the Yellow-nosed albatross is limited as this species is only found on Prince Edward Island which is visited infrequently. Small flying birds Fourteen prion, petrel and tern species, as listed earlier, are included in the group. Population estimates for all species have been limited to broad ranges of 1000s, 10s of 1000s or 100s of 1000s, with only the most recent surveys providing more definitive estimates on the populations found at the islands. The biomass estimate calculated for this study is approximately 282 t, which is similar to that in the review by Ryan and Bester (2008) of around 300 t. The Salvin?s prion, Blue petrel, Great-winged petrel and White-chinned petrel dominate the small flying seabirds in both population numbers (abundance) and also in biomass as together they constitute 89% and 85.7% respectively of these two estimates. The majority of these species are summer breeders with the two exceptions being the Great-winged and Grey petrels, which breed in winter (Table 2.4, from Cooper and Brown 1990). 2.1.7. Mammals Ten cetacean species have been observed in the vicinity of the islands including three Mysticeti species (the Southern Right whale, Humpback whale and Blue whale), and six Odontoceti species (Sperm whales (Physeter macrocephalus), orcas (Orcinus orca) and various small cetaceans which include the Long-finned pilot whales (Globicephala melas), Hourglass dolphins (Lagenorhynchus cruciger), Dusky dolphins (Lagenorhynchus obscuris), Southern Right whale dolphins (Lissodelphis peronii) and the Strap toothed Beaked whale (or Layard's beaked whale) (Mesoplodon layardii). Of these, only the orcas are considered resident at the islands (Condy et al. 1978; Keith et al. 2001; Pistorius et al. 2002; Tosh et al. 2008).  A review of orca populations based on opportunistic observations from 1973 to 1996 shows the average pod size is 3.56 individuals and the maximum size is 28 individuals (Keith et al. 2001). 35  Populations peak October to December, decrease in January, and have a small increase in late April to early May after which killer whale sightings, during most years, decrease to almost nothing (Skinner et al. 1978; Keith et al. 2001; Pistorius et al. 2002; Tosh et al. 2008). No time series data on changes in the population have been documented for the islands. At any one time, if up to 28 individuals are present, the biomass can be as great as 63 t (using Trites and Pauly 1998) average weights, assuming a sex ratio of 1:1), though the biomass estimated for this study is considered to be around 11 t for the year. The Pinnipeds that breed at the islands are the southern elephant seal (Mirounga leonina), the Subantarctic fur seal (Arctocephalus tropicalis) and Antarctic fur seal (A. gazella). Rare sightings of three additional pinnipeds have been made in the area (Leopard seal (Bester et al. 2006), Weddell seal and the South African fur seal)).  The largest of the three species of seal at the islands is the Southern elephant seal Mirounga leonina. Extensive research has been conducted on this seal population which comes to the islands to breed, moult and over-winter (Skinner et al. 1978; Bester and Pansegrouw 1992; Pistorius et al. 1999a; Kirkman et al. 2003; Pistorius et al. 2004). This species has been in decline at the Prince Edward Islands (PEIs) since the 1960s (See Fig 2.4.M). While early total population estimates of around 10 000 individuals (Rand 1962) are not considered accurate, decline rates were estimated at 69.5% of the population based on these earliest observations to 1977 (Pistorius and Bester 1998). Declines since 1977 have been well documented with a further decrease of 66.3% from 1977 to 2004 recorded (Condy 1977; Condy 1978a; Skinner et al. 1978; Condy 1981; Pistorius et al. 1999a; Bester and Hofmeyr 2005; Ryan and Bester 2008) (Appendix 2.B.).  The Southern elephant seal population at the PEIs is part of the Kerguelen stock, which includes populations at neighbouring Crozet and Kerguelen, and the population of this stock as a whole is in decline, as are the populations at other sub-Antarctic islands (McMahon et al. 2005).  The reasons behind the declines are unknown but various hypotheses have been examined. In the past, Condy (1978) considered competition with man for fish in winter feeding grounds, predation by orcas at the islands, and competition for local food resources with fur seals to have led to the population decline. Competition with fisheries and human activity was discounted in 36  the late 1980s (Wilkinson and Bester 1988) but has since been cited as being a potential reason for the decline in population numbers at other Sub-Antarctic islands (e.g., Green et al. 1998; Goldsworthy et al. 2001). More recent studies suggest inter-specific competition and environmental change to be the most plausible explanations (McMahon et al. 2005).  For the PEIs, population a reproductive response to declining populations has been documented. Pistorius et al. (2001b) found that age at maturity declined and fecundity rates increased as the population declined, indicating a compensatory response. These authors propose that relative increase in food availability because of the population decline promoted earlier sexual maturity correlated with more rapid growth of juveniles when population abundance lowered (Pistorius et al. 2001b).  The low pup mortality rate observed between 1990 and 1999 suggested that this was not the major population regulating agent at Marion (Pistorius et al. 2001a; Pistorius and Bester 2002).  Generating biomass estimates for this population involved a number of assumptions regarding the social structure of the elephant seal population. Total population estimates were made using female to pup ratios of 1.05 (Pistorius et al. 1999a) and population to pup ratios of 3.15 (Pistorius et al. 1999a), in contrast to other island populations where a value of 3.50 is traditionally used (Laws 1977) with census data of the respective groups from published literature (Rand 1962; Condy 1977; Condy 1978a; Skinner et al. 1978; Condy 1981; Pistorius et al. 1999a; Ryan and Bester 2008; See Appendix 2.B.). Because male to female ratios are known to vary, two biomass estimates were made using male to female ratios of 1:11 (Rand 1962; Condy 1978a) and 1:16 (Skinner and van Aarde 1983) (who provide adult sex ratios for each year between 1973 and 1982 and they range between 8.07 and 20 with the average = 16.42) combined with seal mass estimates (taking into account sex and maturity) from Ryan and Bester (2008). A third estimate was calculated using the ?average seal mass? (value of 353 ? 137kg) as determined by Condy (1981). All estimates from 1952 to 1999 compared favourably (2.4. M), with standard deviations within 4% of the average value of the three estimates. Population estimates for elephant seals on Prince Edward Island are scarce with estimates from 1970s and 2000s showing the population to be between 30 and 35% of that found on Marion (Condy 1978a; Ryan and Bester 2008) and a value of 32.6% was used to extend the Marion estimate to include the biomass on Prince 37  Edward. Estimates for the elephant seal biomass made in this study are lower (approximately 637 t for Marion and PE combined) than the biomass of 2500 t put forward by Ryan and Bester (2008), but the estimates here are in line with the rates of decline reported in the literature.  Elephant seals do not remain at the islands year round. For Marion, bulls arrive in mid August and cows follow in September. Maximum harem size is reached in mid October, and females leave 4 weeks later. Yearlings start the moult haul-out in mid November, are then joined by adult females from mid December to mid March, and by adult males from late December to mid April (Condy 1979). For the purposes of this study, Elephant seals are considered present at the islands for 8 months of the year.  Two species of fur seals are found on the islands, the Sub-Antarctic fur seal (Arctocephalus tropicalis) and the Antarctic fur seal (A. gazelle). Both were the target of a seal fishery through the 19th century and numbers are thought to be returning to pre-exploitation levels today. Sealing began at the islands in the early 1800s with the earliest recorded sealing activity in 1803, but by 1860 sealing was no longer economically viable. In 1909 an attempt to revive the industry was made without success and finally all sealing is believed to have stopped in the 1930s. Populations have since increased exponentially, and numbers today are believed to be close to pre-exploitation figures. Early seal population estimates date back to the 1950s when the population was estimated at approximately 500 Subantarctic seals and less than 100 Antarctic fur seals (Rand 1956). Population increases have been well documented since (Condy 1981; Kerley 1983a; Wilkinson and Bester 1990a; Hofmeyr et al. 1997; Hofmeyr et al. 2006; Bester et al. 2009) with the Subantarctic seal population currently at ~150000  and still increasing at 5.2% per annum (half the rate of the 1950s ? 2003/4 rate) (Hofmeyr et al. 2006). The Antarctic fur seals are less common at the islands (population estimate of ~5800) and are increasing at 17% per annum (Hofmeyr et al. 2006).  Biomass estimates of 4774 t for Subantarctic fur seal and 226 t for the Antarctic fur seal populations were made using the seal weights provided by Condy (1981), population estimates summarized in Bester et al. (2009), and conversion ratios of population numbers to pup numbers of 4.8 from Kerley (1983a).  This results in a total biomass contribution for the fur seals of 5000 38  t as in Ryan and Bester (2008). The biomass trends over time for these two species are illustrated in Figure 2.4.N.  Along with the marine mammalian fauna found at the islands, a mention should be made of the precence of two land-based mammals introduced to the islands: the house mouse (Mus musculus) and the domestic cat (Felis catus). The mice are thought to have been introduced to the islands when the sealers visited the islands in the early 1800s. In the late 1940s, when South Africa set up the meteorological station, the base soon became infested with mice. As a solution to this problem, five cats were brought to Marion Island. Unchecked, these cats rapidly bred, resulting in a population estimated at in excess of 3000 cats by the late 1970s (Bester et al. 2000). The cats were found to have serious deleterious effects on the breeding populations of birds at the islands, resulting in local extinction of at least one speices. Finally eradicated in the early 1990s, this historical effect of this alien invasion resulted in serious conservation issues for some of the breeding bird populations. These were ultimately resolved with the eradication of the cats, though not without long term implications, including the local extinction of one species.  2.2. ?Carrying capacity? for the islands? The concept of ?carrying capacity? for an ecosystem can be described as the maximum biomass that can be sustained by the available resources (Odum 1983). It is interesting to note that the total top predator biomass that the islands have supported through the documented history has been similar through time (Table 2.3). Biomass estimates made in this study show total biomass at the islands of ~ 16 300 t in the 1950s and ~ 14 700 t in the 2000s, though the species composition has changed. Overall, the contribution of seals to the total biomass has increased from around 24% to up to 38%, although when these estimates are adjusted according to time spent at the islands, the contribution of the seal community for the islands remains at about 20% for the whole period. The penguins, as already noted, form the largest contribution to the overall biomass of top predators at the islands, though the percentage contribution initially increased from 71% to 74% between the 1960s to the 1980s but then drops to 58% in the 2000s. However, if reassessed using the adjusted biomass estimates for time spent at the islands, the contribution remains relatively constant (at around 75%). For all the top predators the most noticeable change between time periods is the change in the seal community from one dominated in terms of 39  biomass by the Southern elephant seals, to one which in 2010 was dominated by the fur seal population (Table 2.3 and Figure 2.5). The total seal biomass in the 1960s was quite similar to that in the 2000s, though the community has changed in composition. Declines in both the Macaroni and Southern Rockhopper biomass estimates are also clearly evident (Figure 2.5). It is well documented that many of the top predators are not limited in their foraging range to the area around the islands, and many spend much of the year elsewhere when not breeding or moulting. However, it seems evident that the islands themselves appear to have a ?carrying capacity? of around 15 000t which can be supported. Whether this biomass limitation is from resources in terms of food or in terms of suitable habitat for breeding and moulting on land is not clear but further investigations of this phenomenon would be interesting to pursue. If one considers the data that have been adjusted to account for the time spent at the islands, there is a downward trend (from 11 980 to 9700) over the assessed time period (1960s to 2000s) in total biomass at the islands and it remains to be discovered if this pattern can be linked to particular drivers in the system.   40  Table 2.1. List of codes used to identify functional groups and the corresponding species names (or Phylum/Class/Order, whichever appropriate) included in this study of the marine ecosystem of the Prince Edward Islands.    41   Table 2.1. Continued. List of codes used to identify functional groups and the corresponding species names (or Phylum/Class/Order, whichever appropriate) included in this study of the marine ecosystem of the Prince Edward Islands.   42   Table 2.2. Area (in km2) considered to be between the contours between 0 and 1500m, as calculated using the GEBCO_08 Grid Version 20100927 from www.gebco.net .    43  Table 2.3. Total biomass (B, t) of all land based top predators for 3 time periods, including adjusted estimates for time spent away from the islands (BTA, t). 44   Table 2.4. Breeding cycles for the seabirds at the Prince Edward Islands (adapted from Cooper and Brown 1990)  45    Figure 2.1. Satellite derived chorophyll-a estimates from the region surrounding the Prince Edward Archipelago (Southern Ocean). Rectangles illustrate area from which satellite remote sensing data was extracted a) Large rectangle represents the 1km SeaWiFS LAC data extracted from January 1998 to December 2004, b) Medium rectangle represents the 1km MODIS LAC data extracted from January 2005 to December 2008; c) Smallest rectangle represents the 1km area LAC data extracted from SeaWiFS (Jan 1998 ? Dec 2004) and MODIS (Jan 2005 ? December 2008), used to quantify the island associated blooms.   46    Figure 2.2. Summary of satellite derived chlorophyll-a data for the open ocean biomass and local area biomass in the vicinity of the Prince Edward Islands. Estimates for 1998 to 2004 were derived from the SeaWiFS Satellite, while estimates for 2005 to 2008 were derived from the MODIS Aqua satellite (see text for details). Error bars represent 1 standard deviation.    47   Figure 2.3. Bathymetry contour lines in the vicinity of the Prince Edward Islands estimated from the GEBCO_08 Grid Version 20100927 (www.gebco.net) bathymetry data.48    Figure 2.4.A-L. Figures illustrating bird biomass time series for penguins, albatross, Giant Petrels and Skua species at the Prince Edward Islands from the 1960s to 2010; 49    Figure 2.4.M-N. Figures illustrating seal biomass time series (M). Elephant seal biomass time series and (N). All seal biomass estimates for the Prince Edward Islands from 1950s to 2010.   50   Figure 2.5. Land based predator biomass for three time periods (1960, 1980, 2000) for the Prince Edward Islands.   51  Chapter 3 Consumption estimates of marine resources by top predators at the Prince Edward Islands 3.1. Introduction Population changes, as observed from long term monitoring programs, play a key role in the changes in consumption of marine resources (Guinet et al. 1996). Consumption estimates of birds and mammals have been made for many marine systems at regional (Sigurj?nsson and V?kingsson 1997) and global scales (e.g., Karpouzi 2005; Karpouzi et al. 2007). Studies have been motivated in some instances by investigations of interactions between top predators and fisheries (e.g., Goldsworthy et al. 2001; Karpouzi et al. 2007). For instance, seal population recovery following the end of sealing in many systems has lead to the hypothesis that the return of the seal population has contributed to the decline or collapse of fisheries stocks. In systems where long term changes have been observed in resident populations of breeding seals and seabirds, the consideration of dietary overlap of species with fisheries, or between groups has been investigated (Goldsworthy et al. 2001). In order to assess the potential impacts that groups may have on each other, or that fisheries may have on groups, assessments of the consumption requirements for the studied groups are required. For this study, the diet composition of the top predators is summarised, and a consumption model put forward to estimate the consumption for the seabirds and seals at the islands. The data are compared between time periods, and with previous estimates of consumption at the islands. In bioenergetic approaches to consumption estimates, detailed information is provided on population structure, differentiation of energetic requirements for different activities, as well as changes in dietary requirements depending on life stages and foraging patterns associated with such developments. Local diet information along with specific energetic content of prey is routinely incorporated into single species bioenergetic models (see Murie and Lavigne 1991; Adams et al. 1993; Perez and McAlister 1993; Boyd 2002; Mecenero et al. 2006; Halsey et al. 2008). When considering all species in an ecosystem, it is not feasible to construct such detailed assessments for each species in a system, but incorporation of the specific local diets and their 52  energetic density can be achieved relatively easily and the difference that this amendment can make to the consumption estimates can be significant (see Chapter 7). For this section, a consumption model which takes into account local diet information is used. The consumption model adopted here is similar to that defined by the ICES Working Group on Seabird Ecology (ICES 2000) and is used in combination with the long term data of the land based top predator populations to provide insight into the change in the consumption of prey species between time periods.  3.2. Methods Only species (or groups) that are considered resident at the island and are land based were included in this analysis. These included the three breeding seals (Southern elephant seal, the Antarctic fur seal and the Sub-Antarctic fur seal), four breeding penguins (Kings, Macaronis, Southern Rockhoppers and the Gentoos), five albatross (Wandering, Grey-headed, Indian Yellow-nosed, Light- and Dark-mantled Sootys) as well as the Giant petrels (the Northern and Southern Giant petrels considered together) and the small flying birds, here called the ?Prions and Petrels?, which includes 14 species (as for section 2.1). Consumption rates for individuals in terms of their energetic requirements were based on Field Metabolic Rates (FMR) for both the mammals and birds as summarised in Nagy et al. (1999, with amendments to selected equations from Ellis and Gabrielsen 2001) and provided in kJ.d-1 (Table 3.1.) In the case of the mammals the equation for ?All mammals? was used (FMR=4.82M0.734 kJ.d-1). For the bird species, order specific equations were used (penguins (Sphenisciformes C= 4.53M0.795 kJ.d-1); albatross, Giant petrels, and prions and petrels (Procellariiformes C=18.4M0.599 kJ.d-1 ? adjusted by Ellis and Gabrielsen 2001 to C=17.9M0.6 kJ.d-1) and for the two terns (Charadriiformes C=8.13M0.77 kJ.d-1 ? adjusted in Ellis and Gabrielsen 2001 to C=8.49M0.77 kJ.d-1); Nagy et al. 1999). Estimates of body mass were taken from the literature. For the three seal species Condy?s (1981) ?average seal mass? was used, and for the bird species/groups data from Ryan and Bester?s (2008) summary of pelagic predators was used. In the two instances where more than one species was considered in a group (i.e. for the ?Giant Petrels? and ?Prion and Petrel? groups), mass estimates are provided based on 53  weighted contributions (according to population numbers) for each contributing species, and the consumption rates were estimated for each species individually, but the consumption estimates used for the group as a whole are calculated from weighted averages (of the biomass contributions) of all species considered in the group. By dividing the daily individual consumption rates of energy required estimated above by the amount of energy in a gram of the diet, it is possible to convert the consumption rates from energy-based (kJ.d-1)  into mass-based estimates (g dry weight.d-1). In order to do this, it is necessary to have the specific diets of each species/group under consideration, along with the energetic content of the prey, and the assimilation efficiencies of each consumer.  To achieve this, a literature review of the unique diets and assimilation efficiencies of species/groups from the islands was compiled. Following this, diet matrices for each species/group were collapsed into six broad prey group categories: vertebrate prey, general fish, mesopelagic fish (dominated by myctophids), cephalopods, crustaceans and a final group which incorporated all remaining prey categories (including benthos and other zooplankton) (Table 3.2). The energetic content of these six prey categories for the region was established from the literature and summarised (Appendix 3.B). An ?average? energetic value per gram of diet was calculated for each species/group using the relative contributions of the prey categories, weighted by their contribution to the energetic content of the diet (kJ.g-1 dry weight) (Table 3.2). Because a fraction of this energetic content is lost during digestion (i.e. not assimilated) the actual energetic value of each gram of diet is decreased in proportion to the assimilation efficiencies of each species/group (Table 3.2). The resulting reduced energetic content of the prey (in kJ.g-1 dry weight) was divided into the estimates of energetic requirements from the consumption estimates (kJ.d-1), resulting in a consumption rate in terms of mass instead of energy (g dry weight.ind-1.d-1). In order to estimate the biomass consumed daily, a dry weight to wet weight conversion of 3.33 was used (water content of prey was assumed to be 70%; Nagy 1987; Nagy et al. 1999). The method described here is the same as the bioenergetic model created by the ICES Working Group on Seabird Ecology (ICES 2000) and used in Karpouzi et al. (2007): 54  ???? =  ???? ???? ? ?????=1 x 1??? Where DFIi is the daily food intake for each species i (g.ind.d-1), ERi is the energy required for each i (kJ.d-1), DCij is the fraction of prey item j in the diet of each i, EDj is the mean energy density of each prey j (kJ.g-1 ). AEi is the mean food assimilation efficiency for each i, and G the total number of prey categories considered. In this study, the ERi was based on the FMR equations of Nagy et al. (1999, with adjustments from Ellis and Gabrielson 2001) for mammals and birds, the composition of the diet (DCij, or fraction of the food prey item) was determined from a literature review of the diets of the top predators at the islands, and then consolidated into six principle prey groups (i.e. G=6). The mean energy density of each prey (EDj) was determined from the literature as was the food assimilation efficiency (AEi) for each consumer.  To extrapolate from individuals to consider the annual consumption of each species/group as well as the land based predators as a whole, data on population estimates along with the proportion of the year that each species/group spends at the islands is required. These calculations were completed for three time periods for which there are data (1960s, 1980s, and 2000s; see section 2.1) and values of total consumption (in metric tonnes) of each of the principle prey groups calculated.  3.3 Results Daily energetic requirements as calculated from Nagy et al.?s (1999) FMRs with amendments as per Ellis and Gabrielsen (2001), for each of the land based top predators are provided in Table 3.1 along with a summary of average mass, the population estimates (details of estimates are provided in Chapter 2) and the percentage of time each species/group spends at the islands. Diets A full review of the diets for all top predators is provided in the Appendix 3.A. In summary, the Southern elephant seals consume a mix of fish and cephalopods, while both fur seal diets have diets dominated by myctophids. Of the penguins, two are principally fish eaters: the Kings with a diet dominated by myctophids and the Gentoos with a principally mixed fish diet, along with 55  some crustaceans, notably the benthic decapod. Both the Macaronis and the Rockhoppers have predominantly zooplankton diets. All albatross species have a mix of fish and squid. The Giant Petrel group have a diet dominated by vertebrates (mainly penguins) while the Prions and Petrel group diet is dominated by crustaceans with fish and cephalopods contributing. A summary of the diet composition of the land based top predator species/groups, summarised into 6 prey categories is provided (Table 3.2). 3.3.1. Assimilation efficiencies The assimilation efficiency for all mammal groups was set to 0.90 (i.e. 0.1 is not assimilated) which was based on digestive efficiencies recorded in the literature for fur seals (93% Miller 1978, Mecenero et al. 2006) and juvenile Steller sea lions 92-96% (Rosen et al. 2000). These figures are similar to those found for harp seals which ranged between 93.5 and 96.6% (Lawson et al. 1997). It is noted that digestive efficiency may vary with age (Rosen and Trites 2000) but this is not considered for the purposes of this model. Assimilation efficiencies for the King penguins, Macaroni and Rockhopper penguins were all set to 76% based on a summary of assimilation efficiencies for other penguins (African penguins (previously ?Jackass? penguin) 74% on a diet of fish (Cooper 1977) and King penguins 81% on a cephalopod diet (Adams 1984), fish diet (75.5%) and squid diet (73%) (Adams 1984, and Adams unpublished, both in Adams et al. (1993)). For the Gentoo penguin a value of 80% was used based on Clark and Prince (1981) and Abrams (1985). For Prions and Petrel group, a value of 76% was used as in Brown (1989), which was based on a study of White-chinned petrels fed on crustaceans, squid and fish (Jackson 1986). A summary of assimilation efficiencies is given in Table 3.2.  3.3.2. Energetic content of prey Six prey categories were nominated based on the difference in energetic density and their importance in the diets of the species/groups considered: vertebrate, general fish (all excluding small pelagic fish), mesopelagic fish (i.e. small pelagic, dominated by Myctophids), cephalopod, crustacean and other. A literature review of the energetic density of prey which occurs at the islands was conducted. The highest energetic content of prey (kJ.g-1 dry weight) was for vertebrate prey (30.7 kJ.g-1 dry weight) which was based on only two wet weight data values 56  (seal blubber, DW Doidge pers comm. and penguin carrion, Burger 1981; both in Hunter 1985). The next most energetically rich prey group considered is that of the mesopelagic fish (dominated by myctophids) at 27.4 kJ.g-1 dry weight (n=17). Data from Tierney et al. (2002) show that energetic content of myctophid fish from Macquarie Island ranges between 22.6-59.3 kJ.g-1 dry weight, which suggests that this value may be an underestimate. A value for all other fish (excluding the mesopelagic fish) was based on much smaller data set (n=5), and a value of 21.3 kJ.g-1 was settled on. This value is in kJ.g-1 dry weight of fish, and is higher than values used in previous studies (Appendix 3.B.) which have a kJ of wet weight value of around 3.98 kJ.g-1 wet weight for fish (Croxall and Prince 1982) which when converted to dry weight (using a factor of 4, which from the literature appears to be most appropriate for this group; see Cherel and Ridoux 1992) gives a lower value of approximately 15.92 kJ.g-1 dry weight.  Cephalopods had the lowest energetic content at 10.3 kJ.g-1 dry weight, which was also based on wet weight estimates with conversion to dry weight values based on water content of 70%. Crustaceans were set to have an energetic value of 14.8 kJ.g-1 dry weight, calculated from a wet weight value of 4.45 kJ.g-1 assuming a wet to dry weight conversion of 3.33 (ie. water content 70%) and are in line with all estimates available (e.g., 4.35 kJ.g-1 for krill; Croxall and Prince 1982). All remaining diet categories were put together in a group named ?Other?, which consisted of benthic prey along with non crustacean zooplankton, and arbitrarily given the energetic value of 12 kJ.g-1 dry weight. Using these data, in conjunction with the relative proportions (in terms of mass) that each prey category is found in the diet of each species/group, an average energetic value of a unit of mass of each diet for each species/group was calculated (Table 3.2).  When dry weight estimates were available they were used, when unavailable wet weight conversions assuming 70% water content were used. A summary of the energetic content of prey for the six selected categories is given in Appendix 3.B.  Land based top predator community consumption Using estimates for time spent at the islands based on the breeding seasons for all species/groups considered (Condy 1981; Cooper and Brown 1990), along with published population estimates (see Section 2.1), calculations were made for three time periods for which there are data (1960s, 57  1980s, and 2000s). An approximation of total consumption (in metric tonnes) by each of the land based predators of each of the principle prey groups was made and is provided in Table 3.3. 3.4. Discussion The total consumption for the system was 589 601 t in the 1960s. This declined to 510 750 t in the 1980s and subsequently increased to 520 899 t in the 2000s (Table 3.3). In the 1960s the mesopelagic fish and the crustaceans were the two most important prey items representing 42.8% and 43.1% of the total consumption respectively. By the 2000s estimate, the mesopelagic fish dominate the consumption (56.2%) with the crustaceans reduced (31.7%). The ?cephalopod? contribution remained fairly constant through time (at ~ 11%), while the ?vertebrate?, ?general fish? and ?other? groups all made low contributions to the overall consumption (less than 2%). The highest contributor to the change was the consumption of mesopelagic fish by the Subantarctic fur seals which increased exponentially in their population during this time (from about 4000 individuals in the 1960s to over 140 000 by the year 2000; see Table 3.1).  The most important component of the seabird consumption was the mesopelagic fish and the crustaceans. In the 1960s estimate the contribution from each of these two groups was approximately equal (43.4% and 45.1% respectively), but by the 2000s assessment the mesopelagic fish had a higher contribution to the overall consumption (Fig 3.1). This was a result of the decline in both the Macaroni and Rockhopper populations (predominantly crustacean eating) coupled with the steady King penguin population estimate (which had a diet dominated by mesopelagic fish).  For the seal consumption, the early estimates from the 1960s  show cephalopods to be most important (51.5%), followed by mesopelagic fish (29.7%), and fish in general (19.8%), reflecting the Southern elephant seal diet. By the 2000s the mesopelagic fish were the most important prey group (93.6%), with the cephalopods decreased to 4.4% and the general fish group to 1.1%. These changes reflect the decline in the Southern elephant seal population and the recovery of the fur seal populations between 1960 and 2000 with the seal community consumption in 2000 reflecting the diet of the fur seals which is comprised predominantly of myctophid fish (see Table 3.2, Figure 3.1.B and Appendix 3.A). In terms of total consumption, the estimates for the 58  Southern elephant seals in the early period is around 24 125 t, which decreased to approximately 4 061 t by the 2000s. In contrast the consumption by the fur seals initially was low (1 660 t) but by the 2000s estimate reached around 59 555 t. Dietary overlap between these two groups is thought to be limited: both are believed to take myctophid fish, however, the contribution to the Southern elephant seal diet is considered to be limited (<10% of the diet) while this prey dominates the fur seal diets.   When all land based top predators are considered together, the pattern observed in the seabird community was clearly evident, which was due to the seabirds having the highest contribution to the overall biomass of this group. In the 1960s assessment, mesopelagic fish and crustaceans were equally important, but by the 2000s the mesopelagics dominated followed by crustaceans, a trend evidenced in the seabird analysis but emphasized by the changes in the seal community consumption as well. Cephalopods were the third most important prey item and remained between 10 and 12% of the overall consumption for all years. Comparison to other estimates An assessment of the consumption of penguins at the PEIs was made by Adams et al. (1993) for the 1980s period. Using a bioenergetic model, the total penguin consumption was estimated to be 880 000 t. In comparison, estimates from this study (considering only penguins for the same period) show the consumption to be approximately half of this estimate (at 442 083 t). The energetic content of prey, however, for this study was based on very similar energetic content of prey for all groups, excepting the fish, which were estimated to have a value of 3.97 kJ.g-1 wet weight, which is equivalent to 15.88 kJ.g-1 dry weight (just over half the value used in this study for myctophid fish, and also less than the value used for all other fish). Recalculating the consumption for the penguins using these data, the consumption for the 1980s increases to 649 778 t, an increase of 1.35 times the estimate in this study, but still 1.35 times less than the estimate from Adams et al. (1993). Differences in consumption between the two studies include a higher contribution of crustaceans to the diet of the penguin community in this study (38% versus 18% in Adams et al. 1993), and Adams et al. (1993) having a higher contribution of mesopelagic fish (70% versus 51% in this study). The contribution of cephalopods is similar 59  between studies (10% this study, 11% Adams et al. 1993) with all other groups less than 1.5% for both studies. For the smaller birds found at the PEIs, food consumption estimates for Marion Island only had estimates for Whitechinned petrels at 5240 t; Great-winged petrels 12 576 t; Blue petrels 1798 t; and Salvin?s prions 28 685 t with most of the consumption taking place in summer except for the Great-winged petrels which occur and feed at the islands in winter. The total consumption for these groups was estimated to be 48 299 t, which is higher than the estimate here of 35 304 t despite this estimate being for both Marion and Prince Edward Islands. Seal consumption for the system was estimated here to be approximately 59 555 t per year for the Subantarctic fur seals, and 1 574 t for the Antarctic fur seals for the 2000s period. Makhado (2002) provided consumption estimates for the fur seal populations from the 1990s. Estimates of 185 986 t and 11 314 t per year for the Subantarctic and Antarctic fur seals respectively were found and far exceed the calculations done in this study. Differences in the assessments include that Makhado (2002) uses a different consumption equation (Nagy 1994 versus Nagy et al. 1999), an age structured model, and bases the calculations for the whole year (365, instead of pro-rata for time spent at the islands). The calorific content of the diet (particularly the myctophids as they dominate both fur seal diets) is similar in both studies (7.00 kJ.g-1 in Makhado 2002), though an attempt to recalculate the assessments of the model used in Makhado?s (2002) study suggest that this value may have been used as a dry weight and not wet weight when converting from kJ required to grams consumed per day.   The accuracy of the estimates in this study is determined by the various input parameters, which include population estimates, diet matrices, energetic density of prey, assimilation efficiencies and the field metabolic rates. In bioenergetic studies, care is taken to consider the different energetic requirements needed for different phases of development (juvenile/adult, breeding/ non-breeding) and these factors have not been taken into account in this study. No formal sensitivity test has been carried out on these results and therefore the uncertainty is not addressed. In previous studies on consumption rates of communities, population estimates and energetic requirements (existence metabolism values, flight or swimming activity levels and their energy costs) were found to have the greatest impact on the consumption estimates (Furness 60  1978, seabird community, Shetland Islands). The early population estimates for most of the groups, which were taken from the literature and used in this study, are not considered accurate, but more recent estimates may be considered more accurate. Regarding the estimates of energetic requirements, the estimates are made using widely accepted field metabolic rates (Nagy et al. 1999), though the study lacks the detail that would be included in bioenergetic assessments of single species. The objective is to provide an overview of the consumption of the community with the focus on the changes observed between time periods, therefore any assumptions are made across all time periods, providing some insight as to how the system has changed over time (as argued in Guinet et al. 1996). 3.5. Conclusions In conclusion, the consumption estimates here are, in general, lower than earlier previous estimates (for the same time periods) which are due in part to a variety of reasons including, (a) the choice of consumption model, (b) use of a different calorific value of the prey, and (c) consumption estimates in this study including consumption only for the duration of the time spent at the islands. Future consultation to improve the population estimates and dietary habits for all the land based predators would improve the consumption assessment. Estimates from this study show that there has been a change from a system with both crustaceans and myctophid fish being equally important to one where myctophid fish are the dominant prey item for the land based top predators. This reflects the population changes with the increasing fur seal population (diet dominated by myctophids) and the declining Southern Rockhopper and Macaroni penguin populations. 61   Table 3.1. Species list (common names), average weight, field metabolic equations used, daily consumption rates for individuals (in terms of wet weight), population estimates for the 1960s, 1980s and 2000s and proportion of time spent at the islands.     62  Table 3.2. Energetic content of six prey categories provided with summarised diet matrix, average energy density of diet and assimilation efficiencies for each species.   63   Table 3.3. Consumption (in t per year) of each species/group by prey category for A)1960s, B)1980s and C)2000s.   A B C 64   Figure 3.1. Community consumption estimates of the six prey categories for A) Seabirds, B) Mammals, C) All land based top predators   65  Chapter 4 Quantifying local and advected resources: Relative importance of the producers in the system at different spatial scales 4.1. Introduction Oceanic islands are known to have an effect on the surrounding waters, resulting in increased productivity. This phenomenon is known as the ?island mass effect? (Doty and Oguri 1956). The Southern Ocean is, in general, characterised as a high nutrient, low chlorophyll (HNLC) region. However, many sub-Antarctic islands, including the Prince Edward Islands (PEI), are hotspots of biological productivity and are host to millions of seabirds and seals which use the islands as a breeding ground and refuge (Pakhomov and Chown 2003).  A conceptual model called the ?Life Support System? (LSS) has been created for the PEIs to describe how these islands are able to support the high biomass of top predators on the islands (Pakhomov and Froneman 1999b; Pakhomov and Chown 2003). The LSS as it stands now has 'offshore' and 'inshore' components that are named after the categories for the source of the supporting production. The ?offshore? component may be considered the allochthonous, or oceanic input, production that is advected into the system. This open ocean productivity is generally low (Hempel 1985; Pakhomov and Froneman 1999b). The ?inshore? component is the autochthonous production of the system and includes both island-associated phytoplankton blooms and near-shore macrophyte production. It is this production that is associated with the ?island mass effect?. Islands create the ?island mass effect? in a number of ways. It has been demonstrated that, in the HNLC Southern Ocean, input of iron from islands and their shallow shelves into the surrounding waters increases productivity (Atkinson et al. 2001; Blain et al. 2001; Blain et al. 2008). Islands also create a disturbance in the flow of the incident current, resulting in turbulence and mixing which may provide suitable conditions for elevated production associated with the islands, also known as ?island stirring? (Mann and Lazier 1996). At the neighboring Crozet archipelago (Atkinson et al. 2001; Bakker et al. 2007) and the Kerguelen plateau (Armand et al. 2008), as well as at the PEIs (Perissinotto et al. 2000), observed increases in productivity in the vicinity of, or downstream of the islands has been attributed to the ?island mass effect?. Island shelf regions 66  provide substrata for near-shore macrophyte production, and runoff from the islands may carry nutrients from top predator populations that reside on the islands, aiding the benthic production and resulting in surface stratification, potentially enhancing local phytoplankton productivity as observed at the PEIs (Perissinotto and Duncombe Rae 1990). The oceanography around the PEIs is complex because of a combination of the frontal features, the bathymetry and the interaction with the islands themselves (Ansorge and Lutjeharms 2002; Ansorge and Lutjeharms 2003; Ansorge and Lutjeharms 2005). Two frontal features are found in the vicinity of the islands: the Sub-Antarctic Front (SAF) to the north and the Antarctic Polar Front (APF) to the south (Deacon 1983; Lutjeharms 1985; Nowlin and Klinck 1986) (Figure 4.1). The position of these fronts is dynamic and variable (Lutjeharms and Valentine 1984; Duncombe Rae 1989a; Lutjeharms et al. 2002) and their proximity (the SAF in particular) is hypothesized to have a significant effect on the hydrodynamics at the islands. When the SAF is close to the islands, the increased current velocities associated with this front are predicted to lead to a flow-through system between the islands (Pakhomov and Froneman 1999a; Ansorge and Lutjeharms 2000) and advected sources of primary productivity dominate the system. Conversely, when the SAF is further north, lower velocities of the inter-frontal zone are predicted to result in water retention over the inter-island shelf and, as a consequence, little exchange of inshore/offshore waters over the inter-island region occurs (Perissinotto and Duncombe Rae 1990). Cross advection over the PEI shelf when the SAF is far to the north of the islands has been observed (Hunt et al. 2008), however there is little doubt that phytoplankton blooms can and do occur in the vicinity of the islands (Allanson et al. 1985; Duncombe Rae 1989b; Perissinotto and Duncombe Rae 1990). A number of mechanisms for on-shelf water retention have been postulated in the past, including upwelling (Grindley and David 1985), eddy formation (Allanson et al. 1985) and Taylor column formation (Perissinotto and Duncombe Rae 1990). However oceanographic data to support such findings are lacking. More recent studies have found mesoscale eddies up- and downstream of the islands (Ansorge and Lutjeharms 2002; Ansorge and Lutjeharms 2003; Ansorge et al. 2004; Durgadoo et al. 2010). Observed elevated phytoplankton production may be the result of the trapping of these eddies over the island shelf, as has been observed at other Southern Ocean islands, e.g., South Georgia (Whitehouse et al. 1999). A summary of knowledge points to the dynamics of frontal systems as the drivers of 67  shelf-water retention, and consequently of the relevant contributions of allochthonous and autochthonous phytoplankton (Perissinotto et al. 2000; Ansorge and Lutjeharms 2002). The second form of autochthonous production is composed of benthic macrophytes. Two kelp species dominate the biomass, the endemic Macrocystis laevis (Hay 1986) and Durvillaea antarctica (Chamisso) (Hariot 1892). Macrocystis laevis occurs along the lee shore of Marion island, and is generally found between 5 and 20m depth (Attwood et al. 1991) but may be found at depths of up to 68m in areas where the substratum is favourable (Perissinotto and McQuaid 1992a). Durvillaea antarctica occurs in the infralittoral fringe of the islands (Beckley and Branch 1992). The aim of this study was to quantify, for the first time, the relative importance of the three sources of primary production to the PEI food web. The relative contribution of each of the three primary production sources will differ, depending on the spatial scale at which the system is considered and we set out to test this by developing Ecopath models with four different boundary sizes. A preliminary investigation into the ecosystem boundary size, as determined by the centrally placed foragers, found that all energetic requirements for the inhabitants of the PEI could be met at the scale of the Economic Exclusion Zone (EEZ) (see Chapter 6). For this reason the EEZ i.e. a circular area of 200 nautical mile (nm) radius centred on the islands was chosen as the largest model considered, and a series of sequentially smaller models, the smallest of which was chosen to represent the island shelf area (radius of 20nm) was used. Using the top down approach of assessing consumer impacts on primary producers and ecosystem mass balance, we aimed to quantitatively evaluate the previous conceptual models of the PEI LSS. 4.2. Methods The ecosystem model A mass-balanced network model (Ecopath) was used to construct an ecosystem model of the Prince Edward Islands. The basic model is a closed system formulation of the functional groups in an ecosystem and the full methodology can be found in the user's guide (Christensen et al. 2004, Christensen et al. 2008). Each group (functionally related group or a single species) is 68  represented by their biomass and the groups are linked through their trophic interactions. The Ecopath model is based on two fundamental equations, one to describe the production term of the system:  Eq 1: Production = predation mortality +fishing mortality + biomass buildup + net migration + other mortality;  and another that satisfies the energy balance within each functional group:  Eq 2: Production = Consumption - Respiration - Unassimilated food.  Although the Ecopath model has the capability to include flows into or out of the system (Eq 1), for the purposes of this model, no net migration, biomass build-up, nor fishing was considered. The first equation deals with balancing the production terms between groups for the model while the second equation ensures that each group is balanced within. The equations are linked through the common production term. For a system with n groups, n linear equations can be written and a series of simultaneous equations are set up for each group and solved using a generalized method for matrix inversion, which is described by Mackay (1981). The routine solves for one of four parameters for each group, which are biomass (B), production/biomass ratio (P/B), consumption/biomass ratio (Q/B) or ecotrophic efficiency (EE), a measure of how much of a group?s production is used within the system.  The data for the life-support system model The ecosystem model was produced to represent the system in the 1980s. The dataset used for this study is an early version of the data and in condensed form (see section 3.1). Biomass estimates for all biota and rate measurements (production and consumption), along with diet compositions and assimilation rates, were made from the published literature of PEI and other sub-Antarctic systems. Trophic linkages were made from diet and stable isotope signatures. The 21 functional groups were divided into 15 consumers, 4 primary producers and 2 detrital groups. Where possible, data from the PEI were used; parameters were otherwise taken from ecosystem models of similar systems (Bradford-Grieve et al. 2003; Cheung et al. 2005; Pruvost et al. 2005). 69  Biomass estimates The mammalian and avian top predators are represented by 5 groups, the seals (comprising the Southern elephant seal Mirounga leonine (Linnaeus 1758), sub-Antarctic fur seal Arctocephalus tropicalis (Grey 1872) and the Antarctic fur seal Arctocephalus gazella (Peters 1875)), the penguins (Kings, Aptenodytes patagonicus (Miller 1778), Macaroni Eudyptes chrysolophus (Brandt 1837), Southern Rockhopper Eudyptes chrysocome filholi (Hutton 1878) and Gentoo Pygoscelis papua (Forster 1781)), the albatross (Wandering Diomedea exulans (Linnaeus 1758), Grey-headed Thalassarche chrysostoma (Forster 1785), Yellow-nosed T. carteri (Rothschild 1903) and both Light-mantled Sooty Phoebetria palpebrata (Forster 1785) and Dark-mantled Sooty P. fusca (Hisenberg 1822)), the Giant Petrels (both Northern Macronectes halli (Mathews 1912) and Southern M. giganteus (Gmelin 1789)) and a group representing the majority of the small flying birds (Petrels and Prions) that breed at the islands (Appendix 4.A.). Kelp gulls Larus dominicanus (Lichtenstein 1823), Subantarctic skuas Catharacta antarctica lonnbergi (Lesson 1831), the Crozet shag Phalacrocorax melangogenis (King 1828) and the lesser sheathbill Chionis minor marionensis (Peters 1934) were not included in the model. Population estimates are available for breeding populations of the top predators on the islands and biomass estimates were based on published estimates from the 1980s (Hanel and Chown 1998). Adjustments in biomass estimates were made for all groups to account for the period of time that each group is considered resident at the islands (Condy 1979; Cooper and Brown 1990). Biomass estimates for mammalian groups include the full population, while for the avian fauna, only breeding populations are included.  The fish groups were divided into two demersal and two pelagic groups, with the separation between large and small fish based on the maximum length attained by each species (small groups total length <50cm, large groups total length >50cm). Two families are considered the most important at the islands, the nototheniids and the myctophids. The demersal groups include benthopelagic and demersal species. The large demersal fish group consists of 21 species. This group includes three nototheniid fish, (the grey rockcod Lepidonotothen squamifrons (G?nther 1880), black rockcod Notothenia coriiceps (Richardson 1844) and the marbled rockcod Notothenia rossii (Richardson 1844)) all bathydemersal and benthopelagic species, as well as the 70  sharks and rays. The small demersal fish group consists of 13 species and incorporates the inshore and continental slope fish species. Important species in this group include inshore nototheniid species, the lobe-lip notothen Gobionotothen marionensis (G?nther 1880) and Gobionotothen acuta (G?nther 1880), as well as the painted notie (Lepidonothothen larseni Lonnberg 1905), which is also found on the continental slope. The large pelagic group contains 11 species and is dominated by the nototheniid Patagonian toothfish Dissostichus eleginoides (Smitt 1898). The small pelagic fish are dominated by the Myctophidae family which comprise 17 of the 35 species. Myctophids are the most abundant pelagic fish in the Southern Ocean (Gjosaeter and Kawaguchi 1980; Sabourenkov 1991; Kozlov 1995) and are an important food source for many of the top predators (Adams and Klages 1987; Brown et al. 1990; Cherel et al. 1993; Lea et al. 2002). Pelagic system biomass estimates for phytoplankton, zooplankton and for the pelagic fish groups were from the PEI, while biomass estimates for the small pelagic fish group and the cephalopods were taken from general Southern Ocean estimates. The PEI system is also notably different from many of the other sub-Antarctic islands in that it lacks a large shelf area as is, for instance, found at the Kerguelen archipelago. Macrozooplankton densities over the plateau at the Kerguelen Islands are higher (Hunt et al. 2011) and many of the top predator diets are dominated by zooplankton. While the PEIs do support locally high zooplankton biomass around the fringe of the islands, summaries of the diets of the most abundant predators indicate that myctophids are the most important prey species for the vertebrate predators found in the island system. Fish fauna studies carried out in the 1970s and 1980s (Gon and Klages 1988), along with the recent 2001 survey (Pakhomov et al. 2006), have provided a comprehensive list of species present at the islands but quantitative assessments are still lacking. Biomass estimates for all fish groups were made using CPUE data from Brandao et al. (2002) and Pakhomov et al. (2006). Similarly, species composition of cephalopods is known (largely from diet analysis of the top predators at the islands) but quantitative estimates of biomass are not known and estimates for the model were based on assessments from other sub-Antarctic systems. The PEI have a rich benthic community and biomass estimates for the benthic components of the ecosystem were taken from the PEI data for all groups (Perissinotto and McQuaid 1990; 71  Attwood et al. 1991; Beckley and Branch 1992). Studies on the inter-tidal and benthic fauna of the islands were conducted in the 1980s, providing a comprehensive list of species present and a review of the ecology of the community (Blankley 1984; Blankley and Branch 1984; 1985; Blankley and Grindley 1985; Arnaud and Branch 1991; Branch et al. 1991a; Branch et al. 1991b; Beckley and Branch 1992; Branch and Williams 1993; Branch 1994). The benthic community comprises approximately 550 species with seven benthic community groups. Nauticaris marionis (Bate 1888), the benthic decapod, has the second highest crustacean biomass and numerous studies have focused on this benthic shrimp because of its perceived key role in the ecosystem (Perissinotto and McQuaid 1990; Kuun et al. 1999; Pakhomov et al. 1999; Pakhomov et al. 2000; Pakhomov et al. 2004). The decapod is consumed by some of the top predators on the islands, notably the penguins (Brown et al. 1990), and therefore provides a link between benthic autochthonous production and higher vertebrates. The benthic fauna are represented in this study by 2 functional groups, one that represents the benthic fauna as a whole, with a second one that represents the decapods.  An illustration of the spatial distribution of the three forms of primary production is given in Figure 4.2. The first comprehensive study of the phytoplankton and zooplankton of the area was conducted in the late 1970s (El-Sayed et al. 1979a; El-Sayed et al. 1979b; Grindley and Lane 1979). Of all the pelagic groups in the system, the zooplankton have received the most attention. Studies describing the zooplankton fauna through the 1980s (Miller 1985; Boden and Parker 1986; Perissinotto and Boden 1989) were followed by community assessments through the 1990s (Perissinotto 1989; Perissinotto and Boden 1989; Perissinotto and McQuaid 1992b) and early 2000s (Hunt et al. 2001; Hunt et al. 2002; Bernard and Froneman 2003; Hunt and Pakhomov 2003). For the purposes of this study, the zooplankton were divided into three functional groups: the large crustaceans (dominated by euphausiid biomass), small herbivorous crustaceans (dominated by copepods) and all remaining zooplankton.  The open ocean phytoplankton was divided into two groups, one to represent the microphytoplankton and one the nano- and pico phytoplankton. As summarized in chapter 2 in situ measurements of chl-a in the vicinity of the open ocean range from <0.1 - 0.52 mg.m-3 (4 studies) (Froneman and Balarin 1998; Froneman and Pakhomov 2000; Bernard and Froneman 72  2005; McQuaid and Froneman 2008). Remotely sensed ocean colour satellite data at 1 km resolution over a six by six degree area centered over the islands (44?S to 50?S, and 35?E to 41?E) from SeaWiFS (1998 to 2004) and a two by two degree area (45.8?S to 47.8?S and 36.8?E to 38.8? E) from MODIS (2005-2008) satellites show a total annual average value of 0.22 mg chl-a.m-3 using standard processing. This value was used in conjunction with a euphotic depth of 100m (Cheung et al. 2005), a chorophyll-a to carbon ratio of 1:43.9 and a carbon to wet weight ratio of 1:9 (Christensen and Pauly 1995). A total biomass for open ocean primary producers was estimated to be 8.69 t.km-2. A summary of the contribution of microphytoplankton to the whole phytoplankton community was estimated at approximately 20% (El-Sayed et al. 1979b; Froneman et al. 1998; Read et al. 2000). No explicit additional production associated with elevated productivity of the fronts was incorporated despite these features falling within the considered area.  The island-associated blooms are usually the result of increased production of diatoms (mainly the chain-forming Chaetoceros radicans (Sch?tt 1895) (Boden et al. 1988), Rhizoselena curvata (Zacharias 1905) and Dictyocha speculum (Ehrenberg 1837) (Perissinotto 1992) or Fragilariopsis spp. (McQuaid and Froneman 2008)). In situ studies of chl-a from within the vicinity of the islands range from 0.01 - 2.8 mg chl-a.m-3 (8 studies) (El-Sayed et al. 1979a; Miller et al. 1984; Allanson et al. 1985; van Ballegooyen et al. 1989; Perissinotto et al. 1990b; Froneman et al. 2000; Perissinotto et al. 2000). The phytoplankton blooms are dominated by diatoms and chl-a concentrations exceed 1.5 mg.m-3 (measurements of up to 2.8 mg.m-3 have been made) (Boden 1988; Duncombe Rae 1989b). Under non-bloom conditions, chl-a concentrations range between 0.05 and 0.45 mg chl-a.m-3 with the composition dominated by nano- and/or picophytoplankton (Perissinotto et al. 2000; Bernard and Froneman 2002). Remotely sensed ocean colour satellite chl-a data at a 1km resolution from a subarea centred on the islands (46.5?S to 47.1?S and 37.5?E to 38.3?E) from 1998 to 2008 were processed (SeaWiFS data 1998 ? 2004, MODIS data 2005-2008). Monthly averages during summer months ranged between 0.4 and 1.4mg chl-a.m-3 showing clear seasonal blooms in the vicinity of the islands. The annual average concentration for the sub-area was 0.27 mg chl-a.m-3. An average annual increase of 0.05 mg chl-a.m-3 over and above the open ocean value of 0.22 mg 73  chl-a.m-3 was attributed to the elevated production associated with the islands. A conservative euphotic depth of 20 m for the blooms was assumed (Perissinotto et al. 1990c). Macrocystis laevis and Durvillea antarctica are the two macrophytes that dominate the system. Quantitative estimates of the macrophytes were made for Marion Island from both photographs and from diving surveys in the 1980s, with estimates of 63 500 t for M. laevis (Attwood et al. 1991) and 3 300 t for D. antarctica (Haxen and Grindley 1985). An extrapolation of these estimates to include macrophyte beds around Prince Edward island based on the percentage of the perimeter of Marion to Prince Edward, resulted in a total biomass of 87 495t for the system (Attwood et al. 1991).  Detritus estimates were made using the empirical equation of Pauly et al. (1993). Using a primary production estimate of 17.155 gC.m-2.yr-1 (lower estimate) (Pakhomov and Froneman 1999b), the annual estimate of detritus resulting from the open ocean phytoplankton productivity was 3.184 g.m-2 (equivalent to t.km-2 ). Using a conversion of carbon to wet weight of 1:9, detrital input was estimated to be 28.7 t.km-2. Rate measurements P/B data were taken from the published literature for the PEIs where possible or otherwise from similar systems. Open ocean production was estimated to be between 94 - 442 mgC.m-2.d-1 (Boden 1988; Balarin 2000). Island associated production ranged from 84 to 3000 mgC.m-2.d-1 (five studies) (El-Sayed et al. 1979b; Allanson et al. 1985; van Ballegooyen et al. 1989; Perissinotto et al. 1990b; Balarin 2000), resulting in P/B ratios for open ocean production of 150 yr-1and for island-associated blooms of 200 yr-1. Production rates for M. laevis were measured in both April and August and the mean productivity was 9.6 gC.m-2.d-1, which resulted in an estimated P/B of 5.22 yr-1 (based on a biomass of 11.5kg.m-2) for macrophytes (Attwood et al. 1991).  Q/B data was estimated using empirical equations for all the top predators and based on energetic demands (Nagy et al. 1999), using the body mass of each group (Trites and Pauly 1998; Ryan and Bester 2008) and calorific content of diets (Burger 1981; Croxall 1984; Abrams 1985; Clarke 1985; Doidge and Croxall 1985; Brown and Klages 1987; Tierney et al. 2002), 74  with an assumption that water content was 70%, along with the specific diet for each group. Q/B data for fish were calculated using the empirical equation of Palomares and Pauly (1998) and taken from the literature for the benthic and zooplankton groups. Diet matrices and assimilation efficiencies used in the calculations were estimated from the literature and are presented in tables 4.1 and 4.2. Nested models to assess primary production at different scales Four models of different sizes were created. The areas for each model were calculated with four different radius lengths (200nm, 100nm, 50nm, 20nm), each of which was centered at the midpoint between the two islands that make up the archipelago (46?46?S, 37?51?E). For each of the four models, the biomass per unit area (t.km-2) was scaled according to the size of each ecosystem i.e. higher biomass per unit area with smaller model size for all top predators as they were assumed to fill the entire area of each model (see Table 4.1). The estimates for all benthic components, including demersal fish, the benthic decapods, benthos, inter-island blooms and the macrophyte biomass, were scaled according to the area that their habitat occupied of the total area for each model i.e. higher biomass per unit area for each smaller model. Pelagic system biomass estimates were assumed to be uniformly distributed throughout the areas considered and therefore the biomass per unit area for all pelagic groups remained unchanged for each of the four models. Input data for all functional groups is presented in Table 4.1, and the diet matrix used for the model is presented in Table 4.2.  4.3. Results The trophic linkages for the purposes of this model have been based on the available data for the island system and these relatively new trophic pathways have now been incorporated into this quantitative assessment, which is illustrated in Figure 4.3 at the scale of the EEZ. Only at this scale does the model balance, i.e. the energetic requirements for all the inhabitants are met. For the smaller size models, reductions in the land-based top predator biomass estimates are necessary for the models to balance. As none of these groups feed directly on the primary producers, the adjustments have no effect on how much of the production is directly used in the system, and for the purposes of this study, will not be considered further. The relative biomass 75  contributions of each of the producers at this scale are evident, with the open ocean nano- and picoplankton contributing the most per unit area (Figure 4.3, 4.4). A quantitative assessment of the primary producers at the islands at each of the four scales considered clearly shows the increased productivity per unit area with proximity to the islands demonstrating the basic principle of the ?island mass effect? (Figure 4.4). The model constructed at the scale of the EEZ has a total producer biomass of 8.89 t.km-2, while the smallest model has biomass contributions greater than three times this value at 29.395 t.km-2. The open ocean production dominates at all but the smallest scale, where the macrophytes become important. Also evident from this assessment is the relatively small contribution the island associated blooms make to the system at all spatial scales. The greatest contribution of these blooms occurs in the smallest scale model, but even then contributes only 1.4% of the total primary producer biomass. If the assumption of the depth of the bloom is increased from 20 m to 120 m (Perissinotto et al. 1990), this contribution would have a maximum contribution of 7.5% at the smallest scale considered here. These data provide an assessment of the biomass of the producers, but how much of this production is consumed within the system? The Ecopath routine, when provided with biomass, production and consumption rates, produces an Ecotrophic Efficiency (EE), given as a proportion, which is a measure of how much of each group is used within the system (Figure 4.5).  Of the two size fractions of open ocean phytoplankton production, the larger microphytoplankton portion is utilized the least. EEs for this group increase with decreasing model size, but all EEs are relatively low (maximum EE at the smallest model size of 0.14). The smaller open ocean phytoplankton size fraction (the nano and picoplankton) has consistently high EEs (between 0.84 and 0.87), suggesting that this source is well utilized in the system at all spatial scales.  As discussed, the contribution of the island associated blooms to the total production is low. However, of this production approximately 30% is consumed within the system. This EE is consistent because the consumers of this production are the benthic fauna, which are scaled with the model size, as are the blooms. This form of production is thought to enter the benthic sub-76  system through fallout following the blooms. Such a detrital component has not been explicitly separated in this model, the inclusion of which would add to the relative importance of this input.   At the smallest scale, while the macrophytes contribute approximately 69.1% of the available primary producer biomass, much of the production is not directly consumed, as shown by the relatively low EEs (range 0.11 ? 0.22). Even at the largest scale, relatively little of the macrophyte production is grazed on directly. Macrophyte detritus on the other hand is well used in the system: minimum EE of 0.60 in the smallest model and a maximum of 0.68 for the largest. This pattern is due to the diet of some of the pelagic components of the community consuming this form of detritus. The general detritus group follows the opposite trend, with increasing EE with decreasing model size (0.35 to 0.63), largely attributed to the benthic fauna which directly consumes the macrophyte detritus.  4.4. Discussion Although the ?island mass effect? was recognised early on as a driving force behind enhanced local production at the PEI, the mode of operation and subsequent transfer of primary production up the PEI food web continued to be a subject of research and source of contention for a number of years. Since the earliest investigations at the PEI, the open ocean allochthonous production was found to be low (average approximately 0.2 mg chl-a.m-3) and studies were initially focussed on the autochthonous production, i.e. that of the island-associated blooms and the macrophyte production. During bloom conditions, chl-a concentrations were 5 to 10 times higher than average (between 1 and 2 mg chl-a.m-3) and this enhanced productivity was the result of seasonal blooms of diatom species (El-Sayed et al. 1979b; Boden 1988; Perissinotto et al. 2000). Studies through the 1990s that focussed on zooplankton grazing showed no evidence of feeding on this microphytoplankton size fraction (>20 ?m), which dominates during such blooms, but grazing occurred in the nano- and pico- size fractions (<20 ?m) (Perissinotto et al. 1990a). From these results, it appeared that the zooplankton community was adapted to consume the smaller size fraction, available year round, and likely did not make use of the seasonal blooms. Production from the blooms was therefore thought to sink out of the surface waters and provide a direct transfer of primary production to the rich benthic subsystem, with the pelagic community not directly benefiting from this form of production. A quantitative assessment of the 77  macrophyte production in the vicinity of the islands was made in the 1980s and found to be greater per unit area than that of the local phytoplankton production, as was to be expected (Mann 1973). However, this production was thought to contribute less to the seas around PEI because of its limited spatial coverage and because it was suspected that almost all of the macrophyte production was exported to the open ocean pelagic environment (Attwood et al. 1991).  At this point, the two forms of autochthonous production were thought to have different fates: the blooms providing input into the benthic community and the macrophyte production being exported from the system. The issue of scale was not directly addressed in these assessments, but publications at the time estimated the ecosystem to be delineated by a 300 km radius, thought to be the extent of the foraging distance of the island?s top predators (Adams et al. 1993).  In the early 2000s, stable isotope studies conducted at the islands provided new insight into the trophic pathways in the PEI system. Unique carbon isotope signatures (?13C) for each the three forms of production at the islands were identified (Kaehler et al. 2000) and allowed for an assessment of the relative importance of each of the producers to the system. Input from macrophyte production, primarily as particulate carbon, to the nearshore benthic community in particular was shown to be substantial (Kaehler et al. 2006), contradictory to earlier perceptions. Following from this, diet studies conducted in conjunction with stable isotope analysis (e.g., on euphausiids (Gurney et al. 2001), decapod Nauticaris marionis (Pakhomov et al. 2004) and the fish community (Bushula et al. 2005; Pakhomov et al. 2006)) contributed to a better understanding of the trophic pathways for many of the key consumers. Two co-occurring near-shore fish species, for instance, were found to have contrasting sources of production, one allochthonous and one autochthonous macrophytes (Bushula et al. 2005). These data provided new insights which were incorporated into the LSS model and allowed for a reassessment of the island system. The relative importance of each of the producers is affected by the scale at which the system is assessed, a consideration seldom included in previous discussions of the PEI system. At all but the smallest spatial scale (where it contributes 23.8% to total available production), the small fraction (nano- and pico-) of the open ocean production dominates in terms of biomass per unit 78  area (between 78.7% and 58.3% of total available producer biomasss). This allochthnous production is the most important of the three sources at the PEI and may be considered the driving force of the island ecosystem, particularly at the larger spatial scales. When the producers are assessed in terms of what is actually consumed at the islands (the biomass combined with the Ecotrophic Efficiency outputs from the Ecopath routine), this component contributes in excess of 92.0% of the production at all but the smallest scale, where it contributes 69.5%. When all first trophic level contributors are included in the assessment (i.e. the two detrital groups along with the primary producers), the general detrital group dominates and the contribution of the allochthonous small size fraction of the open ocean production is reduced to between 32 and 37% for the three larger models, and 21% for the smallest model. Of the two autochthonous contributors, the macrophytes were found to contribute the most to the ecosystem. When the total biomass of the producers in the system was assessed, the macrophytes contributed, in decreasing order of model size, 2.3%, 8.5%, 27.1% and 69.1%. Direct consumption of the macrophytes, however, was limited. Therefore, even at the smallest scale considered, when the available macrophyte biomass dominated the system, the amount consumed directly was 26.4% of the total producers consumed. Detritus generated from this production was well utilized in the system at all spatial scales (Ecotrophic Efficiencies between 0.60 and 0.68), however in terms of its contribution to the system when all first trophic level contributors are assessed (all producers and detrital groups), it is greatest at the smallest spatial scale at 7.70% of the total, which is similar to the contribution of the macrophytes themselves at 7.98%. The combined consumption of these two groups is substantial (15.68%), though it ranks third in terms of what is used in the system behind the general detrital component (62.10%) and the small open ocean phytoplankton contribution (20.99%).  The absolute biomass of the island-associated blooms is small compared to all other contributors, even at the smallest scale (less than 2% for all models). Approximately 30% of this production was consistently consumed within the system and its contribution, particularly in its detrital form, could be important to the benthic community. Inclusion of this component through an explicit contribution of its detrital form may provide further insight into the fate of this production, but from this assessment it appears to be limited. 79  A weakness of this assessment is the omission of the microbial loop. Microbial studies on microheterotrophs in the PEI system have been conducted (Froneman and Balarin 1998; Froneman and Bernard 2004) and grazing impact of protists on phytoplankton may be between 47 and 71% of potential primary production per day (Froneman and Balarin 1998). Bradford-Grieve et al. (2003) produced an ecosystem model of a sub-Antarctic system (Southern Plateau, NZ) which paid close attention to the microbial loop with 5 functional groups of the 19 considered dedicated to this section of the food web. No account was made in their assessment of the macrophyte production, though its input, particularly through the detritus, was acknowledged (Bradford-Grieve et al. 2003). Inclusion of this important component of the food web will further redefine models of the system. 4.5. Conclusions For the time period considered in this study (the 1980s), allochthonous production, in the form of nano- and picophytoplankton, should be considered the most important contributor to the PEI system, and the driver of the island food web at all but the smallest spatial scale. This production reaches the top predators of the PEI ecosystem via zooplankton and small pelagic fish. Elevated phytoplankton productivity associated with the ?island mass effect? at PEI was found to be of little importance, even at the smallest scale considered here. The alternate form of autochthonous production, the macrophyte production, was found to be utilized by both direct consumption and in its detrital form, and provides an important contribution to the system, particularly when considered at the smallest scale. The benthic fish populations and near-shore top predators (e.g., Gentoo penguins) are dependent on this autochthonous production; bentho-pelagic coupling (macrophyte production, benthic decapods and small demersal fish) thus represents an important trophic pathway in the system. Studies on the PEI ecosystem in the past decade have highlighted the importance of the autochthonous macrophyte contribution to the islands system. Indeed, we believe that this autochthonous component contributes towards a unique benthic habitat between the islands which warrants further detailed investigation. Overall, however, the findings of this study confirm previous perceptions that the system is largely reliant on pelagic allochthonous production. It was only when the model was considered at the scale of the PEI EEZ that the energy demands of the current population of PEI top predators was supported.   80  Table 4.1. Input parameter estimates for the four Life-Support System Models (P/B, Production to biomass ratio; Q/B, Consumption to biomass ratio; U/Q, Unassimilated consumption). 81   Table 4.2. Diet Matrix for the four Life-Support System models 82      Figure 4.1. Map of the Indian sector of the Southern Ocean showing the position of the Prince Edward Islands and main frontal features (STC, Subtropical Convergence; SAF, Sub-Antarctic Front; APF, Antarctic Polar Front).   83    Figure 4.2. Map of the Prince Edward Islands, indicating the spatial distribution of the three primary producers.   84   Figure 4.3. Flow diagram with relative importance of each primary producer at 200nm scale. Box size is proportional to the square root of the biomass of the functional group.   85   Figure 4.4. Relative biomass of each primary producer for the four model sizes (model sizes equivalent to circles with given radii).   Figure 4.5. Ecotrophic efficiencies for functional groups of the first trophic level of the four ?Life-Support System? models.   86  Chapter 5 An ecosystem model of the Prince Edward Island Archipelago 5.1. Introduction The Prince Edward Islands (PEIs), like all Sub-Antarctic islands, are hotspots of biological activity. The combination of the alteration of oceanographic dynamics by the interception of an island system and the provision of substrata for both land based top predators and benthic organisms combines to provide a site of elevated productivity in an otherwise relatively low productivity region (See Chapters 1 and 3). These islands are host to millions of seabirds and seals that use the islands as a seasonal breeding ground and refuge during moulting. The shelf areas around the islands are known to support increased levels of benthos and fish populations which aid in supporting the seasonal residents.  At the PEIs the population dynamics of many of the seabirds and seals have been well documented over the past 60 years. Disparate trends have been observed. In some instances the changes can be linked to known drivers. For instance, the Sub-Antarctic fur seals (Arctocephalus tropicalis) population increased exponentially from a few hundred individuals in the 1950s to over 150 000 by 2010 (Condy 1981; Kerley 1983a; Wilkinson and Bester 1990a; Hofmeyr et al. 1997; Hofmeyr et al. 2006; Bester et al. 2009), which is believed to be a population recovery following past exploitation. In other instances the changes, while studied, are not well understood. The Southern elephant seals (Mirounga leonina) have been in decline since the 1970s (Condy 1977; Condy 1978a; Condy 1981; Bester and Hofmeyr 2005) with no conclusive understanding of the reasons behind the decline, though interspecific competition, competition with fisheries and environmental changes have all been cited (Wilkinson and Bester 1990b; Green et al. 1998; Goldsworthy et al. 2001; McMahon et al. 2005). Population fluctuations for penguin species have also been shown, with the Southern Rockhopper penguins (Eudyptes chrysocome filholi) and the Macaroni penguins (Eudyptes chrysolophus) experiencing significant declines between 1994/5 and 2008/9 of 70% and 30%, respectively, (Crawford et al. 2009) raising conservation concerns. Along with the land based top predators, the only fish species to be targeted by a fishery has also experienced a collapse in the population. The Patagonian toothfish (Dissostichus eleginoides) has 87  diminished to a fraction of its former population status following unsustainable catches being harvested from a pristine stock (largely taken illegally in the initial phase of the fishery in the 1990s) (Brandao et al. 2002; Brandao and Butterworth 2009; CCAMLR 2011). Fishery related by-catch of seabirds is documented as resulting in population related fluctuations for some of the albatross species (e.g. Wandering albatross, Diomedea exulans) that breed at the islands (Nel et al. 2002a; Nel et al. 2002c; Nel et al. 2003) though mitigation measures have been successful in reducing these events (CCAMLR 2011). Still, conservation concerns at the islands have been highlighted as all five of the populations of albatross that breed at the islands have been classified as having special status (?near-threatened, vulnerable or endangered?) by BirdLife International (2011) (IUCN Red List for birds. Downloaded from http://www.birdlife.org on 27/09/2011). In addition, the presence of cats at the islands between 1949 and 1992 resulted in considerable predation on the breeding bird populations (Bester et a. 2000) adding to conservation concerns at the islands.  With the islands therefore having both fishery and conservation concerns, there is a need to develop a single platform from which these considerations can be assessed together. A search for a better understanding of the processes that underlie the observed changes and also the linkages between the various constituents of the foodweb is required. The construction of an ecosystem model provides such a framework where interactions between components of the ecosystem can be identified. The ecosystem state, structure and function in both past and present can be assessed and used to develop a better understanding of the ecosystem as a whole.  The wealth of scientific research that has been conducted at the PEIs provides a unique data set from which to build an ecosystem model. This work describes the first ecosystem model of the marine component of the PEIs. Through the construction of the model, a dataset has been compiled, providing a useful summary of existing data for the system as well as highlighting data gaps. This, combined with an assessment of data quality and model sensitivity provides a decision support platform for assessing where scientific research efforts should be focussed in future to improve the model parameters and thereby the quality and usefulness of the model. Finally, an assessment of the ecosystem in terms of its trophic structure, biomasses, flows through the food web and relevant ecosystem indicators is presented. 88  5.2. Methods 5.2.1. Study area For the purposes of the creation of the model, the study area has been set to be the Exclusive Economic Zone (EEZ) of the Prince Edward Islands (See Figure 5.1). This area is demarcated with a circle with a radius of 200nm, centered between the two islands (46?46?S, 37?51?E). The total area is equal to 431015km2. 5.2.2. Modelling approach The approach used for this study is the mass balanced network model known as Ecopath (Christensen et al. 2008). The original 'Ecopath', first proposed by Polovina and Ow and developed by Polovina (Polovina 1984; Polovina and Ow 1985; Polovina 1986) combines a system of simultaneous linear biomass budget equations, which balance biomass production and loss. Since its original formulation (Christensen and Pauly 1992; Christensen and Pauly 1993) the model has been developed so that it no longer relies on a steady state, and it incorporates a network analysis component from theoretical ecology for detailed assessment of the trophic flows (Ulanowicz 1986; Christensen et al. 2004).  These developments allow comparisons between ecosystems. Ongoing developments to the software have meant that this model is the fore-runner in its field. It has the capacity to represent all trophic levels. It is the most widely used ecosystem model and is appropriate for addressing broad ecological questions (Plaganyi 2007). A summary of the strengths and weaknesses of the model is discussed in Chapter 1.  In Ecopath, species can be considered individually, or pooled together into functionally related groups. Each group is represented by their biomass and the groups are linked through their trophic interactions. The basic Ecopath model is a closed system mass-balanced formulation of the functional groups in an ecosystem and the full methodology can be found in the user's guide (Christensen et al. 2008).  There are two principal equations in the Ecopath model, the first describing the production term within a group (i) (Eq 1) and the second describing the production term of the group (i) within the system (Eq 2).  89  Equation 1: Production = Consumption ? Respiration ? Unassimilated mass Pi = Qi ? Ri ? Ui Equation 2: Production = Predation + Other mortality + Fishing mortality Pi = Bi (M1i) + Bi (M2i) + M3i Where: Pi is the total production rate of i  Qi is the consumption rate of i Ri is the respiration rate of i  Ui is the rate of the unassimilated mass of i  Bi the biomass of i M1i is instantaneous predation rate of i M2i is the ?other mortality? rate of i   M3i is the total fishery catch rate of i M1i, the instantaneous predation rate of i, can be expressed as the sum of all n predator groups? consumption rates (Q) on the prey (i) as:  M1i =?Qj(DCji)/Bi??=1 Where Qj is the consumption rate, in this instance for predator j, and DCji is the fraction of the prey i in the average diet of predator j.  Adjusting the equation to include the Q/B ratio for predator j results in  M1i = ?Bj (Q/B)j(DCji)/ Bi 90  Substitution of this into equation 2 gives: Pi = ?Bj (Q/B)j (DCji) + Bi (M2i) + M3i Ecopath uses the term 'Ecotrophic Efficiency' (EE), which is the proportion of the production of a group that is utilized in the system.  Substituting M2i (other mortality) with Pi (1-EEi) / Bi results in:  Pi - ?Bj (Q/B)j (DCji) - Pi (1-EEi) - M3i = 0 Adjusting the format of this equation to reflect production to biomass (P/B)i ratios, this equation can be expressed as: Bi (P/B)i - ?Bj (Q/B)j (DCji) - Bi (P/B)i  + Bi (P/B)i (EE) - M3i = 0 Or Bi (P/B)i (EE) - ?Bj (Q/B)j (DCji) - M3i = 0 This equation indicates that the diet and consumption rate of a predator can be used to determine the predation mortality term for the prey. Or alternatively, if the mortality for a given prey is known then the equation can be used to estimate the consumption rates for one or more predators (Christensen and Walters 2004). Based on the last equation, for a system with n groups, there are n linear equations. From the series of n simultaneous equations created (equivalent to the number of groups in the system), solutions for the equations are calculated based on the assumption of mass balance within the system using a generalised method for matrix inversion (see Christensen et al. 2008). The routine solves for one of four parameters for each group: biomass, production/biomass ratio (P/B), consumption/biomass ratio (Q/B) or ecotrophic efficiency (EE). Three of the four parameters must be entered to calculate the one unknown. This means that the basic model parameters required include biomass estimates, diet compositions, assimilation efficiencies, catch rates 91  (where applicable) and three rate measurements (consumption, production and mortality). If all of these data are available, the EE of each group can be calculated by the model.  For the model as a whole, the energy input and output of all living groups must be balanced (Christensen et al. 2008). In Equation 2, only the production term of a group is included. To ensure balance within each group, Equation 1 is used. In Equation 1, respiration is the one term conventionally not measured and so it is left to the model to estimate (though if desired, this can be entered using an alternative input structure in Ecopath). The two master equations of Ecopath (Eq 1 and 2) can be considered filters for mutually incompatible estimates of flow with the result providing a possible picture of the energetic flows, the biomasses and their utilization (Christensen et al. 2008). 5.2.3. Data Three models have been compiled for the PEIs to represent the ecosystem in three different decades: the 1960s, the 1980s, and the 2000s. Model parameter estimates were made to represent the island system for a time period of one year. Biomass (B), production/biomass ratio (P/B) (equivalent to the total mortality rate Z, (Allen 1971), consumption/biomass ratio (Q/B) (food consumed by a biomass unit of the group) as well as respective diet data were entered into the Ecopath framework and ?Ecotrophic Efficiencies? (EE) (how much is used within the system) were computed. Although Ecopath has the capability to consider flows into or out of the system (immigration, emigration, or biomass build up, see Chapter 4), for the purposes of this study the ecosystem was assumed to be in a steady state and closed. Seasonal migration into and out of the defined area was dealt with by adjusting the biomass of each group according to the proportion of the year that they spend at the islands. For the purposes of the models, all major marine biological components were considered and the system was divided into 37 functional groups as described in Chapter 2 (Table 2.1). Species were aggregated into functional groups based on a combination of systematics, size, habitat, and diet preference. Groupings were also based on fishery or conservation considerations and were influenced by data availability. Guidelines for aggregation of groups were followed as recommended in the literature (Fulton et al. 2003a; Pinnegar et al. 2005; Quince et al. 2005a; 92  Raick et al. 2006). Overall the model incorporated four mammal groups, four penguin groups, seven flying bird groups, nine fish groups, two cephalopod groups, one all encompassing benthos group, a benthic decapod group, three zooplankton groups, four primary producers, and two detrital groups. The names of these functional groups are given in italics from here on throughout the text and 3 letter codes for each group provided for use in figures (see Table 2.1 in Chapter 2). Where italics are not used, the reference is to the species or groups in general and not specifically referring to the model functional group. Data for all biological components of the marine ecosystem of the PEIs were collated from a full review of both the published and grey literature for the islands. The biological data available for the marine component of the PEI ecosystem is extensive and the model construction benefitted a great deal from the wealth of primary data that are available. For many of the required parameters, local data from population surveys, growth and reproduction studies, diet assessments as well as stable isotope analyses were found. In those instances where PEI ecosystem specific data could not be found, data from similar systems (Sub-Antarctic or Antarctic) were used. A simple overview of the data compilation for each of the primary parameters is described below, with a complete description given in Appendix 5.A.  Biomass estimates For many of the mammals and seabirds, only abundance estimates were available, therefore assumptions were necessary to extrapolate these estimates to produce biomass estimates. For mammal population estimates, conversion factors included incorporating sex ratio estimates, female to pup ratios, pup ratios to population estimates, or males to population estimates. Population to biomass estimates were calculated using average weights for different components of the populations derived from Trites and Pauly (1998) or Condy (1981). Biomass estimates for the breeding bird populations were calculated from published data on population sizes and weight estimates provided in Ryan and Bester (2008). Adjustments of biomass estimates were also made to account for the proportion of time a group spends in the system in one year based on Cooper and Brown (1990). For all the seabird and seal groups, biomass estimates were reconstructed for each of the three different time periods selected for construction of the models. 93  Data on the fish fauna from the islands were limited due to the lack of reliable survey data. Biomass estimates were made by combining quantitative data from the commercial fishery with qualitative assessments made of the fish community from the PEIs (Brandao et al. 2002; Pakhomov et al. 2006; Brandao and Butterworth 2009) with additional information from the nearby Kerguelen (Duhamel and Hautecoeur 2009). Broadly the fish were divided according to their habitat preference (benthic or pelagic) and divided by size into large and small groups (the division made at Linf = 50cm). Within these divisions, groups perceived to be of particular importance to the system (for example the myctophid fish, the Patagonian toothfish, and a subset of three other Nototheniid species) were considered separately. The demersal fish were divided into three categories, Small Inshore Demersals, Small Continental Slope Demersals and Large Demersals. These divisions were made by taking known depth preferences into account. Sharks and rays were considered in a separate group. The benthic community was lumped together into one primary group excluding the benthic decapod, which was considered separately because of its perceived importance in the island system. Three functional groups represent the relatively well studied zooplankton community.  The primary producers were also considered in four separate groups. The open ocean phytoplankton was divided into a large size fraction, the microphytoplankton (greater than 20?m) and a small size fraction, the nano- and picophytoplankton together (less than 20?m), the plankton associated with the blooms that occur in the vicinity of the islands and the macrophyte community. Estimates for the open ocean productivity and the island associated blooms were made from remotely sensed satellite data (1997 to 2008). The macrophyte biomass estimate was based on in situ data from surveys at the islands in the 1980s. Two detrital groups were included, one representing all detritus generated in the system, and a separate one to account for the macrophyte detrital contribution.  Biomass estimates for the model were not made spatially explicit. The biomass estimates are entered as biomass per unit area calculated for the total EEZ. A detailed account of all estimates are provided in Appendix 5.A. Biomasses were estimated in absolute terms (tonnes), usually for land based or benthic groups, or in relative terms (per km2) for all pelagic groups. Biomass input data provided in both formats are summarised in Table 5.1. 94  Across the three time periods, biomass estimates for the land based top predators, for which there are time series data available, have been provided. For all pelagic groups, excluding fish groups where data exist, the biomass estimates between time periods are unchanged due to lack of time series data availability. Rate estimates (P/B; Q/B) Production to biomass rate estimates were obtained using the Ecopath software guidelines (Christensen et al. 2008). For populations that have not been exploited, the production to biomass ratio was considered to be equivalent to total mortality (Z) (Allen 1971) and estimates were either taken from the literature or calculated using empirical equations. P/B for the fish groups in particular were calculated using the equation of Palomares and Pauly (1998). Rates remain unchanged for all groups for all years, excepting for the Patagonian Toothfish, which has a higher P/B estimate for the 2000s model, which is considered appropriate following its exploitation. In ecosystem models the importance of food quality is often overlooked when the currency of the model is biomass. For consumption to biomass estimates of the mammals and seabirds used in this model, the food quality was incorporated in the estimates. Local diet preferences for each group were coupled with the local energetic content of prey items. Assimilation efficiencies were taken into consideration and water content of prey items was assumed to be 70%. Combining this information with annual energy requirements estimated by Nagy et al. (1999) from Field Metabolic Rates (FMR) (with adjustments as per Ellis and Gabrielsen (2001)), it was possible to provide Q/B estimates that take into consideration the food quality derived from the local diet preferences of these groups, a consideration not usually incorporated into such models. A full description of the methodology is provided in Chapter 7.  Incorporating the diet quality into the consumption estimates in such a way is valid and useful for the static models built here, but it is important to note that the consumption rate estimates in the model are not linked to the diet matrices as the calculations are done independently. Therefore, when the model is run in its temporally dynamic form (Ecosim), where prey switching is permitted to occur (linked to prey availability), the energetic value of the diet 95  content should be reassessed to ensure that the consumption estimates correspond to the new diet preferences. A future development of the ecosystem modelling software where this feature could be incorporated into the model would be beneficial. All rate estimates are given in Table 5.1. Diets Trophic linkages were made from published information and included traditional data on diet as well as information on trophic linkages from stable isotope and fatty acid signatures (See Appendix 5.A). The input parameters for the proportion of the diet that is considered ?Unassimilated? are provided in Table 5.1. The input parameters for each functional group for the diets are provided in Table 5.2. Diets for different time periods for most functional groups were assumed to be the same. In instances, however, where the diet included a stipulated contribution from a broader group (for instance, the diet for Orcas was determined as being 40% of the higher vertebrates (seals and seabirds)), the diets for these groups were adjusted to represent consumption of the functional groups according to the contribution each group made to the total biomass of the ?broader group? for each specific time period. This approach was used for the diet matrices of Orcas, Southern Elephant Seals and for Giant Petrels for each of the three different time periods. This same approach of assigning relative proportions of a diet according to the available biomass was carried out in all instances where diet information was provided at a higher level than that of the functional group. Data Quality Pedigree The EwE model provides a framework to assign a measure of data quality to the parameters used in the model: the ?Pedigree?. Using this framework, each parameter is assigned a value to measure its perceived quality (Table 5.3.). A qualitative (descriptive) data pedigree was used to describe the data source for each parameter (Biomass, P/B and Q/B) for each functional group and thereby establish the model ?Pedigree?.  96  Sensitivity Analysis A sensitivity analysis to test which parameters were most important in terms of changing the outcome of the model was also performed on the data. (Due to the unavailability of this tool in the EwE version 6, the model was constructed in EwE version 5 for this analysis). In the sensitivity analysis, each input parameter (B, P/B, Q/B) for each functional group is changed in turn in 10% steps from -50% to 50% and the fourth parameter to maintain mass-balance then calculated. A table of the magnitude of the changes (as a percentage) of the fourth parameter for each functional group according to the change in the percentage of variation of the input parameter was extracted from Ecopath. The results of the sensitivity analysis were summarised by calculating a simple index of sensitivity per component following the methodology outlined in Olson and Watters (2003). The index is a count of all the parameters affected by ?30% or more for each component. This index was further modified in this study by removing all counts that were the result of the change in a parameter having an effect on itself. Ecosystem Network Analysis The Ecopath routine has incorporated in it a variety of indicators and a network analysis tool to summarise the system in terms of biomasses and flows (Lindeman 1942; Odum 1971; Finn 1976; Ulanowicz 1986; Christensen 1995b; Ulanowicz 1995). Transfer efficiencies estimated for each trophic level are calculated as the percentage of throughput entering a trophic level that is subsequently passed on to the next trophic level (or harvested) (Christensen and Pauly 1998). Biomasses and flows are also assessed and provided in the form of a Lindeman spine, which is a summary of the transfer efficiencies (TEs) and biomasses summarised by trophic level (TL). The mixed trophic impact (MTI), derived from economic theory (Ulanowicz and Puccia 1990), is provided and allows for the quantification of direct and indirect trophic interactions among groups with the positive and negative effect that a hypothetical increase in the biomass of one group would have on another. This index can be positive or negative, and from it, the relative total impact (RTI) can be calculated as an absolute overall effect (Libralato et al. 2006) 97  ??? = ???????2???? where the effect on itself is not included (therefore MTIii is not included in the calculation). Summary statistics include a summary of the consumption in the system, the total system throughput, the mean trophic level of the catch, the net primary productivity, the total primary production/ total respiration, the net system production, the total primary production/ total biomass, the total biomass estimate and the total biomass divided by the total throughput as well as the connectance index, and system omnivory index.  5.3. Results 5.3.1. Balancing the model The balancing procedure for all three models was done manually without the aid of the automation feature available in Ecopath. Through the process of constructing the model many lessons were learned. One important factor, which contributed to easing the balancing process, was assigning the contribution of prey to predators based on the preys? local availability in instances where diet preferences were not specified at species/ functional group level, but at a higher level. An additional important step was the reassessment of the consumption rates. This issue was originally highlighted in the 1980s model where the Giant Petrels? (BGPs) high consumption rates, leading to high mortality rates on their prey (the penguins in particular) pushed the associated EEs of these groups over 1 and proved difficult to resolve. Initially, a ?dead penguin? functional group was created to feed the BGPs and balance the model. This solution was considered unsatisfactory by experts on the system (pers. comm. P. Ryan) and hence led to an in depth consideration of the consumption rates used. Initially, the investigation was solely for this group, but ultimately resulted in a reworking of the Q/B rates for all the mammal and bird functional groups. A review of the consumption rates traditionally used in ecosystem models was carried out, and the method used to estimate this parameter was revised to take local information into account. The inclusion of local diets and associated energetic content 98  of the prey into the consumption estimates made a significant contribution to the model parameterisation (see Chapter 7) for all three models for the different time periods. Following the inclusion of this adjustment, only relatively minor adjustments for a limited number of groups were required. For the 1960s model, adjustments to the Southern Elephant Seal diet formulation had to be made. An initial run of the model highlighted that consumption by Southern Elephant Seal on both the Large Pelagic Fish and the Patagonian Toothfish was too high (EE of 1.07 and 1.65 respectively). An adjustment to increase the contribution of Large Cephalopods by 5%, combined with a reduction by 1% for Large Pelagic Fish and 4% for Patagonian Toothfish was able to resolve the model balancing procedure.  For all other time periods the contributions were left in proportion to their biomass contributions as originally assigned. An alternative solution would have been to increase the biomass of the fish groups, as biomass data for these groups are uncertain. For the models of all three time periods (1960s, 1980s and 2000s) minor adjustments were also made for the diet matrix of the Giant Petrel group. The contribution of Prions and Petrels to the diet had to be reduced by 1% in the 1960s model, and by 5% in the 1980s and 2000s models to allow the model to balance. In each case diet was reassigned to the penguin contribution (weighted according to relative biomasses of the 4 species for each of the 3 time periods). The most logical explanation for this issue is related to the probable underestimate of the Prions and Petrels as the census results for these time periods are considered underestimates (Ryan and Bester 2008) and the effect of the cat predation on the on the small flying bird populations would also have played a role at these time periods. Population changes resulting from the cat predation are explored using temporal simulations of this model and discussed later in this work (see Chapter 8).  Finally, specific adjustments had to be made to the Patagonian Toothfish data. Initial biomass estimates for this group were derived from an estimate made from the data from the voyage of the Iris, using swept area, for the Patagonian Toothfish (Brandao et al. 2002). An estimate of 1168 tonnes, with coefficient of variation of 213%, was made. For the model area under 99  consideration here (the EEZ of the PEIs), this amounts to a biomass of 0.00271 t.km-2. Considering the coefficient of variation of 213%, the upper limit of this estimate may be considered to be 0.00577 t.km-2. This biomass estimate is presumed to be a small percent of the original biomass that would have been in the system prior to the fishery crash of the mid 1990s. If the current estimate is 5% of original biomass, estimated biomass for the 1960s and 1980s model are between 0.05420 and 0.11544 t.km-2.  In balancing the model, these ranges were taken into account and the values chosen from within this range to satisfy the system?s requirements. The highest value of 0.09000 t.km-2 was used for the 1960s model, which was required to meet the energetic requirements of the Elephant Seal population. An intermediate value of 0.07500 t.km-2 was used for the 1980s model. This value was determined following simulations of the Patagonian Toothfish fishery and was the biomass estimate that produced the best fit to the available fisheries data (Chapter 8). A value of 0.00429 t.km-2 was used for the 2000s model, which is just over twice the estimate from the survey data from that time period (0.00271t.km-2). The final balanced model output data are provided in Table 5.1 with a flow diagram for the 1980s model provided in Figure 5.2.  5.3.2. Data quality  An assessment of the data quality for each parameter is provided in Table 5.3. For the mammal and bird biomass data that are derived from population estimates of land based groups, local sampling contributes to a relatively high estimate of data quality. For the Orcas and all the fish groups, approximate or indirect methods were used and the data are relatively poor in comparison. Exceptions are for the Patagonian Toothfish, where trawl survey data combined with fishery statistics improves the estimate of this group. Locally derived estimates for the small demersal fish groups are available and, although the estimates are considered to be of low precision, the quality of these estimates can be considered slightly better. No local quantitative data exist for the cephalopods for the islands system and all data were taken from other estimates of models from similar systems. All the biomass estimates for all invertebrate groups were obtained from local data with low precision, as well as for the primary producers. The quality of the open ocean phytoplankton data may be considered improved by the inclusion of the satellite derived biomass estimates.  100  Production to biomass ratio data for the mammals and birds were derived in almost all cases from values available for similar / the same species from similar / the same locality. Exceptions include the Orcas, for which an empirical equation was used, and the Southern Elephant Seals for which an estimate was made from local data. For all other functional groups the production to biomass ratio was taken from other models and does not receive a high pedigree index. As already documented, much effort was put into improving the estimate of the consumption to biomass ratio for the mammals and bird functional groups. Even so the grading given to this parameter estimate was left as an empirical relationship (a relatively low ranking in terms of pedigree). For all other groups, the data were taken from other models and hence have even lower rankings in terms of pedigree. The pedigree index calculated by the Ecopath routine for this model was 0.303. The range of pedigree indices for 50 models analysed was from 0.164 to 0.676 (Morissette 2007), and a value of 0.303 would place the pedigree for this model in the intermediate range. If the adjustments to the consumption rate estimates made in this model for the land based top predators are considered to be locally derived (they are still from an empirical equation, but have local data incorporated) the model pedigree improves to 0.402, considerably higher (and in the upper range for EwE models). In any event, much of data for many of the pelagic groups for these models is uncertain and an effort should be made to improve these estimates in particular. Sensitivity analysis  A sensitivity analysis provides an assessment of which parameters have the greatest effect on the system. A summary of the counts of which of the three parameters were most important showed that changes in Biomass (t.km-2) had the greatest effect for all three models (between 100 and 101 counts). Consumption to biomass (Q/B, yr-1) ratios were found to be the second most important parameter with counts between 64 and 68, and production to biomass (P/B, yr-1) rates had the fewest occurrences (34 ? 37). When this sensitivity index was amended to exclude those occurrences where the change of a variable for a functional group affected the output of the same functional group, the counts contributing to the index for each parameter were reduced, with no 101  counts found for the P/B ratio. Biomass and Q/B ratios were found to be equally important, with counts ranging from 63 ? 67 for the three models. The sensitivity index is plotted in Figure 5.3 for the three models. Across all three time periods, the sensitivity tests highlight the importance of the estimates of the Orcas for all years. The second most important group identified in the sensitivity analysis was the Giant Petrels, with high counts in all years, with the 1980s period the highest (64 counts). The third most important group identified were the Southern Elephant Seals, though the counts decline through time, as does the biomass of this species. Across all the time periods, the Small Pelagic Fish are found to be important, along with the Benthos and both Large and Small Crustacean Zooplankton groups. These are followed by the Myctophid Fish. In the 1980s model, along with those already mentioned, the Patagonian Toothfish as well as the Gentoo Penguins are also found to be important. 5.3.3. Model output Trophic level The Ecopath routine provides an output of trophic level (TL). These are given as fractional outputs which have no units (Table 5.1). Trophic Levels are based on the relative contribution and the TL of the diet components. Primary producers and detritus groups are by default of TL 1. All consumers are TL 2 or higher, depending on their diets (see Christensen et al. 2008, pg 92). The TL outputs are based on diets that were identical between years for all functional groups excepting the Orcas, Southern Elephant Seals and the Giant Petrel groups. The difference in TL estimates between time periods for these three groups were minor, with differences in estimates less than 0.15 between all years. Orcas have the highest TL at 5.11 for the 1960s model, with the Giant Petrels close behind at 5.10. The Orcas TL declines through the time periods to 5.09 in the 1980s model and 4.98 in the 2000s model. The Southern Elephant Seals show a similar trend of declining TL from 4.74 in the 1960s and 1980s models, to 4.67 in the 2000s estimate. In contrast the Giant Petrels increase from a TL of 5.10 in 1960s, to 5.15 in the 1980s and finally to 5.18 in the 2000s model.  102  All the seals occupied the 4th TL and ranged between 4.38 and 4.74, with the Southern Elephant Seals the highest (and ranging from 4.67 to 4.74 across time periods). Of the penguins, the King Penguins were the highest at 4.41 with the Gentoo Penguins next at 3.91. The relatively high values for these two species reflect their predominantly fish diets. Penguins eating mainly crustaceans, the Macaroni Penguins and the Southern Rockhopper Penguins, had lower TLs at 3.80 and 3.59 respectively. All albatross were similarly ranked, with trophic levels between 4.30 and 4.40. The TL of the small flying birds was estimated to be 3.78. The mean trophic level for these land based top predators was 4.38.  The highest trophic level calculated among the fish was for the Patagonian Toothfish at 4.63. Sharks and Rays followed at 4.56. Large Pelagic Fish were the only other functional group of the fishes to occupy the fourth trophic position with a value of 4.31. All the remaining fish groups (all large and small demersals, as well as the Small Pelagic Fish and Myctophid Fish) were ranked at the third order trophic level with TLs estimated at between 3.14 and 3.79. It should be remembered that individual fish species within each group may occupy higher or lower TLs, but the diet matrix that reflects the prey of the group as a whole was used and dictates the resulting TLs found here.  Large Cephalopods were ranked relatively high at 4.05, with Small Cephalopods at 3.27. All remaining groups, including the Benthos group (2.27), the Benthic Decapod (2.21), and the three zooplankton groups (Large Zooplankton Crustacean 2.73, Small Zooplankton Crustacean 2.33, and All Other Zooplankton 2.00) occupied the 2nd trophic level.  Ecotrophic Efficiencies The Ecotrophic Efficiencies (EE) output of the models provided measures of how much of a group is used within the system (Figure 5.4). The results were as to be expected, with most of the top predators having EE values that were relatively low. Most of the seal and albatross groups were below values of 0.2, but the penguins and small flying birds had higher predation rates and therefore higher EEs. In many ecosystem models it is assumed that most of what is produced in the system is consumed and, when automated routines are adopted, values of 0.95 are suggested as being default EE estimates. In this construction, all EEs are estimated by the model as all 103  other parameters (B, P/B and Q/B) are entered. The result is that some of the EE estimates are less than might be expected and highlight areas where more information may be needed for the model construction. For instance, all the demersal fish have EEs at 0.6 or lower, which, particularly for the smaller sized demersal fish may be considered low. In terms of the Patagonian Toothfish the EE is high, at close to 1, for both the 1960s and the 2000s model, while for the 1980s this value is relatively low (0.3). The EE for the Small Pelagic Fish is too low (~0.2) for a group that is primarily a forage fish group and suggests that the biomass for this group should be reconsidered (reduced). In contrast, the Myctophid Fish, which trophically occupy the same niche as the Small Pelagic Fish have a high EE (between 0.9 and 1.0 for all three models) which is as expected. Partitioning of biomass between these two groups should be revised. Both the cephalopod groups have high EEs for the 1960s model, reduced through time in the later models. The EEs for the crustacean groups (both Large and Small) have EEs that are not as high as might be expected for groups which form such an important component of the diets of many predators. The patchy distribution of the zooplankton, both spatially and temporally, may be a contributing factor to the lack of direct consumption on these groups. Finally the low EEs of the Large Open Ocean Phytoplankton as well as the Island-associated blooms show that this production does not enter the system through direct consumption, but their contribution to the detrital food chain should not be overlooked. Explicit tracing of the contribution of these inputs to the detrital food chain would provide a way to quantify the importance of this production to the ecosystem and should be considered for future improvements of the model.  5.3.4. Ecosystem network analysis Transfer efficiencies For the system as a whole, the transfer efficiencies (TEs) averaged for the trophic levels (TLs) II ? IV are approximately 11.2 % (range 10.7 to 11.7%) across all three time periods (Table 5.4). When the efficiencies are calculated from the primary producers only, the values are slightly lower (10.7 ? 10.9%) than for the system when the transfer efficiency including the detritus is included in the calculations (11.6 ? 11.7%). Assessed per trophic level, the lower trophic levels have relatively high TEs (between 15.8 (producer) to 19.4% (detritus)), decreasing with 104  increasing TL showing good coupling between the lower trophic levels and their predators. Interestingly, among the three models the TE differs with TLs IV and higher for the three time periods, with the highest TEs found in the early model (1960s), and declining efficiency in the coupling of the system in the subsequent two models. The network analysis for the system is shown diagrammatically in the Lindeman Spine (Figure 5.5). The Lindeman Spine shows the consumption, predation, exports/catches, respiration and flow to detritus summarised for each trophic level (Baird and Ulanowicz 1993). The highest flows are in the lower trophic levels with minor differences between the time periods for trophic levels III, IV, and V owing to the differences in the biomasses of the top predators, also evident in the total system throughput. A summary of the biomass by TL is provided (Figure 5.6), clearly showing the highest biomass estimates are in TLs I ? III with TL II having the greatest overall for all three time periods.  Model structure Each of the balanced models constructed here is one of a number of possible models. To explore the basic structure of the model formulation, an assessment of the mixed trophic impact (MTI) provides a summary of the diet structure of all functional groups and assesses the impact each group has on all other groups. The impact can be positive (a prey has a positive impact on its predator) or negative (predation impact of predator on prey) and these patterns are seen in the data (Appendix 5.B). Interesting results include the effect of groups with relatively low biomass, which have impacts on other groups. These include the Orcas and Giant Petrels, which have both positive and negative effects on many of the land based mammals and birds. The Gentoo Penguins, despite their relatively low biomass contribution to the system as a whole, have a measureable effect on the demersal fish groups. King Penguins, with their relatively high biomass contribution to the system have a negative effect on most of the land based top predators, except for the Giant Petrels.  Along with the MTI, the Ecopath routine provides the relative total impact (RTI) which is an index of the trophic impact, consolidating the impact by removing whether it is negative or positive and considering only its magnitude. Results of this assessment are provided in Figure 105  5.7 and show that Orcas, Southern Elephant Seals and Giant Petrels are the vertebrate predators with the highest RTI. The Orcas consistently have the highest impact of 1. The Southern Elephant Seals also rate high at around 0.5, even in the 2000s model when the biomass for this group is substantially reduced, and they score higher than the Sub-Antarctic Fur Seals, even in the 2000s when their biomass estimates are far lower than that of the fur seals (less than 1000 t versus approximately 5000 t). The Giant Petrels are shown to have a high RTI. Following these groups, the King Penguins also consistently rank as a key group for all time periods with an intermediary biomass contribution to the system. As expected, the importance of the nekton groups (Small Pelagic Fish, Myctophid Fish, both cephalopod groups) are identified through the RTI and all have corresponding high biomass contributions. The Patagonian Toothfish is also identified, though this is only true for the two models prior to the start of the fishery (1960s and 1980s), following which the index becomes much reduced with the reduction in its biomass contribution. Finally, the Benthos and Large Zooplankton Crustaceans are also identified as having relatively high RTI scores highlighting the importance of these two groups in the system.  In this assessment it is useful to discuss what species might be termed ?keystone? in the ecosystem. A keystone species can be defined as a species that plays an important role in an ecosystem while at the same time making a relatively low biomass contribution to the ecosystem (Power et al. 1996).  By plotting the RTI against biomass values for each of the functional groups, an indication of which species might be considered ?keystone? can be provided. In Figure 5.8 the plot of this index against each group?s relative biomass for all functional groups for the three time periods is provided. It highlights the findings noted above that, while some groups with high biomasses are found to have key roles in the foodweb (e.g. the nekton groups), the groups with low biomasses fit the description of ?keystone species?. Those that can be highlighted as such (as defined here) in this ecosystem are the Orcas, Southern Elephant Seals and the Giant Petrels (labelled on the figure), with all three groups contributing less than 0.01% to the living biomass of the ecosystem as a whole yet having a high RTI. The Gentoo Penguins are also worth mentioning, as although the RTI for this group is not particularly high (around 0.2), its RTI is relatively high compared to the low contribution this group makes to the overall biomass. 106  5.3.5. Summary statistics  Statistics and flows of the three models constructed for this study are given in Table 5.5 along with the summary statistics of an additional 8 ecosystem models of other Sub-Antarctic and Antarctic systems for comparative purposes. The models include that of the Kerguelen Plateau (Pruvost et al. 2005), South Georgia (Hill et al. 2012), the South Shetlands (Bredesen 2003), the Falklands (Cheung and Pitcher 2005), three different assessments of the Antarctic Peninsula (Erfan and Pitcher 2005; Cornejo-Donoso and Antezana 2008; Hoover 2012), and the Southern Plateau (NZ) (Bradford-Grieve et al. 2003). The summary statistics provide an overview of the system, with information on the magnitude of the ecosystem and a summary of the flows within it. Comparisons of the summary statistics for the three different time periods constructed of the PEIs show minor differences between model formulations suggesting that, while the constituents of the system have changed over time, these elements are not captured in these metrics.  The net primary production of the system was estimated to be 1278 t.km-2.yr-1, with the total primary production divided by the respiration being 1.56. This indicator is an index of ecosystem maturity, with systems close to one being considered mature as the production and respiration tend to be balanced (Christensen et al. 1993; Christensen 1995a). For the PEI system, with the indicator greater than 1, more is produced than respired and this suggests that the system is intermediate in its maturity. The net system production was 459 t.km-2.yr-1 and total production/biomass, which may be considered an indirect measure of organism size, was 30.38. The total biomass (excluding detritus), which can be used as a measure of ecosystem size was 42.08 t.km-2, indicating a relatively small system, while the total biomass/ total throughput of 0.012 yr-1 suggests a relatively complex system. Regarding the fisheries and the metric provided to assess it, the data for the mean trophic level of the catch for the PEI 2000s model (the only model for which there is a fishery) shows a value of 4.63, which is most similar to that of the Southern Plateau model (4.48, Bradford-Grieve et al. 2003). In both instances, the Patagonian toothfish is the only fishery. For the other models summarised here, mixed fishery catches (South Georgia and South Shetlands) or principally krill catches (Antarctic Peninsula) result in lower trophic level of catches. The connectance and 107  omnivory indices for the PEI models were similar to the average of all systems summarised here and could be considered relatively low. 5.4. Discussion This construction of a mass balanced model of the PEIs attempts to integrate many sources of scientific data. The model formulation is only one possible construction representing the ecosystem, and while the data search was extensive, improvements from consultation with experts for each parameter would certainly benefit the model estimates. It is hoped that the full transparency on the data preparation that is included here (see Appendix 5.A) will ease future contributions and improvements to the model construction.  It is evident from the compilation of data, and the assessment summarised in the pedigree index of the data quality, that many of the pelagic groups would benefit greatly from quantitative assessments. In particular, assessments of all the large fish groups (the Patagonian Toothfish, Large Demersal Fish, Large Nototheniid Demersal Fish and Large Pelagic Fish) are needed. With only one trawl survey of the fish population to date, a quantitative survey would greatly improve the data of these groups. Secondly, assessments of the nekton are almost entirely lacking with no dedicated sampling efforts having been applied to this group in the vicinity of the islands. Considering the importance of the nekton (small pelagic fish and cephalopod groups) to the ecosystem as a whole, they must be highlighted as requiring urgent attention in terms of data requirements. For rate estimates, attention should be focused on the lower trophic levels, the values for which have been derived in all instances from other models. Future scientific data collection efforts should not only address the areas in which there are data gaps, but should also focus on those parameters that have the most impact on the system. The results of the sensitivity tests highlight groups that can be considered important. These range from the top trophic levels (Orcas, Southern Elephant Seals) through the middle order groups (Small Pelagic Fish, Myctophid Fish, Benthos) and also include some of the lower trophic levels (Large Zooplankton Crustaceans and Small Zooplankton Crustaceans) for all three models. In addition, for the 1980s time period, both the Giant Petrels and the Patagonian Toothfish were identified as having significant impacts on the system. The results of these sensitivity tests are 108  interesting as they highlight that, while some groups with high biomass contributions (e.g. the small pelagic fish and the crustacean groups) are identified as being important, this is not necessarily always the case (e.g. Orcas, Southern Elephant Seals). A range of TLs was represented by the species identified to be of importance through the sensitivity test, as was found by Olson and Watters (2003). In their study some low TL groups (primary producers, secondary consumers, epipelagic fish) and some high TL groups (marlins, sharks, toothed whales) along with some middle order TL groups (cephalopods) were all identified. In contrast, Allain et al. (2007) found that the results of the test identified the most important groups to be in the lower trophic levels only. To assist in identifying where future efforts should be focussed, a qualitative summary of the relative urgency for attention has been compiled (Table 5.6). The table indicates which groups were highlighted, not only through the process of identifying data gaps/-data quality issues (Pedigree index), but also through an assessment of the impact the group had on the model (through the sensitivity index, mixed trophic index, relative trophic impact and the identification of ?Keystone? species). The Orcas, Southern Elephant Seals and Giant Petrels were identified through this assessment as being of primary importance. These groups were followed in rankings by the Myctophid Fish, Small Pelagic Fish groups and Gentoo Penguins. The King Penguin had the next highest ranking, equivalent to the cephalopod groups. Finally, all the large fish groups, benthic and pelagic and including the Sharks and Rays were identified. A noticeable omission from the species identified as having an impact on the system is the decapod shrimp, Nauticaris marionis. Numerous studies have identified this species as having a key role in the system and its absence in this assessment may be due to the scale of the system adopted here.  In summarising the data, are we able to capture the differences for the three separate time periods in the network analysis? Input data into the models for these different time periods differ only for the groups for which there are time series data. In total this is 15 of the 37 functional groups. Fourteen of these are mammal and seabird groups (the only land based top predator functional group without time series data is the Prions and Petrels). The only other group for which the biomass estimates are adjusted for each time period is the Patagonian Toothfish. In terms of the biomasses, all these groups combined count for less than 0.4% of the total consumer biomass for 109  the system. So, although the changes in the ecosystem are relatively significant for some of the functional groups (the reduction in the elephant seals, the increase in the fur seals, the changes in the penguin populations, the collapse of the Patagonian toothfish stock) they are not reflected in most of the metrics which capture the biomasses and flows for the system as a whole, showing only minor differences between periods. A challenge is how to capture these changes using the ecosystem indicators. One indicator that does reflect some differences is the TE for the higher TLs (greater than IV), which declines in each subsequent time period. One interpretation of the reduced TE in subsequent models may be related to a decline in the efficiency of the system following a perturbation. Christensen and Pauly (1998) refer to systems becoming ?leaky? following exploitation. Transfer between trophic levels becomes ?sloppy? and energetically the coupling declines. This finding is also illustrated here through the decline in the TEs for the higher trophic levels. Alternatively the decrease in efficiency may be linked to the decrease in biomass of the higher TLs resulting in less absolute transfer (fewer predators) and a lower TE. Regarding the summary statistics for the ecosystem as a whole in comparison to the other 8 Southern Ocean/-Antarctic systems (Table 5.5), the PEI statistics were found to be most similar to those of the Kerguelen Islands (Pruvost et al. 2005) for most metrics (sum of consumption and production, flows, and total system throughput). The Southern Plateau model, with units of carbon (as opposed to biomass), is not directly comparable excepting for non-dimensional measures. In general, the Falklands, South Shetlands and South Georgia model ecosystems were found to have the highest values for most of the metrics, with the three Antarctic Peninsula models spread across the range of systems summarised here. As an example, the total biomass (excluding detritus), which can be considered an estimate of ecosystem size, was lowest for the PEI models (42 t.km-2) with the Kerguelen Islands model twice as high (at 82.69 t.km-2). The three Antarctic Peninsula models ranged from 156 to 557 t.km-2 and the Falklands, South Georgia and South Shetlands between 244 and 358 t.km-2. 5.5 Conclusions This work provides the first quantitative network model of the PEIs and highlights areas for which more research is needed to address data gaps while focusing on those elements of the system that have been identified to play an important role in the system. Through further 110  development and improvement this model can provide a platform for an ecosystem approach to marine resource management, where options for both fisheries and conservation objectives can be explored. The extension of the model to include temporal simulations through the fitting of time series data will allow for hindcasting of past populations dynamics (Chapter 8), as well as investigating forecasting using potential climate change scenarios (Chapter 9). Developing the model further to be spatially explicit will also allow assessment of policy options, such as the implications following the recent declaration of the PEI Marine Protected Area (April, 2013) (https://www.environment.gov.za/content/princeedwardislands_declaredmarineprotectedarea). Clarity of the data compilation will also provide a useful dataset for the development and use of other ecosystem models which can be constructed to compare and contrast management scenarios.111  Table 5.1. Model parameters used for input for all three time periods. Includes Biomass (t or t.km-2) calculated for the Exclusive Economic Zone (EEZ) of the Prince Edward Islands (PEIs) for 1960s, 1980s and 2000s; Estimates of Production to Biomass (yr-1); Consumption to Biomass (yr-1) and Unassimilated Consumption (UC) is provided. Model outputs of Trophic Level (TL) and Production to Consumption rates (P/Q) given. Black text indicates the original estimates (done in t or t.km-2, whichever was appropriate). Blue text is the conversion to relevant units (from the black text figures). * 0.08 for 1960s & 1980s models, 0.13 for 2000s model; **5.11 for 1960s, 5.09 for 1980s, 4.98 for 2000s; ***4.74 for 1960s & 1980s, 4.67 for 2000s; 5.10 for 1960s, 5.15 for 1980s, 5.18 for 2000s.   112  Table 5.2. (A) Diet matrix for each functional group except Orcas, Southern Elephant Seals and Giant Petrels with contributions summing to 1 for all consumers for all three time periods (1960s, 1980s, and 2000s); (B) Diet matrix for Orcas, Southern Elephant Seals and Giant Petrels with unique diets provided for each time period (1960s, 1980s, and 2000s) (see text for details).    113    114  Table 5.3. Index of Data Pedigree generated for each functional group based on the data quality of three input parameters (Biomass (B), Production to Biomass (P/B) and Consumption to Biomass (Q/B) estimates). Key for generating the index provided.     115  Table 5.4. Transfer efficiencies summarised by Trophic Level (TL) for flows from the producers, the detritus and all flows combined for all three time periods. A summary of the flows for TLs 2 to 4 for each case also provided.    116  Table 5.5. Summary statistics of the Prince Edward Island model for each time period (1960s, 1980s and 2000s) as compared to 8 other Southern Ocean/Antarctic Ecopath models.    117  Table 5.6. Qualitative assessment to highlight where scientific research efforts should be focussed in future. 118    Figure 5.1. Study areas. Map showing the Exclusive Economic Zone (EEZ) of the Sub-Antarctic Prince Edward Islands situated southeast of South Africa.   119    Figure 5.2. Schematic representation of the food web. Each functional group is represented by a circle which is scaled to the square-root of the biomass (t); groups are distributed with increasing Trophic Level (TL) on the y-axis and all trophic linkages indicated by grey lines. 120    Figure 5.3. Results of the sensitivity analysis. The index is the count of estimated parameters of the model affected by at least 30%, given the changes (between -50% and 50%) in the input parameters of each functional group (listed on the y-axis). Effects within groups have been eliminated from the analysis.121    Figure 5.4. Plot showing the model results of the Ecotrophic Efficiencies (EE) (0-1), an index of how much of the production for each functional group is used in the system, for each of the three time periods (1960s, 1980s and 2000s).122     Figure 5.5. Diagram showing the trophic flows as summarised in the Lindeman spine for each of the three time periods for which models were constructed, A) 1960s, B) 1980s  and C) 2000 (TST = Total System Throughput, TE ? Transfer Efficiency).     A B C 123   Figure 5.6. Bar plot showing the Biomass (B) by trophic level for each of the three time periods (1960s, 1980s, and 2000s).   124   Figure 5.7. Plot showing the relative trophic impact (RTI) of each functional group for each time period (1960s, 1980s, and 2000s).   Figure 5.8. Scatter plot of the relative trophic impact (RTI) plotted against biomass (B), providing an indicator of ?keystone? species.   125  Chapter 6 Exploring ecosystem boundary size  6.1. Introduction In ecological theory, an ecosystem is an area within which the energy flow between community production and consumption is balanced (Odum 1969). Such a statement suggests that there are boundaries to ecosystems, and therefore ecosystem boundaries are intrinsic to ecology theory. But setting the boundary for a marine ecosystem can be difficult because, unlike with terrestrial systems where physical boundaries may be obvious, in marine systems the transition between ecosystems is often not discreet. Exploring the boundary size of a marine ecosystem through energetic requirements can provide insight into the scale at which the system should be considered. An island-centered marine ecosystem provides a system that is easy to conceptualize for the exploration of ecosystem boundaries, particularly when many of the constituents of the ecosystem are bound, for one reason or another, to the island or island shelf. A study on the Privolov Islands (Cianelli et al. 2004) explored the issue of ecosystem boundary size through an energetic approach, where the boundary of the system was determined in terms of the balance between predatory demand and prey production. The study focused on the centrally placed foragers in the system and assessed at what extent the system should be considered. The Cianelli et al. (2004) study inspired this case study of the Prince Edward Islands (PEIs) to address the same question. Many of the top predators at the PEIs are island-based for the duration of their residency in the region considered here and the scale of the system can therefore be investigated through similar means. In order to conceptualize an ecosystem as an entity around which to draw a boundary and consider the system of flows to be mass balanced, one has to assume that i) the system is closed, ii) the imports balance the exports, or iii) explicitly account for both the imports to and exports from the system using net migration and biomass accumulation/loss terms if required.  The scale of the formulation of the ecosystem model will depend on which option is selected. For the PEIs, it would not be appropriate to consider the system closed as it lies in an easterly flowing Antarctic Circumpolar Current with a ?conveyor belt? of open ocean production flowing into the system. At the same time, this flow-through environment means that the elevated 126  productivity, which is a direct result of the interaction of the oceanography and the islands themselves, also flows out of the system. The premise of the model used here is that the inflow and outflow of the imports and exports are in balance when the system is considered at the scale of the Exclusive Economic Zone (EEZ).  In the case of the PEIs, the pelagic system is a flow through system associated with the Antarctic Circumpolar current, flowing from west to east. For this model, the assumption is made that the pelagic imports and exports may be considered equivalent. Hypothetically, the predation impact of the top predators on the incoming pelagic resources (as has been documented by Hunt et al. (2001)) could be balanced by the increased pelagic production which is stimulated following the island mass effect, with the additional production associated with the islands resulting in downstream increases in pelagic productivity.  In addition to addressing the question of imports/ exports, in order to assume a ?closed? system one needs to adjust the parameters of the model to account for the residence time of the species that do not spend the entire year within the system. This can be relatively easily achieved by adjusting the biomass (t) of the respective groups according to the time they spend at the islands. This approach has been used in the formulation of the original PEI ecosystem model and is preferable to making adjustments to consumption and production rates (to achieve the same goal) as these rates are in many instances based on field based estimates associated with the biological demands of the life stage of the residents when based at the islands. For example, consumption rates measured for many of the land based top predators are derived from data collected during the breeding season, the only period in the year when the data are accessible. These consumption rates are higher than the rates for other times of the year due to the increased energetic demands associated with breeding, but as the model is attempting to capture the rates for the periods when the top predators are in the system, these elevated rates are appropriate for the model construction.  The objective of this study was to consider the Prince Edward Island ecosystem at various scales to investigate the size at which the energy requirements of all functional groups could be met, and explore the ecosystem boundary size at which the various functional groups should be considered.  127  6.2. Methods The mass balanced Ecopath model of the ecosystem representing the EEZ of the Prince Edward islands (PEIs) for the 1980s time period was used (see Section 5.2). This model was then adjusted to create an additional three smaller models of the system. The areas for each model were calculated with four different radius lengths (the original remaining with a radius of 200nm; the additional models with radii of 100nm, 50nm and 20nm) (Table 6.1, Figure 6.1). Each model was centered at the midpoint between the two islands that make up the Archipelago (46?46?S, 37?51?E). Data for biomass, production and consumption rates as well as the diet matrices used are provided in Chapter 5 (see Table 5.2 and 5.3; Appendix 5A).  For each of the additional three models, the biomass per unit area (t.km2) was scaled according to the size of each ecosystem (i.e. higher biomass per unit area with smaller boundary size for all top predators, as they were assumed to fill the entire area of each model). Biomasses per unit area of pelagic groups, however, remained the same between models. Biomass estimates for the benthic components of the ecosystem (i.e. all demersal fish, the benthic shrimp, Nauticaris marionis, the ?Benthos?, ?Macrophytes? and ?Macrophyte Detritus? groups) were scaled according to the area that the benthic component occupied of the total area for each model.  Following the calculation of the basic estimates for each of the output parameters of the model, the ecotrophic efficiencies (EEs, a measure of how much of each group is used in the system) and consumption mortalities were used to explore the ecosystem size. Finally, a balanced model was constructed by making the necessary adjustments to the parameters of the model to assess which of the above four models was most appropriate for each of the functional groups. For comparative purposes, the above exercise was then repeated for the 2000s model (Input parameters summarized in Section 5.2 and provided in Table 6.2). 6.3. Results The results of the outputs following the basic estimate routine of the Ecopath model are provided in Table 6.3.   128  Model 1 (200 nm radius):  This model represents the PEI EEZ and balances without adjustment, as is described in Chapter 5 for the 1980s version of the model.  Model 2 (100 nm radius):  When the model is run, it does not balance. The ecotrophic efficiencies (EE) for 4 groups exceed 1, i.e. the production of each of these groups is not sufficient to supply the consumption. These groups are the Patagonian Toothfish (EE = 1.20), the Myctophid Fish group (EE = 2.11) and the two Cephalopod groups (Large Cephalopods EE = 2.25, and Small Cephalopods EE = 2.14). Ecopath provides a breakdown of the mortality estimates caused by each functional group on each functional group, and highlights which groups are ?responsible? for the over-consumption of the groups. The Southern Elephant Seals are identified as being the primary consumers of the Patagonian Toothfish, while the King Penguins are identified as the primary consumers of the Myctophid Fish. In the case of the two cephalopod groups, the greatest consumers of the Large Cephalopods are many of the land based top predators, including the Southern Elephant Seals, the King Penguins, and the Macaroni Penguins, and for the Small Cephalopods the Prions and Petrels. This result suggests that the foraging range for these groups is greater than the 100nm model constraint.  Model 3: 50 nm model On the first run, the groups that had an EE >1 were the Sharks and Rays (EE = 2.63), Large Pelagic Fish (EE = 2.34), Patagonian Toothfish (EE = 4.77), Small Pelagic Fish (EE = 1.140), Myctophid Fish (EE = 6.74) and the Large and Small Cephalopods (EE = 7.96 and 7.01, respectively). Investigation into the predation mortalities showed that the Southern Elephant Seals were responsible for the consumption on the Sharks and Rays, the Large Pelagics and the Patagonian Toothfish. The King Penguins consumed the Myctophid Fish and the Large and Small Cephalopods, and finally the Macaroni Penguins and the Prions and Petrels consumed the Cephalopods (Large and Small respectively).   129  Model 4: 20 nm model This model is the smallest one considered, and could never conceivably provide a balanced outcome with all the land based top predators constrained to feed in such a limited area. Accordingly, the ecotrophic efficiencies for the prey of these groups are greatly over consumed by up to 48 times (Table 6.3). The Sharks and Rays (EE = 16.434), Large Pelagic Fish (EE = 12.06), Patagonian Toothfish (EE = 29.74), the Small Pelagic Fish (EE = 6.54) Myctophid Fish (EE = 39.15) and Large (EE = 47.93) and Small Cephalopods (EE = 41.68) highlight this. In addition, some of the smaller prey were also over consumed: the Large Zooplankton Crustaceans (2.568), the Small Zooplankton Crustaceans (EE = 1.103) and the Other Zooplankton (EE = 1.21). Investigation into the mortality rates highlights the ?culprits? as the Southern Elephant Seals on the Sharks and Rays, the Large Pelagic Fish, the Patagonian Toothfish and the Cephalopods (both groups); the Sub-Antarctic Fur Seals on the Myctophids Fish; the King Penguins and Macaroni Penguins on the Small Pelagic Fish, Myctophid Fish as well as both groups of cephalopods; the Rockhopper Penguins on the Small Cephalopods; the Wandering Albatross on the Cephalopods; the Prions and Petrels on the Myctophid Fish and the Large and Small Cephalopods; and the Large pelagic Fish on the Large Cephalopods. Model 5: Mixed Balanced Model (MBM)  For this model, an attempt was made to balance the model at the scale appropriate for each functional group from a top down perspective. Groups identified above were adjusted to see at what scale each group should be considered. Adjustments were made, in an iterative process, based on the information provided above. It is evident from Model 2 (100nm model) that many of the top predators could not satisfy their energetic requirements if restricted to prey inside this theoretical area. Using Model 2 as a starting point, the biomass per unit area for both the Southern Elephant Seals and the King Penguins were reset to their Model 1 (200nm model) values. When the model was rerun, it still did not balance, and a new ?culprit? was identified in the Giant Petrels which were then also reduced to biomass estimates reflecting the Model 1 values. Overconsumption of the Patagonian Toothfish and Myctophid Fish as well as both groups of Cephalopods was still evident (EEs for these groups > 1). Further adjustments to the fur seals (both groups) reduced the EEs to close to 1 (Myctophid Fish 1.07, Large Cephalopods 130  1.07, Small Cephalopods 1.27). With a reduction in biomass of the Macaroni Penguins, the model came close to balancing with only the Myctophid Fish and Small Cephalopods with EEs exceeding 1. A final adjustment of the Prions and Petrels to the Model 1 biomass estimate resulted in a balanced model.  When attempting to balance the model based on the Model 3 (50nm model) parameters, initial amendments were as before, adjusting those groups highlighted by the model to be causing the over-consumption (reflected in the EEs). They were the Southern Elephant Seals, Fur Seals, King Penguins, Macaroni Penguins, Giant Petrels and the Prions and Petrels which were all set to the biomass estimates for the largest model (Model 1). After attempting to balance this model, the Orcas, Southern Rockhoppers, Large Demersal Fish and Large Nototheniid Demersal Fish were all highlighted as being responsible for the continued overconsumption. Adjustments of the first three of these groups to biomasses from Model 2 (100 nm) were made. The Large Nototheniid Demersal Fish biomass estimate was set to be intermediary between Model 2 and 3 (with a value of 0.0440 t.km-2). The albatrosses were then highlighted as being responsible for the unbalanced outcome, and only when these biomasses were reduced to the Model 1 estimates did they stop contributing to the overconsumption. Following these adjustments, to balance the model both the Myctophid Fish and the Large Cephalopod groups had to be increased by small amounts (Myctophid Fish from 3.38 to 3.40 t.km-2; Large Cephalopods from 0.065 to 0.060 t.km-2) to enable the system to balance. Finally, an attempt to balance the smallest model (Model 4) was made. The pattern of identifying those groups that were over-consumed in the model parameterization procedure and identifying the consumers was repeated. As before the biomass estimates for the Seal groups (Southern Elephant, Antarctic Fur Seal and Sub-Antarctic Fur Seal), two of the penguin groups (King Penguin and Macaroni Penguin), all the albatross groups, the Giant Petrels and the Prion and Petrel groups were reduced to reflect estimates used in the largest model (Model 1). Following this, the Orcas and the Southern Rockhopper Penguin groups were reduced to the Model 2 estimates and the Large Demersal Fish group to the Model 3 estimate. These adjustments brought the model close to balancing, with only 4 groups over-consumed (i.e. EE >1): Myctophid Fish (EE=1.49), Large Cephalopods (E=1.18), Large Zooplankton Crustaceans (E=1.49) and All 131  Other Zooplankton (EE=1.19). Adjustments to the Large Nototheniid Demersal Fish (to 0.26 t.km-2, a value between the Model 2 and 3 estimates) and the Benthos (to the Model 3 value of 8.0 t.km-2) brought the model very close to balancing with only the Myctophid Fish and Large Cephalopods still over-consumed (EEs of 1.12 and 1.05 respectively). Redistribution of biomass between the Small Pelagic Fish and the Myctophid Fish groups (increase of Myctophid Fish by 0.225t.km-2 and decrease to Small Pelagic Fish of the same), along with an increase in the Large Cephalopods value from 0.065 to 0.070, brought the model to balance. As is evident from the data supplied in Chapter 2, the composition of the top predators has changed significantly in the 20 years that separate the 1980s model description from that in the 2000s. In particular, the declines recorded in the Southern Elephant Seal population, the recovery of the fur seal populations as well as the collapse of the Patagonian toothfish stock would create a different scenario for this analysis. Therefore, the exercise was repeated for the dataset which represents the islands for the 2000s period, and the outcome showed some differences. In the first reduction in ecosystem size, only the King Penguins were identified to be responsible for over-consumption of the Myctophids resulting in the model not balancing. With a further reduction in ecosystem size (to the Model 3 size) the seals (both the Southern Elephant Seals and the Sub-Antarctic Fur Seals) and the Macaroni Penguins were identified as contributing to the over-consumption (on the Sharks and Rays, Large Pelagics, Patagonian Toothfish, Myctophid Fish and both Large and Small Cephalopod groups). On the final reduction in boundary size (Model 4) the Wandering Albatross and Prions and Petrels were added to the aforementioned groups and, along with the Large Demersals, were responsible for the model not balancing (due to over-consumption of the before mentioned groups and the Small Pelagic Fish).  A difference found between the two time periods was that only the King Penguins were identified as contributing to the overconsumption in the first model size reduction. The Southern Elephants Seals, the Sub-Antarctic Fur Seals and the Macaroni penguins are only identified in the subsequent reduction in model size, and finally the Wandering Albatross, Prions and Petrels and Large Demersals were identified in the smallest model. Another notable difference between the two time periods was that the Southern Rockhopper Penguins have no effect on the over-consumption for the 2000s time period while they are identified as being important to the 132  consumption in the 1980s model of the same size. This reflects the decline in the Southern Rockhopper Penguin population between these two time periods.  6.4. Discussion Due to the scaling of the ecosystem models, the initial absolute biomass of the top predators and the benthic ecosystem effectively stayed the same in each model description, while the pelagic component became reduced with decreasing ecosystem size. Therefore the total absolute quantity of pelagic resources available within the system was dependant on the ecosystem size. In this exercise, with the decrease in ecosystem size, the pelagic resources were reduced. For those consumers that rely on the pelagic component to meet their consumption needs, the ecosystem boundary size would determine the population size that could be sustained.  It is apparent from the formulation of Model 1 (at the scale of the EEZ) that, if the system is considered to be in mass balance (in this instance with the assumption that imports and exports are equivalent), the energetic demands of the functional groups included in the model can be met. However, with each subsequent model that represented a reduction in size of the ecosystem (Models 2, 3 and 4), various functional groups could no longer be sustained within the system. In each instance the model was not able to achieve mass balance and groups that were over-consumed were highlighted (top-down approach). Investigation into those groups that were responsible for the over-consumption provided insight into the scale at which each group should be considered. For Model 2, the Southern Elephant Seals and the King Penguins were the first two groups to be identified as being responsible for the over consumption within the model. Following these two, the Macaroni Penguins and the Prions and Petrels were identified. This result was found in the Model 3 results. For the Model 4 (shelf model) results, in addition to the four functional groups already identified, the Sub-Antarctic Fur Seals, Rockhopper Penguins, and Wandering Albatross were also added to the list of top predators unable to satisfy their energetic demands at the smaller scale.  To gain a better understanding of the knock-on effects of adjustments to the models that would be required to balance the model, Model 5, a mixed balanced model, was constructed. The model provided a summary of the appropriate ecosystem size for the various groups of top predators 133  included in the assessment for the 1980s time period. The outcome of the balancing exercise suggested that the Southern Elephant Seals, the Antarctic and Sub-Antarctic Fur Seals, the King Penguins, Macaroni Penguins, Giant Petrels, albatross groups and Prions and Petrels should all be considered at the largest scale (Model 1, 200nm or EEZ). The Orcas and Rockhopper Penguins could be considered at a smaller scale (Model 2, 100nm) along with the Large Demersal Fish group. In addition, changes to the Large Demersal Nototheniid Fish (to a value of 0.26 which is intermediary between the Models 3 and 4) and a change in the Benthos group to the Model 3 value (of 8.0t.km-2) brought the model very close to balancing. The redistribution of biomass between the Small Pelagic Fish and the Myctophid Fish along with a small increase in the Large Cephalopods (from 0.065 to 0.070 t.km-2) resulted in a balanced model.  This exercise provided a routine that establishes the appropriate scale to consider the various groups of the model. At smaller ecosystem sizes the consumption needs of the larger predators could not be satisfied. The results tie in well with the foraging ranges of the top predators, which are considered to ultimately dictate the size at which particular functional groups should be considered to enable the ecosystem model to balance. Seals are known to have extensive foraging ranges, with the Southern Elephant Seals covering the largest distances (> 1 000km (~540nm) from Marion Island (Bester 1989; Bester and Pansegrouw 1992)). Of the penguins, the Kings travel the furthest (mean foraging range of adults with chicks during summer at Marion Island ranged between 225km (122nm) and 300km (162nm), depending on the size of their chicks (Adams 1987). These groups were identified as the first two to not be able to satisfy their energetic requirements in the reduced ecosystem size. Foraging range estimates for the Macaroni penguins have been recorded as being between 59 and 303 km (32 and 164nm) and between 4 and 157km (2.2 and 84.8nm) for Southern Rockhopper penguins (Brown 1987). In previous studies, the ecological extent of the PEIs system was considered to be an ocean area within a radius of approximately 300km (or 161 nm) from the islands, based on the foraging range of the seabirds nesting at the islands (Cooper pers. Comm. in Gon and Klages 1988). This estimate is similar to the finding in this model, where the energetic requirements of most of the avian fauna could not be met at the Model 2 (100nm radius) scale but were satisfied at the Model 1 (or EEZ) scale. The only land based top predator able to satisfy its energetic requirements in the smallest rendition of the model (Model 4) is the Gentoo penguin, and this concurs with its known 134  foraging range, which is thought to be less than 40km (21.6nm) and matches closely with the boundary of the smallest model (20nm radius).  While the outcome of this exercise is intuitive, it is encouraging to see that the model output supported the observations of the foraging distances of many of the top predators. The assumption is that each group would only travel as far as is necessary to satisfy the energetic demands for itself and its offspring. It would not make sense for these centrally placed foragers to travel great distances if the food availability within a smaller area was sufficient. However, in this exercise, there is no consideration of the heterogeneity of the system, temporally or spatially. The biomass estimates for example are assumed to be homogenous and spread evenly throughout the area under consideration which does not capture the reality of the often patchy environment. The question then arises whether an ecosystem model, without explicit spatial mapping, could provide insight into the foraging distances of these land based top predators. The results provide some confidence that despite the lack of explicit consideration in the spatial context of the ecosystem in the model construction (such as elevated productivity associated with oceanic fronts or shelf regions) the model is able to capture such patterns. Such patterns are, however, only evident when the groups considered are at their carrying capacity within the ecosystem, as when the numbers are below those that the ecosystem can support, the results are not clear. This is highlighted for example in the repeat of the exercise for the 2000s time period. The decline in the Southern Elephant Seals population resulted in this group only being picked up as needing a larger foraging range in the Model 3 configuration, a range well below the observed foraging range of this species. In addition, the Southern Rockhopper Penguins are not identified as requiring an increased foraging range due to their reduced population size during this time period.  Taking both considerations into account (the lack of spatial heterogeneity and the carrying capacity of the populations) the results of the exercise can provide insight into the minimum extent of the ecosystem boundary while the observed foraging distances of the system constituents might be considered the maximum extent as found in the Cianelli et al. (2004) study.  135  6.5. Conclusions In the process of stepwise reduction of the size of the ecosystem, various groups were highlighted as being over consumed, and investigating the consumption rates of the predators allowed the consumers responsible to be identified. It is interesting to see that those groups that were identified are also the groups that have the greatest known foraging distances. This exercise was dedicated to assessing at what scale the model of the PEIs should be created. Is there a boundary size that is appropriate? The results show that the answer ultimately depends on the functional group in which one is interested. If considering the ecosystem as a whole with all the constituents without quantifying the imports/exports or biomass accumulation, the results here suggest that the formulation at the scale of the EEZ (Model 1) would be appropriate. If, however, the focus was to be on functional groups that do not utilize the resources of such a vast area, then, by making biomass adjustments to particular functional groups (reduction of groups that require larger area to satisfy their energetic demands) one can reduce the total area considered in the model and consider the ecosystem at a smaller scale.  Further development of the model to include spatial data could greatly improve this model. The EwE program currently has a habitat capacity model under development which will be able to include spatially explicit preferences for species movements associated with physical features (both spatially and temporally resolved) within the model framework. The use of the existing spatially resolved database generated from the Lombard et al. (2007) study would provide a platform from which the habitat capacity model could be run. This will provide an exciting avenue for exploring future management options with the inclusion of explicit spatial considerations in particular considering the recently (April 2013) declared Marine Protected Area at the Prince Edward Islands and associated management zones.   136  Table 6.1. Description of the model size according to the length of the radius (nm) used, area included in the model (km2) and the proportion of each model that could be considered to represent the shelf region (as a percentage).  137  Table 6.2. Input parameters for each model. MBM = Mixed Balanced Model. B = Relative biomass (t.km-2). P/B = Rate of production to biomass (yr-1). Q/B = Rate of consumption to biomass (yr-1). UC = Unassimilated consumption (between 0 and 1). Numbers in black = Pelagic associated groups. Numbers in grey = Land based or benthic associated groups. * = value = 0.08 for 1980s model and 0.13 for 2000s model.   138  Table 6.3. Ecotrophic Efficiency output estimates for all Model sizes 1-4 for the 1990s and 2000s time period, as well as for the Mixed Balanced Model created for the 1980s. Ecotrophic Efficiency values that exceed 1 are printed in red text.  139      Figure 6.1. Illustration of the theoretical boundaries of each of the four models used in the study. The outermost ellipse represents Model 1 (an area equivalent to a circle with a radius of 200nm centered on the islands, representing the Exclusive Economic Zone of the Prince Edward Islands). The innermost ellipse represents Model 4 (an area equivalent to a circle with a radius of 20nm, representing the shelf region of the islands). Bathymetry data source: The GEBCO_08 Grid, version 20100927, http:www.gebco.net.    140  Chapter 7 The importance of incorporating diet quality into consumption rates for ecosystem modelling studies 7.1. Introduction Ecosystem modelling approaches are becoming more popular as researchers endeavour to find practical solutions that can aid addressing marine resource management objectives to serve both conservation and exploitation needs. Parameterisation of ecosystem models is an important step in creating a model and can be time consuming to complete. Some parameters needed for ecosystem models are considered to be model specific, such as biomass estimates and the trophic linkages between groups (diet matrices), where data unique to the study area are compiled for each model. Other parameters, like production and consumption estimates, are generally considered transferable between systems and hence are often used across systems, without taking system specific considerations into account. These include the use of empirical equations for parameterisation of such rates. For production terms, species specific and site specific information is often incorporated into the estimates, which are calculated for each model (e.g. P/B rates estimated for fish using the equation of Palomares and Pauly (1998), which takes species specific information and local temperature into account). For consumption rates, however, it is common to transfer estimates between systems without taking local differences into consideration.  A variety of approaches to determine consumption rates are commonly used in ecosystem models for mammals and birds (Kleiber 1961; Laws 1977; Innes et al. 1987; Nagy 1987; Trites et al. 1997; Nagy et al. 1999; Ellis and Gabrielsen 2001; Knox 2007b). In some instances, consumption estimates are considered to be directly related to body mass, simply as a percentage of body mass consumed per day  (Laws 1977; Pakhomov et al. 1996; Knox 2007b), and have been used to determine the consumption rates for invertebrates, fish and mammals. More commonly for higher vertebrates (birds and mammals), consumption estimates are based on allometric equations of body mass (M) C= aMb, where C is consumption per day, and ?a? and ?b? are uniquely defined parameters for species or groups. In such empirical equations, consumption is calculated as being determined by the daily requirements of the individual (as estimated from 141  metabolic rates (basal metabolic rates (BMR) or field metabolic rates (FMR)) in terms of biomass of food ingested (kg.d-1) e.g. Innes et al. (1987), or in terms of energy required (kJ.d-1) e.g. Kleiber (1961), Nagy (1987) (which provides energetic equations with biomass conversions), Sigurjonsson and Vikingsson (1997), Trites et al. (1997), Nagy et al. (1999), Ellis and Gabrielsen (2001) or Karpouzi et al. (2007).  In general, when such allometric equations are provided based on biomass, assumptions of the energetic content of the prey are made without consideration of the system. In those instances where estimates of consumption are based on energy, the conversion from units of energy to biomass required per day (as necessary for the model) are usually done based on using generic conversion factors (based on the diet preference of the particular group; see Nagy 1987, Nagy et al. 1999) and part of the value of considering the consumption rate in terms of energy is lost. The lack of consideration of the energetic content of prey is a criticism that has been leveled at ecosystem models that use biomass as their currency (Plaganyi and Butterworth 2004), and addressing such issues would improve the parameterisation of such models. Species specific bioenergetic models are a good starting point for examining how one might improve consumption rate estimates. In such models, detailed information on energetic requirements for the species concerned is provided, including differentiation of needs for different activities, as well as changes in dietary requirements depending on life stages and foraging patterns associated with such developments. These energetic requirements are matched with food requirements, which require information on diets, energetic density of the diet components as well as the digestive assimilation efficiency of the studied species. While it is not usually feasible to do such detailed compilations of data for each species in an ecosystem model, lessons that can be learned from species specific bioenergetic models can be relatively easily incorporated into parameters for ecosystem models. Estimates of energetic requirements from field metabolic rates in particular are widely used to estimate consumption rates for many of the birds and mammals in ecological studies and intrinsically incorporate in them the variety of activities that are undertaken. An important assumption in using such methods is that the daily energetic/food requirements are actually met. Keeping this assumption in mind, for the purposes of the ecosystem model, one can focus on the food requirements and improve the estimates by 142  incorporating local species specific information on diets along with specific energetic content of prey. This is routinely incorporated into single species bioenergetic models (see Murie and Lavigne 1991; Adams et al. 1993; Perez and McAlister 1993; Boyd 2002; Mecenero et al. 2006; Halsey et al. 2008) and has the potential to add value to consumption estimates used in ecosystem models.  For this study, it is proposed that the incorporation of local diet information, where it exists, in terms of composition, energetic content and assimilation efficiencies, should be used to improve the consumption rate estimates used in ecosystem models. In order to do this, a case study based on the data preparation for the ecosystem model of the Prince Edward Islands (PEIs) (Southern Ocean) is presented. A summary of commonly used empirical equations demonstrates the variation in consumption rates available to the ecosystem modeller for estimating consumption to biomass rates for mammals and birds (see Figure 7.1). For this study, the field metabolic rates of Nagy et al. (1999) (as amended by Ellis and Gabrielson 2001) for the bird species at the PEIs were used to demonstrate the difference that can be made by incorporating local diet related information (diets, energetic content of prey and assimilation efficiencies) into the consumption rate estimates and how the inclusion of such data can affect parameterization of the model. 7.2. Method Four breeding penguin species (Kings, Macaronis, Southern Rockhoppers and the Gentoos), five albatross species (Wandering, Grey-headed, Indian Yellow-nosed, Light- and Dark-mantled Sootys) as well as the Giant Petrels (Northern and Southern) and Sub-Antarctic Skua, 14 species of small seabirds (including prions, petrels and terns), the kelp gull and the Crozet Shag (Species listed in Table 7.1) are included in this study. Consumption rates for individuals in terms of their energetic requirements were based on field metabolic rates (FMR) for birds as summarised in Nagy et al. (1999, with amendments from Ellis and Gabrielsen 2001) and provided in kJ.d-1. A full description of the method used is provided in Chapter 3. Consumption rates for each species of bird at the islands are given, with estimates of daily food intake (DFI) and Q/B rate estimates that were calculated by simply 143  multiplying the DFI values by 365 and dividing by the average mass of the species concerned (Table 7.1). 7.3. Results Diets A summary of the diet composition of the land based top predator species/groups, divided into six prey categories, is provided (Table 7.2). Two of the penguins are principally fish eaters: the Kings, with a diet dominated by myctophids, and the Gentoos, with a mixed fish diet with some crustaceans, notably the benthic decapod. Both the Macaronis and the Rockhoppers have predominantly zooplankton diets. All albatross species have a mix of fish and squid. The Giant Petrels have a diet dominated by vertebrates (mainly penguins) while the Prions and Petrels diet is dominated by crustaceans with fish and cephalopods contributing.  Assimilation efficiencies and energetic content of prey A full description of the assimilation efficiencies and energetic content of prey is provided in Chapter 3. The relevant information required for this Chapter is provided in Table 7.2. Consumption rates Taking into consideration local diet preferences, prey energetic content and assimilation efficiencies (Table 7.2), field metabolic rate (FMR) estimates were used to establish the daily food intake (DFI; g.ind-1.d-1) and annual consumption to biomass rates (Q/B; .y-1) for all species (Table 7.1). Although the steps taken to incorporate the local diet are relatively simple, the results, as compared with estimates derived from standard processing, can be quite different. As shown in Figure 7.2, the incorporation of local diet information can alter the estimate of the Q/B rate by as much as 84% or -32% (significant in a paired 2 tailed t-test, p < 0.0002). To illustrate this for the PEI ecosystem, a closer examination of the penguin group and the Giant petrels is appropriate. The amendments in the consumption rates for these groups were instrumental in resolving the balancing of the PEI ecosystem model (see Chapter 5). As already outlined, the King, Gentoo, Macaroni and Rockhopper penguins all co-exist at the PEIs. These 144  species vary in mass (see Table 7.1) and in their feeding preferences (Table 7.2). The diets vary between locations and between seasons, but in general the King penguin is an offshore feeder travelling great distances from the islands on which it occurs and consuming primarily myctophid fish. The Gentoo penguin is, by contrast, a near-shore feeder, and consumes a mix of both inshore fish and zooplankton. The Macaroni, while consuming a limited amount of fish, feeds predominantly on zooplankton, which is the primary source of food for the smaller Rockhopper (Figure 7.3.A). Usually, consumption rates within groups (for birds, it is usually considered at the level of Order) are calculated using a single equation with body mass affecting the difference between groups. However the differences in diet preference and the difference this makes to the energetic content of the prey for each species can and should be taken into account when calculating consumption rates. If a diet is nutritionally more valuable, the consumer will require less mass of the diet to satisfy their energetic requirements and using a standard uniform value for all species may be considered inappropriate. A comparison of what the energetic content is of a single gram of each of the available prey groups (Figure 7.3.B) and a single gram of the diet for each of these species when the local diet information is incorporated into the assessment (with and without the inclusion of the assimilation efficiency) as compared to standard processing (where a default value of 16.2kJ.g-1 dry weight is used, Nagy et al. 1999) is illustrated (Figure 7.3.C). From this summary, for all species, with the exception of the King penguin (where the high energetic content of the myctophid fish, which is the dominant species in the diet, results in a greater average energetic value), the energetic density of the diets within the species is estimated as being less than that for standard processing. This therefore translates into the consumption rate estimated for three of the four penguin species as being higher than when using standard processing, but lower for the King penguin.  Regarding the Giant petrel consumption estimates, the adjustment made to account for the energetic density of the local diets of these species decreased the original Q/B value by one third (from 47.33 to 31.79yr-1 and 49.62 to 35.62 yr-1 for the Southern and Northern Giant petrels respectively). This is due to the relatively high energetic density of the Giant petrel diet, which is dominated by vertebrate prey. Until this amendment was made, there had been no satisfactory way found to resolve the energetic requirements of this group within the ecosystem model (see 145  Chapter 5) and the relative ease with which the model was able to be balanced following these adjustments was considered to be encouraging.  7.4. Discussion When tasked with developing an ecosystem model there are a plethora of parameters that need to be collected (for example, considering only the biomass, consumption, production and diet matrix in the PEI model there are 1517 parameters to be entered for the static model). In general, the focus at the local scale is on preparing site specific species lists and diet matrices, which leaves little time for dedicated work on those parameters that are considered transferable between systems (e.g. consumption and production rates). When searching for transferable parameters, such as consumption rates, in other ecosystem models, there can be a lack of transparency in the parameterization of the model. In some instances the data source is referenced but the method of calculation is not explicitly stated (e.g. Cornejo-Alonso and Antenzana 2008) or, quite commonly, another model is cross-referenced (e.g. Erfan and Pitcher 2005, Cheung and Pitcher 2005) and a search for the original source ensues. In searching for consumption values for this work, a good illustration of the types of issues that arise can be found in the summary for the consumption rate estimates for the King penguins (See Appendix 5.A. for full documentation). A search in the literature returned a number of studies for consumption rates for this species. Some were based on consumption in weight per day and require subsequent estimates on how many days the birds spend at sea. For example, at the Iles Crozet, Putz and Bost (1994) found an average daily intake of food of 2.3 kg.d-1 from an average of 132 food ingestion events per day over a period of 125 days, which would give a Q/B ratio of 65 yr-1 (using Abrams (1985) average weight of 13.0 kg). Bost et al. (1997) calculated the average per day ingested was 2.4 kg resulting in an annual Q/B of 67.3 yr-1 (using mass of 13kg, Abrams 1985) or 73 yr-1 (using mass estimate of 12 kg, Ryan and Bester 2008) if the birds feed each day. For the Kerguelen model, a Q/B value of 67.9 yr-1 (from total prey biomass of 1.99 t?km-2) was calculated (Pruvost et al. 2005), but was considered too high, as was the estimate of  38.0 yr-1 (based on 1985 data) from Cherel et al. (2005). The value of 12.0 yr-1 from the Weddell Sea model (Jarre-Teichmann et al., 1991) was used for the Kerguelen model (Pruvost et al. 2005), which was considered to be in line with other similar systems (e.g. 18 year-1 for the 146  Southern Plateau, New Zealand; Bradford-Grieve et al., 2003). This is in contrast to the Falklands model which has a value of 80 yr-1 for Q/B for all penguins (Cheung and Pitcher 2005). It was following such issues that the search for a better estimate of Q/B for the PEI model was commenced. Nagy et al.?s (1999) field metabolic rate equation for penguins (Order: Sphenisciformes equation C= 4.53M0.795 kJ.d-1) with the average bird weight of 12.0 kg (Ryan and Bester 2008), taking into account diet preferences (see diet section) and energy content of prey, digestion efficiency of 76%, water content of diet assumed to be 70%, returns a Q/B estimate of 42.5 yr-1. This compares well with calculations that result from Ellis and Gabrielson?s (2002) equations of Q/B = 40.9 yr-1 and falls mid way in the range of rates outlined above (from 12 to 80 yr-1). There are instances in ecosystem models where allometric equations with species specific information are used, both with (e.g., Cianelli et al. 2004) and without (Shannon et al. 2003; Coll et al. 2006; Coll et al. 2007; Piroddi 2008) explicit consideration of the potential difference that local energetic content of prey of the diet might make. The work presented here is intended to raise the awareness of the effect such considerations may have on the output of the ecological models and an assessment of whether this is important for each system will need to be made. There is a large variation that is found in the various empirical equations that are available for the estimation of consumption rates for mammals and birds, as illustrated for these groups at the PEIs (Figure 7.1). Results from ecosystem models will vary depending on the method that is used and may be important to consider in instances when a species plays an important role in the ecosystem in terms of its biomass contribution or key role in the system.  7.5. Conclusions This short communication provides an example of where the inclusion of local diet data can improve the assessment of consumption for the purposes of constructing a mass balanced ecosystem model. A simple sensitivity test (Chapter 5) of the PEI ecosystem model used in this case study highlighted the consumption to biomass rate estimates (Q/B, yr-1) to be as important as biomass estimates (B, t.km-2) and more important than production to biomass (P/B, yr-1) rate estimates in the model structure. Local diet assessments of prey and associated energetic density of the prey species can be important contributing factors when improving Q/B estimates. Global 147  datasets of species diet information are invaluable (e.g., Pauly et al. 1998 for mammals), and are heavily relied on for the construction of many ecosystem models. However, such datasets do not always provide information at the local scale and, while incorporation of the diet matrix in the consumption estimates improves the estimate and is appropriate for large scale studies (e.g., Karpouzi et al. 2007), further improvements can be made by using local diet information for evaluation of the energetic value of prey.  Using local diet information, the model parameters of the system were improved and the energetic requirements of the system were subsequently met with relatively minor adjustments. For future work, it is important to quantify the value and the impact such amendments make to the ecosystem model performance and to consider whether the adjustments are necessary in temporally and spatially dynamic models. This work demonstrates the issue of the energetic value of prey species can be incorporated into models that use biomass as their currency and can begin to address this issue that has been raised previously (Plaganyi and Butterworth 2004).   148  Table 7.1. Daily food ingestion rates (g.ind-1.d-1) and annual consumption to biomass rate estimates (yr-1) of the species of birds found breeding at the Prince Edward Islands using Field Metabolic Rates (FMR) as defined by Nagy et al. (1999) (with amendments by Ellis and Gabrielson 2001) calculated using average mass estimates from Ryan and Bester (2008).   149  Table 7.2. Diet matrices of the breeding bird species of the Prince Edward Islands summarised into six prey categories, with energetic density of prey categories given, and average energetic density of diet (kJ.g-1) and assimilation efficiency (as a fraction of the diet) for each bird species provided.  150   Figure 7.1 Illustration of the variation in the consumption to biomass rate estimates (yr-1) as calculated for selected A) mammals, B) large seabirds and C) small seabirds found at the Prince Edward Islands. 151    Figure 7.2. Graph showing the percentage difference between consumption to biomass rate estimates (Q/B, yr-1) calculated using Nagy et al.?s (1999) field metabolic rates with standard processing versus using local diet and energetic density of prey for all breeding bird species found at the Prince Edward Islands.   152    Figure 7.3. Illustration of the A) diets of the penguin species summarised into six prey categories (?fish general?, ?fish ? mesopelagic?, ?cephalopods?, ?crustaceans?, ?vertebrates? and ?other?), energetic density of the B) prey categories and C) penguin diets with and without assimilation efficiencies as compared to standard processing (shown in red).   153  Chapter 8  Population dynamics at the Prince Edward Islands:  Hindcasting of three known events  8.1. Introduction In the past, the communities at the Prince Edward Islands have experienced a number of changes which have been linked in some instances to known direct effects of human impacts. These include exploitation events at the islands such as the fur seal fishery (started in 1800s and terminated in the first half of the 19th century), which resulted in a depletion of the fur seal population from over 100 000 individuals to a few hundred. Another example is that of the Patagonian toothfish fishery (began in the 1990s and is ongoing today, 2013). Initial illegal, unregulated and unreported (IUU) fishing in the region resulted in a collapse of the fishery stock within a few years of the resource being identified. In addition, introductions of alien invasive species have had marked effects on the islands. The introduction of the domesticated cat (Felis catus) had devastating effects on the small bird populations, resulting in the local extinction of one species before the cats were finally eradicated (Bester et al. 2000).  Other changes that have been well documented in the scientific literature are less well understood and the effects of fisheries, interactions between species and environmental changes have all been considered to play a role. In an attempt to untangle the multitude of potential drivers in such complex biological systems, it can be useful to use an ecological model to explore the effects of known drivers, and through such analysis, better understand the system. Simple exploration can provide insight into the performance of the model (and assessment or reconsideration of parameterization) and insights into ecosystem effects as illustrated by the model. Following on from this, investigations into indirect effects and exploration of potential drivers can be carried out through temporal simulations. Ecosystem effects of fisheries on top predators and interactions between the top predators have received attention in the Southern Ocean and have been assessed in terms of their implications for fisheries management, e.g., Mackerel icefish, Champsocephalus gunnari, at South Georgia (Reid et al. 2005) and Patagonian toothfish at Heard (Green et al. 1998) and Macquarie (Goldsworthy et al. 2001) Islands. Concerns regarding such interactions (for instance the increasing fur seal populations and their 154  potential effect on seabirds) have been raised for the PEIs (Hofmeyr and Bester 1993; Guinard et al. 1998). These interactions can be explored through an ecosystem model, which includes consumption by competing fauna. Results from such work can inform management strategies and assist, for example, in setting fisheries quotas for the region. South Africa is a member of CCAMLR (Commission for the Conservation of Antarctic Marine Living Resources) and as such is committed to promote an ecosystem approach to fisheries (EAF) (Constable et al. 2000), yet no ecosystem model has been used for the management of the PEIs to date although the fishery is ongoing. The aim of this work was to create temporal simulations of the Prince Edward Island marine ecosystem to hindcast the ecosystem dynamics directly resulting from three examples of known human impacts: 1) the exploitation of the fur seals, 2) the Patagonian toothfish fishery and 3) the effects of the cat predation on the small flying bird population, and to assess the ecosystem effect in terms of changes of biomass for all functional groups for each scenario. 8.2. Method Dynamic temporal model (Ecosim) The dynamic simulation capability of the model (Ecosim) is described in Walters et al. (1997) and is based on the initial parameters that are defined in the mass balanced model (Ecopath). The basics of the temporal simulation consist of biomass dynamics expressed through a series of coupled differential equations. The equations are derived from the Ecopath production equation for the system: ?Bi?? = ?? ? ???? - ? ???? + Ii ? (M2i + M3i + Ei)Bi Where dBi/dt represents the growth rate during the time interval dt of group (i) in terms of its biomass, Bi, gi is the net growth efficiency (production/consumption ratio), Qji is the consumption of predator j on prey i, M2i the non-predation (?other?) mortality rate, M3i fishing mortality rate, Ei is the emigration rate, Ii is the immigration rate. The first summation estimates the total consumption by group (i), and the second the predation by all predators on the same 155  group (i). The consumption rates Qij, are calculated based on the ?foraging arena? concept where Bi?s are divided into vulnerable and invulnerable components and it is the transfer rate (vij) between these two components that determines if control is top-down (Lotka-Volterra), bottom-up (i.e. donor driven) or of an intermediate type (Christensen et al. 2008). Unless otherwise stated, the vulnerability settings were left as default values (2.0). The set of differential equations is solved in Ecosim using a Runge-Kutta 4th order routine (Walters et al. 1997; Walters et al. 2000, Christensen et al. 2008).  The temporal simulation allows for time series data to be imported into the system and the model may be fitted to this data. The process of improving model fit can be done through an automated routine, which minimizes the sums of squared residuals (SS) by adjusting the vulnerability settings (of particular groups as defined by the user) or through identifying forcing functions that may contribute to creating a better SS for a particular model.  Three scenarios were created to explore how well the model is able to replicate the observed time series data available from the fur seal exploitation (on both the Antarctic and Subantarctic fur seals), the Patagonian toothfish fishery (on the Patagonian toothfish with some associated fish bycatch also included) and the impact of the introduction of the feral cats on the small flying birds (the Prion and Petrel group) at the PEIs. 8.2.1. Fur seal exploitation The history of the fur seal industry was summarised by H?nel and Chown (1999). The earliest recorded sealing on the PEIs was in 1803 and by 1860 sealing was no longer economically viable. In 1909 an attempt to revive the industry was made without success and finally all sealing was stopped in the 1930s. In order to create a simulation of this exploitation, a fur seal fishery was created targeting both the Subantarctic and the Antarctic fur seals. Hunting commenced in the early 1800s, ceased by 1860 and recommenced for a short period in the first part of the 1900s. Data were created to drive the two fur seal populations through exploitation (Appendix 6.A). The Ecopath model from the 1960s time period (see Chapter 5) was used as the base from which to create a model to represent the ecosystem in the first decade of the 1800s. The only amendment made to this model was an adjustment of the fur seal biomass estimates to reflect the 156  most recent biomass estimates that are available from the islands. The estimates are from 2004 for the Subantarctic Fur Seals (Hofmeyr et al. 2006) and from 2008 for the Antarctic Fur Seals (Bester et al. 2009). Following this adjustment, the new 1800s Ecopath model needed to have additional adjustments to attain mass balance. The Ecotrophic Efficiencies (EEs) that were generated from the initial balancing attempt presented the Patagonian Toothfish, Myctophid Fish, Large Cephalopods and Small Cephalopods all to be over consumed with EEs ? 1 (values of 1.02, 1.04, 1.00 and 1.00 respectively). In order to balance the model adjustments were made to each of these groups. The Patagonian Toothfish biomass was adjusted with an increase of 2% (from 0.0900 to 0.0918 t.km2), an increase (of 18%) in Myctophid Fish (from 3.375 to 4.000 t.km2) with a decrease of 12.5% in Small Pelagic Fish from 1.125 to 1.000 t.km2, and small adjustments of less than 2% to both the cephalopod groups (increase in Large Cephalopods from 0.065 to 0.066 t.km2, and an increase in Small Cephalopods from 0.045 to 0.046 t.km2). With these adjustments the model balanced. Data to drive the fur seal exploitation was then imported into the model along with reference biomass estimates from survey data to compare the model outputs with time series from the islands. Following the output of the Ecosim run (FS S1), the automated fitting to time series capability of the EwE software was employed to refine the fit of the model to the data. The vulnerability settings for both the Subantarctic Fur Seals and Antarctic Fur Seals were selected to be adjusted in the search to reduce the sum of squares residuals (SS) to gain a better fit. 8.2.1. Patagonian toothfish fishery The fishery was initiated in the 1990s and is ongoing today. The legal fishery began in October 1996, though IUU fishing is known to have been in operation by 1995 and perhaps earlier (CCAMLR report 2009 Appendix Q). Estimates from a simple age-structured production model (ASPM) showed that the spawning biomass was depleted to, at most, a few percent of its pre-exploitation level during the early phase of this fishery (Brandao et al. 2002). Up to 7 vessels have been licensed to fish at the islands during any one year, however, since 2001 this has been reduced to only two operators (CCAMLR report 2009 Appendix Q). Annual fisheries catches, including IUU fishing, were estimated to be over 20 000 t in the first year of fishing, which rapidly declined to less than 1000 t by 2001 and declined to approximately 323 t by 2006 157  (Brandao and Butterworth 2009). In order to simulate this fishery, the 1980s Ecopath PEI EEZ model was used as a starting point (see Chapter 5). A time series file with the estimated catch data from 1997 to 2006 for the Patagonian toothfish fishery from Brandao and Butterworth (2009) was used to drive the model, along with catch data derived from the CCAMLR report series accounting for bycatch of other groups (Sharks and Rays, Large Demersal Fish, Large Notheniid Demersal Fish and Large Pelagic Fish, data provided in Appendix 7.A.). Two scenarios were run for the fishery. For the initial run (FPT S1), a dataset where no depredation due to cetaceans is considered was used (Brandao and Butterworth 2009, Appendix 1, z = 0). A relative abundance index for toothfish provided by the standardized commercial catch-per-unit-effort (CPUE) series for the PEIs EEZ for the longline fishery (Brandao and Butterworth 2009) was used as reference data to compare the output of the model.  The time series data (of catch for all groups and the CPUE data for the toothfish) were imported into EwE and the model was run.  In order to improve the model performance, the automated fitting to time series procedure was followed by conducting a vulnerability search for the Patagonian Toothfish functional group. Loss of toothfish to toothed cetaceans during landing procedures have suggested very high levels of depredation (up to two fish taken for each fish landed). Therefore, the procedure was repeated with data which included depredation to cetaceans according to the estimates from Brandao and Butterworth (2009) with z=1 (depredation starting in the year 2000, and increasing to a maximum of 1 fish per landed fish by 2002 and remaining at this level to 2006) and run as the second scenario (FPT S2). A vulnerability search was completed for the Patagonian Toothfish group.  8.2.1. Cat predation on small flying birds The house mouse (Mus musculus), which apparently reached Marion island from shipwrecks and sealers? boats in the 19th century (Siegfried 1978; Skinner et al. 1978; de Villiers and Cooper 2008), became a pest at the meteorological station and in 1949 five domestic cats (Felis catus) were taken to Marion to keep the mouse infestation at the base under control (Skinner et al. 1978). The cats soon became feral (Anderson and Condy 1974) and by 1974-1975 an estimated cat population of 2137 was found on Marion Island (Van Aarde 1979; Van Aarde 1980; Bester et al. 2000, est 2139). Crude estimates of 3.7 to 10.6 cats per km2 (Skinner et al. 1978) were made 158  with an estimated intrinsic rate of growth of 23.3% per year (Van Aarde 1978) and a total energy requirement of 9.97 x 108 kJ (Van Aarde 1977) or 10.2 x 108 kJ (van Aarde 1980). It was estimated that between 90 and 95% of the dietary requirements of the cats was satisfied by predation on small petrels. With the main energy content of bodies of petrels equal to 25.4kJ.g-1(dry weight) a minimum of 35 t (dry weight) of birds would have been required. The impact of the predation resulted in the Common Diving petrel Pelecanoides urinatrix, previously reported to breed on Marion Island, made locally extinct and other species (including the Great-winged petrel Pterodroma macroptera (Cooper and Fourie 1991) and the burrowing petrels (Cooper et al. 1995)) were heavily impacted. A control program using a viral disease, hunting, trapping and ultimately poisoning resulted in more than 3000 cats being killed and none being sighted since 1991 (Bester et al. 2000).  An Ecopath model that reflects the island system for the 1940s was generated from the 1960s Ecopath model (See chapter 5). No amendments were made to the functional groups of the model, though a fishery was added and named ?Cats on birds? to introduce the cat population to the model without having to include all biological aspects of this species as an additional functional group. Based on the historical data available, a time series was created to simulate the cat population on Marion Island. An exponential regression line between known data points from Bester et al. (2000) was used to generate a cat population estimate from 1949 to 1992 (Appendix 8.A.). Between 410 430 (van Aarde 1977) and 455 000 birds (Van Aarde 1980; Bester 2000) were estimated to have been consumed each year, which equates to between 192 and 213 birds per cat. Using the rate estimate of 201 birds per cat per year, the biomass of small Prions and Petrels (t) to be taken by cats was set based on the cat population and an average wet weight estimate of 0.324 kg per bird (used by Van Aarde 1977; Van Aarde 1980). These data were then imported into the 1940s Ecopath model and run in Ecosim with the cat predation treated as a ?fishery? removing biomass of the Prion and Petrel functional group. This simulation (CAT S1) resulted in extinction of the Prion and Petrel functional group and reductions in the predation by the cats were explored to find a predation rate from which the group would ultimately recover (partial recovery CAT S2; full recovery CAT S3). 159  8.3. Results 8.3.1. Fur seal exploitation Results from the automated fitting to time series procedure for the fur seal populations improved the sum of squared (SS) residuals from 10.84 at the start of the procedure to 3.87 for the fourth iteration. A figure illustrating the hunting mortality, the model output and the survey data for both species of fur seal in terms of biomass (t.km-2) is provided in Figure 8.1. A and B. The model captures the recent population growth, but does not manage to correctly simulate the rate of recovery and lags in its simulation of the current population biomass estimates.  The ecosystem effect of the fur seal hunting, as reflected in the changes in biomass (t.km-2) for all functional groups for the FS S1 scenario, is illustrated in Figure 8.2. Functional groups which were affected by a change of ?3% compared to their starting biomass at any time during the 211 year simulation totaled 20 (See Table 8.1).  As expected, both fur seal populations experienced the greatest changes, being brought close to extinction (-99.9% of biomass as compared to the initial 1800 biomass) and showing recovery in the final years of the simulation. The Orcas, Southern Elephant Seals, King Penguins and Gentoo Penguins all showed positive responses to the seal culling of between 4 and 14%, and with the recovery of the fur seal populations, these all subsequently declined. The Wandering Albatross showed an initial positive response followed by a decline but the overall effect was negligible in terms of total biomass. For all other albatross groups, increases in relative biomass were between 1 and 6% but all declined in the latter part of the 211 year simulation. Giant Petrels showed an immediate increase in biomass following the start of the seal hunting with a high of 12% increase from the initial starting value followed by a decline. Sharks and Rays showed an initial positive response (maximum increase from initial starting value of 6 %) followed by a decline to initial biomass estimates. This pattern was repeated for almost all fish groups (Large Demersal Fish, Large Nototheniid Demersal Fish, Small Continental Slope Demersal Fish, Small Inshore Demersal Fish, Large Pelagic Fish, Patagonian Toothfish and Myctophids) with initial positive responses of between 2% and 12% followed by declines, returning to starting biomass values. In contrast, the remaining fish group, the Small Pelagics, 160  showed the opposite pattern but of negligible magnitude (0%). The Large cephalopods group initially responded with a peak of 5 % but soon returned to within a percent of original biomass estimates. Responses of all other groups were less than ?3%. 8.3.2. Patagonian toothfish fishery Mortality estimates (from catch data) used to drive the model, excluding (FPT S1) and including (FPT S2) cetacean depredation, are shown along with the catch-per-unit-effort data used as an index of the population biomass in Figure 8.3.A. The simulation of the Patagonian toothfish fishery provided promising results, where the model was able to effectively simulate the fishery and provide a relatively good fit to the CPUE data (Figure 8.3.B) for both simulations. Following the automated fitting to time series for the Patagonian toothfish data, the sum of squared residuals (SS) value of 1.27 was improved to 0.57 and the vulnerability of the Patagonian Toothfish group was adjusted from the default value of 2.000 to 1.493 to provide the best result. Repeating the exercise with the inclusion of the cetacean depredation produced SS of 288.40, which was reduced to 0.81 by the 6th iteration and a vulnerability setting of 1.132. The ecosystem response to this driver, as seen in the changes in biomass in all functional groups is provided in Figure 8.4. Changes in biomass for all groups for both scenarios (FPT S1 and FPT S2) following the application of the Patagonian toothfish fishery as a driver were similar, differing only in the magnitude of change within a functional group of a few percent of the starting biomass depending on the scenario (with or without cetacean depredation) (Table 8.2). The figure shows the results from the first described scenario where no cetacean depredation is used in the driver. The Patagonian Toothfish group was brought to within -92 and -93% of its initial standing stock. Seventeen functional groups experienced a change in biomass greater or less than 3% as compared to their starting biomass due to the Patagonian toothfish fishery (Table 8.2). Most groups responded positively, including the Southern Elephant Seals (increases of between 10 and 14%), Macaroni Penguins, Gentoo Penguins (7 to 8%) and the Southern Rockhopper Penguins, but to a lesser extent (1-2%). The Light-mantled Sooty Albatross showed a decline, while all other albatross groups increased, with the Wandering Albatross having the greatest positive response (of 11-13% as compared to its initial starting biomass). Positive 161  responses were shown by Sharks and Rays (between 29 and 37%), Large Demersal Fish (14 to 17%) and Large Pelagic Fish (52 to 60%).  Small Continental Slope Demersal Fish showed an increase (13 to 15%), but the trend turned towards a decline by the end of the simulation. Small Inshore Demersal Fish initially increased, then declined, but the magnitude of the change was less than 5%. Increases were found for Large Cephalopods (27 to 32%), Benthos, Large Zooplankton Crustaceans and All Other Zooplankton. The Benthic Decapod had a varied response through the simulation. Fish groups that were adversely affected included, along with the already mentioned toothfish, the Large Demersal Nototheniid Fish, the Small Pelagic Fish, and the Myctophid Fish, the latter two showing a slight recovery towards the end of the simulation. The Orcas (-5 to 7%) and both Antarctic and Subantarctic Fur Seals were adversely affected (between 5 and 7%), and the King Penguins and Prions and Petrels both had initial increases followed by declines, but of insignificant magnitude. 8.3.3. Cat predation of small flying birds Following the initial simulation (CAT S1), the cat predation results in extinction of the Prion and Petrel group by 1975 (Figure 8.5). Only when the predation rate on the Prions and Petrels was reduced to 1/7th of its original estimate (30 birds per cat per year) (CAT S2) did the model show a recovery of this functional group following the cat predation (Figure 8.5). Further exploration of the model showed only at a rate of 1/10th of the original estimate (21 birds per cat per year) (CAT S3) would the Prion and Petrel population recover to its initial starting biomass estimate (equivalent to the 2000s estimate).  Ecosystem effects of this scenario on the biomass of all other groups are shown in Figures 8.6 (CAT S1) and 8.7 (CAT S3). The initial simulation which results in the Prion and Petrel extinction has, as would be expected, a larger impact on all the functional groups (as measured by the change in each functional group compared to its starting biomass) and the impact is in most cases sustained. Magnitudes of changes that exceed ?3% of the original starting biomass total 16 (Table 8.1) and include sustained positive responses for the Orcas (3%), Southern Elephant Seals (9%) and King Penguins (5%). The Macaroni Penguins and Rockhopper Penguins both showed positive responses. The Gentoo Penguins had an initial positive response followed by a decline, but not of any appreciable magnitude. The Giant Petrels showed a decline 162  (-8%), while all the albatross species showed positive gains between 6 and 11%. Almost all of the fish groups (Sharks and Rays, Large Demersal Fish, Large Nototheniid Demersal Fish, Small Continental Slope Demersal Fish, Small Inshore Demersal Fish, Large Pelagic Fish, Patagonian Toothfish and Myctophid Fish) show increases of up to 6% of initial biomass followed by declines of various severity, but none greater than -1%. The Small Pelagic Fish group shows the opposite pattern with an initial decline followed by a recovery, but the magnitude seems inconsequential (within 0.2 of a percent). For the cephalopod groups, the Large Cephalopods initially increase by approximately 8% as compared to their starting biomass, but subsequently decline to an intermediary level between their initial and peak biomass. The Small Cephalopods respond with an initial increase of as much as 10% and remain at this elevated level. The responses of all other functional groups were less than ?3%.  The pattern of the response for all groups is similar to that for the scenario where the predation rate is reduced to one seventh of the original estimate (CAT S2), but responses are slower to initialise and the magnitude of the response is reduced (no figure provided), though 14 groups still reach the criteria of a positive or negative response equal to or exceeding ?3%, including the Prions and Petrels which come to -91% of their starting biomass estimate (Table 8.1). A positive response was observed for the Southern Elephant Seals (7%), King penguins (4%) and all albatross species (between 4% and 7%). The large pelagic fish groups (Sharks and Rays, Large Pelagic Fish, Patagonian Toothfish) and both cephalopod groups also showed positive responses (of between 3 and 8%). The Giant Petrels had the greatest negative response (-7%), following the declining Prion and Petrel numbers. In the final simulation (CAT S3), the Prions and Petrels were reduced to half of their original population and the ecosystem effects were limited, with only 7 groups affected positively or negatively by the change and none more than ?3% of their starting biomass. 8.4. Discussion 8.4.1. Fur seal simulation While the hindcasting of the fur seal industry is able to capture some of the qualities of the recovery of the fur seal population, their rate of increase over the final 50 years of the 211 year 163  simulation is not accurate. The discrepancy between the survey data and the model predictions may be due to the failure to create an accurate reflection of the culling patterns, an underestimation of the production term for the fur seals, or possibly, the higher rate of population growth observed is due to migration to the islands from neighboring subantarctic islands. Further exploration of the driver and the response by the fur seal population should be explored to improve the model performance and consultation with mammal experts would be beneficial.  In terms of the ecosystem response, this simulation does generate some patterns in other functional groups that have been observed in the time series data available for the islands. The decline of the Southern Elephant Seals at the PEIs has been well documented, and the results of the simulation suggest that the recovery of the fur seals may impact the Southern Elephant Seals, though the magnitude of this effect in the simulation is limited (the initial increase was only 4% and the subsequent decline does not exceed this amount). It is also interesting to note the simulation shows a similar pattern for the King Penguins, which respond in a positive manner to the seal hunting (up to 8% increase) and follow with declines over the last 50 years of the simulation. A first glance at the Southern Rockhopper Penguin data shows a pattern that is clearly opposite to observations from the islands. The population at the PEIs has been in decline over the past 30 years, yet the simulation suggests that this group should respond in a positive way to the recovery of the fur seal population. The magnitude of these changes, however, is small and possibly inconsequential when other drivers in the system are considered. Other responses that are worth mentioning include what would be a positive response by the Gentoo Penguins to this driver (of 9%), and also a decline in the Patagonian Toothfish following the recovery of the fur seal population (but of only 5% in magnitude). 8.4.2. Patagonian toothfish fishery simulation The low vulnerability settings found during the vulnerability search following the fitting to time series procedure for the Patagonian Toothfish fishery simulation suggest that the Patagonian toothfish stock was close to its carrying capacity prior to the exploitation in the mid 1990s. Anecdotal evidence from over-wintering researchers suggests that this may indeed be the case. However, it is unlikely that the earlier illegal fishing activity in the 1970s and 1980s in other 164  areas of the sub-Antarctic (Goldsworthy et al. 2001) would not have affected the stocks in the vicinity of the PEIs.  In the proposed management procedure for the toothfish for the PEIs, Brandao and Butterworth (2009) put forward four different Operating Models (OMs) which reflect ?Optimistic?, ?Intermediate?, ?Less Pessimistic? and ?Pessimistic? status for the resource. From these models it is possible to compare the estimates for the pre-exploitation spawning biomass (Ksp) for each scenario with that estimated from the ecosystem model developed here. The ?Optimistic? scenario has a Ksp of 138 499t, ?Intermediate? is 88 205t, ?Less Pessimistic? is 45 703t and ?Pessimistic? is 29 723t. The Ksp that allows for the best fit of the model in this study is 32 326t without the cetacean depredation and up to 34 481t with cetacean depredation. This sets the estimates generated here between the ?Less Pessimistic? and ?Pessimistic? estimates of Brandao and Butterworth?s (2009) (See Figure 8.8) OMs and lends confidence that the model is at least generating results that are in line with other assessments. In terms of the ecosystem response of all functional groups, many responses shown are contrary to patterns observed at the islands. The results of the simulation suggest increases in the Southern Elephant Seals, the Macaroni Penguins and the Southern Rockhopper Penguins, and decreases in the fur seal populations and the King Penguins. Survey data from the islands for all these groups show opposite trends. The simulation, however, does show the magnitude of all these responses to be relatively minor and probably insignificant and other drivers in the system apparently overshadow the patterns observed here. In the case of the fur seals, for instance, the recovery from exploitation appears to be the over-riding driver responsible for the observed population patterns. Once the effect of this driver is no longer a consideration, other drivers may begin to play a role.  8.4.3. Cat predation on small flying birds  The initial simulation of the effect of the predation impact of the cat population on the Prion and Petrel population (CAT S1) suggests that the estimates of 290 birds per cat per year would be in excess of what would be sustainable for the PEIs, unless the Prion and Petrel population was up to 10 times greater than present day estimates. It must be stressed that the population estimates 165  for this group are not considered precise, but it seems unlikely that the population estimates would be out by such a large margin. The literature suggests that up to 95% of the cat diet would have been based on birds (Van Aarde 1980), but it seems more likely that mice would have made a portion of the diet, with the small bird populations contributing during the relevant breeding seasons. Seasonality of bird consumption was found by Van Aarde (1980), where the winter breeding Great-winged petrels were only taken from June to August, White-chinned petrel chicks in December, and Salvins prions not found in cat diets between March and May, but found throughout the rest of the year with a peak in August and September. Some birds were also found year round (including Soft-plumaged and Kerguelen terns) (Van Aarde 1980). This would affect the consumption rate of the bird populations, perhaps reducing it to a level that would be more sustainable. The breeding success of the Great-winged and Blue petrel was found to improve following the cat eradication (Cooper et al. 1995) and this is reflected in the results of the second and third simulation runs (CAT S2 and CAT S3) where the predation impact was reduced to 1/7th and 1/10th of the original suggested consumption rate respectively. In terms of ecosystem effects, the pattern observed is in general an increase in most groups as a response to the declining Prion and Petrel group. In the simulations where the Prion and Petrel group are not driven to extinction (CAT S2 and CAT S3), a decline in most populations is observed. The only exception is the Giant Petrels, which track the Prions and Petrels biomass trends, but lag in their response time. These responses are of appreciable magnitude for the first simulation (CAT S1) (as high as 10%), but are much reduced by the third simulation (maximum of 4%).  8.5. Conclusions The first criterion set for using ecosystem models for fisheries management is whether they can replicate historic trends in ecosystems (Christensen and Walters 2011) and, with the exploration of the temporally dynamic version of the PEI model, it was possible to hindcast past known events and to examine the ecosystem effects of the drivers. The findings presented demonstrate that the model in its current form is able to capture and replicate, to some degree, the patterns observed in the field data, particularly in the case of the fur seal populations and the Patagonian toothfish fishery for those groups that are directly affected by the driver. The drivers assessed 166  here do not, however, result in ecosystem effects that are found in the survey data from the islands and the conclusion must be that additional drivers need to be investigated in the search for model outcomes that reflect the observed trends. Further development of the temporal simulations where such drivers are combined and the ecosystem effects are assessed against the time series data that exist for all groups should be explored.    167  Table 8.1. Estimates of the maximum and minimum % difference as compared to the starting biomass (t.km-2) for each functional group in the PEI marine ecosystem model for each hindcasting scenario run.    168   Figure 8.1. Reconstruction of fur seal exploitation driven by hunting with model results compared to time series survey biomass data for the PEI EEZ from 1800 to 2010 for Antarctic Fur Seals A) and Subantarctic Fur Seals B).   169     Figure 8.2. Trends in biomass estimates (t.km-2) from 1800 to 2010 for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the fur seal industry.   Biomass (t.km-2) Time (years, 1800 - 2010) 170    Figure 8.3. Reconstruction of the Patagonian toothfish fishery from 1980 to 2006 presented with fishing mortality ( as catch data, t.km-2) used to drive the model (A) and model output showin in (B), with and without cetacean depredation as compared to time series relative abundance data (from catch-per-unit-effort (CPUE)) data provided by Brandao and Butterworth (2009) for the PEI EEZ.   171    Figure 8.4. Trends in biomass estimates (t.km-2) from 1980 to 2006 for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the Patagonian toothfish fishery.   Biomass (t.km-2) Time (years, 1980 - 2010) 172    Figure 8.5. Reconstruction of the cat predation on small flying birds (Prions and Petrels) from 1940 to 1990 with predation estimates (derived from cat population and consumption estimates from Bester et al. 2000) driving the model (with original estimates (CAT S1), 1/7th original estimates (CAT S2) and 1/10th original estimates (CAT S3)).   173    \   Figure 8.6. Trends in biomass estimates (t.km-2) for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the cat predation on the small flying birds (Prions and Petrels) from 1940 to 1990 based on original consumption estimates (CAT S1).   Biomass (t.km-2) Time (years, 1940 - 2010) 174      Figure 8.7. Trends in biomass estimates (t.km-2) for all functional groups of the Prince Edward Islands marine ecosystem following the system being driven by a reconstruction of the cat predation on the small flying birds (Prions and Petrels) from 1940 to 1990 based on estimates 1/10th of original predation estimates (CAT S3).   Biomass (t.km-2) Time (years, 1940 - 2010) 175    Figure 8.8. Spawning biomass estimates taken from Brandao and Butterworth?s (2009) Figure A.1. for the Patagonian toothfish at the Prince Edward Islands. Original caption from figure: Spawning biomass estimates (note that recruitment can vary prior to the onset of harvesting). Estimates are given for the Optimistic, Intermdeiate, Less Pessimistic and Pessimistic scenarios (details of the conditioning of these scenarios to the data are provided in the text; see also the caption to Table A.4). All results shown assume a cetacean depredation factor z=1, i.e. recent losses to cetacean depredation are equal to the landed longline catch.   176  Chapter 9 Population dynamics at the Prince Edward Islands:  Forecasting of climate change scenarios through forcing of primary producers 9.1. Introduction Climate change and its potential impacts present a challenge for biologists by providing an additional dimension that needs to be considered when interpreting ecosystem dynamics. Along with identifying natural patterns in ecosystems, the effects of climate change are superimposed on systems and these effects need to be teased apart. Investigating how organisms, ecological processes and whole ecosystems respond to a changing climate is a challenge and Subantarctic islands have been identified as having much to offer in furthering our understanding in this regard (H?nel and Chown 1998; Smith 2002). With the threat of climate change becoming more widely accepted, the subantarctic and Antarctic ecosystems have been identified as critical areas for global change research (Smith 2002). These ecosystems, in general, have impoverished biota and occur in harsh environments with relatively simple ecosystems that are sensitive to perturbations and therefore make ideal systems to study (Smith 2002). Ecosystems at higher latitudes are expected to show effects of global climate change relatively early and the Prince Edward Island (PEI) ecosystem therefore lends itself to investigations into exploring climate change scenarios. Long term changes in the populations of a number of the top predators in the subantarctic have been connected with climate warming (Weimerskirch et al. 2003). Interpretation of long term changes in populations will be confounded because of the variety of levels of interaction that may be considered. Understanding biological patterns and how they change across measurement scales is a fundamental conceptual problem in ecology (Levin 1992 in Francis et al. 1998), which holds true for climate variability and its analyses, which act across all temporal and spatial scales and result in difficulties in creating ecosystem models that can accurately predict responses to climate change (Francis et al. 1998; Trites et al. 2007). Observations of climate change at the PEIs have been made since the early 1990s (Smith and Steenkamp 1990; Chown and Smith 1993) with a warming and drying of the islands recorded from observations dating back to 1949 (Smith 1991; Smith 1992). Annual mean surface air 177  temperature at Marion island has increased by 1.2?C between 1969 and 1999 (Smith 2002) and sea-surface temperatures rose by 1.4?C at Marion from 4.5?C in 1949 to 5.9?C in 1998 (Melice et al. 2003). One of the potential effects of climate change is the change in position of the frontal systems. This theoretically could have at least two possible consequences. In the first instance, the position of the Sub-Antarctic Front (SAF) with respect to the islands is thought to determine the type of production that occurs at the islands and a southward shift in the SAF (already documented in Hunt et al. 2001) may have important implications for the PEI ecosystem, potentially affecting the productivity (Pakhomov and Froneman 1999a). A southward shift would increase the ?flow through? mode, bringing allochthonous production to the islands and benefiting the offshore feeders, but reducing the water retention mode, and potentially adversely affecting the benthic community and the inshore foragers (Pakhomov and Chown 2003) by reducing the occurrence of the island-associated blooms. The warming at the islands and the decline in rainfall (Smith 2002) would also contribute to a reduction in the river run off and potentially reduce the occurrence of the island associated blooms, which are thought to be linked to the water column stability that such run-off generates. An investigation into such a phenomenon and its effect on the PEI system would be beneficial for understanding future potential changes at the islands.  With this in mind, satellite remote sensing data, centered on the island area (Latitude 46.5 ? 47.1S; Longitude 37.4 ? 38.4E), were extracted on a monthly basis, initially using the Giovanni tool (NASA) for SeaWiFS data (http://disc.sci.gsfc.nasa.gov/giovanni), and finally reprocessed for this study by F. M?lin (of the Joint Research Centre, EC) for all available marine data (SeaWiFS, MODIS-Aqua, MODIS-terra) (See Figure 9.1). The figure provides the results of chl-a estimates derived from the satellite data. There is evidence of a reduction in the magnitude of the seasonal bloom between 2003 and 2009, though it remains to be known whether this reduction would persist into the future. Preliminary investigations (through collaborations with Sokolov and Rintoul, pers. comm.) into whether the patterns observed in the timing and magnitude of the bloom were linked with the position of the middle branch of the SAF as identified by Sokolov and Rintoul (2009), showed no relationship. Despite being unable to relate the occurrence and magnitude of the blooms to the position of the SAF, the data from the remote sensing does indicate a premise on which to hypothesize that the reduction in magnitude of the 178  spring bloom may be a realized phenomenon and exploration of such a hypothesis could provide useful insights into the system.  In terms of the open ocean productivity in the region, there are a number of scenarios which may be considered. Increased productivity associated with frontal features is well documented and many top predators (e.g. seals, penguins and albatross species) visit the frontal features to feed. The migration south of such features due to global climate change may have an effect on these populations. A southward shift of the Antarctic Polar Front (APF) due to warming may take this highly productive front out of foraging range of these species, effectively reducing the open ocean productivity associated with this feature. Relatively minor shifts in frontal features can have lasting impacts as many of the centrally placed foragers that are island based during their breeding cycle are restricted in the distance they can travel from shore when their offspring are young. Extension of trip duration to reach productive zones can increase the energetic costs of foraging and increase time away from the young, which can both negatively affect breeding success. The scenario of reduced reproductive success for the King penguins due to the southward migration of the APF (Kooyman 2002) was realized, as illustrated by Guinet et al. (1997) where the population at Crozet was unable to forage at the APF as it became out of range when it  moved further south with the season. This particularly affected breeding birds with newly hatched chicks that were limited in the amount of time that they could spend at sea (Guinet et al. 1997).  On the other hand, elevated productivity associated with eddies generated both upstream and downstream of the islands has been observed and could provide additional foraging areas for the top predators (Ansorge and Lutjeharms 2002; Ansorge and Lutjeharms 2003; Ansorge et al. 2004). An increase or decline in the rate at which such features are generated in the vicinity of the islands, the longevity and path that such eddies travel, and their proximity to the islands, would all potentially play a role in affecting the productivity in the region. Evidence that this particular region of the Southern Ocean is experiencing a decline in open water productivity has been shown by Vantrepotte and Melin (2011) who have identified it as a region which has experienced a significant declining trend over the decade between 1997 and 2007.  179  Finally, visual evidence from satellite imagery of the islands (Google Earth 2013) shows an extension of the kelp beds in the vicinity of the islands that exceeds the distribution of the kelp forests when compared to illustrated maps from the 1980s (Attword et al. 1991, See Figure 4.2). A change, whether positive or negative on the open ocean productivity would have an effect on the ecosystem.  Considering these various points raised here, the aim of this work was to produce forecasting scenarios of the marine ecosystem behaviour under the potential effects of climate change by driving the system through changes (both positive and negative) on the primary producers to investigate ecosystem effects of such drivers.  9.2. Method Dynamic temporal model (Ecosim) The model is described in the previous chapter, Chapter 8. Please refer to section 8.2. Forcing functions Two simple forcing functions (F1 and F2) were defined to drive the PEI ecosystem model that represents the 2000s time period. Both were defined as linear functions. F1 increased from zero to 1 over a 100 year period, while F2 decreased from 1 to zero over a 100 year period. The Ecopath model representing the 2000s time period was used as the starting point for the simulations. The simulations therefore began in 2000 and ended in 2099. These two functions were applied to act on each of the four primary producers (see Christensen et al. 2008, pg 45) for individual simulation runs which resulted in 8 different scenarios (S1-S8, see Table 9.1). Model outputs of biomass changes for each functional group were created and percent changes for each functional group as compared to their starting biomass for each scenario were summarised. 9.3. Results Plots showing the changes in biomass (t) for each functional group for each of the 8 climate change scenarios (S1-S8) are presented in Figures 9.2 to 9.9. The response of each functional 180  group in terms of its minimum and maximum change in the biomass for each group (as compared to its starting biomass) resulting from the forcing function are provided in Table 9.2. Scenario 1: Increasing Island Associated Blooms Using the F1 function to drive the system with a gradual increase in Island Associated Blooms (PIA), resulted in this production increasing by three fold by the end of the 100 year simulation. This increase had a limited effect on the ecosystem as a whole. The only other producer affected was the Macrophytes which declined by 3%. All other deleterious effects were less than 3%. The greatest effect overall was on the Benthos group which increased by 18 % from its starting biomass. Other groups that were affected more than 3% from their starting biomass were the Gentoo Penguins (4%), Sharks and Rays (8%); Large Demersal Fish (4%); Small Continental Slope Demersal Fish (6)%, Small Inshore Demersal Fish (6%). Almost all patterns were linear or unidirectional, except the Dark-mantled Sooty Albatross, which showed a sigmoidal pattern in changes of biomass estimates through time, the magnitude of which, however, was inconsequential. Scenario 2: Decreasing Island Associated Blooms The F2 function was used to drive the system with a decline in the Island Associated Blooms (PIA), resulting in effectively no biomass of this producer by 2060. A similar set of functional groups were adversely affected by this driver as were positively affected by F1 on the same producer. The Benthos functional group decreased the most (-9%), followed by the Sharks and Rays (-4.3%), the two small demersal fish groups (-3%) (Small Continental Slope Demersal Fish and Small Inshore Demersal Fish) and the Gentoo Penguins (-2%). Scenario 3: Increasing Open Ocean Large Phytoplankton Using the F1 function which generates a gradual increase in Open Ocean Large Phytoplankton (POL), by the end of the 100 year simulation this group had doubled. Overall, this created large, generally linear responses for almost all groups (31 out of the possible 34 living groups). Those that were negatively affected were the other primary producer groups: the small fraction of the open ocean production was reduced by 10% as compared to its starting biomass, the PIAs by 181  16% and the Macrophytes by 19%. Positive responses were of a high magnitude with maximum values reached by the end of the simulation. The Benthos group showed the greatest response to this input with an increase (as compared to its starting biomass) of 90%. Many groups associated with the benthic system showed strong positive responses, including the Sharks and Rays (69%), the Gentoo Penguins (56%), Benthic Decapod (49%) and the demersal fish groups (Large Demersal Fish (47%), Large Nototheniid Fish (22%), Small Continental Slope Demersal Fish (46%) and Small Inshore Demersal Fish (54%)). All the remaining large top predators also benefited in terms of increasing biomass from this production, with biomasses increasing between 20 and 45% (Orcas (44%), Southern Elephant Seals (37%), Antarctic Fur Seals and Subantarctic Fur Seals (27%), King Penguins (22%), Macaroni Penguins (36%), Rockhoppers (38%) and Giant Petrels (33%)). All albatross groups showed positive responses (Grey-headed Albatross (29%), Wandering Albatross (25%), Indian Yellow-nosed (19%), Light-mantled Sooty (15%), Dark-mantled Sooty (16%)) as did the smaller Prions and Petrels (19%). The pelagic fish (Large Pelagic Fish (40%), Patagonian Toothfish (23%), Small Pelagics (31%) and Myctophid Fish (24%)) all had positive responses as did all nekton and plankton groups (Large Cephalopods (32%), Large Zooplankton Crustaceans (39%), Small Zooplankton Crustaceans (5%), All Other Zooplankton (22%), Small Cephalopods (13%)). Scenario 4: Decreasing Open Ocean Large Phytoplankton Using the F2 function generated a gradual decrease in Open Ocean Large Phytoplankton (POL) and by the end of the 100 year simulation the POL group had decreased to zero. This had limited positive effects on the other primary producers with Open Ocean Small Phytoplankton (POS) increasing by 6%, PIAs 18% and the Macrophytes by 15%. Negative effects were seen throughout the system. Those groups that were affected most by the increasing POL were conversely affected by the greatest magnitude in this simulation, with Gentoo Penguins (-25%), Orcas (-24%), Large Demersal Fish (-23%) and Small Inshore Demersal Fish (-23%) all showing negative responses. Most functional groups (17) registered negative responses of between 10 and 20% by the end of the simulation (Table 9.2). Those least affected (>-10%) were the