ISSN 0429-9345

FAO
FISHERIES
TECHNICAL

477
PAPER

477

Models for an ecosystem

approach to fisheries

Models for an ecosystem approach to fisheries

This report reviews the methods available for assessing the impacts of interactions
between species and fisheries and their implications for marine fisheries management.
A brief description of the various modelling approaches currently in existence is provided,
highlighting in particular features of these models that have general relevance to the field
of ecosystem approach to fisheries (EAF). The report concentrates on the currently
available models representative of general types such as bionergetic models, predator-prey
models and minimally realistic models. Short descriptions are given of model parameters,
assumptions and data requirements. Some of the advantages, disadvantages and
limitations of each of the approaches in addressing questions pertaining to EAF are
discussed. The report concludes with some recommendations for moving forward in the
development of multispecies and ecosystem models and for the prudent use of the
currently available models as tools for provision of scientific information on fisheries in an
ecosystem context.

FAO
Cover:
Illustration by Elda Longo
Models for an ecosystem FAO
FISHERIES
TECHNICAL

approach to fisheries
PAPER

477

by
Éva E. Plagányi
University of Cape Town
South Africa

FOOD AND AGRICULTURE AND ORGANIZATION OF THE UNITED NATIONS

Rome, 2007
The designations employed and the presentation of material in this information
product do not imply the expression of any opinion whatsoever on the part
of the Food and Agriculture Organization of the United Nations concerning the
legal or development status of any country, territory, city or area or of its authorities,
or concerning the delimitation of its frontiers or boundaries. The mention of specific
companies or products of manufacturers, whether or not these have been patented, does
not imply that these have been endorsed or recommended by the Food and Agriculture
Organization of the United Nations in preference to others of a similar nature that are
not mentioned.

The views expressed in this publication are those of the author(s) and do not necessarily
reflect the views of the Food and Agriculture Organization of the United Nations.

ISBN 978-92-5-105734-6

All rights reserved. Reproduction and dissemination of material in this information

product for educational or other non-commercial purposes are authorized without
any prior written permission from the copyright holders provided the source is fully
acknowledged. Reproduction of material in this information product for resale or other
commercial purposes is prohibited without written permission of the copyright holders.
Applications for such permission should be addressed to:
Chief
Electronic Publishing Policy and Support Branch
Communication Division
FAO
Viale delle Terme di Caracalla, 00153 Rome, Italy
or by e-mail to:
[email protected]

© FAO 2007
 iii

Preparation of this document

The increased awareness of the importance of taking into account interactions among
fishery resources and the ecosystem in fisheries management has prompted the need
to improve the knowledge base on how ecosystems function including how they are
impacted by marine capture fisheries. Over time this has led to the development of
different approaches for the modelling of ecological interactions in marine ecosystems
exploited by fisheries. This paper reviews the models available for assessing the
impacts of ecological (indirect) direct interactions between species and fisheries and the
implications these have for fisheries management.
As this is a broad and rapidly-evolving issue, the report provides an overview of
the main types of modelling approaches rather than detail each aspect of the models.
Moreover, it includes a critical analysis of the advantages, disadvantages and limitations
of each modelling approach for representing ecosystem dynamics and interactions
between ecosystems and human activities, including in particular, fisheries. This report
is expected to serve as a useful reference for fisheries scientists and managers seeking an
overall view of the relative merits of the main types of modelling approaches available
for fisheries assessment in an ecosystem context.
The report was funded by the FAO project “Capacity Building for an Ecosystem
Approach to Fisheries” (GCP/INT/920/JPN).
iv

Abstract

This report reviews the methods available for assessing the impacts of interactions
between species and fisheries and their implications for marine fisheries management. A
brief description of the various modelling approaches currently in existence is provided,
highlighting in particular features of these models which have general relevance to the
field of the ecosystem approach to fisheries (EAF). The report concentrates on the
currently available models representative of general types such as bionergetic models,
predator-prey models and minimally realistic models. Short descriptions are given
of model parameters, assumptions and data requirements. Some of the advantages,
disadvantages and limitations of each of the approaches in addressing questions
pertaining to EAF are discussed. The report concludes with some recommendations
for moving forward in the development of multi-species and ecosystem models and for
the prudent use of the currently available models as tools for provision of scientific
information on fisheries in an ecosystem context.

Plagányi, É.E.
Models for an ecosystem approach to fisheries.
FAO Fisheries Technical Paper. No. 477. Rome, FAO. 2007. 108p.
 v

Contents

Preparation of this document iii

Abstract iv
List of tables and figures vii
Acknowledgements viii
Abbreviation and acronyms ix
Executive summary xi
1. Introduction 1
2. Review of current modelling approaches 3
2.1 Whole ecosystem and dynamic system models 10
2.1.1 ECOPATH with ECOSIM (EwE) 10
2.1.2 Biogeochemical models 14
2.1.3 ERSEM and SSEM 14
2.1.4 IGBEM, BM2 and ATLANTIS 15
2.1.5 SEPODYM/SEAPODYM 18
2.2 Minimum realistic models 20
2.2.1 The original MRM 21
2.2.2 ESAM (Extended Single-species Assessment Models) 24
2.2.3 MSVPA approach 25
2.2.4 MULTSPEC, BORMICON and GADGET 27
2.2.5 Multi-species statistical models 29
2.3 Individual-based models 30
2.3.1 OSMOSE 30
2.3.2 INVITRO 32
2.4 Bioenergetic models 33
2.5 CCAMLR model development 34
2.5.1 Predator-prey models 34
2.5.2 KPFM (Krill-Predator-Fishery Model) 35
2.5.3 EPOC model (Ecosystem Productivity Ocean Climate Model) 36
2.5.4 Mori and Butterworth multi-species model 36
2.5.5 SMOM (Spatial Multi-species Operating Model) 37
3. Comparison of models 39
3.1 Level of complexity and realism 39
3.2 Functional response formulations 39
3.3 Whole ecosystem models vs MRMs 42
3.4 Advantages, disadvantages and limitations 43
4. Potential of tools to address multi-species research questions 45
5. Roles for models in operational management procedure development 49
6. Moving models forward – future developments 51
7. Prudent use of the precautionary principle 53
8. Pointers from previous studies and workshops 55
8.1 Modelling interactions between marine mammals and fisheries 55
8.2 Areas of focus 57
8.3 General guidelines 57
8.4 Ecosystem-based management strategies 58
8.5 Practical steps to implementing an EAF 58
vi

9. Summary of model comparisons and recommendations 61

10. References 65
Appendix 81
Tables A 1a-d Model comparison 83
Tables A 2a-d Model comparison 95
Tables A 3a-d Summary of advantages, disadvantages and
limitations of each method 103
Tables A 4 Model comparison to address EBFM questions
107
 vii

List of tables and figures

Table 1 List of model acronyms 5

Table 2 Categorization of models according to

model units, feeding relationships assumed 7

Figure 1 Flowchart summarizing the classification of

various models 4

Figure 2 Schematic summary showing the trophic level

focus of different multi-species models 8

Figure 3 Schematic summarizing the typical (current)

number of modelled species or model
compartments for selected models 9

Figure 4 Schematic comparing consumption rate

formulations for two contrasted cases 40
viii

Acknowledgements

This study was funded by the Fisheries and Aquaculture Department of the Food
and Agriculture Organization of the United Nations (FAO), Rome, with logistical
support provided by the University of Cape Town, South Africa. This report benefited
enormously from frequent discussions and inputs from Prof. Doug Butterworth at
the University of Cape Town. Dr Beth Fulton is thanked for having been the most
thorough critic and contributor, and she also wrote the section on INVITRO.
The other reviewers, Dr Villy Christensen, Dr Mariano Koen-Alonso and
Prof. Gunnar Stefannson and the editors, Drs Kevern Cochrane and Marcelo
Vasconcellos, also provided useful suggestions and additions leading to a greatly
improved manuscript. I am grateful for the assistance of Drs Kerim Aydin, Andrew
Constable, Patrick Lehodey, John Tschirhart and George Watters, who provided me
with necessary material and answers to queries. Sections of this manuscript are drawn
from the author’s PhD thesis and support from the Marine Resource Assessment and
Management Group (MARAM) within the Department of Mathematics and Applied
Mathematics at UCT is gratefully acknowledged.
 ix

Abbreviations and acronyms

(See also the list of model acronyms in Table 1.)

ABM Agent-Based Models

ADMB AD Model Builder
AGGMULT Simplified version of MULTSPEC with only the age structure
retained
ASPM Age-Structured Production Model
BENEFIT Benguela Environment Fisheries Interaction and Training
Programme
BM2 Bay Model 2
BORMICON BOReal MIgration and CONsumption model
CCAMLR Commission for the Conservation of Antarctic Marine Living
Resources
CITES Convention on International Trade in Endangered Species
CPUE Catch per Unit Effort
DEAT Department of Environmental Affairs and Tourism (South Africa)
EAF Ecosystem Approach to Fisheries
EBFM Ecosystem-based Fisheries Management
ECONMULT Model for describing the economy of the Barents Sea fisheries
under different harvesting control rules
ENSO El Niño Southern Oscillation
EPOC Ecosystem Productivity Ocean Climate Model
ERSEM European Regional Seas Ecosystem Model
ESAM Extended Single-species Assessment Model
ESD Ecologically Sustainable Development
EwE ECOPATH with ECOSIM
FAO Food and Agriculture Organization of the United Nations
GADGET Globally applicable Area-Disaggregated General Ecosystem
Toolbox
GEEM General Equilibrium Ecosystem Model
GIS Geographical Information System
GLM Generalised Linear Model
GOTM General Ocean Turbulence Model
IBM Individual-Based Models
ICES International Council for the Exploration of the Sea
IGBEM Integrated Generic Bay Ecosystem Model
IMR Institute of Marine Research, Bergen, Norway
IWC International Whaling Commission
KPFM Krill-Predator-Fisheries Model
LME Large Marine Ecosystem
MCM Marine and Coastal Management, South Africa
MOOVES Marine Object Oriented Virtual Ecosystem Simulator
MP Management Procedure (analogous to OMP and MSE)
MPA Marine Protected Area
MRM Minimally Realistic Model
MSE Management Strategy Evaluation (analogous to MP and OMP)
x

MSFOR Multi-species Forecasting Model

MSM Multi-species Statistical Model
MSVPA Multi-species Virtual Population Analysis
MSY Maximum Sustainable Yield
MULTSPEC Multi-species model for the Barents Sea
NAMMCO North Atlantic Marine Mammal Commission
NMFS National Marine Fisheries Service
OMP Operational Management Procedure (analogous to MP and MSE)
OSMOSE Object-oriented Simulator of Marine ecOSystems Exploitation
P Production
P/B Production:Biomass ratio
PPBIM Port Philip Bay Integrated Model
PVM Parallel Virtual Machine
Q/B Consumption per unit biomass (or per capita biomass)
R A language and environment for statistical computing (R
Development Core Team. 2003)
SAM Single-species Assessment Model
SKEBUB SKEleton BUlk Biomass ecosystem model
SEAPODYM Spatial Ecosystem And Population Dynamics Model
SEASTAR Stock Estimation with Adjustable Survey observation model and
TAg-Return data
SIR Sampling-Importance-Resampling algorithm
SMOM Spatial Multi-species Operating Model
SPM Sequential Population Analysis
SSEM Shallow Seas Ecological Model
SSMU Small-Scale Management Units
SystMod System Model for the Norwegian and Barents Sea
TAC Total Allowable Catch
VPA Virtual Population Analysis
WSSD World Summit on Sustainable Development
 xi

Executive summary

This report reviews the methods available for assessing the impacts of interactions
between species and fisheries and their implications for marine fisheries management.
The focus is on modelling methods and multi-species population dynamics effects,
rather than on the full range of ecosystem aspects of fishing which encompass, for
example, environmental effects and technical interactions (e.g. bycatch issues), although
minor mention of these is made.
The first section takes a broad overview of some of the most commonly applied
multi-species/ecosystem approaches to fisheries management. The next section
summarizes the results and conclusions reached by previous studies and workshops
on the subject, including the ICES/SCOR Symposium on Ecosystem Effects of
Fishing, the Workshop on the Use of Ecosystem Models to Investigate Multi-species
Management Strategies for Capture Fisheries, the International Whaling Commission
(IWC) Modelling Workshop on Cetacean-Fishery Competition, the North Atlantic
Marine Mammal Commission (NAMMCO) workshops and the Workshop on
Ecosystem Approaches to Fisheries in the southern Benguela.
A brief description of the various modelling approaches currently in existence is
provided, highlighting particular features of these models which have general relevance
to the field of the ecosystem approach to fisheries (EAF). Models discussed include:
whole ecosystem/dynamic system models, minimum realistic models, individual-based
models and bioenergetic models.
These models are compared in a series of tables and figures, using the following
criteria:
1. the level of complexity and realism, e.g. the number of modelled species, the
representation of size/age structure of the species, and the types of processes
represented (physical and biological);
2. the types of functional responses of predators to changes in abundance of prey
species and their consequences and limitations;
3. how uncertainties in model structure, parameters and data are treated;
4. how environmental effects and interactions with non-target species (e.g. marine
mammals; sea turtles; sea birds) are incorporated;
5. the spatial representation of species interactions and habitat related processes;
6. model suitability for dealing with migratory species, i.e. species that cross
ecosystem boundaries;
7. where possible, model adequacy to allow the analysis of the different types of
management controls in use, such as effort control, minimum size, total allowable
catch, protected areas and closed seasons;
8. model adequacy to allow the assessment of the effects of short, medium and long-
term ecosystem changes;
9. model suitability to conduct assessment and policy exploration, considering the
model’s potential use to conduct historical reconstruction of resources to describe
the current status of the ecosystem and to evaluate the potential effects of various
kinds of decisions (short and long term);
10. model transparency of operation and ease of use; and
11. data requirements and model suitability for data poor areas.
A description is also given of model parameters, some important assumptions, data
requirements, technical information such as the computing platform, a list of examples
where the approach has been used, notes on the model history as well as any additional
xii

useful features of an approach. Some advantages, disadvantages and limitations of each

of the 20 approaches are listed, together with notes on the ease of presentation of model
outputs and the user-level of programming and mathematical skills required.
The most widely used approach is undoubtedly ECOPATH with ECOSIM (EwE),
which is likely to remain a forerunner given the user friendly interface and on-going
improvements to the software. However, faced with incomplete knowledge of ecosystem
functioning, there has been increasing recognition that definitive conclusions cannot be
drawn from a single model structure. There has thus been a parallel increase in efforts
to modularize models so that different components can be easily substituted. Spatial
considerations are similarly playing an increasingly important role in the development
of ecosystem modelling approaches. Nonetheless, even some of the earliest approaches
such as Multi-species Virtual Population Analysis (MSVPA) are still being used and
improved. A summary is presented of some recent advances being planned for the
different modelling approaches.
A set of commonly asked questions pertaining to EAF is identified and the potential
of the various modelling approaches to address these questions is assessed. This
preliminary analysis suggests that a range of different model constructions are needed;
no one model is necessarily superior to all others in all respects. EwE is capable of
addressing the widest range of topical EAF research questions. The model considered to
show the greatest potential to contribute to practical fisheries management advice (such
as changes to total allowable catch (TAC)) is Globally applicable Area Disaggregated
General Ecosystem Toolbox (GADGET). Although still under development, this is
currently the model with the most rigorous statistical framework for testing multi-
species based management advice. It is also the modelling approach most capable of
detailed sensitivity investigations to alternative growth, consumption and recruitment
formulations. Additionally, it operates within a spatial framework and overcomes
many of the associated computing constraints by running on multiple computers in
parallel. Nonetheless, it too has limitations in that it is capable of representing only a
relatively small component of the ecosystem and is not suitable for all systems. Models
such as EwE and ATLANTIS are more appropriate for addressing broader questions.
The incorporation of ecosystem considerations into current Operational Management
Procedures (OMPs) and other management strategies for marine resources is also
discussed. ATLANTIS is ranked the best operating model within a simulation testing
framework. Unfortunately it seems unlikely that sufficient data will be available to
implement an ecosystem operating model framework in most marine systems. Further
development is encouraged of approaches that take explicit account of uncertainty
and management issues, for example, through the use of a simulation framework
incorporating feedback control rules used in actual management.
Approaches such as the Extended Single-Species Assessment Models (ESAM)
are often a good first step. Similarly, examples are given of equations that provide a
useful starting template for multi-species modelling approaches, being built up slowly
and in synchrony with data availability. Some of the less well-known (in a global
context) modelling approaches are shown to include some additional useful features,
for example, SEAPODYM’s (Spatial Environmental POpulation DYnamics Model)
habitat index and OSMOSE’s (Object-oriented Simulator of Marine ecOSystem
Exploitation) explorations with simple individual predation rules.
This report is a first step towards initiating more detailed discussions of these models,
their uses and their limitations. This process is considered critical in moving forward
the development of methods for assessing indirect ecosystem impacts of fisheries.
Arguments are presented that whereas a good range of models has been developed
for the task of EAF, greater focus is needed on strengthening these approaches and
conducting the necessary data collection and experimentation to underpin confidence
in these approaches. Would-be model developers are encouraged to assess whether
 xiii

they would be adding anything to the current suite of models, given that approaches
such as EwE and GADGET have benefited from an extensive network of collaborators
over a number of years.
Considerable scope exists for significant future developments in multi-species and
ecosystem models, particularly with respect to their use as tools in EAF. Some of the
major areas of current research include:
• investigations pertaining to the effects of model complexity – in particular, the
effect of specific formulations (often feeding functional responses) on model
outputs;
• the treatment of uncertainty;
• representation of socio-economic factors and human behavioural drivers;
• multiple sector dynamics and management (with OMPs being an increasingly
popular method); and
• the effective (and feasible) representation of biodiversity.
 1

1. Introduction

The 21st century has ushered in a new era in fisheries management in which the prevalent
terminology is the ecosystem approach to fisheries (EAF; Garcia et al., 2003) in contrast
to more “dated” terms such as surplus production and single-species models. This is at
least in part attributable to the increasing pressure exerted on species subject to fishing
(and interconnected species in the ecosystem) and a growing realization of the need
to consider broader socioeconomic effects as well as the ecosystem effects of fishing.
Although computational restraints are much less of a problem due to improvements in
modern computing power, progress in this field is still (and may always be!) impeded
by imprecise parameter estimation given limited and noisy data and the associated
limited understanding of ecosystem functioning.
Nonetheless, as powerful new tools such as ECOPATH with ECOSIM (EwE)
(Polovina, 1984; Christensen and Pauly, 1992; Walters, Christensen and Pauly, 1997;
Walters et al., 2000) are further developed and distributed, there is a growing body of
scientists being drawn to this challenging new field. In practice, single species models
are still the dominant tool worldwide for providing timeous and reliable scientific
advice regarding the management of commercially valuable stocks. As single-species
and EAF approaches become increasingly merged in the development of management
advice, it is important that modellers have a good understanding of both single-
species and ecosystem approaches. Multi-species considerations are yet to be formally
included in the stock assessment approaches for the major fisheries resources globally.
However, considerable work has been conducted worldwide to construct multi-species
models and, more recently, in implementing EwE (Walters, Christensen and Pauly,
1997), which is currently the most widely utilized approach worldwide.
The aim of this report is to review the methods available for assessing the impacts of
interactions between species and fisheries, in particular ecological (indirect) interactions
and their implications for fisheries management. A wide variety of different methods
are at hand to address this issue (e.g. Pope et al., 1988; Larska and Wootton, 1998;
Boyd and Murray, 2001; Eisenack and Kropp, 2001; Kaschner et al., 2001; Crawford,
2004; Dalton, 2004; Drapeau et al., 2004; Yemane, Field and Griffiths, 2004; Daan et al.,
2005), but the focus here is specifically restricted to modelling methods. Given that
this is a large topic on its own, the field of ecosystem indicators (e.g. Rice, 2000) is not
discussed and the reader is referred to the International Council for the Exploration
of the Sea (ICES) Journal of Marine Science vol. 62, 2005 for a recent review of this
topic. The scope of this report is on multi-species population dynamics effects, rather
than on the full range of ecosystem aspects of fishing encompassing, for example,
environmental effects and technical interactions (e.g. bycatch issues), although minor
mention of these is made. Although some of the discussions are relevant to freshwater
or estuarine fisheries, this report focuses only on marine fisheries. The potential of
approaches to contribute broadly to fisheries management is discussed as well as their
more specific potential to contribute to practical advice. To achieve the latter, a multi-
species modelling approach should provide at least qualitative and ideally defensible
quantitative guidance as to the management of marine natural resources. One of the
most obvious uses relates to modifications in annual allowable catch levels deemed
necessary because of the predicted effects that fishing on a target species will have on
other components of the ecosystem (Plagányi and Butterworth, 2004), but ultimately
these tools may be called upon to give advice on all potential management levels
(including spatial management, temporal closures, gear restrictions and discarding
2 Models for an ecosystem approach to fisheries

practices).
The first part of this review takes a broad overview of some of the most commonly
applied multi-species/ecosystem approaches to fisheries management. The next section
summarizes the results and conclusions reached by previous studies and workshops on
the subject, including the ICES/SCOR Symposium on Ecosystem Effects of Fishing
(ICES Journal of Marine Science 57, n.3, June 2000), the Workshop on the Use of
Ecosystem Models to Investigate Multi-species Management Strategies for Capture
Fisheries (Pitcher and Cochrane, 2002), the IWC Modelling Workshop on Cetacean-
Fishery Competition (Journal of Cetacean Res. Manage. 6 (Suppl.) 2004) and the
Workshop on Ecosystem Approaches to Fisheries in the southern Benguela (African
Journal of Marine Science 26, 2004).
The need for an EAF is well recognized and indeed mandated. However, there
is still a need for, on the one hand, many ecosystem modellers to better acquaint
themselves with the practical realities of providing reliable management advice and, on
the other hand, for single-species modellers to step back from the often frantic process
of conducting stock assessments and use their expertise to guide the development and
implementation of multi-species management tools. Given the potentially large scope
of this study, the focus has been restricted to the most widely-applied or well-known
approaches as well as those considered by the author to show promise in advancing
this field. This manuscript is not intended as a final authoritative view to compare the
different modelling approaches but is rather a working document to assist and direct
further discussion of the various modelling approaches.
The choice of an appropriate model depends not only on the question to be addressed
but also on other logistical constraints such as the person power and associated
costs. The various modelling approaches discussed will roughly be compared giving
consideration to the above.
 3

2. Review of current modelling

approaches

An overview is given below of some of the current approaches to modelling multi-

species/ecosystem effects in the context of their possible application to fisheries
management. This review is by no means exhaustive but has attempted to capture
broadly the main model types that are either well known and widely available and
show potential as a tool in this context. The aim here was thus not to exactly describe
every multi-species/ecosystem model developed – models such as that by May et
al. (1979), Beddington and May (1982), Skeleton bulk biomass ecosystem model
(SKEBUB) (Bax, 1985) and Pech et al. (2001) were not deemed to meet these criteria
but future revisions will take into account approaches that are sufficiently strongly
supported. Moreover, the purpose of comparing the models is to assist in greater
understanding of the models available and in making informed decisions in instances
where resources are limited and hence it is important to select the best possible model
upfront. It is acknowledged that the choice of method depends on the question and
research objectives and that the ideal (if not always practical) scenario is one in which
a suite of models is developed and compared (Fulton, Smith and Johnson, 2003a).
Moreover, in ideal circumstances the suite of models will be drawn from a wide range
of types, as the model structure (and even its development history) can have significant
implications for the potential range of dynamics displayed (Fulton and Smith, 2004).
Plagányi and Butterworth (2004) outline an increasing hierarchy of multi-species
model complexity to account for biological interactions that pertain to commercially
important species. It is important to appreciate that increasing model complexity to
take better account of biological realism which can lead to an associated increase in
scientific uncertainty, as a result both of lack of knowledge of functional relationships
and of imprecision in estimates of the associated parameter values. The reader is
referred to other texts (e.g. Fulton, Smith and Johnson, 2003a; Raick, Soetaert and
Grégoire, 2006) for further discussions dealing with the important issue of model
complexity. The reader is also referred to the excellent text of Walters and Martell
(2004) for an overview of food web modelling, parameterization of ecosystem models
and strategies for ecosystem management.
The simplest multi-species models explore the question of how to harvest a
target population appropriately, whilst simultaneously accounting for the needs of a
predator dependent on that population as prey. If both predator and prey are subject
to exploitation, it is necessary to simultaneously model both predator and prey
populations as functions of physical variability, catch levels and the strength and nature
of the functional relationship between the two populations. If an intermediate trophic
level species is targeted (in a “wasp-waist” system, see Cury et al., 2000 in particular),
it may be necessary to account for the functional relationships between the targeted
species and its key predators, competitors and prey items. In this case appropriate catch
levels are likely to be affected by variability in both upper and lower trophic levels. The
most complex multi-species models strive to suggest modifications in the catch level of
a species based on the direct and indirect predation and competition effects associated
with the simultaneous removal of other food web components. In addition, it may be
necessary to consider negative feedback loops such as cannibalism. Other factors such
as human and fleet dynamics may also play a role at various levels, but consideration
of these factors was considered beyond the scope of this report.
4 Models for an ecosystem approach to fisheries

Models and their categorization

The different models discussed can broadly be categorized according to the framework
presented in Hollowed et al. (2000) which has been slightly modified and updated as
shown in Figure 1. Models which represent only a subset of the ecosystem are termed
Minimally Realistic Models and typically focus on inter-species interactions only and
hence may also be termed Dynamic multi-species models. They may however also
include some consideration of physical and environmental forcing actors. In contrast,
Dynamic system models incorporate the environment and lower trophic levels,
although this is often at the expense of not representing the higher trophic levels in
sufficient detail (when considered in a fisheries management context). In classifying
models further, it is important to differentiate between models that take age structure
and spatial aspects into account (Figure 1). Finally, the term Whole ecosystem models
is reserved for models that attempt to represent all trophic levels in an ecosystem
in a balanced way. Note further that Figure 1 is necessarily simplistic as it does not
reflect other important details relevant to the organization and regulation of ecological
systems (M. Koen-Alonso, pers. comm.) – for example, modelling predation as size-
dependent produces different results to models assuming age-dependent predation
(de Roos, Persson and McCauley, 2003; de Roos and Persson, 2005).
This review focuses on the following types of models (Figure 1, Table 1):
- Whole ecosystem models: models that attempt to take into account all trophic
levels in the ecosystem, including ECOPATH (Polovina, 1984; Christensen and
Pauly, 1992), ECOSIM (Walters, Christensen and Pauly, 1997) and ECOSPACE
(Walters et al., 2000) and other bioenergetic trophodynamics models (e.g. Yodzis,
1998; Koen-Alonso and Yodzis, 2005);

FIGURE 1
A flowchart summarizing the classification of the various models listed in Table 1.
The flowchart has been modified and updated from that presented in Hollowed et al.
(2000). Boxes with models covered in this report are highlighted

Biological No Technical interaction

interactions models MSYPR
described Murawski 1984

Yes
No
Predators added to single-species
Predator prey
models e.g. SEASTAR
feedback
Gulland 1983; Livingston and Methot
1998; Hollowed et al. 2000; Plagányi
Yes 2004; Tjelmeland and Lindstrøm 2005

No Handles the Yes

environment
and lower
trophic levels

Handles Yes
age/size
No structure
Handles age
structure Yes
No

Handles
spatial Handles Yes
Multispecies Production structure No spatial
No
Models e.g. Horbowy 2005 structure
Yes

Spatial dynamic systems

Aggregate system
Dynamic multi-species models models e.g. ATLANTIS,
models e.g. EwE,
BORMICON, GADGET, MRMs, ERSEM, SEAPODYM
SKEBUB, SSEM
MSVPA& MSFOR, MSM,
MULTSPEC, OSMOSE
Dynamic systems models
e.g. some recent EwE
applications
Spatial aggregate
systems models e.g.
ECOSPACE
Review of current modelling approaches 5

TABLE 1
Alphabetical list of model acronyms, full names and references to primary developers/users
Model Name References

ATLANTIS ATLANTIS Fulton, Smith and Johnson, 2004; Fulton, Smith

and Punt, 2004; Fulton, Smith and Punt, 2005
Bioenergetic/ Multi-species trophodynamic model using Yodzis and Innes, 1992; Yodzis, 1998;
allometric model bioenergetic and allometric approach Koen-Alonso and Yodzis 2005

BORMICON BOReal Migration and CONsumption model Bogstad, Hauge and Ulltang, 1997;
Stefansson and Palsson 1998
CCAMLR models Commission for the Conservation of Antarctic Butterworth and Thomson 1995; Thomson et al.,
Marine Living Resources 2000; Mori and Butterworth 2004, 2005, 2006
EPOC Ecosystem Productivity Ocean Climate model Constable 2005, 2006
ERSEM II European Regional Seas Ecosystem Model Baretta, Baretta-Bekker and Ruardij, 1996;
Baretta-Bekker and Baretta, 1997~;
Download from http://www.ifm.uni-hamburg.
de/~wwwem/dow/ERSEM/
ESAM Extended Single-species Assessment Models - Livingston and Methot, 1998; Hollowed et al.,
Models that are extensions to more conventional 2000; Plaganyi, 2004; Tjelmeland and
single-species stock assessment models Lindstrøm, 2005

EwE ECOPATH with ECOSIM Polovina, 1984; Christensen and Pauly, 1992;
Walters, Christensen and Pauly, 1997; Walters et
al., 2000; Christensen and Walters, 2000, 2004;
Christensen, Walters and Pauly, 2000;
Website: www.ecopath.org

GADGET Globally applicable Area Disaggregated General Trenkel, Pinnegar and Tidd, 2004; Begley and
Ecosystem Toolbox (GADGET); old name was Howell, 2004; Taylor et al., 2004; Taylor and
BORMICON (BOReal Migration and CONsumption Stefansson, 2004; Begley, 2005.
model); Fleksibest is a variant of Gadget. Website: www.hafro.is/gadget,

GEEM General Equilibrium Ecosystem Model Tschirhart and Finnoff, 2003; Tschirhart, 2004;
Eichner and Tschirhart (in press)
IBM Individual-Based Models (e.g. OSMOSE) DeAngelis and Gross, 1992; Shin and Cury, 2001;
Ginot, LePage and Souissi, 2002; Ginot et al., 2006;
Alonzo, Switzer and Mangel, 2003; Colomb et al.,
2004; Kirby et al., 2004
IGBEM Integrated Generic Bay Ecosystem Model Fulton, 2001; Fulton, Smith and Johnson, 2004
INVITRO INVITRO Gray et al., 2004; Gray et al., 2006
KPFM Krill-Predator-Fishery Model (KPFM, also KPFM2) Watters et al., 2005, 2006
MRM Minimally Realistic Model E.g. Punt and Butterworth, 1995
MSM Multi-species Statistical Model Jurado-Molina, Livingston and Ianelli, 2005;
Jurado-Molina, Livingston and Gallucci, 2005
MSVPA and MSFOR Multi-species Virtual Population Analysis and Helgason and Gislason, 1979; Pope, 1979, 1991;
Multi-species Forecasting Model Sparre, 1991; Magnússon, 1995; Vinther, 2001

MULTSPEC Multi-species model for the Barents Sea; simplified Bogstad, Hauge and Ulltang, 1997; Tjelmeland
version is AGGMULT which is also connected to a and Bogstad, 1998
ECONMULT - a model describing the economies of
the fishing fleet

MOOVES Marine Object-Oriented Virtual Ecosystem Colomb et al., 2004

Simulator
OSMOSE Object-oriented Simulator of Marine ecOSystem Shin and Cury, 2001, 2004
Exploitation

SEAPODYM Spatial Ecosystem and Population Dynamics Model Bertignac, Lehodey and Hampton, 1998; Lehodey
(SEAPODYM) - previously Spatial Environmental et al. 1998; Lehodey, 2001; Lehodey, Chai and
Population Dynamics Model (SEPODYM) Hampton, 2003; www.seapodym.org
SEASTAR Stock Estimation with Adjustable Survey Tjelmeland and Lindstrøm, 2005
observation model and TAg-Return data
SKEBUB SKEleton BUlk Biomass ecosystem model Bax, 1985
SMOM Spatial Multi-species Operating Model Plagányi and Butterworth, 2006 a,b

SSEM Shallow Seas Ecological Model Sekine et al., 1991

SystMod System Model for the Norwegian and Barents Sea Hamre and Hattlebakk, 1998
6 Models for an ecosystem approach to fisheries

- Dynamic multi-species models or Minimum Realistic Models: models restricted

to represent a limited number of species most likely to have important interactions
with a target species of interest, for example, Punt and Butterworth (1995). The
term Minimally Realistic Model (MRM) was first coined by Butterworth and
Harwood (1991) in response to recommendations to this effect made at a preceding
international workshop. Other models that fall into this category include Multi-
species Virtual Population Analysis MSVPA and MSFOR (Pope, 1991; Sparre,
1991; Magnússon, 1995; Vinther, 2001); Scenario Barents Sea (Schweder, Hagen
and Hatlebakk, 2000); Systmod (System Model) (Hamre and Hattlebakk, 1998);
MULTSPEC (Bogstad, Hauge and Ulltang, 1997; Tjelmeland and Bogstad, 1998);
BORMICON (A BOReal Migration and CONsumption model) (Stefansson and
Palsson, 1998); SEASTAR; GADGET (Globally applicable Area-Disaggregated
General Ecosystem Toolbox) (see e.g. webpage http://www.hafro.is/gadget;
coordinator G. Stefánsson); CCAMLR predator-prey models (e.g. Butterworth
and Thomson, 1995; Thomson et al., 2000), Individual-Based Models (IBM) and
MSM (Multi-species Statistical Models) (Jurado-Molina, Livingston and Ianelli,
2005);
- Dynamic System Models: models that attempt to represent both bottom-up
(physical) and top-down (biological) forces interacting in an ecosystem, including
Individual-Based Models (IBM), OSMOSE (Object-oriented Simulator of Marine
ecOSystem Exploitation) (Shin and Cury, 2001; Shin, Shannon and Cury, 2004),
INVITRO (Gray et al., 2006), biogeochemical models e.g. IGBEM (Integrated
Generic Bay Ecosystem Model) (Fulton et al., 2004) ATLANTIS (Fulton and
Smith, 2004) and SEPODYM/SEAPODYM (Spatial Environmental POpulation
DYnamics Model) (Bertignac, Lehodey and Hampton, 1998; Lehodey et al., 1998;
Lehodey, 2001; Lehodey, Chai and Hampton, 2003).
- Extensions of single-species assessment models: models that expand on current
single-species assessment models taking only a few additional interactions into
account (e.g. Livingston and Methot, 1998, Hollowed et al., 2000; Tjelmeland
and Lindstrøm, 2005). For convenience, these models are here termed ESAM
(Extended Single-species Assessment Models).
Models can be classified as Minimally Realistic Models (MRM) on the one
hand and “ecosystem” models on the other. A MRM seeks to include only those
species considered likely to have important interactions with the species of primary
interest. The MRM group includes MSVPA and its derivatives which project into the
future (e.g. Vinther, 2001), MULTSPEC, BORMICON/GADGET, Seastar, Scenario
Barents Sea and the original seal-hake MRM of Punt and Butterworth (1995). Shared
characteristics of these models include the following (NAMMCO, 2002):
• they are system specific;
• only a small selected component of the ecosystem is modelled, and
• lower trophic levels and primary production are modelled as constant or varying
stochastically.
In contrast, the ATLANTIS and ECOPATH/ECOSIM models, for example, are
generic and capable of explicitly including most ecosystem components as well as
incorporating lower trophic levels and primary production, though naturally they can
also be applied in a simplified form closer to the MRM concept.
In discussing these different modelling approaches below, it is useful to further
classify models (see Table 2) as either “Efficient predator” models or “Hungry
predator” models (Butterworth and Plagányi, 2004). In the former set of models the
predator is assumed to always get its daily ration (e.g. MSVPA, MULTSPEC), though
the species composition of this ration may change with varying prey abundances over
time. In contrast, in the latter set, predators are assumed to compete with others of
the same (and possibly other) species for limited vulnerable proportions of prey (e.g.
“foraging arena”-based models applied in approaches such as ECOSIM).
Review of current modelling approaches 7

TABLE 2
Categorization of models according to feeding relationships assumed as well as whether the
primary model focus is on the effects of non-target species on a commercial prey species, the
effects of fishing on the population of interest or on effects operating in both directions
Model Model units (biomass “Efficient predator” or Primary model focus
or nutrient pools) “Hungry predator” model

ATLANTIS Nutrient Hungry predator Effects in both

directions
Bioenergetic/allometric models Biomass Both Effects in both
directions
CCAMLR models Biomass Efficient predator Effects of fisheries on
protected or other
species
ERSEM II Nutrient Hungry predator Effects in both
directions
EwE Biomass Hungry predator Effects in both
directions
GADGET Biomass Both Ecosystem effects on
target population
IGBEM Nutrient Hungry predator Ecosystem effects on
target population
Individual-based Models (IBM) Biomass Efficient predator Ecosystem effects on
target population
INVITRO Biomass Efficient/Hungry1 Effects in both
predator directions
KPFM Biomass Efficient predator Effects of fisheries on
protected or other
species
MRM Biomass Efficient predator Ecosystem effects on
(Punt and Butterworth 1995) target population
MSM Biomass Mixed Limited effects in
both directions
MSVPA and MSFOR Biomass Efficient predator Ecosystem effects on
target population
MULTSPEC Biomass Efficient predator Ecosystem effects on
target population
OSMOSE Biomass at different Efficient predator but Effects in both
levels of aggregation can starve directions

ESAM Biomass Efficient predator Ecosystem effects on

target population
SEAPODYM Biomass Efficient predator Ecosystem effects on
target population
SEASTAR Biomass Efficient predator Ecosystem effects on
target population
SMOM Biomass Efficient predator Effects of fisheries on
protected or other
species
SSEM Nutrient Efficient predator Ecosystem effects on
target population
1
Dependent on agent types used

In general, the models presented also differ substantially (Table 2) in terms of

whether they represent:
i. only the effects of non-target species on a commercial prey species (e.g. MSVPA,
BORMICON and other models were originally constructed with the primary
aim of assessing fish stocks);
ii. only the effects of fishing (e.g. resulting in prey depletion) on the population of
interest (e.g. CCAMLR models constructed with this aim in mind); or
iii. effects operating in both directions (e.g. ECOSIM).
8 Models for an ecosystem approach to fisheries

Criteria used to compare models

The tables in the Appendixes can be consulted to further examine the above models.
The models are compared (Tables A1 a-d) based on the following criteria:
1. The level of complexity and realism, e.g. the number of modelled species
(Figures 2, 3), the representation of size/age structure of the species and the types

FIGURE 2
Schematic summary showing the trophic level focus of different multi-species
models a) in general and b) for the Antarctic ecosystem given that the latter has
a relatively simple structure

�����������������
������������

��������������������
��

������������������
��

����������������������
��

�����������
���������������

��������������������������������

�����������������������������

���
�� �����
���
�� �

���������������
��������
����������������
�������������������
������������������
����������

�����

�������������

� ���
�����
�� ���
��
Review of current modelling approaches 9

FIGURE 3
Schematic summarizing (approximately) the typical (current) number of modelled
species or model compartments for selected models as listed in Table 1. The solid
rectangles represent the range whereas the dashed lines indicate either rare/unusual
applications or intended future extensions to the model.

30 +
NO. OF SPECIES / COMPONENTS

EwE

ATLANTIS

20

OSMOSE
INVITRO
ERSEM

10
GADGET

Bioenergenetic
BORMICON

SEAPODYM
5 MULTSPEC

SEASTAR
MSVPA

3 MSM

ESAM
MRM

type
2

of represented processes (physical and biological). Although it is not the subject of

the report, brief commentary is provided regarding the representation of technical
interactions or the direct ecosystem effects of fisheries (e.g. bycatch);
2. The types of functional responses of predators to changes in abundance of prey
species and their consequences and limitations;
3. How uncertainties in model structure, parameters and data are treated;
4. How environmental effects and interactions with non-target species (e.g. marine
mammals; sea turtles; sea birds) are incorporated;
5. The spatial representation of species interactions and habitat related processes;
6. Model suitability for dealing with migratory species, i.e. species that cross
ecosystem boundaries;
7. Where possible, model adequacy to allow the analysis of the different types of
management controls in use, such as effort control, minimum size, total allowable
catch, protected areas and seasons;
8. Model adequacy to allow the assessment of the effects of short, medium and long-
term ecosystem changes;
9. Model suitability to conduct assessments and policy exploration, considering its
potential use to conduct historical reconstruction of resources to describe the
current status of the ecosystem and to evaluate the potential effects of various
kinds of decisions (short and long term);
10. Model transparency of operation and ease of use;
11. Data requirements and model suitability for data poor areas.
112
A second set of tables (Tables A2 a-d) summarizes for each of the 20 models
compared, a description of model parameters, some important assumptions, data
requirements, technical information such as the computing platform, a list of examples
where used, notes on the model history as well as any additional useful features of
an approach. Finally, a summary is presented in a third set of tables (Tables A3 a-d)
of some advantages, disadvantages and limitations of each method, as well as notes
10 Models for an ecosystem approach to fisheries

on the ease of presentation of model outputs and the user-level of programming and
mathematical skills required.
A preliminary comparison is attempted of the potential of the different modelling
approaches to address a range of Ecosystem-Based Fisheries Management (EBFM)
research questions outlined in the text (Tables A4).
Discussion is also provided regarding the incorporation of ecosystem considerations
into current Operational Management Procedures (OMPs) and other management
strategies for marine resources. An OMP is the combination of a prescribed set of
data to be collected and the analysis procedure to be applied to these data, to provide
a scientific recommendation for a management measure, such as a Total Allowable
Catch (TAC), for a resource (Butterworth, Cochrane and Oliveira, 1997; Butterworth
and Punt, 1999; Cooke, 1999). A key aspect of the OMP approach is that the analysis
procedure has been tested across a wide range of scenarios for the underlying
dynamics of the resource using computer simulation. This is to ensure that the likely
performance of the OMP in terms of attributes such as (high) expected catch and (low)
risk of unintended depletion is reasonably robust to the primary uncertainties about
such dynamics. By way of example, this approach is used at present to manage South
Africa’s three most valuable fisheries: for hake, for pilchard and anchovy and for west
coast rock lobster (De Oliveira et al., 1998; Butterworth and Punt, 1999; Geromont et
al., 1999) and initial progress has been made in including ecosystem considerations into
these OMPs (Plagányi et al., 2007).
In what follows, a relatively brief description of the various modelling approaches is
presented with much of the supplementary information given in the Tables. The author’s
discretion has been used in drawing the reader’s attention to aspects of the various
modelling approaches that may be of interest and hence, unlike in the Tables, model
descriptions given in the text hereunder are presented at different levels of details.

2.1 WHOLE ECOSYSTEM AND DYNAMIC SYSTEM MODELS

Such approaches attempt to take all trophic levels in the ecosystem into account, from
primary producers to top predators. Quite sweeping simplifications and assumptions
may need to be made in this process. Examples are the ECOPATH with ECOSIM
(EwE) framework, which is usually applied in this manner and biogeochemical models
such as IGBEM and ATLANTIS (Fulton, 2001; Fulton, Smith and Johnson, 2004;
Fulton, Smith and Punt, 2004).

2.1.1 ECOPATH with ECOSIM (EwE)

Given that the ECOPATH (Polovina 1984; Christensen and Pauly, 1992), ECOSIM
(Walters, Christensen and Pauly, 1997) and ECOSPACE (Walters, Pauly and Christensen,
1999) suite is currently dominating attempts worldwide to provide information on
how ecosystems are likely to respond to changes in fishery management practices,
it is important that the applicability of these approaches to answering questions in
this context be carefully reviewed (Aydin and Friday, 2001; Aydin, 2004; Aydin and
Gaichas 2006; Plagányi and Butterworth, 2004). A description of the ECOPATH with
ECOSIM approach is given below (see also www.ecopath.org):
Briefly, the fundamental ECOPATH mass balance equation is based on that
originally proposed by Polovina (1984). This balance for each functional group i in an
ecosystem (detritus excepted) is described by (Walters and Martell, 2004):
is described by (Walters and Martell 2004):
Bi � �P B �i � EEi � � �Q B � j � DC ij � B j � C i � BAi � NM i (1) (1)
j

B and B areBthe
where where biomasses of i and the consumers (j) of i respectively;
i and Bj are the biomasses of i and the consumers (j) of i
respectively;
 Review of current modelling approaches 11

(P/B)i is the production/biomass ratio for i;

EEi is the fraction of production of i that is consumed within, or
caught from the system (the balance being assumed to contribute
to detritus);
Ci is the fishing mortality (landings + discards) on i;
(Q/B)j is the total food consumption per unit biomass of j;
DCij is the fractional contribution by mass of i to the diet of j;
BAi is a biomass accumulation term that describes a change in biomass
over the ECOPATH base-reference-unit time step (usually one year),
and
NM i is the netnet
is the biomass
biomassmigration
migration(immigration-emigration)
(immigration-emigration)forfor i.

Methods to achieve mass balance in an ECOPATH model include both ad hoc

trial and error adjustments and the use of inverse models to minimize the imbalances
between inputs and outputs (e.g. Savenkoff, Vézina and Bundy, 2001). Inverse methods
attempt to provide an internally consistent description of trophic interactions between
all functional groups by finding a solution subject to the constraints posed by the
available data on prior knowledge of the system (Savenkoff et al., 2004). There are
several studies based on an inverse modelling approach (e.g. Vézina et al., 2000,
Vézina and Pahlow, 2003; Savenkoff et al., 2004). Although they have limited practical
applicability because of their static-flow nature, they are useful in addressing issues
of parameter uncertainty and the weighting of evidence from different sources in a
statistically defensible manner.
The ECOSIM models convert the above “steady-state”1 trophic flows into dynamic,
time-dependent predictions. At basis, for prey i and predator j, Walters, Christensen
and Pauly (1997) model the dynamics of the vulnerable (Vij) and non-vulnerable (Ni-Vij)
components of the prey abundance (by number) of i as:

�
d N i � Vij � � � (2)
� �vij N i � Vij � v'ij Vij (2)
dt
dVij
dt
� �
� � vij N i � Vij � v'ij Vij � aijVij N j (3) (3)

is a ijVij N j ,and
where the total consumption rate Qij of prey i by predator j is andNj
represents the number of predator group j.

Under the assumption that the dynamics of the Vij are much faster than those of the
dVij
han those of the N
Nii, is set to zero, yielding:
dt

�
Vij � vij N i vij � v'ij � aij N j � (4) (4)
and hence
and hence (taking(taking
biomassbiomass to be proportional
to be proportional to numbers) the standard ECOSIM
term form describing trophic
interaction term for describing trophic flows Qij between
between prey
prey group
groupi and predator
group j:
Qij � aij vij Bi B j �vij � v'ij � aij B j � (5) (5)

1
Strictly in applications where some BA term is non-zero, the ECOPATH approach does not reflect
“steady-state”/“equilibrium”. However, the spirit of the approach, even with this adjustment, is to
represent balances in a “steady” (possibly steadily changing) situation, in contrast to modelling the
dynamics fully.
12 Models for an ecosystem approach to fisheries

where aij isisthe

where theraterate of effective
of effective searchsearch fori by
for prey prey and v,ij ,, v'ij are
predator, jand areprey
prey
vulnerability parameters.
This consumption equation has been amended in subsequent versions of ECOSIM
to the form (Christensen and Walters, 2004):
��� ��� �� ������ ��� ��� � � �
���  (6)
���  ����� ���  ��� ��� �� ����� � � �

where Ti is the prey (i) relative feeding time;

Tj is the predator (j) relative feeding time;
Sij are the user-defined seasonal or long-term forcing effects;
Mij represents mediation forcing effects; and
Dj accounts for handling time limitations on consumption rate by predator j
as follows:
h jT j
Dj � (7) (7)
1 � � a kj Bk Tk M kj
k
where hj is the predator handling time.

AsAsinin
the classic
the classicLotka-Volterra formulation ((Qij � aij Bi B j ),), flows
Lotka-Volterraformulation flows are
aredetermined
determined by both prey
by both prey and predator biomasses, but Equation (5) (and its extended form shown
in Equation (6)) incorporates an important modification in that it encompasses a
framework for limiting the vulnerability of a prey species to a predator, thereby
including the concept of prey refugia and also tending to dampen the unrealistically
large population fluctuations usually predicted by the Lotka-Volterra formulation.
Earlier, to overcome the limitations of a biomass dynamics framework, where
relevant, juvenile and adult pools in ECOSIM II were linked using a delay-differential
equation system that kept track of flows in terms of numbers as well as biomass.
However, more recent versions of EwE include a facility to model fully age-structured
population dynamics with multiple life history stanzas and recommend the use of this
approach in favour of the adult/juvenile splitting implemented earlier (see Walters and
Martell, 2004). The multiple-stanza version of ECOSIM is a major advancement and
permits testing of, inter alia, the effects of biomass pool composition on aggregated
consumption estimates, the introduction of greater resolution on size-dependent
interaction rates and evaluation of problems such as growth overfishing (Walters and
Martell, 2004).
In many respects, EwE achieves a good balance in model structure between
simplicity and the level of complexity that often accompanies other ecosystem model
representations. Although users have tended to include a large number of components
in their EwE models, it can also be used in more of a Minimum Realistic Model (MRM)
sense (Butterworth and Plagányi, 2004).
Plagányi and Butterworth (2004) review the basic equations and assumptions,
strengths and weaknesses, some past and possible future applications and hence the
potential of this approach to contribute to practical fisheries management advice.
Strengths include the structured parameterisation framework, the inclusion of a
well-balanced level of conceptual realism, a novel representation of predator-prey
interaction terms, the use of a common framework for making comparisons between
systems studied by different researchers, the rigorous analytical framework provided
by ECOPATH (in contrast to an ad hoc type model) and the inclusion of a Bayes-like
approach (ECORANGER) to take account of the uncertainty associated with values for
model inputs. Somhlaba (2006) suggests that ECORANGER is likely computationally
inefficient and could be improved. Aspects of the actual EwE model structure that
Review of current modelling approaches 13

may merit further attention or are potentially problematic include the need to initiate
projections from “steady state” ECOPATH solutions2 (in standard applications), the
questionable handling of life history responses such as compensatory changes in the
natural mortality rates of marine mammals, possible problems in extrapolating from
the microscale to the macroscale3, as well as some (though not too far-reaching in
practice) mathematical inconsistencies in the underlying equations.
Many of the shortcomings of EwE applications are attributable to user misuse (or
insufficient use) rather than to the actual model structure. Uncritical use of default
parameter settings or setting of vulnerability values to the same constant for all species
is unsatisfactory, because inter alia it assumes the same prior exploitation history for all
species and may result in overcompensatory stock–recruitment relationships. There is a
paucity of systematic and stepwise investigations into model behaviour and properties.
As with all multi-species approaches, the major limitation in applying the EwE approach
lies in the quality and quantity of available data. Plagányi and Butterworth (2004) argue
that current EwE applications generally do not adequately address uncertainty in data
inputs and model structure. Recent improvements to the software that use a computer-
automated iterative technique for mass-balancing Ecopath models are a step in the right
direction in the sense that it incorporates a facility for Monte Carlo–based explorations
of sensitivity to different starting conditions (Kavanagh et al., 2004). Nevertheless such
developments must be used with care as dependence solely on such methods can see
the modeler lose their sense of the model’s driving forces and many useful insights into
system dynamics can be lost (E. Fulton, pers. comm.).

Implications of the ECOSIM interaction representation

Plagányi and Butterworth (2004) argue that models need to be closely scrutinized to
understand the extent to which underlying model assumptions predetermine or have
implications for the results obtained. By virtue of EwE being packaged in a form that
is readily digested by as many people as possible, undiscerning users can more readily
use it as a “black-box”, neglecting to test the appropriateness of default parameter
settings and conferring inadequate consideration to alternative functional relationships.
The modular version currently under development is likely to improve issues of
transparency and accessibility as well as forcing less discerning users to better explore
the robustness of their model predictions.
The ECOSIM “foraging arena” concept (see Walters, Christensen and Pauly, 1997;
Walters and Kitchell, 2001; Walters and Martell, 2004) (see Equations 5 and 6), is a novel
functional response representation that is supported to some extent by studies of fish
populations. However, complications to be borne in mind include the fact that EwE
cannot straightforwardly depict instances where the foraging arena V’s (vulnerability
pools) are used simultaneously by multiple predators. This may be important in
instances such as when a fish predator targets similar prey to those targeted by a marine
mammal, or in which there are overlaps in the vulnerability pools available to marine
mammals and to fisheries. EwE as presently configured implicitly assumes that direct
interference between predator species (which it ignores) is inherently different from
within-species interference (explicitly modelled by Equation (5)).
Caution is advised regarding earlier published results from ECOSIM in which users
adopted earlier default settings. As explained in Plagányi and Butterworth (2004, 2005),

2
As with most modelling approaches, it is problematic to extrapolate to situations far from the initial/
equilibrium state.
3
The point here is that if one has a particular functional form at the microscale and the parameters of that
form vary from place to place, this does not mean that when you integrate that form over space the
resultant functional form will necessarily lie within the set of forms covered by varying the parameters
of the original form. This is a problem that persists with almost all models.
14 Models for an ecosystem approach to fisheries

these early versions of ECOSIM could not yield pure-replacement results when
predicting the effects of a “predator” (a fishing fleet, say, that acts identically in terms of
prey selection) in supplanting marine mammals. Expressed another way, this argument
is that default parameter value selections for the model effectively hard-wired it to
such an extent that they effectively swamped other signals pertinent to predicting the
effects of a marine mammal reduction. Cooke (2002) similarly demonstrated through
the use of a simple model that whether or not the reduction in cetaceans results in
higher fishery yields than would otherwise, other things being equal, be obtained,
depends critically on the assumed vulnerability of the fish to the whales. It is only
under scenarios assuming a high vulnerability of fish to whales that fishery yields
are predicted to be sensitive to the abundance of whales. These results highlight the
importance of exploring robustness to assumptions related to consumption because a
priori assumptions in this regard strongly influence model outcomes in terms of whether
or not they yield pure-replacement results. Values other than default could of course be
selected, for example, Mackinson et al. (2003) showed that particular combinations of
ECOSIM settings can be used to produce alternative “emergent” forms of functional
responses, specifically Type I and II, but not Type III, behaviours. In recent years
Type II and Type III functional responses have been built into the ECOSIM general
functional response, which even permits combinations of these variants and hence is
now extremely flexible.

The current and future EwE

A number of modifications and improvements have recently been added to EwE.
Given fairly recent improvements in terms of age-structure handling, many of the
older models have or are in the process of being modified and this is likely to result
in valuable new insights. EwE has in the past been criticized for inadequate handling
of issues of uncertainty (e.g. Plagányi and Butterworth, 2004) but the more recent
versions include improved capabilities to balance models based on uncertainty,
examine the impact of uncertainty as part of the management process and to quantify
input parameter uncertainty to run ECOSIM using a Monte Carlo approach to fit
to time series (V. Christensen, University of British Columbia, Canada, pers comm.,
Kavanagh et al., 2004). (see also Future Developments section).

2.1.2 Biogeochemical models

This category of models differs from the other models discussed in being nutrient-pool
based rather than biomass-based (Table 2).

2.1.3 ERSEM and SSEM

The European Regional Seas Ecosystem Model (ERSEM) was developed to simulate
the annual cycles of carbon, nitrogen, phosphorus and silicon in the pelagic and benthic
components of the North Sea (Baretta, Baretta-Bekker and Ruardij, 1996). ERSEM
model version II (VII) is described in the special issue of the Journal of Sea Research
Vol. 38 (Baretta-Bekker and Baretta, 1997). The model requires detailed data inputs and
focuses on the phytoplankton and zooplankton groups, with detailed representation of
microbial, detrital and nutrient regeneration dynamics. The model is driven by a wide
range of forcing factors including irradiance and temperature data, atmospheric inputs
of nitrogen, suspended matter concentration, hydrodynamical information to describe
advective and diffusive transport processes and inorganic and organic river load data
(Lenhart, Radach and Ruardij, 1997). The spatial scope of the model encompasses
the entire North Sea. More recently, Blackford, Allen and Gilbert (2004) provide a
mathematical description of ERSEM-2004 (developed from ERSEM II) together with a
description of its application to six contrasting sites within the North, Catalan, Cretan
and Arabian Seas. They conclude that when coupled to high resolution hydrodynamic
Review of current modelling approaches 15

models, ERSEM can be applied over large geographical and temporal scales and is thus
a useful tool for studies focusing on lower trophic levels.
The consumers module of ERSEM includes mesozooplankton, microzooplankton
and heterotrophic flagellates. Consumer uptake is of a Michaelis-Menton form and
depends on both food availability and water temperature. A “food matrix” is used as
an input to describe the relative prey availability or preference of the different food
sources for each consumer (Solé, Estrada and Garcia-Ladona, 2006). A useful feature
described in Blackford, Allen and Gilbert (2004) is the introduction of a Michaelis-
Menton term to prevent excessive grazing of scarce prey based on a lower threshold
feeding parameter.
In the current context, one of the most useful applications pertains to attempts to
link ERSEM to individual growth models for fish (Bryant et al., 1995; Heath, Scott and
Bryant, 1997). The entire North Sea herring population was modelled using an age-
structured cohort model that was linked by adjusting the biomass of groups in ERSEM
to reflect prey uptake by herring and conserving carbon and nutrient balances by
accounting for defecation, excretion and mortality products from the fish (Heath, Scott
and Bryant, 1997). The detailed representation of transport processes within ERSEM
allowed simulation of important juvenile growth processes such as year-specific
dispersal and timing of larval recruitment. The model was useful in demonstrating the
extent to which hydrographic and planktonic conditions are responsible for short-
term year-to-year variability in growth but the model failed to explain longer-term
underlying trends thought to be due primarily to density-dependence.
ERSEM could be adapted for other regions as it is essentially a generic model which
is then coupled to an appropriate physical model for a region, such as the General
Ocean Turbulence Model (GOTM). ERSEM has been shown to be equally applicable in
tropical and warm temperate systems such as the Arabian Sea, Mediterranean and Irish
Seas (Allen, Blackford and Radford, 1998; Allen, Sommerfield and Siddorn, 2002; Crise
et al., 1999). Adapting it to other systems requires a fair amount of data. Given that the
focus of ERSEM is on the lower trophic levels, it is unlikely to be able to contribute
to practical fisheries management but is nonetheless a good tool for understanding
environmental drivers and bottom-up processes impacting fish populations.
The Shallow Sea Ecological Model (SSEM) (Sekine et al., 1991) also includes detailed
representation of processes such as swimming, advection and diffusion and requires
inputs in the form of water temperature, currents and nutrient loads from surrounding
land masses. It has specifically been developed to predict the impact on fisheries of
coastal development activities. It is thus adequately tailored for this use but would not
be suitable for broader questions related to the ecosystem impacts of fisheries.

2.1.4 IGBEM, BM2 and ATLANTIS

IGBEM (Integrated Generic Bay Ecosystem Model) (Fulton et al., 2004) is a coupled
physical transport-biogeochemical process model constructed through amalgamation
of ERSEM II and the Port Phillip Bay Integrated Model (PPBIM) (Murray and Parslow,
1999). Some of its main features are summarized in Tables A1a to A4, but it is not
further discussed here given that this model is essentially superseded by ATLANTIS.
ATLANTIS (Fulton, Smith and Punt, 2004) was developed from the “Bay Model
2” (BM2) ecosystem model of Fulton et al. (2004), first applied to Port Philip Bay,
Australia. Its development has been tightly coupled to efforts to evaluate potential
methods and tools (such as ecological indicators) for use in ecosystem-based fisheries
management using a Management Strategy Evaluation (MSE) approach. This approach
requires not only models of how the management decisions are made (including
associated monitoring activities), but at its core it must have an operating model to
represent the “real world” including the impact of fishing and other anthropogenic
effects. ATLANTIS is arguably currently the best model worldwide to play this role
for some of the following reasons:
 16 Models for an ecosystem approach to fisheries

1. It includes the full trophic spectrum;

2. It has a more simplified representation of physiological processes than most other
biogeochemical models, following a detailed sensitivity analysis to determine the
importance of including various processes (Fulton, 2001). On the other hand,
some processes not considered in other models, such as mixotrophy, are included
as they are considered important;
3. Vertebrates such as fish are modelled using age-structured formulations;
4. Lower trophic level groups are represented better than in most whole ecosystem
models (in that it allows some age structuring at the juvenile-adult level for
potentially important invertebrates such as cephalopods and large crustaceans),
whereas the upper trophic level groups are represented better than in other
biogeochemical models;
5. The model is spatially resolved;
6. Multiple vertical layers can be considered;
7. The modular structure allows the substitution of a wide range of different sub-
models for various components;
8. The nutrient-pool formulation allows testing of effects such as nutrient inputs
from point sources;
9. There is detailed coupling between physical and biological processes
10. Multiple representations of some of the processes are included, thereby allowing
the user to choose the preferred option for their modelled system.
Given the above, it is perhaps of interest to briefly describe the equations used to
model fish populations in particular. The rates of change for a vertebrate group (FX)
are given by (Fulton, Smith and Punt, 2004):
d�FX i ,s � (8)
� G FX i ,s
dt
dt (8)
(8)
d�FX i ,r � (9)
� G FX i ,r
dt (9)
dt (9)
d�FX i ,d �
� TIMM ,FX i � TEM ,FX i � M FX i � � PFX, j �FFX i (10)
dt j � predator
groups (10)
where s represents structural weight (skeletal and other non-reabsorbable material),
r reserve weight (fats and other tissues that are broken down when food is limiting), d
density and i age class (either a single year class or a proportion of the total life span
of the animal). The rate of change includes consideration of the difference between
of the difference between movement into ((TIMM , FX i) and out of ((TEM , FX i)) a cell,
movement into cell and removals
removals due
due to
to natural
natural mortality
mortality M, predation mortality P (see below) and fishing mortality F.
Six alternative functional response representations are currently included, with a
common feature being the use of prey availability terms (discussed below). An example
of one of the most commonly chosen grazing term formulations which describes the
consumption of a particular prey group by CX is given by:

CX � k CX � p prey,CX � � refuge � prey (11)

Pprey ,CX � � �
� CX ��
� � p j, CX � j � �� CX, DL �p DL, CX � � CX, DR �p DR, CX
�
1 � k CX � � j � prey �
�CX
(11)
 Review of current modelling approaches 17

where k CX is the clearance rate of CX;

where
p prey ,CX is preference (or availability) of that prey for the
predator CX;
� refuge isis aa term
termused
usedififthe
thegroup
groupisisdependent
depe on biogenic habitat
refuges;
� CX is the growth efficiency of CX when feeding on live prey;
DL and DR are respectively the labile and refractory detrital pools;
and
� CX represents the maximum temperature-dependent daily growth
rate for the group CX.
Fulton, Smith and Punt (2004) note that the prey availability parameter (Pprey , CX)
is similar to the “vulnerability” parameters in ECOSIM (see Equation (5)) as not all
prey are simultaneously available for consumption by a predator. Both habitat and
size refuges are handled in ATLANTIS. Moreover, it includes the most sophisticated
equations (of which this author is aware) to handle the concept of prey refuges given
that the habitat refuge variable can take account of, for example, degradation of the
physical environment due to coastal developments (see Fulton, Smith and Punt, 2004
for further details).
Short-term spawning and recruitment events are modelled as affecting the various
vertebrate pools. Reproduction is modelled as a pulse each year with the materials
required to do this being removed from a group’s reserve weight and a proportion
of the age class simultaneously ageing into the next age class. The amount of reserve
weight (mg N per individual) used during spawning is given by:

�U FXi � max �0 , �Z FX � �1 � X RS �� FX i,s � YFX �� , FX i,s � FX i,r � �1 � X RS � � FX i,s

��
s FXi �� � �� ZFX���1 � X RS ���FX i,s ��� FX i,s � FX i,r �� �� �
U
� FXi � max � 0, � � � � ���
, FX i,s � FX i,r � �1 � X RS � � FX i,s
�� � �� �YFX ���1 � X RS ���FX i,s ��
�
� � � � � �
(12)(12)

where U FX i is the proportion of age classthat

where i thatis isreproductively mature, Z FX isisthe
reproductivelymature, the fraction of
a group’s
fraction of weight
a group’s weightused spawning, YFX is a spawning function constant and
usedininspawning,
ion constant and X RS is the ratio of structural to reserve weight in well fed vertebrates.
In the current model, recruitment can be represented using one of 15 alternative
stock-recruitment relationships (ranging from standard forms such Beverton-Holt and
Ricker, through to more speculative functions conditioned on plankton biomass or other
environmental drivers). As an example, the recruitment btj in cell j at time t when using
the well known Beverton-Holt recruitment relationship is given in ATLANTIS by:

� � � Ltj �
� �
� � � Ltj �
btj � � � (13) (13)
tx
where α, β are the conventional Beverton-Holt constants, tx is total length of recruit
period; and Lt j represents
representsthe
theoffspring
offspringbiomass
biomassinincell
cell
j at time t, with:

Ltj � �s FX i � FX i,d � �1 � � recruit � � �t �� (14)

i � age class
(14)
The term s FX
Theterm represents
i
thethe
represents spawn
spawnfrom
fromage classi,i, � recruit is an episodic recruitment
ageclass
scalar and � is an impulse function, which controls the pulsed nature of recruitment.
An added feature worth mentioning is that ATLANTIS includes a detailed
exploitation model that deals with the impacts of multiple anthropogenic pressures
18 Models for an ecosystem approach to fisheries

(pollution, coastal development and broad-scale environmental change), with a focus

on the dynamics of fishing fleets. Multiple fleets can be simulated, each with their
own characteristics (in the form of gear selectivity, habitat association, targeting, effort
allocation and management structures). Multiple alternative formulations are available,
with the more complicated capable of explicitly handling economics (including quota
trading), compliance decisions, exploratory fishing and other complicated real world
concerns.
The exploitation model interacts with the biological model and also supplies
‘simulated data’ to the sampling and assessment sub-model. The ‘simulated data’, which
may be sector dependent or independent data (via a user defined monitoring scheme),
include realistic levels of measurement uncertainty in the form of bias and variance.
The simulated data are then input to actual assessment models (to date, these have
included surplus production, ADAPT-VPA and fully integrated assessments) and the
output of these acts as input to the management sub-model that applies a set of decision
rules and management actions (currently only detailed for the fisheries sector). The
management sub-model includes a broad range of possible management instruments
such as gear restrictions, spatial and temporal zoning, discarding restrictions, bycatch
mitigation and biomass reference points.
A negative surrounding the breadth and flexibility of the various sub-models
(and their modular form) is that it can seem a daunting and parameter-intensive tool
that may be associated with large uncertainties (E. Fulton, pers comm.). Supporting
software and methods to make this task easier are under parallel development. In a
data rich situation, ATLANTIS may be well suited to a user’s needs, whereas it may
be argued that in a data poor situation the framework is still quite useful for asking
“what-if” questions. As with all modelling approaches, ATLANTIS is not appropriate
in all circumstances and must be used sensibly.

2.1.5 SEPODYM/SEAPODYM
Tuna fisheries are typically high value multi-species and multi-gear fisheries in
which interactions can occur and hence it is not surprising that considerable effort
has been focused on developing a Spatial Environmental POpulation DYnamics
Model (SEAPODYM, previously SEPODYM) (Bertignac, Lehodey and Hampton,
1998; Lehodey, 2001; Lehodey, Chai and Hampton, 2003). SEAPODYM is a two-
dimensional coupled physical-biological interaction model at the ocean basin scale,
developed for tropical tunas in the Pacific Ocean (Lehodey, Chai and Hampton, 2003;
Lehodey, 2005). The model includes an age-structured population model of tuna
species, together with a movement model which is based on a diffusion-advection
equation such that swimming behaviour is modelled as a function of habitat quality.
The inclusion of spatial structure was essential given the need to account for fishing
effort distribution, the widely ranging swimming behaviour of tuna and environmental
variations (Bertignac, Lehodey and Hampton, 1998). The latter are simulated using
input data in the form of sea surface temperature (SST), oceanic currents and primary
production, predicted either from coupled physical-biogeochemical models such as
OGCM (Ocean General Circulation Model, Li et al., 2001) or satellite-derived data
(Lehodey, Chai and Hampton, 2003).
SEAPODYM has thus far only been run in the Pacific Ocean and the first multi-
species simulation including three tuna species (skipjack Katsuwonus pelamis, yellowfin
Thunnus albacares and bigeye T. obesus) has only recently been completed. However,
there are plans to develop additional modules for other oceanic predators (P. Lehodey,
CLS, Toulouse, France, pers. comm.). Moreover, the model executable, associated
software and documentation, including a manual (Lehodey, 2005) are available on the
website www.seapodym.org. The model structure differs from the other models in
the Dynamic systems model category (Figure 1) in terms of representing only a small
Review of current modelling approaches 19

subset of the species in the ecosystem but it is linked to a physical model and hence
allows investigation of, for example, the relationship between climate variability and
recruitment and biomass fluctuations (Lehodey, Chai and Hampton, 2003).

Habitat index and model equations

SEAPODYM incorporates a number of features which render it useful in a broader
context, particularly to explore the dynamics of upper trophic level predators which
are highly mobile. Several fish and top predator species are likely to distribute
themselves spatially based on the availability of prey and the physical characteristics
of the environment as is the case for tuna (Lehodey et al., 1998). The habitat index Ha
included as part of SEAPODYM is thus designed to preferentially distribute tuna in
regions with large food availability and temperature in a range deemed favourable for
the species in question. Tuna larvae are assumed to be passively transported by surface
currents whereas young and adult tuna movements are constrained by the adult habitat
index. The rate of movement into and out of favourable and unfavourable habitats is
modelled by including a function to increase the diffusion (D) advection (( � )) at
andadvection
) and at
low values of habitat index. Movement is also proportional to the size of the fish such
that:

Da � D � La � �1 � �H a �g 2 � H a ���
(15) (15)
� a � � 0 � La � �1 � �H a � g1 � H a ���
where Da and
where and � a are respectively the diffusion and advection at age a, La the
length of fish at age a and g1 and g2 two coefficients constraining the shape of the
function. Parameterisation is achieved by comparing with the results of tagging studies
(Lehodey, Chai and Hampton, 2003). The above approach is fairly straightforward and
could usefully be applied in other systems/models too provided physical information
is available on sea surface temperature, currents and primary production. Tagging
information is also required to estimate the parameters of the movement model.
The natural mortality rate in the model depends also on an index of habitat quality.
As in more traditional single-species models, the fishing mortality is computed as
proportional to thefishing effort E i , j ,t, the catchability coefficient of the fishery q and
fishingeffort
the gear-and age-specific selectivity coefficients sa, i.e.

Fi , j ,t ,a � s a qEi , j ,t (16) (16)

where Fi , j ,t ,a, is the fishing mortality rate of age class a fish in spatial cell i, j during
where
time period t. A knife-edge selectivity function is assumed.
Recruitment is modelled as independent of the adult population density. Instead
spawning occurs in all cells in which mature tuna are present and SST is above a limit
value. Thereafter the larvae are distributed passively by sea currents. The model has
also been extended to permit investigations of the effect of other environmental factors,
such as food availability and predation, on larval survival and pelagic fish recruitment
(Lehodey, Chai and Hampton, 2003). This aspect of the model is thus suitable for
extending to other pelagic species such as sailfish, swordfish and sharks.
SEAPODYM has several features which suggest that it could be a useful tool if
applied to model marine mammals such as whales, but the recruitment formulations
would need to be modified for this purpose. Another limitation relates to the lumping
of all the tuna forage items into a single model compartment (as was indeed necessary
given the original aims of the model) (Lehodey et al., 1998, Lehodey, 2001). This
means that the model is not suitable for exploring hypotheses in which it is important
to differentiate between the quality and quantity of different types of prey items or
to represent unavailable fractions of this component. The model does not explicitly
20 Models for an ecosystem approach to fisheries

model inter-species and inter-trophic level interactions and hence is not suitable as a
tool to address questions related, for example, to impacts mediated through trophic
interactions.
The population dynamics equations underlying SEPODYM are relatively
straightforward and as such are generally applicable to a wide range of species.
Population size (P) is determined as follows:

�P � � �P � � � �H � (17)
� �D � � � �0P � � ZP � R (17)
�t �x � �x � �x � �x �
where R is recruitment and Z is the total mortality rate. The equation above is
generalized to two dimensions and solved using the finite difference method using
discrete time steps of one month and 1˚-square spatial cells (Bertignac, Lehodey
and Hampton, 1998). Other methods are used to solve the other partial differential
equations and advection terms. In general it appears the numerical solution methods
are slow because computing power is currently the major impediment to adding more
species groups to the model (P. Lehodey, pers comm.).
SEAPODYM is an improved version of SEPODYM in that it incorporates an
improved description of intermediate trophic levels in three vertical layers, as well as
improved handling of multiple predators (Lehodey, 2005). Moreover, an improved
numerical scheme allows the use of spatial stretched grids so that resolution can be
changed (reducing computation time), depending on the level of interest of a region.
The six components of the mid-trophic level included in SEAPODYM are epipelagic,
migrant mesopelagic, non-migrant mesopelagic, migrant bathy-pelagic, highly migrant
bathy-pelagic and non-migrant bathy-pelagic. Given that the most recent version
includes several forage components, revisions were necessary to simulate the coupling
of forage mortality to the density of predators. This has essentially been done by
adding a single mean daily food ration parameter for each predator species, which is
used to compute the total forage required by each predator from the various forage
components (Lehodey, 2005). Potential problems with this simple approach include
the possibility of the combined predator forage requirements exceeding the available
forage biomass.
SEAPODYM thus fits under the “fixed ration” model category defined earlier.
Most of the models in this category do not include any feedback from predators to
prey. SEAPODYM similarly does not explicitly include such feedbacks, but has a
number of potential indirect feedback loops in that changes in foraging mortality can
change both spawning habitat and feeding habitat, with changes in the latter in turn
resulting in changes in natural mortality and fish spatial distribution (Lehodey, 2005).
SEAPODYM is a valuable tool for integrating data from the environment, fisheries
and biology of target species to explore bottom-up forces that affect fish populations.
An example is the use of SEPODYM to explore the biological consequences of an
ENSO (El Niño Southern Oscillation) event in the pelagic ecosystem for the equatorial
western and central Pacific ocean (Lehodey, 2001) as well as to explore global warming
scenarios (Loukos et al., 2003).

2.2 MINIMUM REALISTIC MODELS

Punt and Butterworth (1995) developed the first so-called MRM in response to a need
to quantify the potential effect of seals on hake, the most valuable fishery for both
South Africa and Namibia. The Punt and Butterworth (1995) approach was founded
in the recommendations of a workshop held in Cape Town in 1991 to develop a basis
to evaluate fur seal-fishery interactions off the west coast of South Africa (Butterworth
and Harwood, 1991). This led to the coining of the term Minimum Realistic Model
(MRM) to describe the concept of restricting a model to those species most likely to
Review of current modelling approaches 21

have important interactions with the species of interest.

A critical issue raised in this context relates to the optimal level of complexity for
multi-species models (see e.g. Pinnegar et al., 2005; Quince, Higgs and McKane, 2005).
Reducing the number of species considered, or aggregating similar species into groups,
reduces the number of inter-species links which need to be modelled, but consequently
also reduces the number of weak links included in the model. Yodzis (1998) used a
food web model of the Benguela ecosystem to show that the exclusion of feeding
links representing less than 10 percent of consumption both by and of any species had
minimal effect on model predictions, but that above this threshold for linkage strength
the model predictions started to become unreliable. The reasons why simplified model
outcomes varied drastically from outcomes based on detailed foodweb structure is
likely due to the presence of potentially strong diffuse effects in complex food webs
(Yodzis, 2000).

2.2.1 The original MRM

Off the South African west coast, the fur seal population (Arctocephalus pusillus pusillus)
is estimated to consume about as much hake as is landed by fishers (Butterworth et al.,
1995), begging the question of whether the hake fishery would benefit in response to
a seal cull. The commercially valuable hake consists of two species, a shallow-water
(Merluccius capensis) and a deep-water species (M. paradoxus), with the larger of the
shallow-water species eating the smaller individuals of the deep-water species.
The Punt and Butterworth (1995) model was restricted to the two species
comprising the hake resource, seals, a grouped category of large predatory fish and the
hake fishery. Together these were estimated to account for more than 90 percent of all
mortality of hake. The level of detail taken into account for each component depends
on that considered necessary to capture the key aspects of its dynamics. Thus fully
age-structured models were used for the two hake species (to capture cannibalism and
interspecies predation effects), but the “other” predatory fish components were simply
lumped into either a small or large fish category.
One advantage of the Punt and Butterworth (1995) model is that a realistic population
dynamics model (Butterworth et al., 1995) was used to simulate the seal population,
in contrast to the more usual practice of trying to adapt models originally constructed
to simulate fish dynamics. A summary of the major features and assumptions of this
approach is listed below:
• the model is discrete (with half-year time-steps);
• the dynamics of the two hake species are modelled separately using a (modified)
age-structured production model. The two species are treated as one in a sensitivity
test;
• the model includes both cannibalism and interspecific predation;
• equations (18) and (19) below include noise terms which were ignored for all the
deterministic calculations and handled in a rather ad hoc way for the stochastic
runs. This aspect could be improved, for example, through the use of Bayesian
methods (A.E. Punt, School of Aquatic and Fishery Sciences, University of
Washington, pers. comm.); and
• natural mortality for hake has four sources:
1. Predation/cannibalism by hake: this is affected by three factors: the number
of predators, the number of prey and the “desirability” of different species/age-classes
to a particular predator. The daily hake ration of a predator of species j (either seals,
M. capensis or M. paradoxus) is assumed to be given by a Holling Type II feeding
function relationship, as recommended by Butterworth and Harwood (1991), on the
grounds of simplicity and availability of sufficient data to allow parameter estimation.
The daily hake ration of a predator of species j and age a during the first half of the year
y is thus given by:
22 Models for an ecosystem approach to fisheries

~
� � j �� 2 / 2
R yj, a � Raj 1 � exp(�� ajV yj,a e y , a � ) � (18)
� yj , a from N (0; � �2 )
j
where R y ,a
where is the
is the mass
mass of hake
of hake consumed
consumed each day by predators of species j and
age a during year y;
~
Raj is the maximum daily ration for a predator of species j and age a;
� aj determines the extent of saturation in the feeding function
relationship,
V yj,a is the total biomass of hake which is available for consumption by
predators of species j and age a during the first half of year y; and
�� reflects the extent of the annual variation in the diet.

2. Predation by seals – the same form as above.

3. Predation by “other predatory fish” (e.g. snoek Thyrsites atun, kingklip
Genypterus capensis and sharks): assumed that the number of hake of species i and age
a which are eaten by these fish is related to the abundance of such hake by a Holling
Type II feeding relationship. The number, D, of hake of species i and age a which are
eaten during the first half of the year is given by:

D yi ,,predfish
a � �
� u ai B yopf 1 � exp � v ai wai � 1 N yi , a � e
4
� iy,,predfish
a �� �2 / 2
� (19)

� yi ,,predfish
a from N �0; � �2 � (19)
i
where u
where a is the
is the maximum
maximum number
number of
of hake
hake of
of species
species i and age a per unit
biomass of other predatory fish which could plausibly be eaten
(pre-exploitation level);
exploitation level);
B yopf is the biomass of “other predatory fish”, as a fraction of the pre-
exploitation level;

wai � 1 isisindividual
is the thetheindividual
massmass
individual massof
ofhake
of hake hake of age a � 14 ;;
ofage
of age
4

N yi ,a is the number of hake of species i and age a in year y;

v ai determines the extent of saturation in the feeding function relationship;

and

�� reflects the extent of the annual variation in the diet.

i
Note
Note
Note that u aand
that
that and v ai were
werepre-specified
pre-specified inputs
inputs (sensitivity
(sensitivity to their
to their values
values was was
examined).
4. Basal natural mortality rate (Mb) – mortality attributed to “other causes” not
included in the model. This was somewhat arbitrarily set to 0.1 yr -1.
Of the many factors considered in the sensitivity tests by Punt and Butterworth
(1995), notable changes to the base-case trial were obtained only by increasing the
extent of predation by seals on M. paradoxus. There thus exists a need to examine more
recent data to check the validity of the assumption in the original model that seals feed
mainly in shallow waters and hence that their hake consumption is presumably nearly
all constituted by M. capensis. A second aspect of the Butterworth et al. (1995) seal
model which may need to be revised concerns the model structure lacking any feedback
Review of current modelling approaches 23

between a paucity of hake and a population-dynamic response in (for example) weight-

at-age, survival and/or reproduction of seals, i.e. it was assumed that there was always
sufficient “other” food for such predators.
The hake model used a Holling Type II feeding function relationship. The way in
which the daily ration of daily ration ofis acomprised
a predator predator is ofcomprised of differe
different hake species and age-
classes depends in part on on the “desirability” ((� a ',a - see eqn. App.II.12
the“desirability” j ,i
App.II.12 in Punt and Bu
Butterworth, 1995) that predators of species j and age a’ exhibit for hake of i and age
a, as estimated from available feeding data.
Punt and Leslie (1995) computed estimates of diet composition and daily ration for
the Cape hakes using information on stomach contents collected during demersal trawl
surveys by the SFRI (Sea Fisheries Research Institute – now MCM) between 1988 and
1994. Estimates of evacuation rates for Cape hake were obtained using a model of the
stomach evacuation process and data for juvenile Cape hake and other gadoids. Of
interest is that their estimates of evacuation time were notably larger than those used
in earlier analyses, suggesting that the time to evacuate 90 percent of a prey item ranges
from 2 to 10 days depending on the meal size and the size of the predator. A key feature
of this study was the conclusion that hake meal frequency decreased rapidly with hake
size, so that the largest hake were feeding about once every 10 days only. Without this
low feeding rate, the model produced a perpetual-fishing-machine - large hake would
be so effective at eating small ones, that the harder one fished and removed larger hake,
the more smaller hake escaped such predation and became available to make for even
larger sustainable fishery catches (D.S. Butterworth, UCT, pers. comm.).
The notion that digestion time constraints likely put a cap on the consumption rates
of hakes is important in discussing the appropriate form of the functional response
because, for example, it runs counter to one of the assumptions underlying ECOSIM’s
functional response formulation, namely that “predators with full stomachs are not a
common field observation” (Walters and Kitchell, 2001). Walters and Martell (2004)
note further that studies such as that by Schindler and Eby (1997) (based on 18
freshwater fish species in lakes) suggest that realized growth rates are typically only 26
percent of the maximum possible rate predicted from bioenergetics. Other data such
as that in Table I of Punt and Leslie (1995) suggests predators such as hake regularly
show full stomachs, but there is evidence in the literature in support of both views.
For example, Arrington et al. (2002) showed that across 254 fish species the mean
percentage of empty stomachs was some 16 percent, but this varied from 0 percent to
79.4 percent among individual species. Arrington et al. (2002) suggest that piscivorous
fish in particular regularly experience long periods of empty stomachs.
A potential problem with the “desirability” parameters concerns the fact that these
are assumed to be independent of density. This could be addressed to some extent by a
more intensive stomach sampling exercise, for example by using techniques to smooth
spatial and temporal variability in food composition and predator abundance, such as
the geostatistical approach of kriging (Bulgakova, Vasilyev and Daan, 2001). A further
example of methods used to separate prey size preference from prey availability is
given in Floeter and Temming (2003) (who consider North Sea cod).

Management procedure considerations

A noteworthy feature incorporated in the Punt and Butterworth (1995) approach
involved taking explicit account of uncertainty and management issues through the
use of a simulation framework that incorporated the feedback control rules actually in
place for setting TACs for the hake fishery. The purpose of this approach was to check
whether, even if a seal reduction did increase hake sustainable yields, the management
system applied to compute TACs was such as to be able to take advantage of this. In
a similar context, Cooke (2002) stresses the importance of considering management
 24 Models for an ecosystem approach to fisheries

constraints and issues of uncertainty as integral components of attempts to assess

the effects of changing cetacean abundance on fishery yields. The approach of Punt
and Butterworth (1995) provided a useful framework for further work in this field
and it is encouraging that there are currently a steadily increasing number of multi-
species Management Procedure/MSE studies taking this approach beyond single and
limited multi-species applications to consider much broader aspects of ecosystems or
assemblages.

2.2.2 ESAM (Extended Single-species Assessment Models)

Livingston and Methot (1998) and Hollowed, Ianelli and Livingston (2000) explicitly
modelled predation mortality in a catch-at-age stock assessment model applied to
the Gulf of Alaska walleye Pollock (Theragra chalcogramma). They incorporated
the effect of three predators: arrowtooth flounder (Atheresthes stomias), Pacific
halibut (Hippoglossus stenolepis) and Steller sea lion (Eumetopias jubatus) by defining
predation mortality as a type of fishery. Two important features of this approach were
the use of a flexible functional response form capable of reflecting varying levels of
predator satiation and of statistical methods to fit the model to the data. Tjelmeland
and Lindstrøm (2005) provide a further example of the incorporation of predators
into standard fish stock assessment models. They incorporated predation by northeast
Atlantic minke whales in the SeaStar herring stock assessment model and estimated the
parameters of the consumption formula by directly including the consumption term in
the likelihood function maximized.
A first step to constructing a multi-species model based on a rigorous assessment
model is to include the various predators simply as alternative “fishing fleets”, rather
than estimating their effects as part of a “natural mortality” term. Gulland (1983)
outlined methodology for extending single-species models to take account of multi-
species considerations. Plagányi (2004) similarly applied the “predators as a fishing
fleet” approach to a simple representation that incorporated the two Cape hake
species as two separate species with M. capensis preying on M. paradoxus and both
of the hake species acting as a predator on juveniles of their own species to emulate
the cannibalism known to occur. Seals were included as a separate “fishing fleet” that
preyed on M. capensis. Each predator was ascribed a selectivity function (based on
stomach content data). The two hake species were modelled simultaneously using
an age-structured production model (ASPM) (e.g. Hilborn, 1990; Butterworth and
Rademeyer, 2005) approach and by fitting to GLM-standardized CPUE data.
As in a typical ASPM, the predator-specific catch by mass in year y is given by:

m
C ypred � � wa N y , a S apred Fypred (20) (20)
a �0

where
where wa is the mass of an animal of age a;

N y ,a is the number of animals of age a at the start of year y;

S apred is the fishing selectivity-at-age for a predator pred; and

Fypred is the fishing “mortality” (strictly here that proportion of the fully
selected numbers present which are caught by predator pred).

The proportion of the selected component of the resource harvested each year
source harvested each year (( Fypred ) by predator pred is therefore given by:
Review of current modelling approaches 25

Fypred � C ypred / Bˆ ypred (21) (21)

with the number of animals of age a taken by predators in year yy (C ypred

, a ) given by:

C ypred
,a � S a
pred
Fypred N y ,a (22) (22)

The major challenge in constructing such a model obviously lies in the choice of a
suitable interaction term. The simplest way to estimate the predator-specific catch by
mass in year y is to use a Lotka-Volterra-type interaction of the form:

C ypred � a pred B ypred B yprey (23) (23)

where a pred isis an

where an“availability”
“availability”constant (i.e.(i.e.
constant the the
interaction
in constant). However,
this is a particularly strong interaction form and alternative forms should be explored,
such as:

�
C ypred � a pred B ypred B yprey 1 � b pred B yprey � (24) (24)

which allow for predator satiation. More complicated functional response

formulations (such as the various Holling functional response formulations or
ECOSIM’s foraging arena formulation) can readily be incorporated in a simple model
of this form.
Plagányi (2004) simultaneously estimated biomasses of the two hake species in
the model fitting process and initial attempts were made to fit thet extrathe extra parameters,
parameters,
constants
namely the interaction corresponding
constants to each
corresponding interaction
to each interaction(e.g. estimate a pred describing predation by
(e.g.estimate
describing predation by M. capensis on M. paradoxus). Initial investigations suggested
that the data were not sufficient to support estimation of (all of) these additional
parameters. However, given appropriate data, it may be possible to input estimates of
the predator-specific catch by mass in year y directly, e.g. seal predation on M. capensis
could be fixed in a base-case.
The development of a simple “fishing fleet” type model as described above is a good
starting point to address multi-species issues, particularly because it could be based
upon existing single-species models (preferably length-based). The approach could be
improved by building on length-structured models given that most feeding interactions
are strongly size-based (see discussion under OSMOSE). By building these models in
a stepwise fashion, they could be extended to achieve greater realism, or moulded to
provide greater insight into predation-mediated changes (BENEFIT, 2004).
A further example relating to modifying conventional age-structured assessment
models to investigate multi-species effects is presented in Chouinard et al., 2005.
They investigated the hypothesis that increased predation by a growing number of
Grey seals Halichoerus grypus resulted in increases in the natural mortality (M) of
Atlantic cod Gadus morhua, thereby playing a role in the decline of this species. Rather
than explicitly modelling seals, their approach entailed estimating trends in M using
sequential population analysis (SPM) within an ADAPT framework.

2.2.3 MSVPA approach

Multi-species Virtual Population Analysis (MSVPA) is a technique that uses commercial
fisheries catch-at-age and fish stomach-content data to estimate both the past fishing
mortalities and the predation mortalities on some of the major fish species of interest
26 Models for an ecosystem approach to fisheries

(see e.g. Sparre, 1991; Magnússon, 1995). Unlike VPA (Virtual Population Analysis)
which assumes that the natural mortality rate remains the same over time and usually
also age, here natural mortality is split into two components: predation due to
predators explicitly included in the model (M2) which depends on time and age because
of variations in predator abundance and residual mortality (M1) due to all additional
factors which are customarily taken to be constant. Based on the estimates of M2 that
result, forward-looking simulations (MSFOR) are then used to determine the average
long-term consequences of changing patterns of fishing.
One disadvantage of this approach is that it requires substantial data pertaining to
the predation ecology of the predators included in the model, to the extent that tens of
thousands of stomachs were sampled in the North Sea in 1981 and 1991, the “Years of
the Stomach”, under the auspices of the International Council for the Exploration of
the Sea (ICES). MSVPA applications have mainly focused on the North Sea, with the
considerable data requirements generally impeding the application of this approach to
other areas, although similar approaches have been applied to the Baltic Sea (Sparre,
1991), Georges Bank (Tsou and Collie, 2001), Eastern Bering Sea (Livingston and
Jurado-Molina, 2000; Jurado-Molina and Livingston, 2002) and Barents Sea as well as
to the Gulf of Maine.
A second potential problem with MSVPAs in general is that they concentrate
on the impacts of predators on prey but ignore any potential effects that changing
prey populations may have on the predators themselves (because of the approach’s
constant ration assumption – see below). Nonetheless, the approach has some utility
in quantifying the relative losses in prey biomass attributable to other predatory
fish, marine mammals and commercial fisheries. Moreover, the MSVPA studies
have made a start (e.g. Rice et al., 1991, Rindorf, Gislason and Lewy, 1998, Jurado-
Molina, Livingston and Ianelli, 2005) in trying to determine the extent to which the
consumption of a given prey is a simple linear function of its relative abundance in an
ecosystem (the constant suitability assumption). “Suitability” is an important input to
MSVPA and specifies the relative preference that a predator would have for different
prey species, if all were present in equal abundances.
Although most areas lack sufficient data to permit the application of a full MSVPA
approach (for which collection of all necessary data is exorbitantly expensive [Hilborn
and Walters, 1992]) such as that applied in the North Sea, there is the possibility of
applying a slightly simpler or even hybrid version. The data intensive requirements of
MSVPA could be reduced (obviously at the expense of increasing model uncertainty)
by restricting the focus to a smaller subset within the ecosystem and by making various
assumptions regarding the length of the time period over which data such as age-length
keys and stomach samples are assumed to be adequately representative.

Hybrid MSVPA approaches

Mohn and Bowen (1996) used a hybrid-type approach to model the impact of Grey
seal (Halichoerus grypus) predation on Atlantic cod (Gadus morhua) on the eastern
Scotian Shelf. Their approach involved first running a standard VPA using commercial
landings and research survey data and then adding the consumption of cod by grey
seals to the commercial landings and repeating the VPA which was retuned to take
grey seal predation into account. They incorporated two alternative models of food
consumption by seals (a constant ration predation model in which the fraction of
cod in the diet was assumed constant and a proportional ration model in which the
fraction of cod in the diet was assumed proportional to cod abundance), with these
two predation models yielding substantially different estimates of the amount of cod
consumed by grey seals.
A further limitation for MSVPA in some contexts is that it is age- rather than length-
based and the latter is frequently inescapable for tropical areas for example. However,
Review of current modelling approaches 27

age/length hybrid MSVPA versions have been produced (Christensen, 1995b). These
approaches are based on length-based catch information as well as a number of other
relationships such as the mean weight of length classes, length-age growth parameters
and prey size selection functions.

2.2.4 MULTSPEC, BORMICON and GADGET

These models (and others not described in detail here such as Scenario Barents Sea
(Schweder, Hagen and Hatlebakk, 2000), Seastar (Lindstrøm, Tjelmeland and Haug,
2002) and FLEXIBEST (IWC, 2004a)) are all of Northern Hemisphere origin and
have variously incorporated predation by marine mammals. A common feature is
that they are area-disaggregated which is a definite advantage given the migratory
behavior of many marine mammals and the consequent importance of considering
spatial-temporal overlaps between fisheries, marine mammals and shared prey species.
In brief, MULTSPEC (see Bogstad, Hauge and Ulltang, 1997) is a length-, age- and
area-structured simulator for the Barents Sea that includes cod, capelin, herring, polar
cod, harp seal and minke whales. Predation interactions are modelled only as one-
way in the case of marine mammals, which in the model do not react to changes in
prey availability. BORMICON (A BOReal Migration and CONsumption model)
is another area-structured approach for the multi-species modelling of Arcto-boreal
ecosystems (Stefánsson and Palsson, 1998).
Given that work is not currently continuing on MULTSPEC and that BORMICON
is being incorporated as a special implementation of GADGET, the focus here falls
instead on a brief review of GADGET (Globally applicable Area-Disaggregated General
Ecosystem Toolbox) (Begley, 2005; see also webpage http://www.hafro. is/ gadget;
coordinator G. Stefánsson). Current case studies include the Celtic Sea, Icelandic
waters, southern Benguela hake populations and the North Sea and North Atlantic
herring. Plagányi and Butterworth (2005) note that GADGET is still being developed
but shows great promise for modelling indirect interactions between marine mammals
and fisheries (and has been recommended for such – NAMMCO, 2002).
In GADGET, populations can be split by species, size class, age group, area and
time step. The model platform is flexible in permitting the easy addition/substitution
of alternative model components of biological processes such as growth, maturation
and predator-prey interactions. Thus, for example, there are currently seven growth
functions from which to choose, including forms such as a simplified “MULTSPEC”
type growth equation, a von Bertalanffy equation, two simplified forms of this as well
as an extended version which allows for spatial and temporal growth differences, an
extended form of the Jones growth function which includes the concept of starvation
and a simple power-based growth equation (Begley, 2005). The beta statistical
distribution is then used to distribute the growths around the mean.

GADGET’s consumption formulations

Prey consumption rate Cp is modelled as dependent on the length of both the predator
and the prey p, as well as the relative abundance of the prey (when compared to the
total amount of food available). Values of C can affect predator growth depending on
the growth function selected. The consumption equations are of interest as they are
formulated in a particularly flexible form as follows (Begley, 2005):

N L M L� L F p (l , L)
C p (l , L) � (25) (25)
�Fp
p (l , L)

where Fp (l,L), which governs the amount of prey consumed by a predator, depends
on the product of prey biomass, energy content Ep and the suitability S, such that:
28 Models for an ecosystem approach to fisheries

F p (l , L) � �S p (l , L) E p N lWl �
dp
(26) (26)

and ML , the maximum possible consumption rate by a predator, depends on

temperature and length as follows:

M L (T) = m0e (m 1T−m 2T ) Lm 3

3
(27)

where m1, m2 and m3 are constants.

Finally � ,, the
Finally the “feeding
“feedinglevel”
level”is:is:

�F p (l , L)
�L �
p (28) (28)
HA � � F p (l , L)
p

where:
L is the length of the predator;
l is the length of the prey;
H is the half feeding level (pre-specified value representing density of prey
corresponding to half maximum consumption level);
A is the size of the feeding area;
d is the preference of the predator for the prey;
N is the number of prey in the length cell l, or number of predators in lenght cell L;
W is the mean prey weight in the length cell; and
T is the ambient temperature.
GADGET currently includes five or more suitability functions (Begley, 2005),
ranging from a constant suitability function (the proportion of the prey length group that
a predator can consume is independent of predator length) to the Richards (logarithmic
dependence on both predator and prey length) and Andersen (dependent on the ratio of
predator length to prey length) suitability functions. Similarly, a number of options are
available to model recruitment, with the following four recruitment functions currently
included (Begley, 2005): a fecundity-recruitment function, a simple spawning stock
biomass (simpleSSB), a Ricker relationship and a Beverton-Holt recruitment function.
Fishing fleets are modelled in an analogous manner to predators and hence suitability
functions are defined for fleets to reflect which stocks are caught.
Movement is implemented by either directly specifying migration matrices, or
calculating these based on migration ratio input information describing the proportions
of the stock that will migrate between different areas. These matrices can for example
be used to capture broad seasonal patterns, even if the finer details are not known. A
particularly useful aspect of GADGET is its tagging experiment feature that can keep
track of the number and proportion of fish in an age-length cell that have been tagged.
A number of tags can be lost from the population at each timestep as a consequence of
capture, natural mortality or tag loss.

Statistical fits to data

Appreciable improvements in representing uncertainty are possible given the inclusion
of a range of options in the construction of penalised likelihood functions that are
maximized to obtain parameter estimates and can also serve to provide associated
confidence intervals when fitting to data. There are currently 12 penalised potential
likelihood contributions incorporated in GADGET (Begley, 2005). These cover the
very wide range needed for multi-species models and are as follows (Begley, 2005):
Review of current modelling approaches 29

Data likelihood contributions:

(1) Catchdistribution (age, length or age-length grouped catch data); (2) Catchstatistics
(biological data such as mean length at age or mean weight at age); (3) StockDistribution
(biological properties of different stock components (e.g. immature and mature
components)); (4) Surveyindices (standardized indices of abundance or age-length
indices); (5) SurveyDistribution; (6) StomachContent; (7) Recaptures (data from field
tagging experiments); (8) RecStatistics, and (9) CatchInKilos.

Penalty functions:
(10) Boundlikelihood (assigns a penalty weight to parameters that move outside
pre-specified bounds); (11) Understocking (penalty term for overconsumption by a
predator or fleet), and (12) MigrationPenalty (penalty term for nonsensical values in
the migration matrices).
Formulations are available to deal with data that are aggregated into either age,
length or age-length groups. The “goodness of fit” of the model is assessed using a
weighted sum of penalised likelihoods for a range of individual components. The use
of a powerful algorithm to conduct global maximization of the penalised likelihood is
a definite advantage as is the continuing work to derive improved statistical measures
of uncertainty.
A large range of variants are available to define the type of linear regression equation
(e.g. linear or log-linear regressions with fixed or estimated slope and intercept) to be
used in the likelihood calculations or the choice of assumed statistical distributions
for the error components of the (implicit) models relating data to model variables
(Multinomial, Pearson, gamma or log).
GADGET is thus extremely flexible in terms of methods for fitting to data, being
comprehensive and incorporating state-of-the-art features, with the only disadvantage
of these being that it is foreboding for a novice user! Although the GADGET manual
is fairly comprehensive, it doesn’t always include the underlying equations for some
components making it difficult to follow these. New users will battle to get going on
their own, suggesting the need for more workshop type sessions as is successfully done
for EwE. Advanced users will greatly appreciate the fact that GADGET is capable
of running on multiple computers in parallel using PVM (Parallel Virtual Machine)
(Begley and Howell, 2004).
As with the other modelling approaches, a major impediment to applying this
approach in many cases is the current lack of adequate data to describe feeding
relationships, especially when considering situations where resource abundances and
their ratios differ greatly from those of the recent periods for which data are available.
A strong advantage however is that GADGET incorporates a data warehouse that
provides the flexibility for ready use of data at the different levels of aggregation that
may be required across a number of investigations.
Some of the recent changes (GADGET versions 2.1.01 and 2.1.02) (Begley, 2005)
to the model include the addition of the Richards and Gamma suitability functions,
a capability to deal with catch information by number rather than mass, of a prey
energetic content component and of parameters to allow for a Type III functional
response.

2.2.5 Multi-species statistical models

One of the most well-known and utilized fisheries assessment methods is VPA or
cohort analysis which is a recursive algorithm utilising catch-at-age information
with no underlying statistical assumptions. Hilborn and Walters (1992) distinguish
between this method and so-called “Statistical Catch-at-Age Methods” which rely on
the formal statistical estimation of parameters. Single-species statistical catch-at-age
models are widely used in fisheries management but there have been fewer attempts
30 Models for an ecosystem approach to fisheries

to extend these approaches to multiple species models. Unlike more traditional multi-
species models such as MSVPA, Multi-species Statistical Models (MSM) are forward-
fitting and hence use likelihood maximisation algorithms for parameter estimation.
This is the same general approach as employed by models discussed elsewhere in
this report, such as Punt and Butterworth (1995), Livingston and Methot (1998) and
Hollowed et al. (2000). However, the MSM approach currently being developed by
Jurado-Molina, Livingston and Ianelli (2005) is categorized separately here because
unlike these other statistical catch-at-age models discussed in this report, it includes
predator-prey feedback dynamics. Thus, changes in the prey population can impact
the predator population and vice versa rather than a one-way interaction only in which
the predator ration is fixed and changes in prey abundance have no effect on predator
populations. The initial application includes only walleye pollock and Pacific cod
Gadus macrocephalus (including cannibalism), but there are plans to incorporate more
species in future model versions (Jurado-Molina, Livingston and Ianelli, 2005).
A distinct advantage of the MSM approach is the use of formal statistical methods
for estimating the parameters of multi-species models and quantifying the associated
uncertainty.

2.3 INDIVIDUAL-BASED MODELS

Individual-based models (IBMs) (e.g. DeAngelis and Gross, 1992; Van Winkle, Rose
and Chambers, 1993; Grimm, 1999) follow the fate of individuals through their life
cycle, under the assumption that individual behaviour has an appreciable effect on a
population’s dynamics. They are thus useful in situations in which an understanding
is needed of how individual behaviour might affect the dynamics of a system. These
models are sometimes referred to as “agent-based” models with the “individual/agent”
being represented by either individual animals and plants, or composite units such
as fish schools or fishing fleets. They have typically been applied to investigate the
dynamics of a single population within the marine environment, but a number of
applications extend these analyses to consider multi-species dynamics as well (e.g.
Shin and Cury, 2001; Ginot, Le Page and Souissi, 2002; Ginot et al., 2006; Alonzo,
Switzer and Mangel, 2003; Kirby et al., 2004; Gray et al., 2003). Megrey, Hinckley
and Dobbins (2002) developed a visualization tool that can be useful in analysing the
outputs from IBM simulations, given that these are often voluminous and complicated.
Grimm et al. (2006) propose a useful standard protocol for describing individual-based
and agent-based models, although only minor mention is made regarding higher-level
entities such as communities consisting of populations. Attention is focused here on
the multi-species individual-based model OSMOSE (Object-oriented Simulator of
Marine ecOSystem Exploitation) (Shin, Shannon and Cury, 2004) and the agent-based
ecosystem model INVITRO (Gray et al., 2003; 2006).

2.3.1 OSMOSE
OSMOSE (Shin and Cury, 2001; Shin, Shannon and Cury, 2004) is a spatial
individual-based model that uses simple individual predation rules to model trophic
interactions. It is thus an excellent framework to explore the hypothesis that predation
is a size-based opportunistic process, depending only on size suitability and spatial
co-occurrence between predators and their prey. Given the need as motivated in
this review for alternative representations of species interactions, OSMOSE has a
potentially important role to play as an alternative modelling approach that can help
to identify consistent patterns in attempting to understand the ecosystem effects of
fisheries (Shin, Shannon and Cury, 2004). It is however limited to some extent in this
regard, in that, for example, when comparing model outputs to those produced by
EwE, OSMOSE is initialized using ECOPATH-based estimates of biomass, annual
natural mortality and fishing mortality values (Shin, Shannon and Cury, 2004). This
 Review of current modelling approaches 31

constrains OSMOSE somewhat in the extent to which it can posit an entirely different
ecosystem make-up. Also, estimates from one modelling approach are usually specific
to that approach and hence great caution should be taken when transplanting estimates
into another approach or even when assuming the same inputs.
The focus of OSMOSE is on piscivorous fish species, with fish schools moving in
a two-dimensional square-celled grid with closed overall boundaries. In the model,
fish move to adjacent cells with the highest biomass of potential prey. Plankton and
other invertebrate species are represented through a total carrying capacity term and
top predators such as marine mammals and seabirds are represented simply using an
additional natural mortality term.
As with the other multi-species models discussed, OSMOSE requires a large
number of input parameters in the form of growth, reproduction and survival
parameters. Some of these parameters are common to different species and ecosystems
which facilitates the parameterisation process. However, there are a number of
influential parameters upon which the model is based and the sensitivity of results to
alternative defensible choices needs to be examined. Specifically, the model assumes a
a minimal minimal
predator-prey sizesize
predator-prey ratio((� ) of 3.5 (the theoretical ratio between predator and
ratio
prey body lengths) (from Froese and Pauly, 1998) and that individual fish of all species
require 3.5g of food per body gram per annum (based on Laevastu and Larkins, 1981;
Gislason and Helgason, 1985; Longhurst and Pauly, 1987 – cited in Shin, Shannon
and Cury, 2004). The constant maintenance food ration assumption adopted here
needs to be borne in mind in interpreting model outputs because it does not account,
for example, for differences between species, for effects due to temperature or for
energetic differences of diverse prey types, or the potentially seasonal nature of major
feeding opportunities. However, a useful feature of the model is that the mean fish
growth rate depends on the quantity of food ingested and if this quantity falls below
the basic maintenance requirement, fish are assumed to die of starvation. A predation
efficiency((� i ))coefficient
A predation efficiency coefficient is computed
is computed basedbased
on theon thebetween
ratio ratio between the
the food ingested by
a group and the maximal ration requirement. When this falls below a critical threshold
level, the starvation mortality rate is modelled as a linear function of the predation
efficiency.
The values which are possibly the most problematic and difficult to obtain are those
for therelative fecundity((� S ))parameters
relativefecundity parameterswhich
whichareare input
input forfor each
each species and represent
the number of eggs spawned per gram of mature female. The reproduction formulation
is one of the simplest possible, with the abundance of recruits of species S at time t
(assuming an equal sex ratio) determined by simple linear proportionality:

AS

2 a =��
N S , 0, t �1 � �S SSBS , t with SSBS ,t � 1 BS , a ,t (29) (29)
a
MS

where a M is the age at maturity, A the terminal age for a species S, SSB is spawning
where
biomass and B is biomass. The current formulation does not permit exploration
of scenarios in which fecundity is a non-linear function of size. Instead of directly
modelling recruitment levels, these emerge from the annual survival of eggs and juveniles
based on modelled predation pressure and the carrying capacity term in the model. By
explicitly modelling predation pressure on fish larval stages, the model provides a useful
comparison with the results obtained from other modelling approaches. However,
without further development, it seems unlikely that OSMOSE will be accepted into the
realm of models contributing to practical fisheries management advice.
A similar age- and size-structured individual-based model termed MOOVES
(Marine Object-Oriented Virtual Ecosystem Simulator) (Colomb et al., 2004) is being
applied to the ecosystem of Guinea.
32 Models for an ecosystem approach to fisheries

2.3.2 INVITRO
Traditionally two main types of ecological models have been used: aggregate state
models (like EwE) and individual based models (such as OSMOSE). Formal separation
of these model types is not always easy. For instance, within the latter form of model,
the individuals may represent schools, patches of homogeneous ground cover, flocks,
patches of reef, or some other subset of a population that could be treated as equivalent
to an entity. From this it is clear that most aggregate state models can be seen as a special
case of an individual (or more properly agent) state model. Consequently, we can treat
aggregate state models as agents within an Agent-Based Model (ABM) system. This is
the approach that has been taken in INVITRO (Gray et al., 2006), which is currently
used as the basis for MSE-based studies focusing on the multiple-use ecosystem-level
management questions within the coastal waters of Australia (e.g. on the Northwest
Shelf of Australia, Little et al., 2006).
Until recently decision-based ABMs have usually been tightly focused on a small
subset of a system (e.g. a single fish in DeAngelis et al., 1991, or a small part of the
food web, as in Van Nes, Lammens and Scheffer, 2002). Advances in the use of hybrid
models, has (within the last five years) seen the incorporation of a wide variety of
ecosystem components into ABMs - facilitated by the coupling of classical dynamic
models, using differential equations and decision-based agents. In this way, the
best means of representing each ecosystem component can be used - for example in
INVITRO classical metapopulation models are used for habitats while IBMs are used
for higher trophic levels or species of conservation concern, such as whales.
To make this conjunction of aggregate state and individual-based models seamless,
INVITRO embeds them in a time-sharing universe. With each model-type (i.e. each
instance of an agent) allowed operating at the most appropriate time and space scales
– the scales that match the native resolution of the processes and their associated data
sets. Seasonal cycles, for example, do not adhere to time steps appropriate for tidal
larval migration. This treatment does have its consequences, not least of which was that
it demanded the development of a sophisticated (operating system-like) scheduler.
INVITRO includes a range of alternative agent types, which can be modularly
combined to create the final ecosystem (the open source nature of the code means
additional modules can also be written by interested users). Currently it contains
modules for three dimensional physical and environmental forcing (not just of
typical fields like temperature, light and currents, but also more unusual fields such
as catastrophic storms), larvae, mobile and sessile fauna from many trophic levels
(including top predators), primary producers, biogenic habitat (such as reefs, seagrass
beds and mangrove forests) a wide range of human activities (including commercial
and recreational fishing, nutrient pollution, salt extraction, shipping, tourism, coastal
development, conservation and oil and gas exploration) and their associated assessment
and management tools (including standard options like spatial management, but also
more hypothetical structures such as alternative management institutions that may
be confined to single sectors or span across multiple sectors). The behaviour and
representation of each agent is specific to its type. Consequently, mobile agents may be
represented as individuals (e.g. turtles and sharks), or small groups (e.g. schools or sub-
populations of fin-fish and prawns), while sedentary habitat-defining agents represent
entire patches (e.g. an entire reef complex).
While this array of agent types is fairly comprehensive (and allows for immense
flexibility) the computational costs of constructing an ecosystem in this way mean that
in practice an MRM approach is taken to model structure, with only a subset of the
ecosystem that incorporates the dominant system components included explicitly in the
model. To date this has meant that only the commercially valuable fish and crustaceans,
top predators, species of special interest (e.g. vulnerable species such as turtles), benthic
communities (or forage communities if in the pelagic system) and primary producers
Review of current modelling approaches 33

have been included. Ongoing work will see a wider set of “supporting” species
included, but it is unlikely that the complete coverage offered by EwE or ATLANTIS
will ever be possible. ABMs are also faced with all the same complexity, uncertainty
and interpretation issues as the other forms of ecosystem models.

2.4 BIOENERGETIC MODELS

A separate suite of models include those based on bioenergetic and allometric
reasoning, which involves parameterising a model using power functions of individual
body mass (Yodzis and Innes, 1992). Yodzis (1998) used a 29-species foodweb model
incorporating allometric reasoning to investigate the effects of a reduction of fur seals
on fisheries in the Benguela ecosystem. However, the model structure implemented
was arguably too linear and lacked age-, spatial- and seasonal structure.
More recently, an improved bioenergetics model has been constructed to describe
interactions between squid, anchovy, hake and sea lions off the Patagonian shelf (Koen-
Alonso and Yodzis, 2005). They used a system of four ordinary differential equations,
with basal equations to model squid and anchovy and consumer equations for hake and
sea lions. The form of equation used for a consumer is very general and could readily
be adapted for other systems:

dBj  
 B j   T j   ekj Fkj    Bi F ji  m j B j   j B j j  H j �
v
� ����������������
(30)
dt  k  i
where:
Bj is the biomass of consumer species j;
�0.25
(modelled as T j � aT j w j
Tj is the mass-specific respiration rate of species j (modelled with
with aT j an allometric coefficient and wj the mean individual biomass
with
of species j);
ekj is the assimilation efficiency for species as a predator j when feeding on prey k;
Fkj is the functional response (i.e. amount of prey species k consumed
by predator species j per unit of time);
mj is the “other natural mortality” rate of species j (due to species not
explicitly included);
Hj is the harvest rate of species j; and
u j , v j are constants specifying the density dependence in other natural mortality.

The density-dependent mortality form can be used to represent strong nonlinearities

in mortality rate, for example as a function of density due to overcrowding of sea lion
colonies during the breeding season (Koen-Alonso and Yodzis, 2005). A particularly
useful feature of the differential equation (30) above is that it is easy to substitute
different functional response variants using the general form derived by Koen-Alonso
and Yodzis (2005):
C ij (31)
Fij � (31)
1 � � hij C ij
i

where
hij is the handling time per unit of prey i and
Cij is the capture rate of prey i by predator j, the formulation of which
varies depending on the functional response assumed.

Difficulties in achieving management-quality multi-species models

Koen-Alonso and Yodzis (2005) stressed the importance of correctly specifying the
form of the functional response and experimented with five different formulations (see
34 Models for an ecosystem approach to fisheries

Table A1a-d). Apart from the allometry-derived parameters, they estimated the model
parameters by minimising the negative log-likelihood for observed (from a database
compiling all the time-series data) biomasses. Particularly commendable is that, unlike
most of the multi-species models presented, they attempted a detailed analysis of
parameter uncertainty using the sample-importance-resample (SIR) algorithm (Punt
and Hilborn, 1997; McAllister et al., 1994). The major contribution of this approach
thus far resides in it having highlighted the dangers of drawing definitive conclusions
from a single model structure.
The Koen-Alonso and Yodzis (2005) multi-species trophodynamic modelling
approach is both time-consuming and data intensive, but is a useful tool in systems where
biomass (and catch) estimates are available for a subset (at least) of the ecosystem. Bjørge
et al. (2002) present another data intensive approach that uses a combined Geographic
Information System GIS and energetics modelling approach. They used radio-tracking
data to construct an energetics simulation model of a population of harbor seals in
Norway. By integrating their results into a GIS model, they were able to analyse the
co-occurrence of fishing operations and seals. They showed that harbor seal predation
probably negatively impacted some fisheries but had a positive effect on shrimp catches
due to the removal of benthic-feeding fishes by seals. More recently, Cornick, Neill
and Grant (2006) used a bioenergetics modelling approach to project Steller sea lion
(Eumetopias jubatus) population trends under various scenarios of walleye Pollock
harvest. Their model included a sea lion life history component, a sea lion bioenergetics
component and a groundfish energetic component. The last component did not explicitly
model the groundfish population – instead it converted randomly-drawn standing stock
biomass into energy available to the Steller sea lions. It provides an interesting example
of a tailored approach including only as much detail as required to address a specific
question. Their simulations were unable to produce energy deficits sufficient to account
for the observed declines in the western US stock of the Steller sea lion.

2.5 CCAMLR MODEL DEVELOPMENT

2.5.1 Predator-prey models

The adoption of the Convention for the Conservation of Antarctic Marine Living
Resources (CCAMLR) and particularly Article II thereof (for a discussion of the
implications see, e.g. Butterworth, 1986), was a crucial step forward in acknowledging
the importance of maintaining the ecological relationships between harvested,
dependent and related populations of marine resources. Krill is the primary food
source of a number of marine mammal species in the Antarctic and concern has been
expressed that a rapidly expanding krill fishery might negatively impact (retard) the
recovery of previously overexploited populations such as the large baleen whales of
the Southern Hemisphere.
Predator-prey modelling procedures have been developed through CCAMLR
to assess the impact of Antarctic krill harvesting on krill predator populations and
to explore means of incorporating the needs of these predators into the models that
are used for recommending annual krill catch levels. Initial modelling procedures
estimated the level of krill fishing intensity that would reduce krill availability and
hence the population of a predator to a particular level (Butterworth and Thomson,
1995; extended in Thomson et al., 2000). More recently models such as KPFM, EPOC
and SMOM have been developed to consider these krill predation issues (see more
details below). Hill et al. (2006) also present a recent review of models pertaining to
the Southern Ocean.
A particular concern in CCAMLR has been the potential negative effects of
concentration of krill fishing in the vicinity of land-based predator breeding colonies,
for which the foraging ranges of parents are necessarily restricted. Mangel and Switzer
Review of current modelling approaches 35

(1998) developed a model at the level of the foraging trip for the effects of a fishery
on krill (Euphausia superba) predators, using the Adelie penguin (Pygoscelis adeliae)
as an example. Their approach of incorporating advection and diffusion processes in
a spatio-temporal framework to model krill availability in relation to the location of
breeding colonies could usefully be extended and applied to situations involving seal
populations. Given the large interannual fluctuations observed in krill biomass, these
models may also need to include the capacity to incorporate physical forcing of prey
dynamics (Constable, 2001; Atkinson et al., 2004). Alonzo, Switzer and Mangel (2003)
have developed a model using individual-behaviour to predict the indirect effects of
Antarctic krill fisheries on penguin foraging.
In general, initiatives such as these pursued under CCAMLR recognize the need to
balance the needs of predators with the socio-economic pressures underlying fishery
harvests.

2.5.2 KPFM (Krill-Predator-Fishery Model)

The krill–predator–fishery model (KPFM) of Watters et al. (2005, 2006) is being
developed specifically to address options for subdivision of the precautionary krill
catch limit in the Antarctic Peninsula region (Statistical Area 48) amongst SSMUs
(Small Scale Management Units) with areas in the range 104 to 93x104 km2. The model
is a whole ecosystem model in that it can be used to investigate the roles of transport,
production, predation and harvesting, but it also resembles a MRM in some aspects as
it focuses on aspects considered to be most important rather than fully specifying the
entire spectrum of ecosystem processes and species. The model is spatially resolved
to the level of SSMUs and surrounding oceanic areas and it uses a transition matrix
approach to model the transport of krill between areas (Watters et al., 2005). Spatially-
explicit delay-difference models are used to describe krill and predator population
dynamics. In the model krill populations are split into juvenile and adult stages and
predators are split into juveniles, breeding adults and non-breeding adults. The model
is currently set up to include from one to four stocks of predators per spatial cell. These
are typically generic seals, penguins, whales and fish, but specific rather than generic
groups may be included instead. Recent modifications (KPFM2) include extensions to
represent seasonality and a structure for allowing predators to move between SSMUs
(Watters et al., 2006). The model has an interesting formulation pertaining to the
way in which predator recruitment (but not survival) depends on krill consumption.
Associated work has focused on compiling data and input parameters for ecosystem
dynamics models of the region (Hill et al., in press), facilitating the comparison of
outputs from different modelling approaches (see below).
A notable feature of the approach is that a Monte Carlo simulation framework is being
used to integrate the effects of numerical uncertainty (Watters et al., 2005, 2006). Multiple
simulations employing alternative assumptions are run to assess structural uncertainty.
Performance measures are being developed both to evaluate catch-allocation procedures
and to assess tradeoffs between predator and fishery performance.
The model thus has a number of very useful features, but also some disadvantages
such as that krill in transit between SSMUs do not suffer predation and fishing
mortalities and the delay-difference dynamics do not capture full age-structured
complexity. An important assumption that is being tested and is a big unknown in the
model is the extent to which predators and the fishery are equal competitors and hence
are equally efficient at competing for limited resources.
The KPFM will permit evaluation of a wide range of management options that
account for the needs of other species when deciding krill catch limits in different
regions. This modelling work is being complemented to some extent by a krill flux
model (Plagányi and Butterworth, 2005b) that is currently being developed to quantify
the flux of krill past islands in the Antarctic Peninsula region and by the SMOM
36 Models for an ecosystem approach to fisheries

described below. CCAMLR (2006) noted the broad agreement in trajectories between
SMOM and KPFM2 in simulation trials when the parameterisation of the two models
was consistent, increasing confidence in these modelling approaches for evaluating
different fishing options.

2.5.3 EPOC model (Ecosystem Productivity Ocean Climate Model)

An Ecosystem Productivity Ocean Climate (EPOC) model (Constable, 2005, 2006),
initially applied only to krill, is being developed using an object-oriented framework
built around the following modules: (i) biota; (ii) environment; (iii) human activities;
(iv) management; (v) outputs, and (vi) presentation, statistics and visualization. Each
element within a module is an object carrying all its own functions and data. It is thus
designed to be a fully flexible plug-and-play modelling framework in response to a
need to easily explore the consequences of uncertainty in model structures as well as
widely varying knowledge on different parts of the ecosystem. The model is being
set up to easily examine the sensitivity of outcomes to changes in model structures,
not only in terms of the magnitudes of parameters but also in the spatial, temporal
and functional structure of the system. An added advantage is that within the same
simulation, different species can be modelled at different spatial and temporal scales as
well as with different biological and ecological levels of complexity (Constable, 2005).
The model is currently being used for developing a Heard Island whole ecosystem
model that will also include oceanographic features (A. Constable, Australian Antarctic
Division, pers. comm.).

2.5.4 Mori and Butterworth multi-species model

Mori and Butterworth (2004, 2005, 2006) developed a model to investigate whether
predator– prey interactions alone can broadly explain observed population trends
in the Antarctic ecosystem since the onset of seal harvests in 1780. The final model
components include krill, four baleen whale (blue, fin, humpback and minke) and
two seal (Antarctic fur and crabeater) species in two large sectors of the Antarctic.
The Atlantic/Indian and Pacific sectors are differentiated because of much larger
past harvests in the former, which consequently shows far greater changes in species
abundances in the model outpat. Unlike most of the other models discussed, the
Mori and Butterworth krill-whale-seal model is fitted to available data on predator
abundances and trends, whilst acknowledging that these data are not without their
problems. The model is successful in explaining observed population trends in the
Southern Ocean on the basis of predator– prey interactions alone, though some
difficulties were encountered.
Early model versions (Mori and Butterworth, 2004) considered baleen whales and
krill only, but an important finding was that it is necessary to also consider other
species in order to explain observed trends. In particular, crabeater seals appear to play
an important role.
The model equations were constructed to be as simple as possible whilst still
capturing the important population dynamics features. The dynamics of krill are
described by (Mori and Butterworth, 2006):

� � B ya � � � �N
� j B ya
n j ,a
B ya ��1 � � �� �
� �Bj � � �B �
y (32)
B ya�1 � B ya �r a
(32)
� �K a � �� a n a n
� � �� j y

where:
where:
B ya isis the
the biomass ofof krill
krill in
in region
regiona in year y;
r aa isisthe
theintrinsic
intrinsic growth
growth rate of of krill
krill in
in region
regiona;
is the intrinsic growth rate of krill in region
K a is the carrying capacity of krill in region a;
� j isisthe
themaximum
maximum per capita consumption
consumptionrate rateofofkrill
krillby
bypredator
predatorspecies
species
j;
 Review of current modelling approaches 37

Bj a is
is the krill
krill biomass
biomasswhen
whenthe
theconsumption
consumption and
and hence
hence alsoalso birth
birth raterate of species
of species j
drops to half
in region of itstomaximum
a drops half of its level, and level; and
maximum
N yj , ais the number of predator species j in region a in year y.
The same basic equation is used to describe each of the predators:

N yj�, a1 � N yj ,a �
� �
� j N yj , a B ya
n

� M j N yj ,a � � j ,a
�N �
j ,a 2
� C yj , a (33) (33)
�Bb � � �B �
y
a n a n
y

where
where
� j isisthethemaximum
maximum birth
birth rate of predator
predator species
speciesj;
M j is the natural mortality of predator species j in the limit of low population
size;
� j ,a is
is aa parameter
parameter governing
governing the
the density
density depende
dependence of natural mortality and
birth (and calf survival) rate for predator species j in region a;
n is a parameter that controls whether a Type II or a Type III functional response
is assumed (n=1 for Type II and n=2 for Type III), and
C yj ,a is
is the
the catch of predator
predator species
speciesj in region a in year y.
j
A likelihood function was maximized d totoestimate
estimate the
the parameters j ,a
parameters M jj,, N1780
1780,,
� j,, � j for all the
a
for all the predator species, jand
and r for krill. Ka can be calculated analytically from the
for krill.
relationship
between the other parameters underunder
between the other parameters the assumption
the assumption th that all the species
considered in the model were in equilibrium (balance) in year 1780, which corresponds
to the co-existence equilibrium level for the species considered. An intra-specific
a-specific density-dependent parameter ((� )) for
density-dependentparameter for each
eachpredator was input to admit a non-trivial
coexistence equilibrium of the species considered. These terms essentially reflect the
impact of limitations of breeding sites for seals and intra-species competition effects for
whales (Mori and Butterworth, 2005). Through taking account of density dependent
effects on feeding rates, model results suggest that Laws’ (1977) estimate of some
150 million tons for the krill “surplus” resulting from the heavy depletion of the larger
baleen whale species in the middle decades of the 20th century, may be appreciably too
high.
The Mori and Butterworth model structure is reproduced here because it is a simple,
pragmatic and self-consistent method that could be adapted for other systems as a
useful starting point to understand trophic interactions. It could also be linked to an
environmental effects module. One disadvantage of the model in its current state is
that it is age-aggregated rather than age-structured, which can, inter alia, result in use
of inappropriate input values for some parameters, as these likely better correspond to
age-structured model constructs (Mori and Butterworth, 2004). The model also focuses
on broad trends and hence lacks the smaller scale spatial structure that is required to
address questions concerning options for subdivision of the precautionary krill catch
limit amongst SSMUs.

2.5.5 SMOM (Spatial Multi-species Operating Model)

The Spatial Multi-species Operating Model (SMOM) (Plagányi and Butterworth,
2006 a&b) builds on the modelling work of Thomson et al. (2000) and Mori and
Butterworth (2004, 2006) described above. The model includes 15 SSMUs and uses
an annual timestep to update the numbers of krill in each of the SSMUs, as well as
the numbers of predator species in each of these areas. The model currently includes
four predator groups (penguins and seals, fish and whales) but is configured so that
there is essentially no upper limit on the number of predator species which can be
included. Given the numerous uncertainties regarding the choice of parameter values,
a Reference Set is used in preference to a single Reference Case operating model (see
e.g. Plagányi et al., 2007, Rademeyer, Plagányi and Butterworth, 2007). The initial
38 Models for an ecosystem approach to fisheries

Reference Set used comprises 12 alternative combinations that essentially try to bound
the uncertainty in the choice of survival estimates as well as the breeding success
relationship. Stochastic replicates are produced to explore different hypotheses such as
those related to the transport of krill.
SMOM is intended for use as an operating model in a formal Management Procedure
(MP) framework. Different MPs are simulation tested with their performances being
compared on the basis of an agreed set of performance statistics which essentially
compare the risks of reducing the abundance of predators below certain levels, as
well as comparing the variability in future average krill catches per SSMU associated
with each MP. CCAMLR (2006) has encouraged the further development of spatially-
explicit management frameworks and the development and evaluation of operating
models and decision rules for adjusting fishing activities (e.g. catch limits) based on
field data in the future.
 39

3. Comparison of models

3.1 LEVEL OF COMPLEXITY AND REALISM

There is a wide range in the levels of complexity of the 20 modelling approaches
considered here (Tables A1-A4, Figures 1-3). Most of the models may be categorized
as of the MRM-type, with only EwE and ATLANTIS representing the full trophic
spectrum (Figure 2). There is typically a trade-off between the range in trophic levels
considered and the corresponding detail with which each group is represented – for
example, in practice EwE models cannot represent the full age-structure of all groups
whereas models built using a restricted subset only of the ecosystem may include very
detailed length/age structure information (e.g. GADGET).
It was not considered practical or feasible to list model parameters in detail for all
20 modelling approaches. However, entries in Table A2 are intended to give a rough
idea of the sorts and numbers of parameters required for each model. By their nature,
ecosystem models are parameter- and data-hungry. It is sometimes argued that single-
species assessment models contain as many or more parameters. However, these
parameters are typically estimated by fitting to data and it is relatively straightforward
to test sensitivity to alternative values. The difficulty with considering multi-species
effects is that the field is still wide open in terms of understanding of the functional
forms of interaction and the availability of data to specify or estimate many of the
parameter values is limited. In the synthesis presented here, attention is drawn to
selected parameter values to which it is difficult to ascribe values conclusively.

3.2 FUNCTIONAL RESPONSE FORMULATIONS

The different functional form of interactions in EwE’s foraging arena (per-capita
consumption by a predator decreases with the overall abundance of that predator)
compared to MSVPA’s (and other models’) constant ration model (per-capita
consumption is set equal to the predator’s required daily ration) for predator feeding
has important implications for model behaviour and predictions. It tends (desirably) to
damp the large amplitude oscillations in population size that are frequently predicted
by multi-species models (see, for example, Mori and Butterworth, 2004). However, this
has additional consequences as detailed below.
Butterworth and Plagányi (2004) contrast the assumptions of the MSVPA (and its
associated derivatives that provide projections) and ECOSIM approaches, which they
categorize as “efficient predator models” and “hungry predator models” respectively.
MSVPA assumes that a predator is always able to consume its desired daily ration of
food.IfIf N j isisthe
daily ration of food. thenumber
number of
of predators of species j and the number of their prey species
ii (( N i ) is
is kept
keptfixed,
fixed,then
thenFigure
Fig. 4a
4a shows the implication of the MSVPA assumption
tion for for
how how thethe
total
totalconsumption rate Qij of prey i by predator j grows as the number of
consumptionrate
predators increases: linear proportionality.
On the other hand, ECOSIM is based upon the foraging arena model (Walters,
Christensen and Pauly, 1997) (Equation 5) which leads to the form of relationship
lationship between consumption rate Qij and thenumber
between total consumption predators N j as shown in
numberofofpredators
Figure 4b. When used in combinations, MSVPA and ECOSIM can possibly make a
first attempt at bounding the likely impact on a fishery of, for example, a reduction in
seal numbers in that, based on the assumed forms of interaction, the former approach
is likely to overestimate the effect and the latter to underestimate it (at least when
using default or low vulnerability settings) (Plagányi and Butterworth, 2005). The data
40 Models for an ecosystem approach to fisheries

FIGURE 4
Schematic showing how the total consumption rate of prey species i by predator
species grows as the number of predators increase for the two contrasted cases: a)
pecies � gr
MSVPA (and its derivatives providing projections) showing a linear proportionality
nd its de
relationship and b) ECOSIM’s foraging arena-based model (shape shown typical for
default parameters) in which the total rate saturates at a constant level for high
numbers of predators (from Plagányi and Butterworth, 2004, 2005a)

(a)
CONSUMPTION RATE

(b)

NUMBER OF PREDATORS Nj

hungry nature of MSVPA does not necessarily preclude the use of MSFOR to predict
forwards, provided the model is initialized using sensible assumptions based on at least
some data (see IWC, 2004a).
Walters et al. (2000) advance two arguments to support the foraging arena over
the constant ration model, namely that satiation is rare in nature “predators with
full stomachs are not a common field observation” (Walters and Kitchell, 2001) and
that handling time effects are trivial in the field because if animals increased their rate
Comparison of models 41

of effective search to the extent where handling time became an issue, they would
be exposed to additional risk of predation hence they avoid doing this. Walters and
Martell (2004) explain further that the basic idea of EwE’s foraging arena theory is that
marine species have limited access to prey resources because of spatial habitat-choice
behaviours aimed at moderating their predation risk. The IWC (2004a) describes the
biological underpinnings of the foraging arena model as “controversial and uncertain”
because there appears to be little observational evidence to distinguish the two
models.
One of the key issues in moving the development of multi-species models forward
is thus the appropriate form of the functional response formulations to be considered
in the models. At opposite extremes, formulations such as that used by ECOSIM
depict per-capita consumption by a predator as decreasing with the overall abundance
of that predator, whereas constant ration formulations (such as that used in MSVPA
approaches) set per-capita consumption as equal to the predator’s required daily ration.
It is strongly recommended that effort be focused on appropriate data collection
and/ or experiments to assist in shedding light as to the most appropriate choice of
model form to represent feeding behaviour. Fenlon and Faddy (2006) argue that rather
than using mechanistic models to interpret data from predator-prey systems, simple
logistic regression analyses are more consistent with the data and take stochastic
variation into account. They present some models for dealing with over-dispersion,
including one based on the beta-binomial distribution which is shown to provide a
better fit to experimental data.
However, extrapolations from the microscale to the macroscale require integrating
the form of a functional response over the area concerned and independent estimates
of parameters at the microscale will not necessarily remain appropriate if the same
functional form is assumed to govern macroscale behaviour. Experimental estimates
of suitability often refer only to the microscale, but multi-species models require
parameter values that reflect effective responses at the macroscale level (Lindstrøm
and Haug, 2001). Reliable integration of microscale estimates of suitabilities over
the spatio-temporal distributions for both predators and prey to provide macroscale
parameter values, is likely a realistic objective for the longer term only; in the shorter
term, regression approaches will probably be needed to attempt to relate macroscale
changes in diet to variations in prey abundance. Studies comparing the performance or
predictions of models representing processes at different scales and/or with different
levels of spatial aggregation can also be informative (Fulton, Smith and Johnson,
2003a).
Most multi-species models utilize a hyperbolic (Type II) functional relationship
(Jeschke, Kopp and Tollrian, 2002; Mackinson et al., 2003). Although difficult to
implement because additional parameters need to be estimated, a sigmoidal (Type
III) functional response is likely more appropriate when modelling generalist
predators, such as whales (Mackinson et al., 2003). This is because these predators
are generalists and hence exert less of a strong effect on depleted prey stocks, as can
be depicted using a sigmoidal relationship. Given model structural uncertainty due
to a paucity of knowledge on functional responses, definitive conclusions cannot be
drawn from models based on a single structure (Koen-Alonso and Yodzis, 2005).
However, the biomass of available food is often such that it spans a limited section
of the functional response curves where they are all very similar so that it is hard to
differentiate between alternative representations, unless there exists some form of
extreme or transient conditions either temporally or spatially (Walters 1986, Fulton,
Smith and Johnson, 2003b). Ideally, evaluations to provide advice on the impact of,
say, the effects of fishing a predator on fisheries for prey species should not be based
on a single representation of species interactions; but rather the robustness of results
across a range of plausible functional forms needs to be considered. Bayesian methods
42 Models for an ecosystem approach to fisheries

are a useful tool for taking account of variability in and uncertainty about feeding
relationships.

3.3 WHOLE ECOSYSTEM MODELS VS MRMS

As highlighted by an international review panel at the 2004 BENEFIT Stock
Assessment Workshop (BENEFIT, 2004), the choice of which multi-species models to
use needs to be linked to scientific goals and/or management objectives. For objectives
related to broad-scale questions regarding the structure of the ecosystem, ECOPATH/
ECOSIM models might be used; other models may be more appropriate for more
specific questions. Unlike EwE, individually tailored approaches such as MRMs have
more flexibility in modelling the dynamics of marine predators, but usually ignore any
potential effects that changing prey populations may have on the predators themselves.
Fulton and Smith (2004) strongly recommend that ideally a suite of different
“minimum-realistic” ecosystem models should be constructed and their results
compared. However, given limited person-power and pressure to produce results, it is
important first to engage in discussions regarding which are the preferred modelling
approach/es to be pursued in each context. Thus, for example, as a first attempt to
address hake multi-species interactions, the 2004 BENEFIT Workshop recommended
that existing models should be adapted to provide estimates of the predation mortality
on hake that is generated by the two hake species. Similarly, CCAMLR has tended to
consider simpler predator-prey type models for the Southern Ocean (e.g. Thomson et
al., 2000).
Nevertheless, whole ecosystem models clearly have an important role to play, given
that few of the other models discussed are suitable for exploring broader ecosystem
questions (Figure 3, Table A4). While predictive multi-species population models may
have limited impact on management decisions in the short-term, if only because of
considerations of lack of data, model complexity and uncertainty and research costs,
there are some initiatives that are being pursued with the information that is at hand
at present. It may be instructive to investigate possibilities of closer links between
ECOPATH data inputs and single-species stock assessment models. In considering
ECOPATH’s potential to contribute to single-species models, there is a need to pursue
the question of whether the constraints provided by the ECOPATH mass-balance
equation appreciably reduce uncertainties associated with single species models. The
mass-balance relationships of the ECOPATH approach (Christensen and Pauly, 1992)
provide some information beyond that conventionally incorporated in single-species
assessments and do so essentially independent of concerns about how best to model
the functional forms of species interactions. Preliminary computations (Somhlaba et
al., 2004; Somhlaba, 2006) suggest that for the Benguela system, the precision of single-
species assessment estimates is unlikely to be improved through taking account of mass-
balance constraints. On the other hand, outputs from single-species stock assessment
models may have some utility for improving biomass and productivity estimates (and
their associated variance estimates) used as inputs to ECOPATH and hence ECOSIM.
Recent additions to the EwE software (Christensen and Walters, 2004) mean that it is
possible to include more life history stages in ECOSIM models.
Butterworth and Plagányi (2004) suggest that until “Whole Ecosystem” approaches
have been shown to demonstrate adequate robustness in their predictions to
uncertainties in input data and alternative plausible choices for the functional forms of
interactions between species, they should have lower priority than the development of
Minimum Realistic Models, given an aim of providing inputs on say catch levels of a
target species. They argue that in the context of providing fisheries management advice,
MRMs would seem the obvious first step to take in the process of moving from single-
species models to the extremely ambitious and demanding aim of a reliable predictive
model for all major ecosystem components. On the other hand, depending on the
Comparison of models 43

nature of the question, whole ecosystem models may be the only suitable tool to use,
particularly when management strategies other than simple TAC application are being
considered. Ecosystem-based management is still in its infancy and hence there is as
yet no consensus on what are the most appropriate management tools. In many areas
there is the realisation that TACs spatial or temporal are unlikely to be appropriate (or
feasible) for all species and that other tools such as closures and gear mitigation devices
may need to be called upon (E. Fulton, pers. comm.). In this context, multi-species
and ecosystem models have a large role to play in assessing the utility of these tools
and even the effectiveness of proposed monitoring schemes or indicators (e.g. Fulton,
Smith and Punt, 2005).

3.4 ADVANTAGES, DISADVANTAGES AND LIMITATIONS

Selected advantages, disadvantages and limitations of the 20 modelling approaches
considered are listed in Table A3. This is by no means a comprehensive list and it would
be instructive for future studies to expand this list. In its current form, it provides a
rough overview of some of the strengths and weaknesses of the different approaches.
 45

4. Potential of tools to address

multi-species research questions

In reviewing the methods available for assessing the impacts of ecological interactions
between species and fisheries, it is important not to lose sight of the aims of the various
approaches. In the current (fisheries management) context, most of the questions
to be addressed by multi-species/ecosystem models fall under one of the following
headings.
1. Understanding ecosystem structure and functioning, e.g. relative roles of top-
down and bottom-up processes.
2. What is the impact of a target fish species on other species in the ecosystem? For
example, does the removal of the target species negatively impact other species
which depend on it as prey (e.g. Gislason, 2003)? Bycatch issues are dealt with
separately under 13 below.
3. What is the effect on top predators of removing their prey? This question is
listed separately given that it is the focus of many multi-species studies. The
classic example is CCAMLR’s focus on the possible impacts on Southern Ocean
predators of an expanding krill fishery.
4. What is the extent of competition between marine mammals and fisheries (see
e.g. Trites, Christensen and Pauly, 1997; Harwood and McLaren, 2002; Kaschner,
2004; Plagányi and Butterworth, 2002; 2005a)? This includes consideration of both
“direct competition”, which involves reduction (by consumption or utilisation) of
a limited resource, but with no direct interactions between the competing species
(Clapham and Brownell, 1996), as when a marine mammal eats a fish that could
otherwise have been caught by a fisherman and “indirect competition” (e.g. Pauly
and Christensen, 1995) in which the competitors may target different resources but
these are linked because of a foodweb effect (e.g. when a marine mammal consumes
a fish that is an important prey species of a commercially desirable fish species).
5. What ecosystem considerations need to be taken into account to rebuild depleted
fish stocks?
6. Is the single-species-based assessment of the status and productivity of a target
species severely biased because of a failure to consider multi-species interactions
(e.g. Pope, 1991; Walters and Kitchell, 2001; Walters et al., 2005)?
7. Is there an ecologically or economically better way to distribute fishing effort in
an ecosystem? The focus here is, for example, on the extent to which different
species should be targeted so as to optimise use of the ecosystem both ecologically
and economically.
8. Are there relatively unexploited species in an ecosystem which could be targeted
without having a detrimental effect on other components of the ecosystem?
9. Is fishing on particular stocks driving the ecosystem to a less productive/less
desirable state (e.g. a new stable state or an adverse shift in marine communities
(Trites et al., 1999, Scheffer, Carpenter and de Young, 2005)?
10. Is the spatial and temporal concentration of fishing negatively impacting the
longterm viability of species such as land-breeding marine mammal predators and
seabirds? Should the spatial distribution of fishing effort be altered to account
for the needs of e.g. land-breeding predators. This includes consideration of, for
example, fishing exclusion zones and MPAs (see e.g. Dalton, 2004; Hilborn et al.,
2004).
46 Models for an ecosystem approach to fisheries

11. Effects of physical/environmental factors on the resources on which fisheries

depend.
12. Effects of habitat modification. This includes consideration of effects such as
trawling damaging benthic habitats and hence having an indirect negative effect
on fish stocks.
13. What are the impacts of bycatch?
14. Effects of the introduction of non-native species.
Naturally there is a large number of very specific questions that models have been
constructed to address and every (good) model is useful in the context for which it
has been designed. The list above is far from complete, but encompasses most of the
commonly phrased questions.
In terms of a broad overview of the usefulness of the different modelling approaches
discussed here, some preliminary suggestions are presented in Table A4 which
highlights those models considered by the report’s author to show the most potential
to address each of the questions above. This is not intended as the final word on the
subject, but rather as a starting point to compare the models with slightly more specific
aims in mind. Given that it can be argued that any ecosystem model contributes to
one’s understanding of the system, the models have been categorized as either showing
the potential to contribute to an understanding of the functioning of the ecosystem as
a whole or to a subset only, recognizing that both these aims are important in different
contexts. Glancing across the 20 modelling approaches considered in Table A4, it is
evident that collectively they cover all the research questions posed here, but that there
are fairly large gaps in the suitability of specific approaches to address subsets of the
questions posed above. Although the finer details of Table A4 can and indeed should
be further debated, the schematic presented here may be useful as a first step to assist in
choosing between models given specific EBFM research questions. Note that although
EwE, ATLANTIS and INVITRO emerge as the clear “winners” in terms of the range
of questions they are capable of addressing, a word of caution is necessary here because
that feature alone does not guarantee that they necessarily provide the best approaches
to address a specific issue.
The research question that emerged as most poorly addressed across all models was
that of the effects of habitat modification, with only ATLANTIS rating highly as a tool
in this regard (Table A4). ECOSPACE can also be used to evaluate the effects of habitat
modification and EwE has some potential for indirectly exploring aspects of this issue,
through trophic mediation. Although there are fairly straightforward examples of this
issue, less direct cases can be rather intractable (see e.g. Sainsbury et al., 1997, Auster
and Langton 1998). On the other hand, the deleterious effects of trawling have long
rung alarm bells (e.g. McConnaughey, Mier and Dew, 2000) and this may point to
a need for more focussed attention to address this issue – naturally in combination
with empirical studies. In contrast, Table A4 suggests that there has been a definite
increasing trend towards constructing models capable of being driven by physical and
other environmental variables. This may be in response to the indication that trophic
interactions are limited in the extent to which they can explain observed trends and
changes in the ecosystem.
A separate category altogether pertains to ecosystem models constructed with
the primary purpose of being used for model testing (e.g. Yodzis, 1998), comparison
purposes (e.g. Fulton et al., 2004) or in a simulation testing framework. An example of
the latter is the use of ATLANTIS as an operating model (see next section) to evaluate
the performance of ecological indicators (Fulton, Smith and Punt, 2004). These are
critical issues to be addressed and it is hoped that in future as much effort will be
focused on these questions as on the further development of new or existing models.
It is particularly useful to test ecosystem models such as EwE by generating simulated
data with known parameters using an operating model such as ATLANTIS. In testing
Potential of tools to address multi-species research questions 47

ECOPATH in this way, it was found that while useful for capturing snapshots and
giving great insight into ecosystem structure and potentially counter-intuitive system
responses in a “what-if” context, it was ill suited in the role of an assessment model
(Fulton, Smith and Punt, 2005; E. Fulton, pers. comm.). This was due to changing error
structures through time, the potential problems with data compatibility (particularly
when diet data was collected at a point in time that is distant from the time the biomass
estimates are made) and the potential to miss once rare links that can become important
if conditions change substantially (E. Fulton, pers. comm.). These are the same sorts
of problems likely to afflict most ecosystem models, highlighting the importance of
seeking the same thorough understanding of the limitations of ecosystem models as is
the case for single-species assessment models.
 49

5. Roles for models in operational

management procedure
development

Operational Management Procedure (OMP) (Butterworth, Cochrane and De Oliveira,

1997; de Oliveira et al., 1998, Butterworth and Punt, 1999), or Management Strategy
Evaluation (MSE) approaches (Smith, Sainsbury and Stevens, 1999), provide scientific
recommendations for management measures such as TACs, closures, gear modifications
and monitoring schemes. The OMP approach has the potential to complement
multi-species approaches through its focus on the identification and modelling of
uncertainties, as well as through balancing different resource dynamics representations
and associated trophic dependencies and interactions (Sainsbury, Punt and Smith,
2000). It has already been used in this role in Australia (Little et al., 2006; Smith et al.,
2004) and a spatial and multi-species MP is being developed for the Antarctic Peninsula
krill-predator-fishery system (Plagányi and Butterworth, 2006, a&b). Elsewhere in
the world attempts are increasingly been documented to incorporate bycatch, stock
structure and spatial aspects into MPs (e.g. Punt, Smith and Cui, 2002; Dichmont et
al., 2005).
OMPs typically involve both “Decision Models” and “Operating Models” (also
termed “Testing Models”). The former essentially integrate resource-monitoring
information (e.g. CPUE, survey indices of abundance) together with a control rule
to provide a scientific recommendation for management such as a TAC and thus do
not necessarily provide an accurate representation of the possible underlying resource
dynamics (Butterworth and Plagányi, 2004). In contrast Operating Models should
accurately reflect alternative possibilities for the true underlying dynamics of the
resource or resources under consideration. They may seek a high degree of realism and
hence may be quite complex (e.g. IWC, 2003; Fulton, Smith and Punt, 2004). Operating
models provide the basis for simulation testing to assess how well alternative candidate
Decision Models achieve the objectives sought by the management authority.
Butterworth and Plagányi (2004) speculate that there is clearly an immediate role
for ecosystem models as Operating Models, but that the development of tactical
ecosystem models as the basis for computing harvest limits within the OMPs
themselves still seems some time off. This is primarily because of the uncertainty
surrounding appropriate choices for the numerous parameter values and the functional
forms to describe species interactions. Cochrane (1998, 2002) and Sainsbury, Punt
and Smith (2000) note that it remains to be seen whether or not the associated
levels of uncertainty can be adequately constrained to yield scientifically defensible
and practically useful conclusions. Prior to the work of Fulton, Smith and Punt
(2004), the inclusion of ecosystem effects in OMP evaluation exercises was generally
implicit only. For example, rather than using a full multi-species operating model in
simulation testing of its Revised Management Procedure, the Scientific Committee of
the International Whaling Commission (IWC) used a simpler approach that allowed
for time-dependence in the intrinsic growth rate and carrying capacity parameters
to mimic the typical impacts on that population of changing levels of other predator
and prey species (IWC, 1989). OMP testing procedures for some key South African
resources have similarly used changes in single species parameters (such as K) as a
surrogate for ecosystem effects (Rademeyer, Plagányi and Butterworth, 2005) and
50 Models for an ecosystem approach to fisheries

attempts are underway to incorporate functional relationships between seabirds

and their prey into the operating models for sardine (Sardinops sagax) and anchovy
(Engraulis encrasicolus), with these in turn augmented by population dynamic model/s
for the predator/s of concern (Plagányi et al., 2007).
 51

6. Moving models forward – future

developments

This report has focused on describing many of the multi-species and ecosystem models
in their current form. However, in several cases, these are constantly evolving and there
is currently a global increase in the effort directed at developing ecosystem models.
This ranges from increasing attempts to extend single-species assessment models to
include additional important prey or predator species, to extending ecosystem models
to evaluate policy options for management.
The MSVPA/MSFOR class of models, initially applied to the ICES areas, were
some of the first multi-species approaches to be developed but are still being applied
and adapted (e.g. Livingston and Jurado-Molina, 2000). Hybrid versions (e.g. Mohn
and Bowen, 1996) have been developed and more recently MSFOR is bring rewritten
as MSM (Multi-species Statistical Model) (Jurado-Molina, Livingston and Ianelli, 2005;
Jurado-Molina, Livingston and Gallucci, 2005). Lewy and Vinther (2004) (see also
Lewy and Nielsen, 2003) are similarly developing a stochastic multi-species model that
takes account of uncertainties in catch-at-age, stomach content and other data.
Regarding other MRMs, there are plans to revise the original Punt and Butterworth
(1995) MRM of hake-seal interactions in the southern Benguela. The BORMICON
model has evolved into GADGET and the latter is currently still being developed
with case-studies having commenced only recently. A Mediterranean Sea model is
being developed and is the first attempt at including a very large number of species
in a GADGET model (see e.g. http://www1.uni-hamburg.de/BECAUSE/content/
case_study_5.html).
The pelagic ecosystem model SEAPODYM has evolved from the earlier SEPODYM.
Recent work has focused on running simulations at a global scale (with a resolution
of one month x 1° latitude x 2° longitude) and preliminary predictions have been
produced for the mid-trophic (forage) components, with a run covering 1860-2100,
using a Intergovermental Panel on Climate Change (IPCC) climate change scenario
for the coming century. New modules are on the table to be developed, first for marine
turtles and then for sharks, marine mammals or even small pelagics such as anchovies
and sardines. Similar advances are being made in other biological models tied to global
ocean models, such as NEMURO (Nishikawa and Yasuda, 2005; Kishi, Nakajima and
Kamezawa, 2005).
EwE has evolved considerably over the past few years and a large project is currently
underway to develop a new generation of EwE (see www.lenfestoceanfutures.org)
that will be fully modularized. A building-block version is to be created that will
facilitate construction of individually tailored versions (V. Christensen, University of
British Columbia, Canada, pers. comm.). The new version is scheduled for release by
September, 2007 and may substantially advance ecosystem-based fisheries management
by providing a readily accessible and easy to use tool capable of producing predictions
based on user inputs by managers and others.
Several hybrid EwE versions have already been constructed to date (e.g. Aydin
et al., 2002) and are being used as sensitivity analyses in stock assessments, for example
to address questions such as the potential impacts of a single-species TAC on other
species (K. Aydin, pers comm.). Given a growing appreciation of the need to consider
economic factors, one encouraging development is that of the GEEM (General
Equilibrium Ecosystem Model) (Tschirhart and Finnoff, 2003; Tschirhart, 2004,
52 Models for an ecosystem approach to fisheries

Eichner and Tschirhart in press) which combines multi-species and economic sector
modelling. The starting base is the same as ECOPATH, but GEEM incorporates
a novel approach to predict functional responses by allowing predators to make
“rational economic choices” based on the expected energetic gain from different prey
types (K. Aydin, pers. comm.).
The bioenergetic-allometric modelling approach of Koen-Alonso and Yodzis
(2005) is being extended to permit investigation of some of the potential effects of
temperature, with a longer term goal being the integration of economic considerations
into ecosystem-based management (Koen-Alonso, Northwest Atlantic Fisheries
Centre, Fisheries and Oceans Canada, pers comm.). Temperature-dependence is
being introduced into the dynamics based on the framework developed by Vasseur
and McCann (2005). This will permit initial investigations of the potential effects of
global warming through an analysis of, inter alia, the effects of temperature on basic
metabolic pathways.
Substantial progress has been made in coupling physical models to biological models.
Taking this one step further, others have argued for the importance of considering the
coupling between ecosystems – the meta-ecosystem approach (Loreau, Mouquet
and Holt, 2003; Varpe, Fiksen and Slotte, 2005). This is particularly important when
considering species such as salmon which migrate from oceanic feeding grounds to rivers
and lakes and species such as herring which migrate between feeding, overwintering
and spawning areas (Varpe, Fiksen and Slotte, 2005). In a similar vein, Vidal and
Pauly (2004) recently demonstrated how a number of local ECOPATH models can
be combined into a single integrated, spatially explicit large marine ecosystem (LME)
– scale model.
This idea of linking across systems is also helping to drive the current development
path of the Australian models ATLANTIS and INVITRO. While both are benefiting
from collaborative work that is expanding the ecological potential of the model, there
has been a growing focus on developing the socio-economic components and the links
to other ecosystem types (such as river catchments) so that broad flow-on and multiple
use management questions can be considered (E. Fulton, pers. comm.).
Nevertheless the development of moderately easy to use full meta-ecosystem
approaches that are useful to management seems some way off. Rather, it is likely that
there will be an increase in the trend to incorporate greater spatial detail into models, as
has been done in ECOSPACE and is being achieved with GADGET and ATLANTIS
for example. Considerable efforts need to be devoted to compile spatially-explicit or
GIS-based data to meet this aim. Parallel increases in computing power and efficiency
of numerical and optimisation methods seem a necessary prerequisite for further
developments on this front. GADGET appears to be a forerunner in terms of the use of
multiple computers to speed runtime as well as attempts to base multi-species models
on a robust statistical framework comparable to that used in single-species assessment
models.
There is an increasing interest in the use of ecosystem models as Operating Models
used to test OMPs. This is an excellent approach to providing a strategic and practical
framework for developing an operational ecosystem approach to management.
However, data limitations are likely to restrict the number of multi-species models that
reach the stage of being considered viable operating models to assist in the management
of target species. At the current level of development, most multi-species models
cannot provide quantitatively reliable predictions. However, if a variety of alternative
plausible models yield qualitatively similar predictions, this could provide a basis for
management response.
 53

7. Prudent use of the precautionary

principle

Given the difficulties of providing definitive scientific advice on stock status and
ecosystem “quality” and interactions, managers are increasingly called upon to apply
the precautionary principle or approach (FAO, 1995). The “Precautionary Principle”
(Principle 15 of the UNCED Rio Declaration (Agenda 21) of 1992) requires that
“where there are threats of serious or irreversible damage, lack of full scientific
certainty shall not be used as a reason for postponing cost-effective measures to prevent
environmental degradation” (FAO 1995) (see also Hilborn et al., 2001). However,
Plagányi and Butterworth (2005) argue that naive application must be avoided
because unsubstantiated claims and overstatements can damage scientific credibility.
Acknowledging the difficulties of providing definitive scientific advice on ecosystem
effects, arguments based on best scientific evaluations, rather than upon unsubstantiated
impressions of the state of a resource, may better safeguard the interests of scientific
credibility (and hence resource conservation) in the long run. Notwithstanding, it is
increasingly being recognized that at least some ecosystem-based management may
need to be based on qualitative considerations only.
 55

8. Pointers from previous studies

and workshops

Several factors have contributed to the current worldwide boom in developing multi-
species and ecosystem models to advise fisheries management decisions, with interest
in this topic evinced by a number of recent conferences on ecosystem considerations,
including the ICES-SCOR, 1999 ecosystem effects of fishing symposium in Montpelier,
France (ICES, 2000), the 2001 FAO expert consultation on ecosystem-based fisheries
management held in Reykjavik, Iceland (FAO, 2003b, see also Sinclair and Valdimarsson,
2003), the Workshop on the Use of Ecosystem Models to Investigate Multi-species
Management Strategies for Capture Fisheries (Fisheries Centre Research Reports
Vol. 10, no. 2, 2002), the IWC Modelling Workshop on Cetacean-Fishery Competition
(IWC, 2004a) and the 2002 Workshop on an Ecosystem Approach to Fisheries
Management in the Southern Benguela, held in Cape Town, South Africa (African
Journal of Marine Science 26, 2004). A number of policy documents have attempted
to set targets, establish universal definitions of terms such as an “ecosystem approach
to fisheries” or EAF (Garcia et al., 2003) and formulate guidelines to operationalise
EAF by suggesting ways of implementing it at a practical level (FAO, 2003a, b).
These initiatives date roughly from the 1982 UN Convention on the Law of the Sea,
to the influential 1995 FAO Code of Conduct for Responsible Fisheries and finally
to the somewhat ambitious 2002 World Summit on Sustainable Development which
“encourage (d) the application by 2010 of the ecosystem approach.” and set as a target
to “Maintain or restore stocks to levels that can produce the maximum sustainable yield
with the aim of achieving these goals for depleted stocks on an urgent basis and where
possible not later than 2015” (WSSD, 2002). Unfortunately the socio-economic reality
in most cases of resources well below their MSY level is that the large short-term catch
reductions needed to achieve anything other than a relatively slow rate of recovery are
very unlikely to be politically acceptable in many countries.

8.1 MODELLING INTERACTIONS BETWEEN MARINE MAMMALS AND FISHERIES

Butterworth and Punt (2003) argue that consideration of the indirect interactions
between marine mammals and fisheries is an appropriate starting point for developing
and testing multi-species models because of the lesser number of foodweb linkages
for apex predators. It is thus instructive to begin discussion with a fairly narrow
focus, namely that of bodies interested predominantly in a small subset of ecosystem
interactions, as it should in theory be easier to reproduce these than the full
spectrum of ecosystem interactions. The North Atlantic Marine Mammal Commission
(NAMMCO) has focused for a number of years on marine mammal-fisheries
interactions. For example, workshops have been convened to investigate the role of
minke whales, harp seals and hooded seals in the North Atlantic (NAMMCO, 1998),
the economic aspects of marine mammal–fisheries interactions (NAMMCO, 2001),
the main uncertainties in extrapolating from feeding behaviour or stomach contents
to annual consumption (NAMMCO, 2002) and to model marine mammal-fisheries
interactions in the North Atlantic (NAMMCO, 2003). Given the conclusion of the
first of these workshops, namely that marine mammals have substantial direct and
indirect effects on commercial fisheries in the North Atlantic (NAMMCO, 1998),
attention was focused on studies related to competition and the economic aspects of
marine mammal-fisheries interactions (e.g. NAMMCO, 2001).
56 Models for an ecosystem approach to fisheries

In light of uncertainties in calculations of consumption by marine mammals,

concrete recommendations were sought with regard to estimating this consumption
in the North Atlantic (NAMMCO, 2002). The next step was to review how available
ecosystem models could be adapted to quantify marine mammal-fisheries interactions
in the North Atlantic. The lessons learnt in this exercise provide a useful framework
in terms of assessing different multi-species models. NAMMCO (2003) listed the
following requirements as being particularly relevant in identifying the desirable
features of a multi-species modelling framework:
1)flexibility of functions for prey selection;
2)flexibility of age structuring (from fully age-structured to fully aggregated);
3)accessible code and transparent operation (not “black-box”);
4)able to be tailored to area and species of concern;
5)includes interactions accounting for most of the natural mortality, M, for species
of concern;
6)spatial and temporal resolution able to be tailored for target species; and
7)uncertainty in data and model structure reflected in results.
One of the conclusions arising from the most recent in this series of workshops
(NAMMCO, 2003) was that while the output from a model such as GADGET was
not expected to be able to predict all aspects of future states of the ecosystem, the
model was seen to have potential utility for management through testing scenarios
where abundances of target species are manipulated. In addition, the workshop
recommended the development of a generic (or “template”) North Atlantic model,
based on GADGET and including major fish and marine mammal species. The main
use of such a model was seen to be to identify the inputs which had the greatest effect
on model predictions and hence to guide research priorities in different regions each
subject to different deficiencies in data.
Plagányi and Butterworth (2005) assessed a number of models in terms of these
seven requirements, as well as the additional requirement that marine mammals be
explicitly included, rather than treated as exogenous components. They concluded
that GADGET and Minimally Realistic Models (MRM), such as the approach of Punt
and Butterworth (1995), show the most promise as tools to assess indirect interactions
between marine mammals and fisheries. Bioenergetic/allometric modelling approaches
such as that of Koen-Alonso and Yodzis (2005) have a role to play too in attempting
to characterize the finer details of these interactions. Given that the Antarctic marine
ecosystem could be viewed as a case on its own, further development of the suite of
CCAMLR predator-prey models (essentially also MRM-type models) is considered
the most appropriate approach for this region. The importance of applying different
modelling approaches to the same system is stressed (provided that appropriate
resources, in terms of both person power and data, are available). This is particularly
useful for qualitative cross-checking to determine whether different approaches give
similar results and therefore gauging how much confidence can be placed in their
reliability. Furthermore, given the importance of comparing the outputs of different
modelling approaches as well as the need to test model predictions both against
simulations and against reality, the suggestion has been made that there needs to be an
internationally-coordinated effort to provide a structure within which model testing
can take place (I. Boyd, University of St Andrews, pers. comm.).
An appreciation for the need to understand the assumptions underlying each
model considered emerged from both the NAMMCO workshop on modelling
marine mammal–fisheries interactions (NAMMCO, 2003) and the IWC workshop on
cetacean-fishery competition (IWC, 2004a). Both meetings stressed the need for:
• careful consideration as to whether or not underlying model assumptions are
appropriate for the case under investigation;
Pointers from previous studies and workshops 57

• tests of the sensitivity of predictions to alternative assumptions, particularly

regarding interaction terms (e.g. Vasconcellos and Gasalla, 2001, Mackinson et al.,
2003); and
• addressing uncertainty, in particular by focusing research on the discrimination of
alternative assumptions that yield appreciably different predictions.

8.2 AREAS OF FOCUS

A further pragmatic recommendation from the IWC workshop (IWC, 2004a) was that
modelling efforts should focus on specific areas/systems where there is the greatest
chance of success. Given a choice of systems to model, it does seem sensible to start
with the “easier” cases, but naturally practical realities may mean that analyses are
needed for more “difficult” areas/systems. Key characteristics of systems proposed
for initial focus included reasonable data availability, relatively simple foodwebs,
strong species interactions, relatively closed system boundaries and low (or obvious)
environmental forcing (IWC, 2004a). One ideal ecosystem for such investigations is
the Barents Sea, where there is evidence of relatively tight predator-prey coupling
with only a few fish species (herring, cod and capelin) playing key roles. Systems
characterized by strong physical forcing (bottom-up control) are likely to show little
or no response to the removal of predators because even strong trophic interactions
may be insufficient to increase the spatial and temporal variability in the abundance of a
species in systems characterized by high residual variabilities as a result of such physical
forcing (Benedetti-Cecchi, 2000). Navarrete et al. (2005) demonstrated that in benthic
communities, the strength of species interactions depends to some extent on regional
discontinuities in oceanographic conditions. The Antarctic ecosystem has often been
proposed as a suitable starting point for developing ecosystem models because it is a
relatively simple ecosystem that has suffered large impacts from overfishing (e.g. Mori
and Butterworth, 2004). However, as with other high-latitude regions with short links
to high trophic levels, it is subject to large physical variability that may need to be better
understood before reliable conclusions can be drawn regarding trophic interactions.
The agreed conclusion of the IWC’s Scientific Committee following discussion of
the report of its workshop (IWC, 2004b) provides some useful insights and reads:
“for no system at present are we in the position, in terms of data availability and model
development, to provide quantitative management advice on the impact of cetaceans on
fisheries, or of fisheries on cetaceans. However, this does not rule out the possibility of providing
qualitative advice if a number of different approaches yield qualitatively similar results.”

8.3 GENERAL GUIDELINES

General guidelines stressed by most of the previous studies and workshops include:
• the overriding importance of further investigations regarding the appropriate
form for functional responses (the prey-predator interaction terms) and feeding
selectivities/suitabilities;
• the need to consider operational (i.e. management) issues;
• the need for further systematic investigations (presumably through simulation
studies) of the numbers of links that have to be included in a non-trivial ecosystem
model for reliable predictive ability.
There is a growing realisation that substantial progress towards implementing
reliable ecosystem models is still some way off given the need in most regions for
considerable data collection and complex analysis. On the other hand, progress in this
field has likely developed faster than anticipated given the encouraging number of
researchers drawn to the field, the necessary legislation having been put in place, the
availability of funding for ecosystem research and the development of tools that are
widely accessible as a first step to explore the issues. Given the resource-hungry nature
of ecosystem investigations, it is nonetheless important that research priorities in this
58 Models for an ecosystem approach to fisheries

area be carefully and realistically chosen and weighed against other research needs
(Butterworth and Plagányi, 2004).

8.4 ECOSYSTEM-BASED MANAGEMENT STRATEGIES

The objective of the 2000 UBC Workshop on the Use of Ecosystem Models to
Investigate Multi-species Management Strategies for Capture Fisheries (Pitcher and
Cochrane, 2002) was to explore the impact of different multi-species harvesting
strategies, with a view to searching for fishing rates and patterns that would maximize
ecological, social or economic goals (Cochrane, 2002). A wide range of EwE models
were used by participants to identify the management strategies which would come
closest to achieving the objectives for each of the ecosystems considered, as well as
estimating the consequences of the various management strategies. This was made
possible following the development of routines within EwE to assist the user in
exploration of fisheries strategies or policies (Walters, Christensen and Pauly, 2002),
effectively using the EwE models in a similar manner to operating models (Cochrane,
2002).
The workshop also stressed the importance of investigating the sensitivity of these
policies to uncertainties in trophic dynamics (e.g. by considering a range of vulnerability
settings). The workshop stressed the dangers of not using the software cautiously and
thoughtfully (Cochrane, 2002). Model results obtained at the workshop were useful in
highlighting the types of tradeoffs encountered in trying to simultaneously maximize
economic, social and ecological goals, and identifying the need for better economic
and other data (e.g. on prices per species and fleet operational costs) before trade-offs
can be computed with any confidence (e.g. Bundy, 2002; Vasconcellos, Heymans and
Bundy, 2002).

8.5 PRACTICAL STEPS TO IMPLEMENTING AN EAF

The 2002 Cape Town workshop “An ecosystem approach to fisheries management in
the southern Benguela: introducing the concept and looking at our options” had two
objectives, stated as:
(i) to introduce the concept of ecosystem-based fisheries management to South
African fisheries scientists and to present modelling tools to achieve this, in
particular the ECOPATH/ECOSIM approach; and
(ii) to propose a framework of practical ways in which the incorporation of ecosystem
considerations (potentially using information from ECOPATH/ECOSIM and
other types of multi-species modelling approaches) into current Operational
Management Procedures (OMPs) and other management strategies for South
Africa’s marine resources could be attempted.
Consensus was reached that an EAF would be highly desirable and should be
implemented immediately using an incremental approach (Shannon et al., 2004). As
a step in this direction, a project being implemented by the Benguela Current Large
Marine Ecosystem project and the FAO held “Risk Assessment for Sustainable
Fisheries” Workshops for a range of stakeholders in each of Angola, Namibia and
South Africa. These used the method of Ecological Risk Assessment developed under
the National Ecologically Sustainable Development (ESD) framework for prioritizing
issues across valuable Western Australian fisheries (Fletcher et al., 2002; Fletcher,
2005). Initial work identified issues surrounding fisheries and the management thereof
and ranked these according to the likelihood that an issue occurs and the severity
of its consequence (Nel, 2005). This has at least made a first attempt at highlighting
important areas to focus modelling efforts. In South Africa, as presumably in many
other areas of the world, many of the major ecosystem issues identified are non-
trophic (Shannon et al., 2004), emphasizing that biological models may often have a
Pointers from previous studies and workshops 59

relatively small role to play in an EAF. Alternatively, this may be a consequence of a

lack of information and knowledge about the way trophic (indirect) interactions affects
fisheries (M. Vasconcellos, FAO, pers. comm.).
The conclusions of the Cape Town workshop overlapped considerably with
those discussed elsewhere in this document, namely that the following are important
shortcomings of ecosystem modelling studies to be borne in mind (Shannon et al.,
2004):
• It may be important to consider the effects of short-term variability;
• Models need to improve their representation of regime shifts and other longer
term ecosystem dynamics;
• Predator-prey functional responses are in need of further investigation;
• Increased attention should be focused on assessing the robustness of a model to
a range of major uncertainties, acknowledging that full sensitivity testing is not
always possible.
The workshop stressed that the long term benefits of an EAF need to be strongly
emphasized and clearly explained. This follows particularly given that in the short-
term at least, it may result in less fish being made available to fishers (D.S. Butterworth,
University of Cape Town, South Africa, pers. comm.) and is likely to result in increased
political and social pressures as well as stretching limited capacity and resources
(Cochrane et al., 2004). An important consideration is that efforts towards this end are
impeded by the fact that there is a current paucity of examples of successful case studies
to show that an EAF is successful and beneficial (Cochrane et al., 2004).
 61

9. Summary of model comparisons

and recommendations

Attention worldwide is increasingly being concentrated on establishing frameworks

for fisheries management that are ecosystem-oriented, notwithstanding that the
operational aspects of this goal are fraught with difficulty (Hall and Mainprize, 2004).
This field is still very new and major gaps still exist between single-species and multi-
species or ecosystem approaches to practical fishery management.
Three particularly important areas requiring attention are the following:
1. Review of underlying shortcomings and assumptions of available multi-species/
ecosystem approaches
This aspect is seen as critical to advancing attempts to incorporate ecosystem
considerations in practical fisheries management. Unfortunately endeavours in
this regard appear to be lagging considerably behind the ever-growing number of
documented applications of ecosystem models. Critical reviews of methods assist
in highlighting weaknesses and hence ultimately in strengthening applications
of an ecosystem approach. Where applied most effectively, conventional single-
species modelling approaches used to inform the management of commercially
important stocks are typically subject to intense scrutiny. Ecosystem models are
likely to be subject to a similar level of scrutiny when they reach the state of being
used as the basis for management recommendations or decisions (with implications
for economically valuable and socially important fisheries in particular). There is
therefore a need for parallel processes of model development, application and
scrutiny – otherwise the danger exists that considerable time and effort will
have been wasted in developing ecosystem models that are later rejected out of
hand when they attempt to enter the management arena or that bad management
decisions, with potentially serious consequences, will be made on the basis of poor
scientific advice.
2. Systematic analyses of alternative functional response formulations to be
considered in models
Although progress in this field is primarily impeded by a lack of suitable data and
experimental studies (noting that the focus here is on recommended modelling
endeavours), simulation and modelling studies can nevertheless contribute. This
issue is critical and hence attention should be focused both on the need to carefully
check model robustness to alternative interaction representation hypotheses and
on simulation exercises to systematically and thoroughly explore this issue.
3. Consideration of uncertainty in model structure, parameter estimates and data.
Models need to account for key levels of uncertainty, preferably within a
strategic and practical framework. This aspect of multi-species/ecosystem models
has lagged unsatisfactorily behind other aspects of model development, given
(understandable) arguments to the effect that detailed sensitivity analyses are
a major undertaking for these models and there are typically inadequate data
available for fitting purposes. While many studies are currently underway
(E. Fulton and F. Pantus, CSIRO, pers. comm.), the most prominent published
example is that of Ginot et al. (2006), which demonstrates the usefulness of
ANOVA-based global sensitivity analyses for exploring which parameters
(in models with only a moderate number of parameters) have an impact on model
output and the interactions between the parameters.
62 Models for an ecosystem approach to fisheries

It is also important to remember that an assessment method (however rigorously

applied) and associated recommendations are unable to successfully achieve conservation
when management fails. As stressed by Parma et al. (2003), sustainability of a fishery
is likely to be achieved only when the right incentives are provided, such as in the
form of secure long-term access rights. The correct incentives and management
structures need to be firmly in place if success is to be achieved. To reach this goal it is
insufficient simply to perfect existing models. Stakeholder participation and dialogue
need to be seen as integral components of multi-species fisheries management and
scientists need to avoid the temptation to use loosely constructed ecosystem models
to justify a preferred point of view. Moreover, although the discussion throughout has
focused on specific modelling perspectives, it is important to bear in mind that in some
cases the best approach would likely depend on experimental studies and an adaptive
management approach (e.g. Walters 1986; Hilborn and Walters 1992; Sainsbury, Punt
and Smith, 2000). For example, an actively adaptive management strategy applied to
the Australian multi-species fishery was successful in resolving key uncertainties about
resource dynamics and sustainable resource use (Sainsbury et al., 1997). The approach
involved identifying four different plausible hypotheses and adopting an experimental
process involving the sequential closure of areas to trawl fishing. After a period of
a few years, the experiment was successful in discriminating among the competing
hypotheses (Sainsbury et al., 1997; Sainsbury, Punt and Smith, 2000). The success of
this earlier work has lead to its extension into the multiple use realm (Little et al.,
2006).
In summary, this report has aimed to document all of the well-known, as well as
several of the less well-known multi-species and ecosystem modelling approaches
used in Ecosystem-Based Fisheries Management (EBFM). Some 20 approaches have
been described (Tables A1-A3), ranging from ESAM (which entails no more than
the addition of one or two species to current single-species assessment models) to
ATLANTIS (covering the full trophic spectrum) at the opposite extreme. The most
widely used approach is undoubtedly EwE, which is likely to remain a forerunner
given the user-friendly interface and on-going improvements to the software. Faced
with incomplete knowledge of ecosystem functioning, there has been increasing
recognition that definitive conclusions cannot be drawn from a single model structure.
There has thus been a parallel increase in efforts to modularize models so that different
components can be easily substituted. Spatial considerations are similarly playing an
increasingly important role in the development of ecosystem modelling approaches.
Nonetheless, even some of the earliest approaches such as MSVPA are still being
used and improved. To give an idea of directions being taken in on-going model
development, a summary has been presented of some other recent advances being
planned for the different modelling approaches.
This preliminary analysis of the potential of the various modelling approaches to
address specific EBFM research questions suggests that a range of different model
constructions are needed; no one model is superior to all others in all respects. This
review has stressed several times that ideally a range of models should be applied, but
this is not always possible because of limitations on resources available to undertake
such analyses. Nonetheless, it may be argued that the model with the greatest potential
to contribute to practical fisheries management advice in regions with reasonable
data availability is GADGET, although as stated throughout, the preferred approach
is parallel development of different models. Although still under development,
GADGET is currently the model with the most rigorous statistical framework for
developing multi-species based management advice. It is also the modelling approach
most capable of detailed sensitivity investigations to alternative growth, consumption
and recruitment formulations. Additionally, it operates within a spatial framework
Summary of model comparisons and recommendations 63

and overcomes many of the associated computing constraints by running on multiple

computers in parallel using PVM. Nonetheless, it too has its limitations in that it is
capable of representing only a relatively small subset of the ecosystem and may be less
useful in tropical regions with much higher species diversity. Models such as EwE and
ATLANTIS are more appropriate for considering broader questions. In particular,
EwE is capable of addressing the widest range of topical EBFM research questions.
The multiple-stanza version of ECOSIM is a major advancement and greatly expands
the potential of this approach to investigate important questions such as the effects
of biomass pool composition on aggregated consumption estimates as well as being
able to represent cannibalism through size-dependent interaction rates (Walters
and Martell, 2004). ATLANTIS is ranked here as the best operating model within a
simulation testing framework. Although it seems unlikely that sufficient data will be
available to achieve such testing in most marine systems, some argue that “what-if”
approaches are becoming more acceptable such that progress could be made on this
front. Approaches that have more recently followed in the footsteps of the Punt and
Butterworth (1995) MRM approach also deserve a closer look in that such Management
Procedure approaches take explicit account of uncertainty and management issues
through the use of a simulation framework incorporating feedback control rules used
in actual management.
As discussed, simple extensions to current single-species assessment models,
termed ESAM approaches here, are often a good first step. Similarly, equations such
as those presented in Mori and Butterworth (2005) are a useful starting template for
multi-species modelling approaches being built up slowly and in synchrony with data
availability. Some of the less well-known (in a global context) approaches have been
shown to include some additional useful features, for example, SEAPODYM’s habitat
index, OSMOSE’s explorations with simple individual predation rules and Koen-
Alonso and Yodzis’s (2005) approach for substituting different functional response
variants.
This report is a first step towards initiating more detailed discussions of these
models, their uses and their limitations. This process will be critical in moving forward
the development of methods for assessing indirect ecosystem impacts of fisheries.
Whereas the modelling tool-box is reasonably well developed and diverse, high levels
of uncertainty around the nature and consequences of most ecosystem interactions
will hinder the efficient application of an EAF. Greater focus is needed on reducing
these uncertainties and conducting the necessary data collection and experimentation
to strengthen confidence in these approaches. Indeed, before embarking on the
construction of a new ecosystem-type model, would-be model developers should
assess whether they would be adding anything to the current suite of models, given
that approaches such as EwE and GADGET have benefited from an extensive network
of collaborators over a number of years. Hopefully, a review such as this will assist
in selecting the most appropriate general form of model to match the question of
interest.
 65

10. References

Allen, J.L., Blackford, J.C. and Radford, P.J. 1998. A 1-D vertically resolved modelling
study of the ecosystem dynamics of the middle and southern Adriatic sea. J. Mar. Sys.,
18: 265-286.
Allen J.I., Somerfield P. and Siddorn, J. 2002 Primary and Bacterial Production in the
Mediterranean Sea: A modelling study. J. Mar. Sys., 33-34: 473-496.
Alonzo, S.H., Switzer, P.V. and Mangel, M. 2003. An ecosystem-based approach to
management: using individual-behaviour to predict the indirect effects of Antarctic krill
fisheries on penguin foraging. J. Appl. Ecol., 40: 692-702.
Arreguín-Sánchez, F. 2000. Octopus-red grouper interaction in the exploited ecosystem of
the northern continental shelf of Yucatan, Mexico. Ecol. Model., 129: 119-129.
Arrington, D.A., Winemiller, K.O., Loftus, W.F. and Akin, S. 2002. How often do fishes
“run on empty”? Ecology, 83: 2145-2151.
Atkinson, A., Siegel, V., Pakhomov, E. and Rothery, P. 2004. Long-term decline in krill
stock and increase in salps within the Southern Ocean. Nature, 432: 100-103.
Auster, P.J. and Langton, R.W. 1998. The Effects of Fishing on Fish Habitat. Report
submitted to Committee on Fisheries, Ocean Studies Board, National Research
Council.
Aydin, K.Y. 2004. Age structure or functional response? Reconciling the energetics of
surplus production between single-species models and ECOSIM. In L.J. Shannon, K.L.
Cochrane S.C. Pillar (eds). Ecosystem approaches to fisheries in the Southern Benguela.
Afr. J. Mar. Sci., 26: 289-301.
Aydin, K.Y. and Friday, N. 2001. The early development of ECOSIM as a predictive
multi-species fisheries management tool. Paper IWC SC/53/E WP2, presented to
the International Whaling Commission Scientific Committee, July 2001. 5 pp.
Aydin, K.Y. and Gaichas, S.K. 2006. Towards a representation of uncertainty in multi-
species models. IWC SC/58/E48.
Aydin, K.Y., Lapko, V.V., Radchenko, V.I. and Livingston, P.A. 2002. A comparison
of the eastern Bering and western Bering Sea shelf and slope ecosystems through the
use of mass-balance food web models. U.D. Dept. Commerce, NOAA Tech. Memo.
NMFS- AFSC-130. 78 pp.
Baretta, J.W., Baretta-Bekker, J.G. and Ruardij, P. 1996. From EMS to ERSEM, towards
generic coastal ecosystem models. Aquabiol., 104: 197-209 (translated from Japanese).
Baretta-Bekker, J.G. and Baretta, J.W. (eds). 1997. Special Issue: European Regional Seas
Ecosystem Model II. J. Sea Res., 38: 169-436.
Bax, N.J. 1985. Application of multi- and univariate techniques of sensitivity analysis to
SKEBUB, a biomass-based fisheries ecosystem model, parameterized to georges bank.
Ecol. Model., 29: 353-382.
Bax, N.J. 1998. The significance and prediction of predation in marine fisheries.
ICES J. Mar. Sci., 55: 997-1030.
Beddington, J.R. and May, RM. 1982. The harvesting of interacting species in a natural
ecosystem. Sci. Am., 247:42-50.
Begley, J. 2005. Gadget User Guide. Available from website www.hafro.is/gadget. 95 pp.
Begley, J. and Howell, D. 2004. “An overview of GADGET, the Globally applicable Area-
Disaggregated General Ecosystem Toolbox”. ICES CM 2004/FF:13: 15 pp.
Benedetti-Cecchi, L. 2000. Variance in ecological consumer-resource interactions. Nature,
407: 370-374.
66 Models for an ecosystem approach to fisheries

BENEFIT. 2004. Report of the BENEFIT stock assessment workshop, University of

Cape Town, 12-17 January 2004. BENEFIT Programme, Swakopmund. 48 pp.
Bertignac, M., Lehodey, P. and Hampton, J. 1998. A spatial population dynamics
simulation model of tropical tunas using habitat index based on environmental
parameters. Fish. Oceanogr., 7: 326-334.
Bigler, B.S., Welch, D.W. and Helle, J.H. 1996. A review of size trends among North
Pacific salmon (Oncorhynchus spp.). Can. J. Fish. Aquat. Sci., 53: 455-465.
Bjørge, A., Bekkby, T., Bakkestuen, V. and Framstad, E. 2002. Interactions between
harbor seals (Phoca vitulina) and fisheries in complex coastal waters explored by
combined Geographic Information System (GIS) and energetics modelling. ICES J. Mar.
Sci., 59: 29-42.
Blackford, J.C., Allen, J.I. and Gilbert, F.J. 2004. Ecosystem dynamics at six contrasting
sites: a generic modelling study. J. Mar. Sys., 52: 191-215.
Bogstad, B., Hauge, K.H. and Ulltang, O. 1997. MULTSPEC: a multi-species model for
fish and marine mammals in the Barents Sea. J. N. Atl. Fish. Sci., 22: 317-341.
Bogstad, B., Tjelmeland, S., Tjelta, T. and Ulltang, Ø. 1992. Description of a multi-species
model for the Barents Sea (MULTSPEC) and a study of its sensitivity to assumptions
on food preferences and stock sizes of harp seals and minke whales. Int. Whal. Comm..
Sci. Comm. Pap., 44 (09). 47 pp.
Boyd, I.L. and A. Murray,W.A. 2001. Monitoring a marine ecosystem using responses of
upper trophic level predators. J. Anim. Ecol., 70: 747-760.
Bryant, A.D., Heath, M.R., Broekhuizen, N., Ollason, J.G., Gurney, W.S.C., and
Greenstreet, S.P.R. 1995. Modelling the predation, growth and population dynamics
of fish within a spatially-resolved shelf-sea ecosystem model. Neth. J. Sea Res., 33: 407-
421.
Bulgakova, T., Vasilyev, D. and Daan, N. 2001. Weighting and smoothing of stomach
content data as input for MSVPA with particular reference to the Barents Sea.
ICES J. Mar. Sci.. 58: 1208-1218.
Bundy, A. 2002. Exploring multi-species harvesting strategies on the eastern Scotian Shelf
with Ecosim. In T.J. Pitcher and K. Cochrane (eds). The Use of Ecosystem Models to
Investigate Multi-species Management Strategies for Capture Fisheries. Fisheries Centre
Research Reports, 10: 112-117.
Bundy, A. 2005. Structure and functioning of the eastern Scotian Shelf ecosystem before
and after the collapse of groundfish stocks in the early 1990s. Can J. Fish. Aquat. Sci.,
62: 1453-1473.
Burnham, K.P. and Anderson, D.R. 2002. Model Selection and Multi-Model Inference:
a Practical Information-Theoretic Approach. Springer, New York. 353 pp.
Butterworth, D.S. 1986. Antarctic marine ecosystem management. Polar Record, 23: 37-
47.
Butterworth, D.S. and Harwood, J. (eds). 1991. Report on the Benguela Ecology
Programme Workshop on seal-fishery biological interactions. Rep. Benguela Ecol. Progm.
S. Afr., 22: 65 pp.
Butterworth, D.S. and Plagányi, É.E. 2004. A brief introduction to some multi-species/
ecosystem modelling approaches in the context of their possible application in the
management of South African fisheries. In L.J. Shannon, K.L. Cochrane and S.C. Pillar
(eds). Ecosystem approaches to fisheries in the Southern Benguela. Afr. J. Mar. Sci., 26:
53-61.
Butterworth, D.S. and Punt, A.E. 1999. Experiences in the evaluation and implementation
of management procedures. ICES J. Mar. Sci., 56: 985-998.
Butterworth, D.S. and Punt, A.E. 2003. The role of harvest control laws, risk and
uncertainty and the precautionary approach in ecosystem-based management. In
M. Sinclair and G. Valdimarsson (eds). Responsible Fisheries in the Marine Ecosystem.
References 67

FAO and CABI Publishing. pp. 311-319.

Butterworth, D.S. and Rademeyer, R.A. 2005. Sustainable management initiatives for the
southern African hake fisheries over recent years. Bull. Mar. Sci.: 76(2): 287-319
Butterworth, D.S. and Thompson, R.B. 1995. Possible effects of different levels of krill
fishing on predators: some initial modelling attempts. CCAMLR Science, 2: 79-97.
Butterworth, D.S., Cochrane, K.L. and De Oliveira, J.A.A. 1997. Management
procedures: a better way to manage fisheries? The South African experience. In E.L.
Pikitch, D.D. Huppert and M.P. Sissenwine (eds). Global Trends: Fisheries Management.
Bethesda, Maryland: pp. 83-90. American Fisheries Society Symposium No. 20,
Butterworth, D.S., Punt, A.E., Oosthuizen, W.H. and Wickens, P.A. 1995. The effects of
future consumption by the Cape fur seal on catches and catch rates of the Cape hakes.
3. Modelling the dynamics of the Cape fur seal (Arctocephalus pusillus pusillus). S. Afr.
J. Mar. Sci., 16: 161-183.
Caddy, J.F. 1999. Fisheries management in the twenty-first century: will new paradigms
apply? Rev. Fish Biol. Fisheries, 9: 1-43.
CCAMLR [Commission for the Conservation of Antarctic Marine Living Resources].
2004. Report of the meeting of the working group on ecosystem monitoring and
management, Siena, Italy, 12 to 23 July 2004. 87 pp.
CCAMLR [Commission for the Conservation of Antarctic Marine Living Resources].
2005. Report of the workshop on management procedures, Yokohama, Japan, 4-8 July
2005. 47 pp.
CCAMLR [Commission for the Conservation of Antarctic Marine Living Resources].
2006. Report of the workshop on management procedures, Walvis Bay, Namibia, 17-
21 July 2006. WG-EMM-06/40, 36 pp.
Chouinard, G.A., Swain, D.P., Hammill, M.O. and Poirier, G.A. 2005. Covariation between
grey seal (Halichoerus grypus) abundance and natural mortality of cod (Gadus morhua) in
the southern Gulf of St. Lawrence. Can. J. Fish. Aquat. Sci., 62: 1991-2000.
Christensen, V. 1995a. Ecosystem maturity: towards quantification. Ecol Model., 77: 3-32.
Christensen, V. 1995b. A Multi-species virtual population analysis incorporating
information of size and age. ICES C.M. 1995, D:8
Christensen, V. 1998. Fishery-induced changes in a marine ecosystem: insights from
models of the Gulf of Thailand. J. Fish Biol., 53 (Suppl.) S1: 128-142.
Christensen, V. and Pauly, D. 1992. ECOPATH II: a software for balancing steady-state
models and calculating network characteristics. Ecol. Model., 61: 169-185.
Christensen, V. and Walters, C.J. 2000. Ecopath with Ecosim: methods, capabilities and
limitations. In D. Pauly and T.J. Pitcher (eds). Methods for Evaluating the Impacts of
Fisheries on North Atlantic Ecosystems. Fisheries Centre Research Reports, 8: 79- 105.
Christensen, V. and Walters, C.J. 2004. Ecopath with Ecosim: methods, capabilities and
limitations. Ecol. Model., 172: 109-139.
Christensen, V., Walters, C.J. and Pauly, D. 2000. Ecopath with Ecosim: a User’s Guide,
October 2000 Edition. Fisheries Centre, University of British Columbia, Vancouver,
Canada and ICLARM, Penang, Malaysia. 130 pp.
Clapham, P.J. and Brownell Jr. , R.L. 1996. The potential for interspecific competition in
baleen whales. Rep. Int. Whal. Comm., 46: 361-367.
Cochrane, K.L. 1998. FAO’s interest in and expectations of this workshop. In D. Pauly and
G. Weingartner (eds). Use of Ecopath with Ecosim to evaluate strategies for sustainable
exploitation of multi-species resources. Fisheries Centre Research Reports, 6(2): 6-7.
Cochrane, K.L. 2002. The use of ecosystem models to investigate ecosystem-based
management strategies for capture fisheries: introduction. In T. Pitcher and K.L. Cochrane
(eds). The use of ecosystem models to investigate multi-species management strategies for
capture fisheries. Fisheries Centre Research Reports 10(2): 5-10.
Cochrane, K.L., Augustyn, C.J., Cockcroft, A.C., David, J.H.M., Griffiths, M.H.,
68 Models for an ecosystem approach to fisheries

Groeneveld, J.C., Lipinski, M.R., Smale, M.J., Smith, C.D. and Tarr R.J.Q. 2004. An
ecosystem approach to fisheries in the southern Benguela context. Afr. J. Mar. Sci., 26: 9-
35.
Colomb, A., Shin, Y-J., Gascuel, D. and Le Fur, J. 2004. Using MOOVES, an individual-
based model of the Guinean ecosystem, for assessing the response of Trophodynamic and
size-based indicators to fishing pressure. ICES CM 2004/FF: 14.
Committee on the Alaska Groundfish Fishery and Steller Sea Lions. 2003. Decline of
the Steller sea lion in Alaskan waters: untangling food webs and fishing nets. National
Academies Press, Washington, DC. 204 pp.
Constable, A.J. 2001. The ecosystem approach to managing fisheries: achieving conservation
objectives for predators of fished species. CCAMLR Science, 8: 37-64.
Constable, A.J. 2005. Implementing plausible ecosystem models for the Southern Ocean:
an ecosystem, productivity, ocean, climate (EPOC) model. Workshop document
presented to WG-EMM subgroup of CCAMLR (Commission for the Conservation of
Antarctic Marine Living Resources), WG-EMM-05/33. (20 pp.).
Constable, A.J. 2006. Using the EPOC modelling framework to assess management
procedures for Antarctic krill in Statistical Area 48: evaluating spatial differences in
productivity of Antarctic krill. Workshop document presented to WG-EMM subgroup
of CCAMLR (Commission for the Conservation of Antarctic Marine Living Resources),
WG-EMM- 06/38.
Constable, A.J., de la Mare, W.K., Agnew, D.J., Everson, I. and Miller, D. 2000. Managing
fisheries to conserve the Antarctic marine ecosystem: practical implementation of the
Conservation of Antarctic Marine Living Resources (CCAMLR). ICES J. Mar. Sci.,
57: 778-791.
Cooke, J.G. 1999. Improvement of fishery-management advice through simulation testing
of harvest algorithms. ICES J. Mar. Sci., 56: 797-810.
Cooke, J.G. 2002. Some aspects of the modelling of effects of changing cetacean abundance on
fishery yields. Paper SC/J02/FW10 presented to the International Whaling Commission
Modelling Workshop on Cetacean-Fishery Competition, June 2002: 5 pp.
Cornick, L.A., Neill, W. and Grant, W.E. 2006. Assessing competition between Steller sea
lions and the commercial groundfishery in Alaska: a bioenergetics modelling approach.
Ecol. Model., 199: 107-114.
Cox, S.P., Essington, T.E., Kitchell, J.F., Martell, S.J.D., Walters, C.J., Boggs, C. and
Kaplan, I. 2002. Reconstructing ecosystem dynamics in the central Pacific Ocean, 1952-
1998. II. A preliminary assessment of the trophic impacts of fishing and effects on tuna
dynamics. Can. J. Fish. Aquat. Sci., 59: 1736-1747.
Crawford, R.J.M. 2004. Accounting for food requirements of seabirds in fisheries
management – the case of the South African purse-seine fishery. In L.J. Shannon,
K.L. Cochrane and S.C. Pillar (eds). Ecosystem approaches to fisheries in the Southern
Benguela. Afr. J. Mar. Sci., 26: 197-203.
Crise A., Allen, J.I., Baretta, J., Crispi, G., Mosetti, R. and Solidoro, C. 1999. The
Mediterranean pelagic ecosystem response to physical forcing. Prog. Oceanogr., 44: 219-
243.
Cury, P., Bakun, A., Crawford, R.J.M., Jarre, A., Quiñones, R.A., Shannon, L.J. and
Verheye, H.M. 2000. Small pelagics in upwelling systems: patterns of interaction and
structural changes in ‘‘wasp-waist’’ ecosystems. ICES Journal of Marine Science, 57:
603–618
Daan, N. and M.P. Sissenwine (eds). 1991. Multi-species models relevant to management
of living resources. ICES Marine Science Symposium 193.
Daan, N., Gislason, H., Pope, J.G. and Rice, J.C. 2005. Changes in the North Sea fish
community: evidence of indirect effects of fishing? ICES J. Mar. Sci., 62: 177-188.
Dalton, T.M. 2004. An approach for integrating economic impact analysis into the
References 69

evaluation of potential marine protected area sites. J. Environ. Manage., 70: 333-349.
Daskalov, G.M. 2002. Overfishing drives a trophic cascade in the Black Sea. Mar. Ecol.
Prog. Ser., 225: 53-63.
DeAngelis, D.L. and Gross, M.J. 1992. Individual-based models and approaches in
ecology: populations, communities and ecosystems. Chapman and Hall, New York.
DeAngelis, D.L., Shuter, B.J., Ridgeway, M.S., Blanchfield, P., Friesen, T. and Morgan,
G.E. 1991. Modeling early life-history stages of smallmouth bass in Ontario lakes.
Trans. Amer. Fish. Soc., September: 9-11.
De Oliveira, J.A.A., Butterworth, D.S., Roel, B.A., Cochrane, K.L. and Brown,
J.P. 1998. The application of a management procedure to regulate the directed and
bycatch fishery of South African sardine (Sardinops sagax). S. Afr. J. Mar. Sci.,
19: 449-469.
De Roos, A.M., Persson, L. and McCauley, E. 2003. The influence of size-dependent
life-history traits on the structure and dynamics of populations and communities.
Ecology Letters, 6: 473-487.
De Roos, A.M. and Persson, L. 2005. The influence of individual growth and development
on the structure of ecological communities. In P.C. de Ruiter, V. Wolters and J.C. Moore
(eds). Food webs. dynamic food webs- multi-species assemblages, ecosystem development
and environmental change. pp. 89-100, Academic press.
Dichmont, C.M., Deng, A.R., Venables, W.N., Punt, A.E., Haddon, M. and Tattersall,
K. 2005. A new approach to assessment in the NPF: spatial models in a management
strategy environment that includes uncertainty. Report of FRDC 2001/002. CSIRO
Marine Research, Cleveland, Australia.
Doedel, E.J., Champneys, A.R., Fairgrieve, T.F., Kuznetsov, Y.A., Sandstede, B. and
Wang, X. 1998. AUTO 97: continuation and bifurcation software for ordinary
differential equations. Available from ftp://ftp.cs.concordia.ca/pub/doedel/auto.
Drapeau, L., Pecquerie, L., Fréon, P. and Shannon, L.J. 2004. Quantification and
representation of potential spatial nteractions in the southern Benguela ecosystem. In
L.J. Shannon, K.L. Cochrane and S.C. Pillar (eds). Ecosystem approaches to fisheries in
the Southern Benguela. Afr. J. Mar. Sci., 26: 141-159.
Eichner, T. and J. Tschirhart. (in press). Efficient ecosystem services and naturalness in an
Ecological/Economic Model Environmental and Resource Economics.
Eisenack, K. and Kropp, J. 2001. Assessment of management options in marine fisheries
by qualitative modelling techniques. Marine Pollution Bulletin, 43: 215-224.
FAO. 1995. Precautionary approach to fisheries. Part 1: Guidelines on the precautionary
approach to capture fisheries and species introductions. FAO Fisheries Technical Paper.
No. 350, Part 1. Rome, FAO. 52 p.
FAO. 2002. The state of world fisheries and aquaculture. FAO, Fisheries Department,
Rome. 149 pp.
FAO. 2003a. The ecosystem approach to fisheries. FAO Technical Guidelines for
Responsible Fisheries. No. 4, Suppl. 2. Rome, FAO. 112 pp.
FAO. 2003b. Report of the expert consultation on ecosystem-based fisheries management,
Reykjavik, Iceland, 16-19 September 2002. FAO Fisheries Report, No. 690. Rome,
FAO. 31 pp.
Fenton, J.S. and Faddy, M.J. 2006. Modelling predation in functional response.
Ecol. Model., 198: 154-162.
Fletcher, W.J. 2005. The application of qualitative risk assessment methodology to
prioritize issues for fisheries management. ICES J. Mar. Sci., 62: 1576-1587.
Fletcher, W.J., Chesson, J., Fisher, M., Sainsbury, K.J., Hundloe, T., Smith, A.D.M. and
Whitworth, B. 2002. National ESD Reporting Framework for Australian Fisheries:
the “How To” Guide for Wild Capture Fisheries. Fisheries Research and Development
70 Models for an ecosystem approach to fisheries

Corporation Project 2000/145, Canberra, Australia. 120 pp.

Floeter, J. and Temming, A. 2003. Explaining diet composition of North Sea cod
(Gadus morhua): prey size preference vs. prey availability. Can J. Fish. Aquat. Sci.,
60: 140-150.
Froese, R. and Pauly, D. 1998. FishBase 98: Concepts, Structure and Data Sources.
Manila: International Center for Living Aquatic Resources Management.
Updates: <http://www.fishbase.org>
Frøysa, K.G., Bogstad, B. and Skagen, D.W. 2002. Fleksibest – an age-length structured
fish stock assessment tool with application to Northeast Arctic cod (Gadus morhua L.).
Fish. Res. 55: 87-101.
Fulton, E.A. 2001. The Effects of Model Structure and Complexity on the Behaviour and
Performance of Marine Ecosystem Models. Ph.D. thesis, School of Zoology, University
of Tasmania, Hobart. 429 pp.
Fulton, E.A. and Smith, A.D.M. 2002. Ecosim case study: Port Philip Bay, Australia.
In T.J. Pitcher and K.L. Cochrane (eds). The Use of Ecosystem Models to Investigate
Multi-species Management Strategies for Capture Fisheries. Fisheries Centre Research
Reports, 10: 83-93.
Fulton, E.A. and Smith, A.D.M. 2004. Lessons learnt from the comparison of three
ecosystem models for Port Philip Bay, Australia. In L.J. Shannon, K.L. Cochrane and
S.C. Pillar (eds). Ecosystem approaches to fisheries in the Southern Benguela. Afr. J. mar.
Sci., 26: 219-243.
Fulton, E.A., Smith, A.D.M. and Johnson, C.R. 2003a. Effect of complexity on marine
ecosystem models. Mar. Ecol. Prog. Ser., 253: 1-16.
Fulton, E.A., Smith, A.D.M. and JohnsonC.R. 2003b. Mortality and predation in
ecosystem models: is it important how these are expressed? Ecol. Model. 169: 157 – 178.
Fulton, E.A., Smith, A.D.M. and JohnsonC.R. 2004. Biogeochemical marine ecosystem
models. I. A model of marine bay ecosystems. Ecol. Model., 174: 267-307.
Fulton, E.A., Smith, A.D.M. and Punt, A.E. 2004. Ecological indicators of the ecosystem
effects of fishing: Final Report. Report No. R99/1546, Australian Fisheries Management
Authority, Canberra.
Fulton, E.A., Smith, A.D.M. and Punt, A.E. 2005. Which ecological indicators can
robustly detect effects of fishing? ICES J. Mar. Sci., 62: 540-551.
Fulton, E.A., Parslow, J.S., Smith, A.D.M. and Johnso, C.R. 2004. Biogeochemical marine
ecosystem models. 2. The effect of physiological detail on model performance. Ecol.
Model., 173: 371-406.
Garcia, S.M., Zerbi, A., Aliaume, C., Do Chi, T. and Lasserre, G. 2003. The ecosystem
approach to fisheries. Issues, terminology, principles, institutional foundations, implementation
and outlook. FAO Fisheries Technical Paper, No. 443. Rome, FAO. 71 pp.
Geromont, H.F., De Oliveira, J.A.A., Johnston, S.J. and Cunningham, C.L. 1999.
Development and application of management procedures for fisheries in southern Africa.
ICES J. Mar. Sci., 56: 952-966.
Ginot, V., LePage, C. and S. Souissi. 2002. A multi-agents architecture to enhance end-user
individual-based modelling. Ecol. Model., 157: 21-41.
Ginot, V., Gaba, S., Beaudouin, R., Aries, F. and Monod, H. 2006. Combined use of
local and ANOVA-based global sensitivity analyses for the investigation of a stochastic
dynamic model: Application to the case study of an individual-based model of a fish
population. Ecol. Model., 193: 479-491.
Gislason, H. 2003. Effects of fishing on non-target species. In M. Sinclair and G. Valdimarsson
(eds). Responsible Fisheries in the Marine Ecosystem. FAO and CABI Publishing, UK:
pp. 255-274.
Gislason, H. and Helgason, T. 1985. Species interaction in assessment of fish stocks with
special application to the North Sea. Dana, 5: 1-44.
References 71

Gray, R., Fulton, E.A., Little, L.R. and Scott, R. 2006. Ecosystem Model Specification
Within an Agent Based Framework. North West Shelf Joint Environmental Management
Study Technical Report – Vol 16.
Gray, R., Fulton, E.A., Hatfield, B., Little, L.R., Lyne, V., McDonald, A.D., Sainsbury, K.
and Scott, R. 2003. InVitro – Agents in glass. In Proceedings of the CABM/HEMA
Workshop, 11-12 July 2003, Melbourne, Australia.
Grimm, V. 1999. Ten years of individual-based modelling in ecology: what have we learnt
and what could we learn in future? Ecol. Model., 115: 129-148.
Grimm, V., Berger, U., Bastiansen, F., Eliassen, S., Ginot, V., Giske, J., Goss-Custard, J.,
Grand, T., Heinz, S.K., Huse, G., Huth, A., Jepsen, J.U., Jørgensen, C., Mooij, W. M.,
Müller, B., Pe’er, G., Piou, C., Railsback, S.F., Robbins, A.M., Robbins, M.M.,
Rossmanith, E., Rüger. N., Strand, E., Souissi, S., Stillman, R.A., Vabø, R., Visser, U.
and DeAngelis, D.L. 2006. A standard protocol for describing individual-based and
agent-based models. Ecol. Model., 198: 115-126.
Gulland, J.A. 1983. Species interactions. In J.A. Gulland (ed.). Fish Stock Assessment: A
manual of basic methods, pp. 186-201. John Wiley and Sons, Chichester: 223 pp.
Hall, S.J. and Mainprize, B. 2004. Towards ecosystem-based fisheries management.
Fish and Fisheries, 5: 1-20.
Hamre, J. and Hattlebakk, E. 1998. System model (Systmod) for the Norwegian Sea and
the Barents Sea. In T. Rødseth (ed.). Models for Multi-species Management, pp. 93-115.
Springer-Verlag.
Harvey, C.J., Cox, S.P., Essington, T.E., Hansson, S. and Kitchell, J.F. 2003. An ecosystem
model of food web and fisheries interactions in the Baltic Sea. ICES J. Mar. Sci.,
60: 939- 950.
Harwood, J. and McLaren, I. 2002. Modelling interactions between seals and fisheries:
model structures, assumptions and data requirements. IWC document SC/J02/FW4.
Heath, M., Scott, B. and Bryant, A.D.1997. Modelling the growth of herring from four
different stocks in the North Sea. J. Sea Res., 38: 413-436.
Helgason, Th. and H. Gislason. 1979. VPA – analysis with species interaction due to
predation. ICES Committee Meeting, 1979/G: 52.
Hilborn, R. 1990. Estimating the parameters of full age-structured models from catch and
abundance data. Bull. Int. N. Pac. Fish. Comm., 50: 207-213.
Hilborn, R. and Walters, C.J. 1992. Quantitative Fisheries Stock Assessment: Choice,
Dynamics and Uncertainty. Chapman and Hall, New York.
Hilborn, R., Maguire, J.-J., Parma, A.M. and Rosenberg, A.A. 2001. The Precautionary
Approach and risk management: can they increase the probability of successes in fishery
management? Can. J. Fish. Aquat. Sci., 58: 99-107.
Hilborn, R., Stokes, K., Maguire, J.-J., Smith, T., Botsford, L.W., Mangel, M.,
Orensanz, J., Parma, A., Rice, J., Bell, J., Cochrane, K.L., Garcia, S., Hall, S.J.,
Kirkwood, G.P., Sainsbury, K., Stefansson, G. and Walters, C. 2004. When can
marine reserves improve fisheries management? Ocean and Coastal Management,
47: 197-205.
Hill, S., Murphy, E.J., Reid, K., Trathan, P.N. and Constable, A. 2006. Modelling
Southern Ocean ecosystems: krill, the food-web and the impacts of harvesting. Biol. Rev.,
81: 581- 608.
Hill, S., Reid, K., Thorpe, S., Hinke, J. and Watters, G. (in press). A compilation of
parameters for ecosystem dynamics models of the Scotia Sea - Antarctic Peninsula
region. CCAMLR Sci.
Hollowed, A.B., Ianelli, J.N. and Livingston, P.A. 2000. Including predation mortality in
stock assessments: a case study for Gulf of Alaska walleye pollock. ICES J. Mar. Sci., 57:
72 Models for an ecosystem approach to fisheries

279-293.
Hollowed, A.B., Bax, N., Beamish, R., Collie, J., Fogarty, M., Livingston, P., Pope, J. and
Rice, J.C. 2000. Are multi-species models an improvement on single-species models for
measuring fishing impacts on marine ecosystems? ICES J. Mar. Sci., 57: 707-719.
Horbowy, J. 2005. The dynamics of Baltic fish stocks based on a multi-species stock
production model. J. Appl. Ichthyology, 21: 198-204.
Hulse, G. 2001. Modelling habitat choice in fish using an adapted random walk. Sarsia,
86: 477-483.
ICES [International Council for the Exploration of the Sea]. 2000. Ecosystem effects of
fishing. ICES Journal of Marine Science, 57(3): 791.
IWC [International Whaling Commission]. 1989. Comprehensive Assessment workshop
on management procedures. In The Comprehensive Assessment of Whale Stocks:
the Early Years. Rep. Int. Whal. Comm. Special Issue, 11: 29-44.
IWC [International Whaling Commission]. 2003. Specifications of the North Pacific
minke whaling trials. J. Cetacean Res. Manage., 5 (Suppl.): 133-149.
IWC [International Whaling Commission]. 2004a. Report of the Modelling Workshop
on Cetacean-Fishery Competition, La Jolla, 25-27 June, 2002. J. Cetacean Res. Manage.,
6 (Suppl.):413-426.
IWC [International Whaling Commission]. 2004b. Report of the Scientific Committee.
J. Cetacean Res. Manage., 6 (Suppl.):1-60.
Jeschke, J.M., Kopp, M. and Tollrian, R. 2002. Predator functional responses: discriminating
between handling and digesting prey. Ecol. Monogr., 72: 95-112.
Jurado-Molina, J., Livingston, P.A. and Ianelli, J.N. 2005. Incorporating predation
interactions to a statistical catch-at-age model for a predator-prey system in the eastern
Bering Sea. Can. J. Fish. Aquat. Sci., 62(8): 1865-1873.
Jurado-Molina, J., Livingston, P.A. and Gallucci, V.F. 2005. Testing the stability of the
suitability coefficients from the eastern Bering Sea. ICES J. Mar. Sci., 62(5):915-924.
Jurado-Molina J. and Livingston, P.A. 2002. Multi-species perspectives on the Bering Sea
groundfisheries management regime. N. Amer. J. Fish. Manage., 22:1164-1175.
Kaschner, K. 2004. Modelling and mapping of resource overlap between marine mammals
and fisheries on a global scale. PhD thesis, MMRU, Fisheries Centre, Department of
Zoology, University of British Columbia, Vancouver, Canada.
Kaschner, K., Watson, R., Christensen, V., Trites, A.W. and Pauly, D. 2001. Modelling
and mapping trophic overlap between marine mammals and commercial fisheries in
the North Atlantic. In D. Zeller and R.D. Pauly Watson (eds). Fisheries impacts on
North Atlantic ecosystems: catch, effort and national/regional datasets. Fisheries Centre
Research Reports, 9(3): 35-45.
Kavanagh, P., Newlands, N., Christensen, V. and Pauly, D. 2004. Automated parameter
optimization for Ecopath ecosystem models. Ecol. Model., 172: 141-149.
Kirby, D.S., Allain, G., Lehodey, P. and Langley, A. 2004. Individual/Agent-based
modelling of fishes, fishers and turtles. Working Paper SCTB17. 17th Meeting of the
Standing Committee on Tuna and Billfish, Majuro, Marshall Islands, 9-18 August 2004.
Kishi, M.J., Nakajima, I. and Kamezawa, Y. 2005. Fish growth comparisons around
Japan using NEMURO.FISH. North Pacific Marine Science Organisation Fourteenth
Annual Meeting, Program Abstracts, 19 September–9 October 2005, Vladivostok,
Russia. 79 pp.
Koen-Alonso, M. and Yodzis, P. 2005. Multi-species modelling of some components of
northern and central Patagonia, Argentina. Can. J. Fish. Aquat. Sci., 62: 1490-1512.
Laevastu, T. and Larkins, H. 1981. Marine Fisheries Ecosystem: Its Quantitative Evaluation
and Management. Farnham, England, Fishing News Books. 162 pp.
Larska, M.S. and Wootton, J.T. 1998. Theoretical concepts and empirical approaches to
measuring interaction strength. Ecology, 79: 461-476.
Laws, R.M. 1977. The significance of vertebrates in the Antarctic marine ecosystem.
References 73

In G.A. Llano (ed.). Adaptations Within Antarctic Ecosystems: Third Symposium on

Antarctic Biology. Scientific Committee for Antarctic Research.
Lehodey, P. 2001. The pelagic ecosystem of the tropical Pacific Ocean: dynamic spatial
modelling and biological consequences of ENSO. Prog. Oceanogr., 49: 439-468.
Lehodey, P. 2005. Reference manual for the Spatial Ecosystem and Population Dynamics
Model SEAPODYM. 58 pp. Available from: <http://www.seapodym.org>.
Lehodey, P., André, J-M., Bertignac, M., Hampton, J., Stoens, A.C., Menkes, L. and
Memery, G.N. 1998. Predicting skipjack forage distributions in the Equatorial Pacific:
preliminary results using a coupled dynamical bio-geochemical model. Fish. Oceanogr.,
7: 317-325.
Lehodey, P., Chai, F. and Hampton, J. 2003. Modelling climate-related variability of
tuna populations from a coupled ocean-biogeochemical-populations dynamics model.
Fish. Oceanogr., 12: 483-494.
Lenhart, H.J., Radach, G. and Ruardij, P. 1997. The effects of river input on the ecosystem
dynamics in the continental coastal zone of the North Sea using ERSEM. J. Sea Res.,
38: 249-274.
Lewy, P. and Nielsen, A. 2003. Modelling stochastic fish stock dynamics using Markov
Chain Monte Carlo. ICES J. Mar. Sci., 60: 743-752.
Lewy, P. and M. Vinther. 2004. Modelling stochastic age-length-structured multi-species
stock dynamics. ICES CM 2004/FF: 20.
Li, X., Chao, Y., McWilliams, J.C. and Fu, L.-L. 2001. A comparison of two
vertical-mixing schemes in a Pacific Ocean general circulation model. J. Climate,
14: 1377-1398.
Lindstrøm, U. and T. Haug. 2001. Feeding strategy and prey selectivity in common minke
whales (Balaenoptera acutorostrata) foraging in the southern Barents Sea during early
summer. J. Cetacean Res. Manage., 3: 239-249.
Lindstrøm, U., Haug, T. and Røttingen, I. 2001. Predation on herring (Clupea harengus)
by minke whales (Balaenoptera acutorostrata) in the Barents Sea. ICES J. Mar. Sci.,
59: 58-70.
Lindstrøm, U., Tjelmeland, S. and Haug, T. 2002. Implementing predation by marine
mammals in fish stock assessments – minke whales and Norwergian spring spawning
herring as case study. Unpublished IWC document SC/102/FW6.
Little, L.R., Fulton, E.A., Gray, R., Hayes, D., Lyne, V., Scott, R., McDonald, A.D. and
Sainsbury, K. 2006. Results and Discussion for the Management Strategy Evaluation
North West Shelf. North West Shelf Joint Environmental Management Study Technical
Report – Vol 14.
Livingston, P.A. and Jurado-Molina, J-J. 2000. A multi-species virtual population analysis
of the eastern Bering Sea. ICES J. Mar. Sci., 57: 294-299.
Livingston, P.A. and Methot, R.D. 1998. Incorporation of predation into a population
assessment model of eastern Bering Sea walleye pollock. Fishery Stock Assessment
Models, Alaska Sea Grant College Program, AK-SG-98-01.
Livingston, P.A. and Tjelmeland, S. 2000. Fisheries in Boreal ecosystems. ICES J. Mar.
Sci., 57: 619-627.
Longhurst, A.R. and Pauly, D. 1987. Ecology of Tropical Oceans. San Diego, Academic
press. 407 pp.
Loreau, M., Mouquet, N. and Holt, R.D. 2003. Meta-ecosystems: a theoretical framework
for a spatial ecosystem ecology. Ecol. Lett., 6: 673-679.
Loukos, H., Monfray, P., Bopp, L. and Lehodey, P. 2003. Potential changes in skipjack tuna
habitat from a global warming scenario: modelling approach and preliminary results.
Fish. Oceanogr., 12: 474-482.
Mackinson, S., Blanchard, J.L., Pinnegar, J.K. and Scott, R. 2003. Consequences
of alternative functional response formulations in models exploring whale-fishery
74 Models for an ecosystem approach to fisheries

interactions. Mar. Mamm. Sci., 19: 661-681.

Magnússon, K.G. 1995. An overview of the multi-species VPA: theory and applications.
Rev. Fish Biol. Fish., 5: 195-212.
Mangel, M. and Switzer, P.V. 1998. A model at the level of the foraging trip for the indirect
effects of krill (Euphausia superba) fisheries on krill predators. Ecol. Model., 105: 235-
256.
Martell, S.J.D., Beattie, A.I., Walters, C.J., Nayar, T. and Briese, R. 2002. Simulating
fisheries management strategies in the Strait of Georgia ecosystem using Ecopath and
Ecosim. In T. Pitcher and K.L. Cochrane (eds). The use of ecosystem models to investigate
multi-species management strategies for capture fisheries. Fisheries Centre Research
Reports, 10(2): 16-23.
Martell, S.J.D., Essington, T.E., Lessard, B., Kitchell, J.F., Walters, C.J. and Boggs, C.H.
Interactions of productivity, predation risk and fishing effort in the efficacy of marine
protected areas for the central Pacific. Can. J. Fish. Aquat. Sci., 62: 1320-1336.
May, R.M., Beddington, J.R., Clark, C.W., Holt, S.J. and Laws, R.M. 1979. Management
of Multi-species Fisheries. Science, 205:267-277.
McAllister, M.K., Pikitch, E.K., Punt, A.E. and Hilborn, R. 1994. A Bayesian approach
to stock assessment and harvest decisions using the sampling/importance resampling
algorithm. Can. J. Fish. Aquat. Sci., 51: 2673-2687.
McAllister, M.K. and Ianelli, N.J. 1997. Bayesian stock assessment using catch-age data
and the sampling-importance resampling algorithm. Can. J. Fish. Aquat. Sci., 54: 284-
300.
McConnaughey, R.A., Mier, K.L. and Dew, C.B. 2000. An examination of chronic
trawling effects on soft-bottom benthos of the eastern Bering Sea. ICES J. Mar. Sci.,
57: 1377-1388.
Megrey, B.A., Hinckley, S. and Dobbins, E.L. 2002. Using scientific visualization tools
to facilitate analysis of multi-dimensional data from a spatially explicit, biophysical,
individual-based model of marine fish early life history. ICES J. Mar. Sci., 59: 203-215.
Mohn, R. and W. Bowen. 1996. Grey seal predation on the eastern Scotian Shelf: modelling
the impact on Atlantic cod. Can. J. Fish. Aquat. Sci., 53: 2722-2738.
Mori, M. and Butterworth, D.S. 2004. Consideration of multi-species interaction in
the Antarctic: an initial model of the minke whale-blue whale-krill interaction. In
L.J. Shannon K.L. Cochrane and S.C. Pillar (eds). Ecosystem approaches to fisheries in
the Southern Benguela. Afr. J. Mar. Sci., 26: 245-259.
Mori, M. and Butterworth, D.S. 2005. Modelling the predator-prey interactions of krill,
baleen whales and seals in the Antarctic ecosystem. Workshop document presented to
WG-EMM subgroup of CCAMLR, WG-EMM-05/34. 51 pp.
Mori, M. and Butterworth, D.S. 2006. A first step towards modelling the krill - predator
dynamics of the Antarctic ecosystem. CCAMLR Sci., 13: 217-277.
Murray, A.G. and Parslow, J.S. 1999. Modelling of nutrient impacts in Port Philip Bay – a
semi-enclosed marine Australian ecosystem. Mar. Freshwater Res., 50: 597-611.
NAMMCO [North Atlantic Marine Mammal Commission]. 2003. Report of the
NAMMCO Scientific Committee Working Group on Modelling Marine Mammal
Fisheries Interactions in the North Atlantic, Annex 2, Report of the 10th Meeting of
the Scientific Committee. In Annual Report 2003. NAMMCO, Tromsø: pp. 259-281.

NAMMCO [North Atlantic Marine Mammal Commission]. 1998. Report of the

Scientific Committee Working Group on the Role of Minke Whales, Harp Seals and
Hooded Seals in the North Atlantic Ecosystems. In NAMMCO Annual Report 1997.
NAMMCO, Tromsø: pp. 125-146.
NAMMCO [North Atlantic Marine Mammal Commission]. 2001. Report of the
References 75

Scientific Committee Working Group on the Economic Aspects of Marine Mammal

– Fisheries Interactions. In NAMMCO Annual Report 2000. NAMMCO, Tromsø:
pp. 169-202.
NAMMCO [North Atlantic Marine Mammal Commission]. 2002. Report of the Scientific
Committee. In NAMMCO Annual Report 2001. NAMMCO, Tromsø, pp. 147-249.
NAMMCO [North Atlantic Marine Mammal Commission]. 2003. Report of the Scientific
Committee. In NAMMCO Annual Report 2002. NAMMCO, Tromsø, pp. 173-259.
Navarrete, S.A., Wieters, E.A., Broitman, B.R. and J.C. Castilla. 2005. Scales of benthic-
pelagic coupling and the intensity of species interactions: from recruitment limitation to
top-down control. PNAS 102: 18046-18051.
Nel, D. 2005. Report on the Risk Assessment for Sustainable Fisheries (RASF) Workshop
for the South African Demersal Hake Fishery, Cape Town, 30 May-3 June 2005.
Department of Environmental Affairs and Tourism: 38 pp.
Nishikawa, H. and Yasuda, I. 2005. Population decline of Japanese sardine and variation
of mixed layer depth in the Kuroshio Extension. North Pacific Marine Science
Organisation Fourteenth Annual Meeting, Program Abstracts, 19 September–9 October
2005, Vladivostok, Russia. 52 pp.
Parma, A.M., Orensanz, J.M., Elías, I. and Jerez, G. 2003. Diving for shellfish and data:
incentives for the participation of fishers in the monitoring and management of artisanal
fisheries around southern South America. In S.J. Newman, D.J. Gaughan, G. Jackson,
M.C. Mackie, B. Molony, J. St. John and P. Kailola (eds). Towards Sustainability of Data-
Limited Multi-Sector Fisheries. Australian Society for Fish Biology Workshop Proceedings,
Bunbury, Western Australia, 23-24 September 2001. Fisheries Occasional Publications No.
6, May 2003, Department of Fisheries, Perth, Western Australia, 186 pp.
Pauly, D. and Christensen, V. 1995. Primary production required to sustain global
fisheries. Nature, 374: 255-257.
Pauly, D., Christensen, V. and Walters, C. 2000. ECOPATH, ECOSIM and ECOSPACE
as tools for evaluating ecosystem impact of fisheries. ICES J Mar. Sci., 57: 697-706.
Pauly, D., Christensen, V., Dalsgaard, A., Froese, R. and Torres, J. 1998. Fishing down
marine food webs. Science, 279: 860-863.
Pech, N., Samba, A., Drapeau, L., Sabatier, R. and Laloë, F. 2001. Fitting a model of
flexible multifleet-multi-species fisheries to Senegalese artisanal fishery data. Aquat.
Living Resour., 14: 81-98.
Pinnegar, J.K., Blanchard, J.L., Mackinson, S., Scott, R.D. and Duplisea, D.E. 2005.
Aggregation and removal of weak-links in food-web models: system stability and
recovery from disturbance. Ecol. Model., 184: 229-248.
Pitcher, T. and Cochrane, K.L. 2002. The use of ecosystem models to investigate multi-
species management strategies for capture fisheries. Fisheries Centre Research Reports,
10(2): 156.
Plagányi, É.E. 2004. Walking the bridge from single- to multi-species approaches
in southern African fisheries management. PhD thesis, University of Cape Town,
Rondebosch, South Africa.
Plagányi, É.E. and Butterworth, D.S. 2002. Competition with fisheries. In W.F. Perrin,
B. Würsig and H.G.M. Thewissen (eds). Encyclopedia of Marine Mammals, pp. 268-273.
San Diego, Academic Press.
Plagányi, É.E. and Butterworth, D.S. 2004. A critical look at the potential of Ecopath
with Ecosim to assist in practical fisheries management. In L.J. Shannon, K.L. Cochrane
and S.C. Pillar (eds). Ecosystem approaches to fisheries in the Southern Benguela. Afr. J.
Mar. Sci., 26: 261-287.
Plagányi, É.E. and Butterworth, D.S. 2005a. Indirect fishery interactions. In J.E.
76 Models for an ecosystem approach to fisheries

Reynolds III, W.F. Perrin, R.R. Reeves, T.J. Ragen and S. Montgomery (eds). Looking to
the Horizon: Future Directions in Marine Mammal Research, pp. 19-45. John Hopkins
University Press, Baltimore, Maryland. 240 pp.
Plagányi, É.E. and Butterworth, D.S. 2005b. Modelling the impact of krill fishing on
seal and penguin colonies. Workshop document presented to WG-EMM subgroup of
CCAMLR, WG-EMM-05/14. 14 pp.
Plagányi, É.E. and Butterworth, D.S. 2006a. A spatial multi-species operating model
(SMOM) of krill-predator interactions in small-scale management units in the Scotia
Sea. Workshop document presented to WG-EMM subgroup of CCAMLR, WG-EMM-
06/12. 28 pp.
Plagányi, É.E. and Butterworth, D.S. 2006b. An illustrative management procedure for
exploring dynamic feedback in krill catch limit allocations among small-scale management
units. Workshop document presented to WG-EMM subgroup of CCAMLR, WG-
EMM-06/28. 17 pp.
Plagányi, É.E. and Butterworth, D.S. (in press). Competition between marine mammals
and fisheries – can we successfully model this using ECOPATH with ECOSIM?
Proceedings of the 4th World Fisheries Congress, Vancouver, Canada, 2004.
Plagányi, É.E., Rademeyer, R.A., Butterworth, D.S., Johnston, S.J. and Cunningham,
C.L. 2007. Making management procedures operational – innovations implemented in
South Africa. ICES J. Mar. Sci, 64.
Polovina, J.J. 1984. Model of a coral reef ecosystem I. The ECOPATH model and its
application to French Frigate Shoals. Coral Reefs, 3: 1-11.
Pope, J.G. 1979. A modified cohort analysis in which constant natural mortality is replaced
by estimates of predation levels. ICES Committee Meeting, 1979/H: 16.
Pope, J.G. 1991. The ICES multi-species assessment group: evolution, insights and future
problems. ICES Mar. Sci. Symp., 193: 23-33.
Pope, J.G., Stokes, T.K., Murawski, S.A. and Idoine, S. 1988. A comparison of fish
size-composition in the North Sea and on Georges Bank. In W. Wolff, C.J. Soeder and
F.R. Drepper (eds). Ecodynamics, Contributions to Theoretical Ecology, pp. 146-152.
Springer, Berlin. 349 pp.
Punt, A.E. and Butterworth, D.S. 1995. The effects of future consumption by the
Cape fur seal on catches and catch rates of the Cape hakes. 4. Modelling the biological
interaction between Cape fur seals (Arctocephalus pusillus pusillus) and Cape hakes
(Merluccius capensis and M. paradoxus). S. Afr. J. Mar. Sci., 16: 255-285.
Punt, A.E. and Hilborn, R. 1997. Fisheries stock assessment and decision analysis: the
Bayesian approach. Rev. Fish. Biol. Fish., 7: 35-63.
Punt, A.E. and Leslie, R.W. 1995. The effects of future consumption by the Cape fur seal
on catches and catch rates of the Cape hakes. 1. Feeding and diet of the Cape hakes
(Merluccius capensis and M. paradoxus). S. Afr. J. Mar. Sci., 16: 37-55.
Punt, A.E., Smith, A.D.M. and Cui, G. 2002. Evaluation of management tools for
Australia’s South East Fishery. 1. Modelling the South East Fishery taking account of
technical interactions. Marine and Freshwater Research, 53: 615-629.
Quince, C., Higgs, P.G. and McKane, A.J. 2005. Topological structure and interaction
strengths in model food webs. Ecol. Model., 187: 389-412.
R Development Core Team. 2003. R: A language and environment for statistical computing.
R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051 00-3, <http://
www.R-project.org>.
Rademeyer, R.A., Plagányi, É.E. and Butterworth, D.S. 2005. Management Procedure
Robustness Tests for the South African Merluccius paradoxus and M. capensis Resources
under Empirical-Based Candidate OMP1a-ii. Marine and Coastal Management document
WG/11/05/D:H:45. 5 pp.
Rademeyer, R.A., Plagányi, É.E. and Butterworth, D.S. 2007. Tips and tricks in designing
References 77

management procedures. ICES J. Mar. Sci., 64: 626-32.

Raick, C., Soetaert, K. and Grégoire, M. 2006. Model complexity and performance: How
far can we simplify? Prog. Oceangr., 70: 27-57.
Rice, J.C. 2000. Evaluating fishery impacts using metrics of community structure.
ICES J. Mar. Sci., 57: 682-688.
Rice, J.C., Daan, N., Pope, J.G. and Gislason, H. 1991. The stability of estimates of
suitabilities in MSVPA over four years of data from predator stomachs. ICES Mar.
Sci. Symp., 193: 34-45.
Rindorf, A. 2002. The effect of stomach fullness on food intake of whiting in the North
Sea. J. Fish. Biol., 61: 579-593.
Rindorf, A. Gislason, H. and Lewy, H. 1998. Does the diet of cod and whiting reflect the
species composition estimated from trawl surveys? ICES C.M. 1998/CC: 5.
Sainsbury, K.J. 1991. Application of an experimental management approach to management
of a tropical multi-species fishery with highly uncertain dynamics. ICES Mar. Sci. Symp.,
193: 301–320.
Sainsbury, K.J., Campbell, R.A., Lindholm, R. and Whitelaw, W. 1997. Experimental
management of an Australian multi-species fishery: examining the possibility of trawl-
induced habitat modification. In E.L. Pikitch, D.D, Huppert and M.P. Sissenwine (eds).
Global Trends: Fisheries Management. American Fisheries Society Symposium, 20: 107-
112. Bethesda, Maryland, USA
Sainsbury, K.J., Punt, A.E. and Smith, A.D.M. 2000. Design of operational management
strategies for achieving fishery ecosystem objectives. ICES J. Mar. Sci., 57: 731-741.
Savenkoff, C., Vézina, A.F. and Bundy, A. 2001. Inverse analysis of the structure
and dynamics of the whole Newfoundland-Labrador Shelf ecosystem. Can.
Tech. Rep. Fish. Aquat. Sci., 2354: viii+56 pp
Savenkoff, C., Castonguay, M., Vézina, A.F., Despatie, S-P., Chabot, D., Morissette, L.
and Hammill, M.O. 2004. Inverse modelling of trophic flows through an entire
ecosystem: the northern Gulf of St. Lawrence in the mid-1980s. Can. J. Fish. Aquat. Sci.,
61: 2194-2214.
Scheffer, M., Carpenter, S. and de Young, B. 2005. Cascading effects of overfishing marine
systems. TREE 20: 579-581.
Schindler, D. and Eby, L. 1997. Stochiometry of fishes and their prey: implications for
nutrient cycling. Ecology, 78: 1816-1831.
Schweder, T., Hagen, G.S. and Hatlebakk, E. 2000. Direct and indirect effects of minke
whale abundance on cod and herring fisheries: a scenario experiment for the Greater
Barents Sea. In G.A. Víkingsson and F.O. Kappel (eds). Minke whales, harp and
hooded seals: major predators in the North Atlantic ecosystem. NAMMCO Scientific
Publications, Vol. 2: 120-132.
Sekine, M., Nakanishi, H., Ukita, M. and Murakami, S. 1991. A shallow-sea ecological
model using an object-oriented programming language. Ecol. Model., 57: 221-236.
Sekine, M., Nakanishi, H. and Ukita, M. 1996. Study on fish mortality caused by the
combined effects of pesticides and changes in environmental conditions. Ecol. Model.,
86: 259-264.
Sekine, M., Imai, T. and Ukita, M. 1997. A model of fish distribution in rivers according
to their preference for environmental factors. Ecol. Model., 104: 215-230.
Shannon, L.J., Cochrane, K.L. and Pillar, S.C. 2004. Ecosystem approaches to fisheries in
the Southern Benguela. Afr. J. Mar. Sci., 26: 328 pp.
Shannon, L.J., Cochrane, K.L., Moloney, C.L. and Fréon, P. 2004. Ecosystem approach
to fisheries management in the southern Benguela: a workshop overview. Afr. J. Mar.
Sci., 26: 1-8.
Shin, Y-J. and Cury, P. 2001. Exploring fish community dynamics through size-dependent
trophic interactions using a spatialized individual-based model. Aquat. Living Resour.,
78 Models for an ecosystem approach to fisheries

14: 65-80.
Shin, Y-J. and CuryP. 2004. Using an individual-based model of fish assemblages to study
the response of size spectra to changes in fishing. Can. J. Fish. Aquat. Sci., 61: 414-431.
Shin, Y-J., Rochet, M.-J., Jennings, S., Field, J.G. and Gislason, H. 2005. Using size-based
indicators to evaluate the ecosystem effects of fishing. ICES J. Mar. Sci., 62: 384-396.
Shin, Y-J., Shannon, L.J. and Cury, P.M. 2004. Simulations of fishing effects on the
southern Benguela fish community using the individual-based model: learning from
a comparison with ECOSIM. In L.J. Shannon, K.L. Cochrane and S.C. Pillar (eds).
Ecosystem approaches to fisheries in the Southern Benguela. Afr. J. Mar. Sci., 26: 95-113.
Sinclair, M. and Valdimarsson, G. 2003. Responsible fisheries in the marine ecosystem.
Rome, Italy and Wallingford, UK. FAO and CAB International.
Smith, A.D.M., Sainsbury, K.J. and Stevens, R.A. 1999. Implementing effective fisheries-
management systems: management strategy evaluation and the Australian partnership
approach. ICES J. Mar. Sci., 56: 967-979.
Smith, A.D.M, Shoulder, P., Sachse, M., Baelde, P., Geen, G., Prince, J., Knuckey.,
Smith, D., and Walker, T. 2004. Alternative Management Strategies for the Southern and
Eastern Scalefish and Shark Fishery: Qualitative Assessment, Stage 1 Report. CSIRO,
Hobart.
Solé, J., Estrada, M. and Garcia-Ladona, E. 2006. Biological conrol of harmful algal
blooms: a modelling study. J. Mar. Sys.
Somhlaba, S., Plagányi, É.E, Butterworth, D.S. and Rademeyer, R.A. 2004. What is the
potential for the constraints associated with ECOPATH to improve estimates of biomass
and productivity in the southern Benguela region? Document presented to BENEFIT-
NRF-BCLME Stock Assessment Workshop, January 2004, University of Cape Town.
BEN/JAN04/SAH/5a: 12 pp.
Somhlaba, S. 2006. Observer Effects on Pelagic Fleet Operations and Bayesian-based
explorations of the mass balance model Ecopath. MSc thesis, University of Cape Town.
109 pp.
Sparholt, H. 1991. Multi-species assessment of Baltic fish stocks. ICES Mar. Sci. Symp.,
193: 64-79.
Sparre, P. 1991. Introduction to multi-species virtual population analysis. ICES Mar. Sci.
Symp., 193: 12-21.
Steele, J.H. and Henderson, E.W. 1994. Coupling between physical and biological scales.
Phil. Trans. R. Soc. Lond. B, Biol. Sci., 343: 5-9.
Stefansson, G. and Palsson, O.K. 1998. The framework for multi-species modelling of
Arcto-boreal systems. Rev. Fish Biol. Fish., 8: 101-104.
Stefansson, G., Sigurjónsson, J. and Víkingsson, G.A. 1997. On dynamic interactions
between some fish resources and cetaceans off Iceland based on a simulation model.
J. Northw. Atl. Fish. Sci., 22: 357-370.
Stephenson, R.L. and Lane, D.E. 1995. Fisheries Management Science: a plea for
conceptual change. Can. J. Fish. Aquat. Sci., 52: 2051-2056.
Taylor, L., Begley, J., Kupca, V. and Stefansson, G. 2004. A simple implementation of
gadget for cod in Icelandic waters. ICES CM 2004/FF: 23.
Taylor, L. and G. Stefansson. 2004. Gadget models of cod-capelin-shrimp interactions in
Icelandic waters. ICES CM 2004/FF: 29.
Thomson, R.B., Butterworth, D.S., Boyd, I.L. and Croxall, J.P. 2000. Modeling
the consequences of Antarctic krill harvesting on Antarctic fur seals. Ecol. Appl.,
10: 1806- 1819.
Thomson, R.B., Butterworth, D.S. and Kato, H. 1999. Has the age at transition of
southern hemisphere minke whales declined over recent decades? Mar. Mamm. Sci.,
15: 661-682.
Tjelmeland, S. 1997. Can the management of the Barents Sea capelin stock be improved
by Multi-species modelling? Forage fishes in marine ecosystems. In Proceedings of the
References 79

International Symposium on the Role of Forage Fish in Marine Ecosystems, pp. 617-632.
Alaska Sea Grant College Program Report No. 97-01. University of Alaska Fairbanks,
1997.
Tjelmeland, S. and Bogstad, B. 1998. MULTSPEC – a review of a Multi-species modelling
project for the Barents Sea. Fish. Res., 37: 127-142.
Tjelmeland, S. and Lindstrøm, U. 2005. An ecosystem element added to the assessment
of Norwegian spring spawning herring: implementing predation by minke whales.
ICES J. Mar. Sci., 62: 285-294.
Trenkel, V.M., Pinnegar, J.K. and Tidd, A.N. 2004. Can Multi-species models be expected
to provide better assessments for Celtic sea groundfish stocks? ICES CM 2004/FF: 05.
Trites, A.W., Christensen, V. and D. Pauly. 1997. Competition between fisheries and
marine mammals for prey and primary production in the Pacific Ocean. J. Northw. Atl.
Fish. Sci., 22: 173-187.
Trites, A.W., Livingston, P.A., Mackinson, S., Vasconcellos M.C., Springer, A.M. and
Pauly, D. 1999. Ecosystem change and the decline of marine mammals in the Eastern
Bering Sea. Testing the ecosystem shift and commercial whaling hypotheses. Fisheries
Centre Research Reports, University of British Columbia. Vol. 7. 106 pp.
Tschirhart, J. and Finnoff, D. 2003. Protecting an Endangered Species while Harvesting its
Prey in a General Equilibrium Ecosystem Model. Land Economics, 79: 160-180.
Tschirhart, J. 2004 A New Adaptive System Approach to Predator-prey Modeling.
Ecol. Model., 176: 255-276.
Tsou, T.S. and Collie, J.S. 2001. Predation-mediated recruitment in the Georges Bank fish
community. ICES J. Mar. Sci., 58: 994-1001.
Van Nes, E.H., Lammens, E.H.R.R. and Scheffer, M. 2002. PISCATOR, an individual-
based model to analyze the dynamics of lake fish communities. Ecol. Model., 152: 261-
278.
Van Winkle, W., Rose, K.A. and Chambers, R.C. 1993. Individual-based approach to fish
population dynamics: an overview. Trans. Am. Fish. Soc., 122: 397-403.
Varpe, Ø., Fiksen, Ø. and Slotte, A. 2005. Meta-ecosystems and biological energy
transport from ocean to coast: the ecological importance of herring migration. Oecologia,
146: 443-451.
Vasconcellos, M. and Gasalla, M.A. 2001. Fisheries catches and the carrying capacity of
marine ecosystems in southern Brazil. Fish. Res., 50: 279-295.
Vasconcellos, M., Heymans, J. and Bundy, A. 2002. The use of Ecosim to investigate
Multi-species harvesting strategies for capture fisheries of the Newfoundland-Labrador
shelf. In T. Pitcher and K.L. Cochrane (eds). The use of ecosystem models to investigate
multi-species management strategies for capture fisheries. Fisheries Centre Research
Reports, 10(2): 68-72.
Vasseur, D. and McCann, K. 2005. A mechanistic approach for modelling temperature-
dependent consumer-resource dynamics. Am. Nat., 165: 184-198.
Vézina, A.F. and Pahlow, M. 2003. Reconstruction of ecosystem flows using inverse
methods: how well do they work? J. Mar. Sys., 40/41: 55-77.
Vézina, A.F., Savenhoff, C., Roy, S., Klein, B., Rivkin, R., Therriault, J.-C. and Legendre,
L. 2000. Export of biogenic carbon and structure and dynamics of the pelagic food web
in the Gulf of St. Lawrence. Part 2. Inverse analysis. Deep Sea Res., 47: 609-635.
Vidal, L. and Pauly, D. 2004. Integration of subsystems models as a tool toward describing
feeding interactions and fisheries impacts in a large marine ecosystem, the Gulf of
Mexico. Ocean and Coastal Manage., 47: 709-725.
Vinther, M. 2001. Ad hoc multi-species VPA tuning applied for the Baltic and North Sea
fish stocks. ICES J. Mar. Sci,. 58: 311-320.
Vinther, M., Lewy, P., Thomsen, L. and Petersen, U. 1998. Specification and documentation
of the 4M Package containing Multi-species, Multi-fleet and Multi-Area Models. Danish
Institute for Fisheries Research, Charlottenlund.
80 Models for an ecosystem approach to fisheries

Walters, C.J. 1986. Adaptive management of renewable resources. Macmillan, New York.
Walters, C.J., Christensen, V. and Pauly, D. 1997. Structuring dynamic models of exploited
ecosystems from trophic mass-balance assessments. Rev. Fish. Biol. Fish., 7: 139-172.
Walters, C.J., Christensen, V. and Pauly, D. 2002. Searching for optimum fishing strategies
for fishery development, recovery and sustainability. In T. Pitcher and K.L. Cochrane
(eds). The use of ecosystem models to investigate multi-species management strategies for
capture fisheries. Fisheries Centre Research Reports, 10(2): 11-15.
Walters, C.J., Christensen, V., Martell, S.J. andKitchell, J.F. 2005. Possible ecosystem
impacts of applying MSY policies from single-species assessment. ICES J. Mar. Sci.,
62: 558-568.
Walters, C. and Kitchell, J.F. 2001. Cultivation/depensation effects on juvenile survival and
recruitment: implications for the theory of fishing. Can. J. Fish. Aquat. Sci., 58: 39-50.
Walters, C. and Martell, S.J.D.2004. Fisheries ecology and management. Princeton
University Press, Princeton. 399 pp.
Walters, C.J., Pauly, D. and Christensen, V. 1999. ECOSPACE: prediction of mesoscale
spatial patterns in trophic relationships of exploited ecosystems, with emphasis on the
impacts of marine protected areas. Ecosystems, 2: 539-554.
Walters, C.J., Pauly, D., Christensen, V. and Kitchell, J.F. 2000. Representing density
dependent consequences of life history strategies in aquatic ecosystems: ECOSIM II.
Ecosystems, 3: 70-83.
Ware, D.M. and Thomson, R.E. 2005. Bottom-up ecosystem trophic dynamics determine
fish production in the Northeast Pacific. Science, 308: 1280-1284.
Watters, G.M., Hinke, J.T., Reid, K. and Hill, S. 2005. A krill-predator-fishery model for
evaluating candidate management procedures. CCAMLR WG-EMM-05/13.
Watters, G.M., Hinke, J.T., Reid, K. and Hill, S. 2006. KPFM2, be careful what you ask
for – you just might get it. CCAMLR WG-EMM-06/22.
Worthington, D.G. and Andrew, N.L. 1998. Small-scale variation in demography and its
implications for an alternative size limit in the fishery for blacklip abalone (Haliotis rubra)
in New South Wales, Australia. In G.S. Jamieson and A. Campbell (eds). Proceedings of
the North Pacific Symposium on Invertebrate Stock Assessment and Management. Can.
Spec. Publ. Fish. Aquat. Sci., 125: 341-348.
WSSD (World Summit on Sustainable Development). 2002. World Summit on
Sustainable Development: Plan of implementation. IUCN document. 54 pp.
Available from <http://www.iucn.org/wssd/files/>
Yemane, D., Field, J.G. and Griffiths, M.H. 2004. Effects of fishing on the size and
dominance structure of linefish of the Cape region, South Africa. In L.J. Shannon,
K.L. Cochrane and S.C. Pillar (eds). Ecosystem approaches to fisheries in the Southern
Benguela. Afr. J. Mar. Sci., 26: 161-177.
Yodzis, P. 1988. The indeterminacy of ecological interactions as perceived through
perturbation experiments. Ecology, 69: 508-515.
Yodzis, P. 1998. Local trophodynamics and the interaction of marine mammals and fisheries
in the Benguela ecosystem. J. Anim. Ecol., 67: 635-658.
Yodzis, P. 2000. Diffuse effects in food webs. Ecology, 81: 261-266.
Yodzis, P. and Innes, S. 1992. Body size and consumer-resource dynamics. Am. Nat., 139:
1151-1175.
Appendixes
Appendixes 83

TABLE A1
Methods available for assessing the impacts of ecological (indirect) interactions between species
and fisheries and their implications for fisheries management. Model comparison including
comparison of level of complexity and realism, functional responses, dealing with uncertainty,
incorporation of environmental effects, spatial representation, handling of migratory species,
adequacy re assessing different management controls and effects of ecosystem changes,
suitability to conduct assessment and policy exploration, transparency of operation and
suitability for data poor areas

TABLE A1a MODEL COMPARISON

Type of model Whole ecosystem Biogeochemical Biogeochemical Dynamic Biogeochemical
models ecosystem ecosystem models multispecies ecosystem models
models models
MODEL Ecopath with IGBEM ATLANTIS INVITRO ERSEM II
Ecosim

1.
Level of
complexity
and realism

a) Can be very Large: 20-30 > 20 typically, 10-20 groups 10-20 groups,
No. of large; typically though to date used typically mostly
modelled around 30 with 15-61 groups (including phytoplankton
species/groups (with multiple stocks habitat groups) and
per group in some zooplankton
cases)

b) Recently full Vertebrates - Vertebrates - age- Detailed Aggregate

Representation age-structure age-structured structured models; representations, biomass pools
of size/age capability for models; invertebrate and including
structure groups invertebrate primary producer age and size
and primary groups (defined structure
producer based on role and
groups - size) - aggregate
aggregate biomass pools; some
biomass pools invertebrate age
structuring

c) Can be included Detailed Detailed Detailed Detailed

Physical/ to limited representation representation of representation representation
biological extent of physical physical processes of physical with e.g.
processes processes, with model driven forces, but light and
input forcing by seasonal variation not nutrients temperature
of nutrients in irradiance and (usually) forcing
and physics temperature, functions
nutrient inputs
from point sources,
atmospheric
nutrient inputs and
exchanges with
oceanic boundary
components
d) Can be included Fishery Excellent Some bycatch No
Technical discards representation; groups, discards
interactions - target includes bycatch and incidental
species. Some groups e.g. impacts are
incidental discarded non-target represented
fishing groups, landed and
mortality marketed non-target
effects on by-product groups
bycatch groups
2. Foraging arena Mixed Flexible e.g. Type II Depending Type II
Functional formulation (Type II, or Type III or other on agent
responses (see text) Type III) types used
By choosing there can be
appropriate explicit feeding
parameter interactions OR
combinations, the state of the
EwE can habitat is taken
generate as a proxy for
a range of foodweb state
functional and fauna is
responses assumed to
including Types be getting its
II and III ration if the
habitat is in
good condition
84 Models for an ecosystem approach to fisheries

TABLE A1a (continued)

Type of model Whole ecosystem Biogeochemical Biogeochemical Dynamic multispecies Biogeochemical
models ecosystem models ecosystem models models ecosystem models
MODEL Ecopath with Ecosim IGBEM ATLANTIS INVITRO ERSEM II

3. ECORANGER Aspects Aspects Aspects considered Explored to a

Uncertainties - although this considered by considered by by bounding using limited extent
in model should/could be Fulton (2001), Fulton (2001), “pessimistic”,
structure, improved; recent Fulton et al. Fulton et al. “middle-of-the-road”
parameters improvements (2004a) (2004a,b) - no and “optimistic”
and data include capabilities formal fitting parameterisations.
to balance to data within Some components
models based on the modelling (in particular target
uncertainty, fitting software, species, fisheries and
to time series though limited biogenic habitat)
and quantifying fitting happens undergo formal
input parameter externally to fitting
uncertainty by the model
running ECOSIM (no feedback
using a Monte Carlo estimation as yet)
approach

4a) Incorporates a Detailed Detailed Forcing is typically Detailed

Environmental facility in the form consideration. consideration currents, winds, consideration -
effects of a (seasonal - light, nutrient, rainfall and light, nutrient,
or longer term) temperature catastrophes temperature
forcing function inputs; long-term inputs; good
routine to represent climate anomaly representation
the mediation of data of river inputs
physical or other and
environmental atmospheric
parameters nutrient inputs

4b) Major focus of More of a focus More of a focus Some consideration, N/A
Interactions approach than target than target but main focus is on
with non- groups groups target, vulnerable
target species and habitat species

5.
Spatial
representation

a) Not explicitly but Spatially explicit Spatially explicit Spatially explicit No

Species implicitly to some representation representation
interactions extent due to
foraging arena
formulation

b) No explicit spatial Detailed Polygonal Three dimensional Good

Habitat related representation representations geometry in continuous space, representation
processes in ECOSIM but matches with explicit habitats of transport
ECOSPACE is geographical (and habitat related processes for
spatially resolved features; multiple processes) plankton groups
vertical water
column layers;
subgrid scale
representation
of physical and
habitat properties

6. Doesn’t handle No - aggregated Movement Movement through N/A

Migratory particularly well; species groups (migration and in/out of the
species ECOSPACE has more and advective modelled area
potential transfer) between
areas and vertical
layers (and also
in/out of the
model domain)
Appendixes 85

TABLE A1a (continued)

Type of model Whole Biogeochemical Biogeochemical Dynamic Biogeochemical
ecosystem ecosystem ecosystem models multispecies ecosystem
models models models models
MODEL Ecopath with IGBEM ATLANTIS INVITRO ERSEM II
Ecosim

7. Good (see e.g. Can be used Can be used to Used to explore None
Model Pitcher and to explore explore alternative alternative
adequacy to Cochrane, 2002) alternative fisheries strategies and
allow analysis fisheries management management
of different management strategies institutional
types of strategies (including both arrangements
management (including both ecologically and (usually in
controls in use
ecologically and economically multiple use
economically motivated policies) management
motivated context)
policies)

8. Good Good Good Good Good for short-

Model term but not
adequacy long-term;
to allow can predict
assessment response to
of effects of short-term
short-, medium- climatic impacts
and long-term
ecosystem
changes
9. Excellent (see No Well suited Reasonable No
Model e.g. Pitcher and
suitability Cochrane, 2002)
to conduct
assessment
and policy
exploration
10. By far the Not very well Good model Documented Model details
Model easiest model documented transparency but no easy published
transparency to use; some (due to but no easy user user interface. and relatively
of operation issues re complexity) and interface and slow Parameterisation easy to use
and ease of transparency presumably not and laborious and calibration for the North
use as code is straightforward calibration. software is under Sea but not
constantly to use Parameterisation development. straightforward
evolving and and calibration to apply to
not always well support software other systems
documented is under
and described development

11. Less data Not suitable Data intensive - Mixed Data intensive
Data intensive than for other than not suitable (dependent on - not suitable
requirements biogeochemical very intensively agent types
and model models but studied systems selected)
suitability requires e.g. Port Philip
for data poor data that Bay, North Sea
areas are difficult
to obtain
such as diet
compositions
and species
abundance
estimates
86 Models for an ecosystem approach to fisheries

TABLE A1b MODEL COMPARISON

MODEL SSEM KPFM MRM e.g. Punt and MSVPA and MSM
Butterworth (1995) MSFOR

1.
Level of
complexity and
realism
a) Lumped Currently 1-4 Typically few e.g. Typically few Thus far
No. of model predator stocks 4 components (6-8) 2 species
modelled components within each (walleye
species/groups e.g. fish, SSMU (Small- pollock
plankton, Scale Spatial and Pacific
nutrients Unit) cod - and
cannibalism)
but could be
extended
b) Aggregate Krill: juvenile Detailed Detailed Fully age-
Representation biomass pools and adult representations representations structured
of size/age components; - age structure - age structure
structure predators:
juvenile,
breeding and
non-breeding
components
c) Detailed Coupled to No physical Not usually None
Physical/ representation physical model represented
biological with e.g. to simulate
processes forcing using transport of
temperature, krill
current and
nutrient loads
from land
d) No No Not included Can be included Not currently
Technical included
interactions
2. Type II Flexible - Type II Fixed ration that Based on
Functional Holling Type is independent Type II
responses II and Type of prey
III functional abundance in
responses forecasts
Appendixes 87

TABLE A1b (continued)

MODEL SSEM KPFM MRM e.g. Punt MSVPA and MSM
and Butterworth MSFOR
(1995)

3. Unknown Monte Carlo Model fits to Explored to Good

Uncertainties simulations to available data. some extent consideration
in model investigate numerical Good initial of these
structure, uncertainty; explorations;
parameters robustness to could perhaps
and data alternative model be improved
formulations using e.g.
explored; no formal Bayesian
fitting to data and methods
hence considerable
uncertainty re some
parameter values
which are input
4a) Forcing Some forcing Not included Can be Not included
Environmental - currents, from e.g. currents included
effects nutrient, and several
temperature formulations linked to
inputs environmental index
4b) Interactions N/A Investigates effects Minor only Minor only Not currently
with non- of krill as target considered
target species species on non-target
predator species
5.
Spatial
representation
a) No Spatially explicit at Not spatial Not spatial No
Species scale of SSMUs but
interactions not at smaller scales
b) No Model’s spatial cells No No No
Habitat related match SSMUs which
processes can have different
physical and biological
features
6. N/A Simulates movements No No Not suitable
Migratory of krill but not
species predators
88 Models for an ecosystem approach to fisheries

TABLE A1b (continued)

MODEL SSEM KPFM MRM e.g. MSVPA and MSM
Punt and MSFOR
Butterworth
(1995)

7. None Designed to Excellent Some Some

Model address options
adequacy to for subdivision
allow analysis of the
of different precautionary
types of krill catch limit
management amongst SSMUs
controls in use
8. Short-term effects Some No No No
Model of changes in
adequacy coastal system
to allow
assessment
of effects of
short-, medium-
and long-term
ecosystem
changes
9. No Designed to Excellent Some No
Model address options contributions
suitability for subdivision
to conduct of the
assessment precautionary
and policy krill catch limit
exploration amongst SSMUs
10. Model details Model still being Detailed Good model Average
Model published ; developed so model descriptions; transparency
transparency easiness of use not generally descriptions moderately but not easy
of operation difficult to assess available yet but easy to use to use
and ease of complicated
use and time-
consuming to
use
11. Data intensive Can be adapted Fairly data Detailed Some
Data but lumped to match level intensive but stomach potential as
requirements components mean of data available focuses on a content data focuses on
and model it may not be as few target input to few/target
suitability bad as some other species only model makes species for
for data poor biogeochemical for which it unsuitable which there
areas models more data for most are typically
usually exists regions, some data
even in data but there
poor areas are hybrid
versions that
require less
data
Appendixes 89

TABLE A1c MODEL COMPARISON

MODEL MULTSPEC GADGET Bioenergetic/ OSMOSE SEAPODYM
allometric
models e.g.
Koen-Alonso
and Yodzis,
2005

1.
Level of
complexity and
realism
a) Typically few Few with From 4 to as 7-20 species Thus far 3 tuna
No. of (3-5) potential for many as 29 species (skipjack,
modelled many yellowfin and
species/groups bigeye) but could
be extended
b) Detailed Detailed Not Detailed Detailed
Representation representations representations represented representations representations
of size/age - species split of age structure
structure by size and age of fish; lumped
plankton forage
components

c) Could be linked Spatial model Not Not represented Time-series of

Physical/ to oceanographic can be coupled represented environmental
biological models; Sea to ocean data in the form
processes temperature circulation of temperature,
affects fish model currents etc;
growth, can be coupled
maximal food to physical/
consumption biogeochemical
and cod stomach models
evacuation rate;
climatological
data used
d) Not represented Included Not Not included Not included but
Technical represented the manual notes
interactions that important
by-catch species
(e.g. marine
turtles, seabirds)
could be included
in future versions
2. Marine mammals Flexible e.g. Tested 5 Fixed ration; Fixed ration
Functional - fixed ration; Type II or Type different starvation model
responses cod: feeding III or other forms: mortality
affected by multi-species component
individual size Holling
at age, prey Type II with
biomass and predator
temperature; interference;
all fish species: multi-species
curvilinear generalized
relationship Holling;
assumed frequency-
between food dependent
abundance and predation,
consumption Evans and
Ecosim
90 Models for an ecosystem approach to fisheries

TABLE A1c (continued)

MODEL MULTSPEC GADGET Bioenergetic/allometric OSMOSE SEAPODYM
models e.g. Koen-
Alonso and Yodzis,
2005

3. Uncertainties Likelihood Uses combined Investigated structural Large Not well

in model function used simulated uncertainty by uncertainties explored;
structure, to estimate annealing and exploring sensitivity not rigorously Statistical
parameters maturation Hooke&Jeeves to alternative dealt with estimation of
and data parameters - optimisation functional response parameters
fit to empirical methods to representations; may be added
maturation estimate best explored parameter
data; also fit parameters uncertainty using the
likelihood according to a SIR algorithm (Punt
function re pre-specified and Hilborn, 1997,
predation likelihood McAllister et al.,
parameters function; 1994).
- based on modular
extensive form permits
stomach sensitivity
content investigation
data; several to range of
explorations alternative
re alternative model
model structures
formulations
and
hypotheses
(e.g. Bogstad
et al., 1992,
Tjelmeland,
1997) but
scope for more
4a) Not explicitly Bottom-up Not included Carrying Detailed
Environmental included but explorations capacity consideration
effects plankton e.g. using constraint of effects of
described using adapted can be varied temperature,
time-varying random walk to simulate currents, etc.;
functions (Hulse, 2001) e.g. random suitable for
with different or periodic investigating
parameters for dynamics climate change
various areas scenarios and
effect of e.g.
ENSO events
4b) Interactions Some Represented Some - sea lions Explicit Considers
with non- representation consideration impacts of
target species e.g. polar cod of non-target these on target
included in fish species but species and
model not other not really the
other way
around
5. Spatial
representation
a) Species Division into Spatially No Spatially Spatially
interactions areas (7 in explicit with explicit with explicit with
Barents Sea) to migration fish schools one degree
describe east- matrices moving to cells
west gradients specifying areas with
in individual movement highest
growth of between areas potential prey
species and biomass
migration
patterns
b) Habitat Minor only Could be No No Good (novel)
related e.g. different tailored by spatial
processes temperatures linking with representation
in different oceanographic of differences
areas models in habitat
quality (see
text for details)
6. Migratory Multiple areas Multiple areas No explicit modelling No Can be
species with migration with migration of migration handled
between areas between areas through
movement
model linked
to habitat
quality
Appendixes 91

TABLE A1c (continued)

MODEL MULTSPEC GADGET Bioenergetic/ OSMOSE SEAPODYM
allometric
models e.g.
Koen-Alonso
and Yodzis,
2005

7. Some e.g. can Excellent Minor No Can be used to

Model explore effects contributions explore impacts
adequacy to of catches e.g. questions of marine
allow analysis from different re culling sea protected areas,
of different areas lions no-fishing
types of areas as well
management as impacts of
controls in use
management
options on
different tuna
(or similar)
species
8. Limited - some Currently No Some Good for
Model climatological minor exploring short
adequacy data input contribution to medium
to allow only possible term changes
assessment in tuna (or
of effects of similar species)
short-, medium- distribution
and long-term
and possibly
ecosystem
changes
abundance
but not more
general
ecosystem
changes
9. Contributes Some Minor Minor Minor
Model to stock contributions contributions contributions
suitability assessment e.g. questions only
to conduct process; re culling
assessment Some policy sea lions and
and policy explorations conversely,
exploration e.g. extent
simulations to which
to explore commercially
scenarios in important
which larger hake fishery
cod catches has a negative
are taken in impact on sea
years with lions
decreased
predation
pressure from
minke whales
10. Good model Excellent Good model Good description Manual available
Model descriptions transparency description of model; ease of with good
transparency but does not but large but not easy use not known but description
of operation appear easy number of to use presumably not of model; An
and ease of to use options, and straightforward executable
use sophisticated version is
software and currently
minimisation available that is
routines, relatively easy to
make it run as requires
moderately changes to
difficult to use parameter file
- more difficult
to change the
model itself.
11. Detailed Model can be Not suitable, Based on Data intensive
Data stomach tailored to but may fairly general hence not
requirements content data available data, be possible parameters suitable for data
and model input required hence good to apply if so could be poor areas
suitability for model for data poor restricted to a applied but some
for data poor makes it areas. few species difficulties
areas unsuitable for
most regions
92 Models for an ecosystem approach to fisheries

TABLE A1d MODEL COMPARISON

MODEL CCAMLR models EPOC SMOM ESAM SEASTAR
e.g. Mori &
Butterworth
2005, 2006

1.
Level of
complexity and
realism
a) Typically few 2 in current Currently 2 Few - typically 2 Few - typically
No. of modelled e.g. 7 example; being predator stocks (and cannibalism) 2 (and
species/groups extended within each -4 cannibalism) - 4
SSMU
b) Not Can select Krill: lumped; Detailed Detailed
Representation represented to include predators: representations representations
of size/age detailed age or juvenile,
structure size-structure; breeding and
Trial example: non-breeding
krill: spatially components
and age-
structured;
predator: age-
aggregated
c) Not Various Can be Not represented Not represented
Physical/ represented formulations coupled to
biological can be physical model
processes accommodated to simulate
e.g. advance transport of krill
and retreat
of sea ice
modelled;
ocean
transport may
be included in
future
d) Not Not currently No Could be Could be
Technical represented represented represented
interactions
2. Type II and Type I Flexible - Holling Type I and II Variable e.g.
Functional Type III relationship in Type II and Type considered Type I, II or III
responses trial; designed III functional
to be flexible responses
Appendixes 93

TABLE A1d (continued)

MODEL CCAMLR models EPOC SMOM ESAM SEASTAR
e.g. Mori &
Butterworth
2005, 2006

3. Likelihood Should permit Reference Set Bayesian Usually

Uncertainties function sensitivity to used comprises methods; considered
in model used to fit alternative 12 alternative considered as rigorously
structure, model to all model combinations as rigorously as in single-
parameters and available data structures, that essentially as in single- species
data and indices of but no formal try to bound the species assessment
abundance; statistical uncertainty in the assessment approaches;
sensitivities testing/fitting choice of survival approaches. uncertainty
to alternative estimates as well evaluated
formulations as the breeding using e.g.
explored; success relationship; bootstrapping
need for Robustness to
a more alternative model
systematic formulations
exploration explored; Some
of sensitivity formal fitting to
to alternative data
input
parameter
choices

4a) Not included Could be Could be linked Not usually Not usually
Environmental linked to to other physical included included
effects other physical oceanographic
oceanographic models but not yet
models but not developed
yet developed

4b) Explicit Could be Investigates effects Focus is on Focus is on

Interactions consideration included of krill as target target species target species
with non-target of krill- species on non-
species whale-seal target predator
interactions species

5.
Spatial
representation
a) Limited (two Spatial Spatially explicit at Not usually Not usually
Species spatial strata) subdivision scale of SSMUs but
interactions into polygons not at smaller scales
(8 in trial
version)

b) No Not currently Model spatial cells No No

Habitat related match SSMUs which
processes can have different
physical and
biological features

6. No explicit Movement Simulates Not usually Not usually

Migratory modelling of matrix can be movements of krill
species migration included but not predators
94 Models for an ecosystem approach to fisheries

TABLE A1d (continued)

MODEL CCAMLR models e.g. EPOC SMOM ESAM SEASTAR
Mori & Butterworth
2005, 2006

7. Mori and Designed to Designed to Good Good

Model Butterworth (2006) achieve this address options
adequacy to not currently but not tested for subdivision
allow analysis sufficiently yet of the
of different developed precautionary
types of krill catch limit
management amongst SSMUs
controls in use
8. No Designed to Some No No
Model achieve this
adequacy but not tested
to allow yet
assessment
of effects of
short-, medium-
and long-term
ecosystem
changes
9. Some potential Designed to Designed to Some Some
Model e.g. to evaluate achieve this address options
suitability possible effects but not tested for subdivision
to conduct of decisions to yet of the
assessment harvest krill or precautionary
and policy particular whale or krill catch limit
exploration seal species amongst SSMUs
10. Model equations Currently Model still Good model Good model
Model very simple but poor model being developed transparency transparency
transparency not easy to use transparency so code not but not easy but not easy
of operation as user requires as still being generally to use to use
and ease of experience re developed available;
use coding and but should be Difficult to use
non-linear moderately by other than
minimisation easy to use experienced
programmer.
11. Requires at least Data intensive Can be adapted Detailed data Detailed data
Data some relative to match level only required only required
requirements abundance data; of data available for few target for few target
and model can be tailored species species
suitability to make the most
for data poor of limited data in
areas data poor area
Appendixes 95

TABLE A2
Model comparison including rough description of model parameters, some important
assumptions, data requirements, technical information, examples where used, model history
and additional useful features of each approach
TABLE A2a
Type of model Whole Biogeochemical Biogeochemical Dynamic Biogeochemical
ecosystem ecosystem models ecosystem models multispecies ecosystem models
models models
MODEL Ecopath with IGBEM ATLANTIS INVITRO ERSEM II
Ecosim

1. For each Requires in excess Many e.g. Many, but Many

Broad group: of 750 parameters phytoplankton basics are parameters e.g.
description of Biomass, P/B, to be estimated production to do with physiological
parameters Q/B, Catch, or input, though parameters such growth, parameters such
(not fully Discards, many ok at as maximum mortality, as maximum
comprehensive Refuge default settings temperature- fecundity growth rate,
as intended to parameters. dependent and speed of half-saturation
give a flavour
Diet growth rate, movement constant, faecal
of the sorts of
parameters)
composition light limitation ratio, excretion
matrix for factors and ratio, respiration
all species. half saturation ratio
Phytoplankton constants;
growth-related Also needs
parameters configuration
such as of foodweb
Michaelis- connections;
Menten uptake More parameters
parameters, needed if complex
maximum representations
P/B ratio for (like temperature
phytoplankton dependent
movement and
spawning) options
selected
2. Trophic Fish migration Functional Dependent Many
Some interactions represented groups describe on agent physiological
important are important; using forcing behaviour of types; habitat and process-
model foraging arena function,fish an “average” as a proxy related
assumptions formulation recruitment individual; in regional
constant spatially predators not applications
and temporally explicitly included (Little et al.
represented 2006)
using quadratic
mortality
terms; not all
prey available
to predators
(availability
parameter)
3. Preferably Very large data Spatially explicit Physical Detailed data
Data data on species requirements. biomass, model data, inputs for
requirements biomass and production, sediments, the North
P/B; spatially consumption, diet initial Sea including
and temporally composition for biomasses and hydrodynamical
appropriate major functional habitat map data re
diet groups, spatial advective
composition and fleet- and diffusive
data; catch disaggregated transport, global
history; time harvest rates; radiation and
series fisheries primary temperature,
data for fitting production rates river nutrient
and processes; loads, fishing
nutrient data; mortality
climate data
4. Runs on C++, could run in Coded in C++, Linux; code is Model coded
Technical Windows PC Linux could run in open source in FORTRAN90
details Linux; Can run on (i.e. available) - both code
(preferrably fast) and executable
PC; Code and exe available and
file available. can be run on
PC; C++ version
developed
96 Models for an ecosystem approach to fisheries

TABLE A2a (continued)

MODEL Ecopath with Ecosim IGBEM ATLANTIS INVITRO ERSEM II

5. Examples globally Port Philip Port Philip Bay Northwest North Sea; see
Examples e.g. Scotian shelf Bay - Australia - Australia; EEZ shelf of Journal of Sea
where used (Bundy, 2002, region for south- Australia Research vol. 38;
2005), Eastern eastern Australia; Mediterranean,
Bering and other continental Irish and Celtic Seas,
western Bering shelf, estuaries and Adriatic; also Catalan,
Sea shelf and slope bays in Australia Cretan and Arabian
ecosystems (Aydin and Tasmania; Seas (Blackford, Allen
et al., 2002), Gulf Northern California and Gilbert, 2004)
of California, Current (western US);
North Sea, Gulf Continental shelf of
of Thailand north-eastern US
(Christensen,
1998), Strait of
Georgia (Martell
et al., 2002),
Southern Benguela
Upwelling
region (Shannon,
Cochrane and
Pillar, 2004), Baltic
Sea (Harvey et al.,
2003), Black Sea
(Daskalov, 2002),
Pacific (Cox et al.,
2002), efficacy
of MPAs in the
central North
Pacific (Martell
et al., 2005) and
many more
6. ECOPATH based Based on Developed from Developed Developed to
History on Polovina amalgamating the “Bay Model 2” to consider simulate the
(1984) model but ERSEM (to ecosystem model of multiple use ecosystem dynamics
developed in user- represent Fulton et al. (2004); management of the North Sea
friendly format; biological first applied to Port questions for
transformed into processes) Philip Bay, Australia the marine
dynamic ECOSIM and PPBIM (especially
version which (to represent inshore/shelf)
has become very physical environment
popular due to processes and
ease of use; freely introduce
available software spatial
with good user structure);
interface and Constructed as
unparalled support a first step in
and training for understanding
users; ECOSPACE effects
developed to of model
handle spatial structure and
aspects such as complexity.
MPAs
7. Includes policy Alternative Includes discarding, Operating Can be linked with
Additional optimisation forms of fish bycatch and system-like models of fish
useful features routine; movement management asynchronous dynamics
ECOTRACER can and migration submodels; Includes time-step
be used to predict investigated alternative fisheries scheduler;
movement and submodels with Hydbrid form
accumulation of alternative bycatch, so best model
contaminants habitat dependency, form (either
and tracers; selelctivity,discarding aggregate
Multistanza and effort allocation state model
populations can - allows representation or IBM/ABM
be designated of effects such as formulation)
as hatchery effort displacement can be used -
populations; due to local stock best match for
Permits evaluation depletion and effect component
of equilibrium of MPAs; novel dynamics can
MSY reference density-dependent be used
points and vertebrate movement
“stock reduction scheme; Includes
analysis”; starvation; Other
ECOSPACE: can sectors represented
analyze impact simply; Socioeconomic
and placement of submodels available
marine protected (e.g. so can consider
areas and explore impacts of quota
fitness-dependent trading); Full MSE cycle
dispersal represented
Appendixes 97

TABLE A2b MODEL COMPARISON

Type of Model Biogeochemical Whole ecosystem Dynamic multispecies Dynamic Dynamic system
ecosystem models models models multispecies models models
MODEL SSEM KPFM MRM e.g. Punt and MSVPA and MSFOR MSM
Butterworth (1995)

1. Many parameters Many parameters; For hake and seal Suitability Initial 2-species
Broad e.g. physiological Krill e.g. background species: total daily parameters, application has
description of parameters such mortality rate, ration, feeding predation 124 parameters
parameters as maximum 4 recruitment function saturation mortality M2, related to initial
(not fully growth rate, parameters including parameter, spawner-recruit age structure
comprehensive half-saturation scalar that mediates parameter parameters, of populations,
as intended to constant, faecal environmental effects reflecting extent terminal fishing recruitment
give a flavour
ratio, excretion on krill, average of annual variation mortality rates, parameters,
of the sorts of
parameters)
ratio, respiration weight, historical in diet; Other residual natural fishing mortality
ratio catches, instantaneous predatory fish: mortality rates parameters and
rate of movement maximum number selectivity
parameter, fraction of hake that
of abundance could be eaten;
available for harvest feeding saturation
and predation; and annual
Predators: natural diet variation
mortality rate, age at parameters;
recruitment to adult Background
stage, 3 recruitment mortality rate.
parameters, 3 Other standard
consumption and age-structured
functional response model parameters
parameters

2. Many Predator recruitment Seals feed mainly Suitability of prey Fixed ration
Some physiological and (but not survival) in shallow waters, remains constant model, constant
important process-related depends on krill and hence according to selectivity
model consumption; krill consume mostly its biomass as a
assumptions in transit between shallow-water proportion of
SSMUs do not suffer hake M. capensis the total biomass
predation and fishing of potential
mortalities; predators prey; constant
and the fishery are M1 (residual
competitors mortality); catch-
at-age measured
without error
3. Input data re Data from a physical Data re historic Stomach content Catch-at-age
Data temperature, model re currents; catches; trends in data to inform re data (landings
requirements currents, nutrient basic biological abundance e.g. predator rations and discards),
runoff from land data for predators; cpue, surveys; and feeding maturity-at-age,
information re length/age preferences; weight-at-age,
predator abundance; composition data; catch-at-age predator ration,
historic catch series; estimates of diet in numbers, predator diet
areas of SSMUs; composition and abundance information, prey
estimates of krill daily ration for indices and mean weight-at-age
density; estimates of each species body weights as in the predator
predator demand; for single-species stomach contents,
time series of models predator annual
environmental ration, residual
anomalies natural mortality

4. Can be run on S-PLUS, also being Fortran model; Runs on Windows Solver routine in
Platform UNIX or Windows recoded in R needs to be PC; typically Microsoft Excel;
PC recoded, possibly recoded by user SIR algorithm
in ADMB (McAllister et al.,
1994; McAllister
and Ianelli, 1997)
implemented in
Visual Basic
98 Models for an ecosystem approach to fisheries

TABLE A2b (continued)

MODEL SSEM KPFM MRM e.g. Punt MSVPA and MSM
and Butterworth MSFOR
(1995)

5. Pesticide Antarctic Southern North Sea, Eastern Bering

Examples inflow and Peninsula Benguela Baltic Sea Sea, central
where used salinity region Upwelling (Sparre 1991), Chile
change in region Georges Bank
drainage (Tsou and
canal (Sekine, Collie, 2001),
Nakanishi and Eastern Bering
Ukita, 1996), Sea (Livingston
Experimental and Jurado-
river system Molina, 2000;
(Sekine, Imai Jurado-Molina
and Ukita, and Livingston,
1997) 2002)
6. Developed to Developed to Developed Developed by Motivated
History predict impact assist CCAMLR in response ICES Multi- by desire to
of coastal in evaluating to debates species working incorporate
development options for whether group; main use predation
activities on subdividing increasing fur was in revising equations
fisheries the krill catch seal numbers predation from MSVPA
among SSMU’s were negatively mortality in a statistical
(Small-Scale impacting the estimates input framework that
Management commercially to single-species allows the fitting
Units) in important management of parameters by
Antarctic hake fishery in models considering how
Peninsula the southern errors enter into
region Benguela region the models
7. Can be used Includes a Takes explicit Includes a Incorporates
Additional to investigate range of account of prediction standard tools
useful features effect of performance uncertainty and model MSFOR such as Bayesian
pesticides measures that management methods and
can be used to issues through decision analysis
evaluate catch- the use of a into a multi-
allocation simulation species context
procedures and framework
assess tradeoffs incorporating
between feedback
predator control rules
and fishery actually in place
performance for setting TACs
for the fishery
TABLE A2c MODEL COMPARISON
Type of model Dynamic multispecies models Dynamic multispecies models Dynamic multispecies models Dynamic system models Dynamic system models
MODEL MULTSPEC GADGET Bioenergetic/allometric models OSMOSE SEAPODYM
Appendixes

e.g. Koen-Alonso and Yodzis,

2005

1. 2 maturation parameters (capelin), Varies depending on model but e.g. For each species: intrinsic Growth: 3 von Bertalanffy von Bertalanffy growth
Broad description 2 predation parameters (cod) growth parameters, maturation, P/B ratio; carrying capacity; growth parameters + 1 condition parameters; length-weight
of parameters (maximum consumption and prey fleet selection, recruitment, initial competition coefficient; factor per species; reproduction: parameters; age at first
(not fully abundance where consumption is population and consumption fraction of maximum annual relative fecundity per maturity; SST limit for
comprehensive half of max. consumption) and 3 physiological capacity for species and age at maturity; preproduction, length of
as intended to migration parameters (capelin) production realized by species; survival: maximal age, age at passive transport phase for
give a flavour allometric coefficients; mean recruitment and additional juveniles; natural mortality;
of the sorts of
individual biomass; “other annual natural mortality; NB initial stock biomass
parameters)
mortality” rate; density- parameter input is predator/prey (equilibrium); several
dependent mortality term size ratio determining minimal foraging, habitat and
threshold for predation to occur; temperature parameters;
also parameter describing food diffusion and advection
biomass to fulfil vital functions coefficients
2. Feeding and growth rate of Range of model assumptions Anchovy and squid prey not Fish predation depends on Movement depends on
Some important predators (minke and harp seal) depends on modules used as e.g. modelled hence assigned size suitability and spatial co- temperature and prey
model assumed to be constant; Curvilinear a number of different growth and carrying capacities and occurrence between a predator availability; recruitment
assumptions relationship between food consumption formulations from competition coefficients (to and its prey; carrying capacity is independent of adult
abundance and fish consumption; which to choose express dietary overlap); constraint; starvation mortality biomass
Mammal predation affects fish but density-dependent mortality impacts fish when nutritional
no feedback from fish abundance of sea lions assumed due resources limited
to marine mammals; Strong herring to crowding-related effects
recruitment simulated two years in during breeding season; prey-
row every 8 years independent digestive pause
(see Jeschke, Kopp and Tollrian,
2002)
3. Large database with stomach Catch data; Length distributions, Catch data and biomass trend Data on mean spatial Detailed data re SST,
Data content data - primarily cod age length keys, mean length/ information for each species distribution of each species currents, prey availability;
requirements stomachs but also e.g. herring weight at age; survey indices by preferably tagging data
and haddock; historical data on length or age, catch CPUE, stomach
capelin catch in numbers by length content data, data on proportion
group, month and area; VPA-based mature at age/length. No catch-
estimates of no. of cod; survey at-age data necessary; Data series
data used re area distribution for do not need to be continuous.
immature cod and other species; Spatially resolved and fleet-specific
estimates of popln sizes of other data required depending on model;
species; data re sea temperature; No limit on no. of data files
climatological data used
4. HP935 Workstation UNIX computing platform tested Fortran 77 run on PC Developed in Java (JdK 1.1.3, Source code in C++ object
Technical for Solaris, Linux, Mac OSX and SunMicrosystems) oriented language with
details Cygwin; also capable of running on executables available
multiple computers in parallel using for Windows and Linux
PVM (Parallel Virtual Machine) platforms. Also parallel
software in Java
99
100 Models for an ecosystem approach to fisheries

TABLE A2c (continued)

MODEL MULTSPEC GADGET Bioenergetic/ OSMOSE SEAPODYM
allometric
models e.g.
Koen-Alonso
and Yodzis, 2005

5. Barents Sea capelin Cod-capelin- Patagonia North Sea, Pelagic

Examples management; shrimp in marine Southern ecosystem of
where used Predation by cod Icelandic community Benguela the tropical
on young cod and waters; Barents (southwest (Shin, Pacific Ocean
haddock taken Sea, North South Atlantic Shannon and (Lehodey,
into account in the Sea, Celtic Sea Ocean); Cury, 2004) 2001, Lehodey,
stock assessment groundfish Newfoundland Chai and
made by the ICES stocks, hake shelf model Hampton,
Arctic Fisheries and key under 2003)
Working Group; pelagic development
Also used to fish species
study impact of interactions
minke whales and in the
harp seals on the Mediterranean
cod, capelin and Sea
herring stocks

6. Developed in Modelling Developed Developed Developed

History response to an marine to explore to explore for tropical
increased demand ecosystems whether a the extent tunas in the
that fisheries in fisheries mechanistically of usefulness Pacific Ocean
interactions should management oriented of local in response
be taken into context; approach can size-based to a need
account, following tailored to shed light predation for a spatial,
the 1983-1986 also examine on some rules in multi- multigear,
capelin collapse; marine common issues species models multi-species
Also, interest mammal in ecosystem model
in Norwegian populations; modelling incorporating
whaling activity flexible in an appropriate
spurred a need other contexts tuna
for models too movement
incorporating fish- model
marine mammal
interactions;
Similar in structure
to BORMICON thus
models merged
to some extent
e.g. by running
MULTSPEC using
BORMICON code

7. Co-operation Can represent Akaike Has been used Numerical

Additional between IMR, predation Information to compare scheme that
useful features Norway and within species; Criterion (AICc) results allows the
PINRO, Russia, maturation; (Burnham and produced use of spatial
resulted in multiple Anderson, by different stretched-
establishment of commercial 2002) used models (e.g. grids so that
stomach content and survey to rank and ECOPATH/ resolution can
data base of 80000 fleets taking select models; ECOSIM); one be increased
cod stomachs catches behaviour of few studies in regions of
from the of models addressing interest
populations; explored using starvation
tagging continuation mortality;
experiments and bifurcation allows
to follow the analysis (Doedel investigation
migration of et al., 1998) of ecosystem
the stock; data size spectra
warehouse (Shin and
Cury, 2004;
Shin et al.,
2005)
Appendixes 101

TABLE A2d MODEL COMPARISON

Type of Model Dynamic Whole Whole Dynamic multispecies Dynamic
multispecies ecosystem ecosystem models multispecies
models models models models
MODEL CCAMLR EPOC SMOM ESAM SEASTAR
models e.g. extension
Mori and
Butterworth,
2005, 2006

1. Krill: intrinsic Krill example: Krill: intrinsic Hollowed, Tjelmeland

Broad growth rate; 2 Natural growth rate; 2 Ianelli and and Lindstrøm
description of consumption mortality rate consumption Livingston, (2000): (2005) example:
parameters parameters; from krill yield parameters; consumption rate, Predation
(not fully Each predator: assessment; Each predator: satiation point and and natural
comprehensive maximum 3 von maximum satiation response mortality
as intended to birth rate; Bertalanffy birth rate; parameters; rates; prey
give a flavour
natural growth natural other typical species-specific
of the sorts of
parameters)
mortality; parameters; 2 mortality; single-species suitability
density- weight-length density- age-structured parameters,
dependent parameters; dependent model parameters prey-specific
mortality or 4 Beverton- mortality or e.g. catchability switching
birth rate Holt spawning birth rate coefficient, several coefficient,
parameter stock recruit parameter recruitment terminal F‘s,
relationship parameters, tagging survival
parameters; residual mortality,
Predator mean body weight,
- abundance proportion mature
and feeding at age, selectivity
function parameters
parameters

2. Density- Model Predators do Hollowed, Ianelli Tjelmeland

Some dependent still being not move and Livingston and Lindstrøm
important mortality developed between (2000): summer example:
model parameters are SSMUs; dietary information assumes weak
assumptions mathematically Predator assumed feedback from
necessary; breeding representative for fish to marine
presumably success entire yr i.e. no mammal
reflect the depends seasonal changes; abundance; prey
impact of on krill abundance of switching of
limitations of consumption alternative prey minke whales;
breeding sites assumed a constant no. of whales
for seals, and proportion of in study area
intra-species predator’s food described by
competition requirements; bell-shaped
effects for Spatial distribution function over
whales of predator and time
prey constant over
time
3. Historic Krill: maturity Basic Hollowed, Ianelli Tjelmeland
Data catch data.; ogive; weight biological and Livingston and Lindstrøm
requirements abundance at age; data for (2000): multi- example: time
trend data matrix of predators; species data - time- series of minke
probabilities information series of predator whales and
of moving re predator abundance, alternative
from origin abundance; annual predator prey, tag-return
to destination historic catch consumption data; other
polygons series: Areas rates and age typical single-
of SSMUs: composiiton of species data;
Estimates of prey consumed; abundance
krill density; other usual: total estimates;
Estimates catch biomass, biomass of cod
of predator bottom trawl input
demand survey estimates
of biomass, egg
production,
fisheries catch-
at-age, survey
size and age
compositions
4. AD Model R statistical AD Model AD Model Builder Developed
Technical Builder run language (R Builder run or other run on PC in user’s
details on PC Development on PC preferred code
Core Team, e.g. SeaStar
2005) extension in
Mathematica
102 Models for an ecosystem approach to fisheries

TABLE A2d (continued)

MODEL CCAMLR EPOC SMOM ESAM SEASTAR
models e.g. extension
Mori and
Butterworth
2005, 2006
5. Atlantic Indian Antarctic Antarctic Peninsula Gulf of Alaska northeast
Examples and Pacific Peninsula region (walleye Atlantic (minke
where used sectors of region - krill; pollock whale - herring
Antarctic Heard Island - flounder interactions)
- halibut - sea
lion)
6. Developed Developed Developed to Developed Developed as
History to test the in response assist CCAMLR in to provide a a first step to
hypothesis to perceived evaluating options framework for incorporate
that species need for for subdividing incorporating multi-species
interaction framework the krill catch predator prey considerations
effects alone providing among SSMU’s interactions into more
can account flexible (Small Scale to account traditional
for likely structure to Management for shifts in single-species
trends in the insert and Units) in Antarctic predation stock assessment
abundances delete model Peninsula region mortality models
of major components; in stock
Antarctic Also to assist assessments
predator CCAMLR in
species over evaluating
the past 50 or options for
so years subdividing
the krill catch
among SSMU’s
7. Inclusion Flexible plug- Developed for use Nonparametric Tjelmeland
Additional of density- and-play as an operating smoothing and Lindstrøm
useful features dependent structure model in a formal treatment example:
parameter MP framework. of selectivity consumption
resulted in Different MPs permitted parameters
some new are simulation greater estimated
insights e.g. tested with their flexibility in as part of
re krill surplus performances representing likelihood term;
hypothesis being compared predator prey-switching
on the basis of selectivity behaviour
an agreed set patterns modelled, tag-
of performance return data
statistics; incorporated
Reference Set
used comprises
12 alternative
combinations
that essentially
try to bound
the uncertainty
in the choice of
survival estimates
as well as the
breeding success
relationship
Appendixes 103

TABLE A3
Summary of some advantages, disadvantages and limitations of each method, as well as notes on the ease of
presentation of model outputs and the user-level of programming and mathematical skills required

TABLE A3a
MODEL Ecopath with IGBEM ATLANTIS INVITRO ERSEM II
Ecosim

Main Ease of use, Detailed Spatially Agent-based so Can be used

advantages large no. of representation explicit biomass uses a targeted to explore
users, structured of processes dynamics representation hydrographic
parameterisation within well- in response across multiple and planktonic
framework, well- studied to different scales and conditions
balanced level temperate fisheries sectors. impacting
of conceptual bay, from management juvenile fish;
realism, novel representation scenarios; includes
representation of sediment Applications as detailed
of predator-prey chemistry an Operating representations
interaction terms to average Model; simpler of the benthic
biomass of fish but adequate system which
representation is important
of processes e.g. in shelf
than most other seas; decouples
biogeochemical carbon and
models; includes nutrient
mixotrophy dynamics; can
which is be coupled
considered to different
important physical models
Main Ease of use can Very detailed Data intensive No easy user Data intensive;
disadvantages lead to poorly representation and no easy user interface Very detailed
constructed of interface representation
models that may physiological of physiological
mislead rather processes; Very processes
than advance data intensive
understanding
Limitations No explicit Birds, marine Base biological Cannot be Not designed
spatial structure mammals and rate parameters easily applied for detailed
in ECOSIM; sharks not are fixed in any to whole-of- representation
equilibrium represented one run ecosystem (in of higher
structure; as dynamic the sense of trophic levels
foraging arena pools but ATLANTIS or such as fish and
formulation rather simply EwE, though top predators
not always as mortality agent types
appropriate; terms on fish; do span all
no allowance Invertebrate trophic levels);
for detailed fisheries not must target its
energetic represented; use carefully
considerations No bycatch
(Aydin and component
Friday, 2001;
Aydin, 2004) and
alternative prey
types treated
as energetically
equivalent;
problems re
modelling
marine mammal
populations
(Plaganyi and
Butterworth,
2004, 2005a&b)
Ease of Excellent Visualisation Visualisation Visualisation Some
presentation of software software (Olive) software and presentation
model outputs (Olive) and Excel and R R analysis software
available analysis support scripts developed
sheets available available
User-level of Entry point Fair level Fair level Fair level Some
programming requires no required required required programming
and programming skills required
mathematical or mathematical although
skills required skills; more explorations
advanced users with currently
can benefit from existing
these skills models should
be relatively
straightforward
104 Models for an ecosystem approach to fisheries

TABLE A3b MODEL COMPARISON

MODEL SSEM KPFM MRM e.g. Punt and MSVPA and MSM
Butterworth (1995) MSFOR
Main Useful for Has attempted Rigorous model Large concerted Provides
advantages exploring to synthesize that fits to effort measures of
effects of state-of-the- data; focuses concentrated parameter
nutrient and art knowledge on groups of on approach uncertainty
pesticide re the system interest only with (e.g. Daan and
runoffs into a relatively these accounting Sissenwine,
in coastal simple model for 90% of hake 1991) with
systems mortality in attendant large
system sampling effort
and studies to
test underlying
assumptions
plus subsequent
efforts to
improve
and modify
approach
Main Data Includes several Difficult to Data hungry, Difficult
disadvantages intensive; not parameters implement Lack of for most to
as well tested that are statistical implement
as other difficult to structure to
models quantify, hence take account
considerable of uncertainty
uncertainty re in parameter
these estimates
Limitations Not Initialized No feedback Age-based Considers only
suitable for from uncertain between rather than small subset of
investigations data; does changes in hake length-based ecosystem
re fisheries not include abundance as required for
other than growth models affecting seal some regions;
coastal and delay- dynamics; predation
impacts difference desirability modelled
dynamics do parameters as one-way
not capture full assumed interaction
age-structured independent with predators
complexity; No of density; No impacting prey
fleet dynamics; explicit inclusion but no effect
no framework of environmental on predators of
for fitting to effects although changing prey
data or formal noise terms population;
statistical included Sensitivity to
testing recruitment
assumptions
Ease of Unknown Useful Not automated Average Unknown
presentation of parameter
model outputs visualisation
and tuning
+ summary
performance
measures
in EXCEL;
Not fully
automated
outputs
User - level of Unknown Not currently Very high - Fairly high; High
programming generally specific examples some user-
and available need to be coded friendly
mathematical although and minimisation packages e.g.
skills required ultimately process is 4M for the
version in R will complex Baltic (Vinther
be accessible et al., 1998)
to users with
moderate
programming
skills
Appendixes 105

TABLE A3c MODEL COMPARISON

MODEL MULTSPEC GADGET Bioenergetic/ OSMOSE SEAPODYM
allometric
models e.g.
Koen-Alonso
and Yodzis, 2005
Main Time-varying spatial Flexibility re model Explores Recognises Attempts to
advantages overlaps between as different modules sensitivity to that size incorporate
predators and prey can be substituted; alternative suitability is environmental
handled; Detailed permits efficient functional fundamental data directly into a
stomach content data optimisation/fitting to response to fish spatial population-
and consumption data; Sensitivity analysis formulations; predation as dynamics
formulations routine identifies detailed well as spatial simulation model;
incorporated; parameters with minor explorations co-occurrence novel movement
Includes cannibalism impacts only which re parameter between a model; level of
can thus be fixed in uncertainty; predator and implication closely
future runs; Possible does not its prey linked to the level
to estimate separate require of information
parameters for each accurate available on each
year e.g. if growth or data re diet aspect
selectivity differences composition
between years
Main Detailed stomach Current lack Requires Includes Insufficient
disadvantages content data required of examples estimation of a relative resolution of mid-
plus spatially-resolved demonstrating its use a large no. of fecundity trophic levels to
information parameters parameter that explore trophic
is difficult to interactions at all
estimate levels
Limitations Model simulates Difficult (but not No physical/ Only fish Tailored very
effects of marine impossible) to apply to environmental dynamics specifically for
mammal predation the whole ecosystem; forcing explicitly tuna; absence of
on fish but no Lower trophic levels considered; modelled a formal fitting
feedback in opposite not well represented age-structure thus e.g. top procedure for
direction; Prey not considered predators the estimation of
selection depends included only parameters
on prey species but as additional
doesn’t account for mortality term
prey or predator size;
Growth depends
on feeding level
and temperature
only, no energetic
considerations;
Model tailored
fairly specifically for
Barents Sea region
Ease of Unknown Good e.g. automatic Not automated Unknown SeapodymView
presentation of sensitivity analysis plots software
model outputs and postscript output includes tools for
files; print files for manipulating and
comparing output visualising data and
outputs
User - level of Fair Intermediate; some High - ability The simulation Low level required
programming initial training to to code plus framework to run executables
and understand basics of experience can be but considerably
mathematical UNIX/Linux; require re nonlinear defined using more to alter
skills required understanding of e.g. minimisation a graphical programs as would
optimisation process interface be needed to adapt
but no need to recode for other regions /
oneself; paramin species
program allows use of
multiple computers to
speed up runtime but is
for the more advanced
user
106 Models for an ecosystem approach to fisheries

TABLE A3d MODEL COMPARISON

MODEL CCAMLR models EPOC SMOM ESAM SEASTAR
e.g. Mori &
Butterworth
2005, 2006

Main Simple but Flexible Relatively Includes ability Focuses on

advantages pragmatic, addition/ simple model to statistically target species of
biologically substraction designed evaluate the fit interest, builds
realistic of modules to produce of the model models in a
equations; fits probability to the data; stepwise fashion
to data distribution results directly starting from
rather than a applicable to simplest possible,
single output; stock assessment fairly statistically
Management e.g. natural rigorous
Procedure mortality shown
framework to vary inter-
annually
Main Age- Still under Considers Considers only Considers only
disadvantages aggregated development only limited limited subset of very limited
and tailored and hence not subset of the ecosystem subset of
fairly tested ecosystem ecosystem
specifically for
krill-centric
ecosystem
Limitations No physical/ No framework Initialized Typically no Typically no
environmental for fitting to from physical/ physical/
forcing data or formal uncertain environmental environmental
considered; statistical data; does forcing but could forcing but could
can’t explicitly testing not include be included; be included;
represent detailed lower trophic lower trophic
observed krill growth levels not levels not
changes in model ; no considered; considered; often
age at sexual seasonality or no feedback no feedback
maturity due fleet dynamics effect of prey effect of prey
to lack of age consumption consumption
structure affecting affecting
predator predator
populations
Ease of Not automated Good Not fully Not automated Not automated
presentation of automated
model outputs outputs
User - level of High - ability Moderate - High - ability High - ability High - ability
programming to code plus knowledge of to code to code plus to code plus
and experience R required plus some experience experience
mathematical in nonlinear experience in nonlinear in nonlinear
skills required minimisation re nonlinear minimisation minimisation
minimisation
TABLE A4
Preliminary comparison of selected different modelling approaches to address a range of EBFM research questions outlined in the text. The shared regions of the table
highlight those approaches currently most appropriate or showing potential to address the aims as indicated, as assessed by the author
Appendixes

Type of model Whole Biogeochemical Biogeochemical Whole Biogeochemical Biogeochemical Whole Dynamic Dynamic Dynamic system
ecosystem ecosystem ecosystem ecosystem ecosystem ecosystem ecosystem multispecies multispecies models
models models models models models models models models models
1 2 3 4 5 6 7 8 9 10
RESEARCH QUESTION/ MODEL Ecopath with IGBEM ATLANTIS INVITRO ERSEM II SSEM KPFM* MRM e.g. MSVPA and MSM
Ecosim and Punt and MSFOR
ECOSPACE Butterworth
(1995)
1a.
Understanding - subset of
ecosystem
1b.
Understanding - complete
ecosystem
2.
Impact of target species
3.
Effect of top predators
4.
Competition: marine mammals
- fisheries
5.
Rebuilding depleted fish stocks
6.
Biases in single-species assessment
7.
Ways to distribute fishing effort
among fisheries
8.
Under-exploited species
9.
Change in ecosystem state
10.
Spatial concentration of fishing
11.
Environmental/physical effects
12.
Effects of habitat modification
13.
Effects of by-catch
14.
Introduction of non-native species
* Still being developed
107
 108

Table A4 (continued)
Type of model Dynamic Dynamic Dynamic Dynamic Dynamic Dynamic Whole Dynamic Dynamic Dynamic
multispecies multispecies multispecies system models system models multispecies ecosystem multispecies multispecies multispecies
models models models models models models models models
11 12 13 14 15 16 17 18 19 20

MULTSPEC GADGET Bioenergetic/ OSMOSE SEAPODYM CCAMLR EPOC* SMOM* ESAM SEASTAR
allometric models
models
1a. Understanding - subset of
ecosystem
1b. Understanding - complete
ecosystem
2.
Impact of target species
3.
Effect of top predators
4.
Competition: marine mammals
- fisheries
5.
Rebuilding depleted fish stocks
6.
Biases in single-species assessment
7.
Ways to distribute fishing effort
among fisheries
8.
Under-exploited species
9.
Change in ecosystem state
10.
Spatial concentration
of fishing
11. Environmental/physical effects

12.
Effects of habitat modification
13.
Effects of by-catch
14.
Introduction of non-native species

* Still being developed

Models for an ecosystem approach to fisheries
 ISSN 0429-9345
FAO
FISHERIES
TECHNICAL

477
PAPER

477

Models for an ecosystem

approach to fisheries

Models for an ecosystem approach to fisheries

This report reviews the methods available for assessing the impacts of interactions
between species and fisheries and their implications for marine fisheries management.
A brief description of the various modelling approaches currently in existence is provided,
highlighting in particular features of these models that have general relevance to the field
of ecosystem approach to fisheries (EAF). The report concentrates on the currently
available models representative of general types such as bionergetic models, predator-prey
models and minimally realistic models. Short descriptions are given of model parameters,
assumptions and data requirements. Some of the advantages, disadvantages and
limitations of each of the approaches in addressing questions pertaining to EAF are
discussed. The report concludes with some recommendations for moving forward in the
development of multispecies and ecosystem models and for the prudent use of the
currently available models as tools for provision of scientific information on fisheries in an
ecosystem context.

FAO