Agent-based modelling of socio-ecosystems: a methodology for the analysis of adaptation to climate change

Abstract

The integrated - environmental, economic and social - analysis of climate change calls for a paradigm shift as it is fundamentally a problem of complex, bottom-up and multi-agent human behaviour. There is a growing awareness that global environmental change dynamics and the related socio-economic implications involve a degree of complexity that requires an innovative modelling of combined social and ecological systems. Climate change policy can no longer be addressed separately from a broader context of adaptation and sustainability strategies. A vast body of literature on agent-based modelling (ABM) shows its potential to couple social and environmental models, to incorporate the influence of micro-level decision making in the system dynamics and to study the emergence of collective responses to policies. However, there are few publications which concretely apply this methodology to the study of climate change related issues. The analysis of the state of the art reported in this paper supports the idea that today ABM is an appropriate methodology for the bottom-up exploration of climate policies, especially because it can take into account adaptive behaviour and heterogeneity of the system's components.

Full text

D S E
Working Paper
Reviewing agent-based modelling
of socio-ecosystems:
a methodology for the analysis of
climate change adaptation
and sustainability
Stefano Balbi
Carlo Giupponi
Dipartimento Scienze Economiche
Department
of Economics
Ca’ Foscari University of
Venice
ISSN: 1827/336X
No. 15/WP/2009

Working Papers
Department of Economics
Ca’ Foscari University of Venice
No. 15/WP/2009
ISSN 1827-3580
The Working Paper Series
is availble only on line
(www.dse.unive.it/pubblicazioni)
For editorial correspondence, please contact:
wp.dse@unive.it
Department of Economics
Ca’ Foscari University of Venice
Cannaregio 873, Fondamenta San Giobbe
30121 Venice Italy
Fax: ++39 041 2349210
Reviewing agent-based modelling of socio-ecosystems:
a methodology for the analysis of climate change adaptation
and sustainability
Stefano Balbi, Carlo Giupponi
Ca' Foscari University of Venice - Department of Economics
Center for Environmental Economics and Management
First Draft: 30/07/09
Abstract
The integrated - environmental, economic and social - analysis of climate change calls for a
paradigm shift as it is fundamentally a problem of complex, bottom-up and multi-agent human
behaviour. There is a growing awareness that global environmental change dynamics and the
related socio-economic implications involve a degree of complexity that requires an innovative
modelling of combined social and ecological systems. Climate change policy can no longer be
addressed separately from a broader context of adaptation and sustainability strategies. A vast
body of literature on agent-based modelling (ABM) shows its potential to couple social and
environmental models, to incorporate the influence of micro-level decision making in the
system dynamics and to study the emergence of collective responses to policies. However,
there are few publications which concretely apply this methodology to the study of climate
change related issues. The analysis of the state of the art reported in this paper supports the
idea that today ABM is an appropriate methodology for the bottom-up exploration of climate
policies, especially because it can take into account adaptive behaviour and heterogeneity of the
system's components.
Keywords
Review, Agent-Based Modelling, Socio-Ecosystems, Climate Change, Adaptation, Complexity.
JEL Codes
Q
Address for correspondence:
Carlo Giupponi
Department of Economics
Ca’ Foscari University of Venice
Cannaregio 873, Fondamenta S.Giobbe
30121 Venezia - Italy
Phone: (++39) 041 2349126
Fax: (++39) 041 2349176
cgiupponi@unive.it
This Working Paper is published under the auspices of the Department of Economics of the Ca’ Foscari University of Venice. Opinions
expressed herein are those of the authors and not those of the Department. The Working Paper series is designed to divulge preliminary or
incomplete work, circulated to favour discussion and comments. Citation of this paper should consider its provisional character.

1. Introduction
1.1 Global change and complex systems
There is an increasing awareness that global change dynamics and the
related socio-economic implications involve a degree of complexity which
is not captured by traditional economic approaches that employ equilibrium
models. In particular, such a top down analysis of the human-environment
system doesn't consider the emergence of social behavioural patterns. This
eventually leads to a flawed policy making process which relies on
unrealistic assumptions (Moss, Pahl-Wostl, and Downing 2001). Yet, the
ultimate source of anthropogenic climate change is the agency of human
individuals grouped in social networks and their interaction. At the same
time, the responses to climate change, in terms of mitigation of greenhouse
gases emissions and in terms of adaptation to climatic variability and slow
changes in mean conditions, have to be found in humans behaviour.
In our global system where human activities prevail and endlessly modify
the environment, climate change is providing the chance to concretely
understand how the environment responds, suggesting a change in human
behaviour, both at a local and global level. Climate change can no longer be
addressed separately from a broader context of systemic sustainability and
adaptation strategies.
The endogenous feedbacks between socio-economic and biophysical
processes and the co-evolution of the human-environment system are
precisely those kind of dynamics included in the notion of social-ecological
systems, or socio-ecosystems (SES). SES are complex and adaptive systems
where social (human) and ecological (biophysical) agents are interacting at
multiple temporal and spatial scales (Rammel, Stagl, and Wilfing 2007).
This definition emphasizes the adoption of a single integrated approach for
the analysis of both social and economical agents and the natural
components of the ecosystem. It postulates the fact that SES are non
decomposable systems, because they emerge from the dynamic interplay
between the social and ecological components. SES show specific
properties such as: (a) non linear dynamics, alternate regimes and
thresholds; (b) adaptive cycles; (c) multiple scales and cross scale effects,
(d) adaptive capacity and transformability (Gunderson and Holling 2002).
Given such properties SES have to be considered as complex and adaptive
systems (CAS). CAS are dynamic networks of many agents (which may
represent cells, species, individuals, firms, nations) acting in parallel,
constantly acting and reacting to the behaviour of other agents. The control
of a CAS tend to be highly dispersed and decentralized. If there is to be any
coherent behaviour in the system, it has to arise from competition and
cooperation among the agents themselves. The overall behaviour of the
system is the result of a large number of decisions made every moment by
many individual agents (Waldrop 1992).
CAS display an ever changing dynamic equilibrium, which fluctuates

between chaotic and ordered states. On the edge of chaos, these systems are
very sensitive to any perturbation from the individual components (Holland
1992). CAS are inherently unpredictable as a whole: “their futures are not
determined and their global behaviours emerge from their local interactions
in complex, historically contingent and unpredictable ways” (Bradbury
2002).
Since the study of CAS is an attempt to better understand systems which are
difficult to grasp analytically, often the best available way to investigate
such them is through computer simulations (Gilbert and Troitzsch 1999). As
a matter of fact, when dealing with CAS, one has to cope with uncertainty
(Perez and Batten 2006). When decision are of major importance and hugely
permeated by imperfect knowledge and deep uncertainty, an improved
understanding of the use models is needed (Funtowicz and Ravetz 1995).
One way is to move towards exploratory modelling, whereby ensembles of
scenarios are used to represent possible futures of the system under study
and criteria such as resilience and stability are used to compare the
robustness of alternative policies (Lempert 2002).
1.2 Introducing agent-based thinking
Past research on computer science (e.g. Wooldridge and Jennings 1995;
Ferber 1999; Huhns and Stephens 1999; Weiss 1999) has shown how CAS
can be represented by means of multi-agent systems (MAS). MAS is a
concept derived from distributed artificial intelligence (DAI), which firstly
used it in order to reproduce the knowledge and reasoning of several
heterogeneous agents that need to coordinate to jointly solve planning
problems. Typically MAS refers to software agents and is implemented in
computer simulations.
According to the DAI derived definition of Ferber (1999) a MAS is a
system composed with the following elements:
1. an environment (E), often possessing explicit metrics;
2. a set of passive, located objects (O). These objects can be located,
created, destroyed or modified by the agents;
3. a set of active agents (A). Agents are particular objects that
constitute the active entities of the system;
4. a set of relationships (R) linking objects and/or agents together;
5. a set of operators (Op) allowing the agents to perceive, create, use,
manipulate or modify the objects.
Agents are virtual entities that demonstrate: (i) autonomous actions within
their environment, (ii) communication with other agents, (iii) limited
perception of their environment, (iv) bounded representation of their
environment (if any) and (v) decision making process based on satisfying
goals and incoming information (Ferber 1999).
Pure MAS, as conceived in DAI, are not fully relevant for modelling SES,
which are real systems based on the law of physics and on human social
interactions. However, including the fundamental contribution of past

research on artificial life (AL) (e.g. Reynolds 1987; Holland 1992; Langton
1992); individual-based modelling (IBM) (e.g. Huston et al. 1988; Grimm
1999) and social simulations (e.g. Schelling 1978; Axelrod and Hamilton
1981; Epstein and Axtell 1996), we are provided with a very promising
framework for the innovative modelling of combined SES and policy-
making in the context of sustainable development (Boulanger and Bréchet
2005).
Although this methodology has assumed many names, we adopted the
umbrella term agent-based modelling (ABM) which we regard as any
systemic and agent oriented modelling approach that employs computer
simulations.
ABM can explicitly represent the sources of social and biophysical
complexity accounting for interdependencies, both in space and time,
heterogeneity and nested hierarchies among agents and their environment
(Parker et al. 2003).
The main advantages of ABM are found in its abilities to: (a) couple social
and environmental systems, linking social and environmental processes; (b)
model individual decision-making entities, taking into account the
interactions between them and incorporating social processes and non-
monetary influences; (c) incorporate the influence of micro-level decision
making into the system dynamics, linking these micro-scale decisions to
macro-scale phenomena; (d) study the emergence of collective responses to
changing environment and policies (Hare and Deadman 2004; Matthews et
al. 2007).
Moreover, agent-based models can be constructed and validated in the
participatory setting, fostering the process of social learning and, while
integrating factual and local knowledge, they can provide assistance for
specific decision making (Barreteau, Bousquet, and Attonaty 2001; Guyot
and Honiden 2006; Pahl-Wostl 2007).
1.3 Further expansion of ABM
To date, ABM has been used to reformulate some main issues of social and
natural science (Bousquet and Le Page 2004). In fact, there exists a
consistent body of work on ABM in sociology and social processes (e.g.
Conte et al. 2001; Macy and Willer 2002; Gilbert and Troitzsch 1999),
economics and finance (e.g. LeBaron 2000; Tesfatsion 2002) and in a set of
environmental issues including land use and cover change (e.g. Parker et al.
2003; Veldkamp and Verburg 2004) ecology and natural resource
management (e.g. Lansing and Kremer 1993; Bousquet and Le Page 2004),
agriculture (e.g. Balmann 1997; Berger 2001), urban planning (e.g. Torrens
and O Sullivan 2001; Batty 2005), and archaeology (e.g. Kohler and
Gumerman 1999). Altogether these various applications constitute the rich
breeding ground for moving towards a new approach to the analysis of
climate change issues.

However, there are limited useful publications on ABM in the arena of
climate change. Some of them stand a very epistemological level stating the
usefulness of the methodology without applying it (e.g. Moss, Pahl-Wostl,
and Downing 2001; Patt and Siebenhüner 2005). Few applications explicitly
aim at analysing climate change at a theoretical level (e.g. Janssen and de
Vries 1998) or at a more empirical level (e.g. Bharwani et al. 2005;
Ziervogel et al. 2005; Berman et al. 2004; Werner and McNamara 2007;
Barthel et al. 2008; Entwisle et al. 2008). In contrast, there are several ABM
applications which generically include climate change elements in their
system modelling (e.g. Dean et al. 1999; Barthel et al. 2008; Hasselmann
2008; Filatova 2009; Mandel et al. 2009; Beckenbach and Briegel 2009).
Such a few available publications are evidence of an immature area of
research. This field of application, only recently, started to rapidly develop,
with many research project forthcoming1, with potential publications in the
future.
The justification of this late development can be found in the intrinsic
characteristics of the methodology.
Given ABM ability to capture complexity and represent detail, the model
has to be built at the right level of description, with just the right amount of
detail to serve its purpose. Therefore, the purpose has to be clearly stated in
order to try hard to limit the model complexity. The reason is that, as
computer models are less constrained technically, their design can still be
too complex compared to classical models (Grimm and Railsback 2005).
General purpose models aiming at representing a system rather than a
problem, which are common in the climate change arena, cannot work.
Moreover, ABM may face challenges of parametrization and validation
(Parker et al. 2001). This is particularly evident when one desires to build an
empirically grounded model. ABM involves soft factors, difficult to
quantify, calibrate and sometimes justify (Bonabeau 2002). Assumptions
necessary for statistical verification and validation, such as normality and
linearity can be at odds with models designed to accommodate complex
behaviours caused by sensitivity to initial conditions, self-organized
criticality, path dependency and non linearities (Arthur 1990; Manson 2001;
Perez and Batten 2006).
Also the communication phase is more difficult, because the model has to
be described in words, other than the universal language of mathematics,
and this turns very often to be less efficient (Grimm et al. 2006).
Finally, ABM is not well suited to make quantitative deterministic
predictions about how a system will function in the future, or about how to
make the system function better in the future, which seems like the issue for
the mainstream climate change economics devoted to top-down “hard
science”. The outcome of a simulation should be interpreted at a more
qualitative level, depending on the degree of accuracy and completeness in
the input to the model (Bonabeau 2002).
1 Global Cities Institute of RMIT University and CSIRO in Australia, University of
Hohenheim, in Germany, Tyndal Centre in UK, Natural Resources Canada, etc., just to quote some of
the institutions with ongoing projects concerning ABM and climate change.

ABM has, to date, had limited impact in policy making, because it has been
predominantly used in deterministic rather than exploratory mode, while it
should be used in conditions of deep uncertainty, where there is no
agreement between stakeholders on correct decisions (Lempert 2002).
Eventually, ABM is object of renewed interested, fuelled by the recent
developments on uncertainty analysis applied to climate change (Barker
2008; Weitzman 2009).
The analysis of climate change calls for a paradigm shift (Bousquet and Le
Page 2004; Martens 2006; Voinov 2008) as it is fundamentally a problem of
complex, bottom-up and multi-agent human behaviour, which involves the
entire socio-ecosystem.
This paper aims at integrating the existent know-how on ABM from
different scientific communities in order to clarify whether or not this could
be an appropriate methodology for modelling the dynamics of SES exposed
to climate change and assessing the related policies within the context of
adaptation and sustainability.
In the second section we define some core notions of ABM, we clarify the
terminology in use, and briefly describe the main scientific domains and
research communities applying this methodology. We approach the different
domains making reference to the three fundamental dimensions of
sustainability identified by the social, the economic and the environmental
systems.
We than go deeper into the subject by elucidating how the concepts of agent,
environment, emergence, interaction, heterogeneity, space and time,
behaviour and validation are treated, including the computing languages,
tools and platforms used for the simulations. We reviewed those applications
that, in our opinion, better fit the idea of modelling SES and include climate
change related elements.
In the fourth section we conclude discussing the main results.
2. ABM and complexity in the three spheres of sustainability
2.1 ABM and the dimensions of complexity
Nowadays ABM constitutes a broad and interdisciplinary movement.
Different terms are used to define subtly different approaches to ABM:
agent-based simulation modelling (e.g Berman et al. 2004), individual-based
modelling (e.g. Grimm and Railsback 2005), multi-agent-based simulation
(e.g. Perez and Batten 2006), agent-based social simulation (e.g. Gilbert
2004), multi-agent simulation (e.g. Bousquet and Le Page 2004), multi-actor
modelling (e.g. Barthel et al. 2008), etc. According to Hare and Deadman
(2004) a key difference, which justifies this terminological diversity, stands
in the complexity of the interactions to be modelled. When emphasis is
placed on modelling complex interactions, and agents are are simplistic, the
AL and ecological roots of ABM prevail. When interactions spawn from the
deliberations of the agents and the deliberative social cognition is most
important, then the DAI roots are prevalent.

In general there are three types of interaction: direct interaction among
agents, which can be physical (grow, push, eat) or by communication, and
interactions mediated by the environment (Bousquet and Le Page 2004). By
means of interactions, interdependencies exist among agents and their
environment, across time and across space.
In the following paragraphs we argue that there are more sources of
complexity, which also influence the terminology in use.
Heterogeneity is another major source of complexity. ABM can consider
heterogeneous system's components situated in dedicated heterogeneous
spaces. Agents' diversity may depend on their experience, values, abilities
and resources but also on their spatial position. In fact, heterogeneity may
also be present across the environment, space and time (Parker et al. 2003).
Complex interactions and heterogeneity combined typically build up a high
degree of spatial and temporal complexity, exemplified in cross-scale
interdependencies and nested hierarchies.
Emergence is a central tenet of ABM and the search for emergence is
explicitly mentioned in most of the modelling efforts (Parker et al. 2001).
An emergent property may be defined as a macroscopic outcome resulting
from synergies and interdependencies between lower level system
components. Emergence characterizes a complex system, the capacities of
whom are greater than the sum of the system. The emergent qualities of a
system are not analytically tractable from the attributes of internal
components (Baas and Emmeche 1997).
The concept of emergence and the concept of cross-scale hierarchies are
related. Identifying emergence, therefore, may require understanding
important cross-scale interactions and deliberately building interactions
across topological, temporal and structural levels, rather than limiting
modelling and analysis to a single scale. Unfortunately this potential to
explicitly represent cross-scale interactions and feed backs, both bottom-up
and top-down, has been minimally exploited in agent-based models to date
(Parker et al. 2001).
Behavioural complexity derives from the agents internal world, their mental
model or architecture, which describe their cognition and learning capacity.
Often agents are endowed with bounded cognition. They have a limited
perception of the environment and derive information from it, which they
use to make assumptions about its state. Agents are not meant to be
omniscient and fully rational utility maximisers as, for instance, the homo
economicus (Gintis 2000). Models of bounded rationality have been used as
an alternative in economics (Simon 1955). Furthermore, borrowing concepts
from psychology, behavioural economics has included dimensions of
economic agents such as emotions, motivations and perceptions (Camerer
2003). In ABM is also possible to incorporate the salient characteristics of
actual human decision-making behaviour (Tesfatsion and Judd 2006),
including the agents capacity of learning from past experiences.
The combination of behavioural complexity with the complexity related to

interactions and heterogeneity allows the representation of adaptation in
agent-based models at both micro and macro scales. The behaviour built
into the decision making structure at the individual agent level, which is
influenced by the system dynamics, is in turn embedded in the systemic
adaptive mechanisms.
2.2 ABM and the triple bottom line of sustainability
In applications of agent-based models to social processes, agents represent
people or groups of people and agent relationships represent processes of
social interaction (Gilbert and Troitzsch 1999). The fundamental assumption
is that people and their social interactions can be credibly modelled at some
reasonable level of abstraction, for at least specific and well defined
processes (Macal and North 2005).
After Schelling (1978), Epstein and Axtell (1996) extended the notion of
modelling human agents to growing artificial societies through agent
simulations, with their ground-breaking Sugarscape model. Social science
computation is now a consolidated subfield of sociology (Gilbert and Abbott
2005). However, sociological ABM is much more concerned with
theoretical development and explanation than with exploratory analysis.
These models do not necessarily aim to provide an accurate representation
of a particular empirical application (Macy and Willer 2002). Instead, their
goal is to enrich the understanding of fundamental processes that might
appear in a variety of applications (Axelrod 1997).
In ABM applications to economic systems agents can be both organization
and individuals, while the design of interactions aims at performing a
natural description of the system, taking into account both the topological
and behavioural dimensions of the components' activities (Bonabeau 2002).
Some of the main classical assumption of microeconomics can be relaxed,
leading to a more realistic representation of economic systems. Firstly,
drawing on behavioural economics, agents are not rational optimizers
(Smith 1989). Secondly, agents are not homogeneous. A key observation of
complexity science is that agents diversity universally occurs in the real
world (Arthur 1999). Thirdly, there can be increasing returns to scale
underlying dynamic processes of rapid exponential growth. Such positive
feedback loops can create self-sustaining processes that quickly take the
system away from its starting point to a faraway state (Arthur 1990). Lastly,
the long run equilibrium state of the system might not be the primary
information of interest. Transient states may be crucial. Furthermore, not all
systems come to an equilibrium (Arthur 2006).
The field of agent based computational economics (ACE) has grown up
around the application of ABM to economic systems. ACE is the
computational study of economies modelled as evolving systems of
autonomous, adaptive and interacting agents (Tesfatsion 2002).
In environmental applications of ABM agents can be both an individual

human or biological organism or, more generically, any biophysical entity,
as a reservoir of a natural resource or a part of it.
In biology, ABM has been used to model the possible emergent structures
resulting from molecular self-assembly (e.g. Troisi, Wong, and Ratner 2005)
and the self-organization of bacterial colonies (e.g. Krawczyk, Dzwinel, and
Yuen 2003) but also to model bacterial behaviour and interaction at multiple
scales (e.g. Emonet et al. 2005).
However, in the environmental domain, ABM applications were initially
developed in ecology at the end of the 1980s following the IBM paradigm
(e.g. Huston et al. 1988; Grimm 1999), which introduced the notion of the
individual to understand the role of heterogeneity. In ecology an agent is
necessarily an individual and scarce emphasis is given to the decision
making process of the agents and to the social organization in which these
individuals are embedded (Bousquet and Le Page 2004).
In contrast, in ABM applications to ecosystem management an agent can
represent any level of organization, while the decision making process and
the social organization are crucial. Frequently, these studies examine
questions of collective problem solving related to the management of a
common natural resource. ABM of ecosystem management is often included
in the categories of agent-based land use models (ABLUMs) (Matthews et
al. 2007) or as multi-agent systems for land use and cover change
(MAS/LUCC) (Parker et al. 2003). In fact, most of the research on ABM
and natural resources management overlaps with ABLUMs. This is because
many of the environmental applications of ABM have a crucial spatial
component and are very often spatially explicit, making use of abstract
grids, cellular automata (CA), and, when case specific, maps from
geographical information systems (GIS). So the landscape very frequently
coincides with the environment where the physical space, the agents and the
resources are represented delineating the system's boundaries and its
organization.
2.3 Shared streams of research in ABM
The short overview of section 2.2 suggests that the definition of agent
cannot be reduced to a specific one, because there are different realms of
applications and processes with different agent characteristics, that can be
successfully modelled with ABM. As suggested by Goldspink (2000) it's
worth defining the minimal agent as “a natural or artificial entity with
sufficient behavioural plasticity to persist in its medium by responding to
recurrent perturbations within that medium so as to maintain its
organization”. The medium is what Ferber (1999) defines as the
environment and can be the background environment, in strictu sensu, or the
substrate of a social system, and may contain active and/or passive agents.
The latter are what Ferber (1999) calls objects. Starting from this any model
can add new agent's features.
In the spirit of the interdisciplinary approach we are interested in the points
of convergence between different scientific disciplines and a framework to
classify them. Building on Macy and Willer (2002), Bonabeau (2002),

Tesfatsion (2003), Bousquet and Le Page (2004) and Janssen (2005) we
identified some main streams of research that can be found in each of the
three scientific domains constituting the triple bottom line of sustainability.
Within self-organization and co-evolution of the system the focus of agent-
based models is on the self-organizing capabilities of the system under
study, in particular how agents' behavioural rules influence their co-
evolution and, ultimately, the system's structure. These models study in
evolutionary terms how the decision making at the micro-level affect the
macro-structure.
The stream of research diffusion processes and networks formation is
interested on how micro-level interactions and transmission of information
lead to the emergence of specific structural phenomena such cultural
convergence, diffusion processes and endogenous formation of networks.
Models often employ learning algorithms like artificial neural networks.
In the stream of research modelling organizations, cooperation, and
collective management the focus of agent-based models is on how the
system's topology and structure influences its behaviour, and in particular
which structure stimulate cooperation in the benefit of the collective.
In parallel experiments we include those applications that compare
computational and empirically observed agents and structures in order to
improve the representation of the system under study. This stream has strong
linkages with the issue of model validation.
Agent's architecture deals specifically with behavioural complexity. The
main issue is how to represent the decision making of the agents and,
ultimately, evolution and learning both at a micro- and macro-level.
Programming is necessarily a main cross-cutting issue given the shared
computer based approach. OOP techniques (Cox 1986) are often advocated
as a crucial mean for constructing an environment in which users can easily
tailor models designed to suit their own particular research agendas. In
general there remains a certain duality between general purpose languages
and more or less specific packages.
While the first three streams define the main research questions of an ABM
application and, therefore, tend to be mutually exclusive, the remaining
three can be understood as necessary accessories and tools among the ABM
movement. We classify some relevant ABM studies belonging to various
disciplines in table 1 in order to show that a huge part of the ABM past
research can find its proper allocation in this framework.
Table 1 – Classification of ABM according to scientific domain and stream of research
2.3.1 Self-organization and co-evolution of the system
In sociology this stream of research is concerned with the emergent
structure in terms of structural differentiation as, for instance, social
segregation (e.g. Schelling 1971). Models often investigate spatial clustering
using CA. Agents can change location and behaviour in response to
selection pressures. Adaptation is based on evolution, which modifies the

frequency distribution of strategies across the population of agents (e.g.
Epstein and Axtell 1996).
In economics this stream of research deals with the self-organizing
capabilities of specific types of market processes and the co-evolution of
firms (Tesfatsion 2002). The most successful studies are those on financial
markets (e.g. LeBaron 2000). Evolutionary models can explain important
stylized facts such as fat tails, clustered volatility, and long memory, of real
financial series (Hommes 2002).
In environmental ABM applications of this stream the focus is on how the
behavioural rules of interacting agents lead to the self-organization of the
ecosystem's structure and to the state of the common natural resource.
2.3.2 Diffusion processes and networks formation
In sociology these models investigate imitation (e.g. Latane 1996) and
diffusion (e.g. Rosenkopf and Abrahamson 1999). Adaptation operates via
social influence and is based on learning, which modifies the probability
distribution of strategies in each agent's repertoire (Nowak et al. 1998).
In economics these models investigate the dynamics of interaction networks
and diffusion processes. Relevant examples of applications focus attention
on the endogenous formation of trade networks (e.g Albin and Foley 1992).
A further kind of network issue is represented by the transmission of
information as occurs with bank panics and stock market crashes (e.g. De
Vany and Lee 2001).
In environmental applications of this stream of research both interaction
networks and diffusion processes are present. Rouchier et al. (2001)
investigated the formation of networks in a field study that focus on
seasonal mobility (transhumance) among nomadic cattle herdsmen. Berger
(2001) studied the diffusion of agricultural technologies based on the
concept of early and late adopters. Deffuant et al. (2002) simulate adoption
of organic farming practices as a consequence of governmental policy.
2.3.3 Modelling organizations, cooperation, and collective management
In sociology, studies dealing with emergent order focus attention on the
ways in which network structures affect the viability of cooperative
behaviour. For example, they can show how egoistic adaptation can lead to
successful collective action without either altruism or global (top-down)
imposition of control, according to the network properties (Macy and Willer
2002).
In economics, organizations can be seen as CAS (Tesfatsion 2002). One can
model the organization's activities by looking at what every actor does.
Therefore, it is possible to model the emergent collective behaviour of an
organization or of a part of it in a certain context or at a certain level of
description (Bonabeau 2002). Studies of firms in the ACE framework have
tended to stress the effects of a firm's organizational structure on its own
result behaviour (e.g. Prietula et al. 1998). Cooperation and coordination are
a prerequisite to achieve an efficient overall performance.

In environmental applications this is a prime issue for the research on
management of common pool resources. These models investigate how the
system topology and structure influences the collective behaviour towards
the common natural resource trying to identify what type of institutional
rules may direct individuals to act in the benefit of the collective (Parker et
al. 2003). The irrigation system in Bali is an early example of the use of
ABM to understand self-governance (Lansing and Kremer 1993).
2.3.4 Parallel experiments
In sociology, organizational life histories generated by simulations are
compared with those observed in empirical populations (e.g. Carley 1996;
Lomi and Larsen 1998).
In economics, human subject behaviour is used to guide the specification of
learning processes of computational agents and computational agent
behaviour is used to formulate hypothesis about the root causes of observed
human agent behaviour (Tesfatsion 2002).
Both the cited sociological and economic applications adopt an a posteriori
approach. In contrast environmental applications tend to adopt an iterative
approach by means of participatory techniques, such as role playing games,
where human subject experimentation is used to test and ameliorate the
computational simulations in an iterative process. In the spirit of adaptive
management (Holling 1978) several researchers2 have developed their
agent-based models together with the stakeholders of the problem under
concern, improving the acquisition of knowledge, the model construction,
the model validation and the model application to decision making (e.g.
Bousquet et al. 1999; Barreteau et al. 2001; Guyot and Honiden 2006).
2.3.5 Agents' architecture
In sociology there seems to be a clear distinction between learning and
evolution. Learning modifies the probability distribution of strategies in
each agent's repertoire. Learning architectures are based on artificial neural
networks (Rumelhart and McClelland 1986). Evolution modifies the
frequency distribution of strategies across the population of agents. In this
case architectures are based on evolutionary algorithms such as genetic
algorithm (Holland 1992).
In economics learning is used as a comprehensive term. The learning issue
is particularly crucial due to the numerous anomalies discovered in
laboratory experiments between actual human-subject behaviours and the
behaviours predicted by traditional rational-agent economic theories (Gintis
2000) A broad range of algorithms is used to represent the agents' learning
processes including reinforced learning algorithms (e.g. Bell 2001), neural
networks (e.g. Luna 2002), genetic algorithms (e.g. Dawid 1996) and
classifier systems (Booker, Goldberg, and Holland 1989), genetic
2 We define these researchers as the “French school” of ABM as they are all more or less
related to the Centre de coopération internationale en recherche agronomique (CIRAD) of
Montpellier and to the Cormas ABM platform.

programming and a variety of other evolutionary algorithms (e.g. Chattoe
1998) that attempt to capture aspects of inductive learning (Tesfatsion
2003). Vriend (2000) put more emphasis on the learning level, which can be
individual, meaning on the basis of own experience, or social, in which
every agent's experience is considered
In environmental applications various agent's architectures are drawn from
computer science in order to represent behavioural complexity (Bousquet
and Le Page 2004). Most are based on the evolutionary metaphor, as the
genetic algorithm (e.g. Manson 2005). Others are defined architectures for
competitive tasks, whereby choices are made by agents when they receive
several stimuli which activates different tasks (e.g. Drogoul and Ferber
1994). Neural networks are employed in order to place emphasis on the
agent's learning capacity: the perception-action relation is modelled by a
network whose connections evolve (e.g. Grand and Cliff 1998). Agent's
decisions may also be expressed in terms of parametrized functions by
means of vector calculation describing the addition of physical forces (e.g.
Reynolds 1987), linear programming describing processes of optimization
more or less bounded (e.g. Balmann 1997), multi-criteria analysis (e.g.
Deffuant et al. 2000), etc. One last way to model cognitive agents is the
belief-desire-intention (BDI) architecture (e.g. Wooldridge and Jennings
1995) where agents memorize the space and the resources in a sort of
mental map but also other agents' reputation when it comes to the moment
of interaction.
2.3.6 Programming
In sociology, applications are more oriented towards ad-hoc platforms.
Gilbert and Bankes (2002), provide a comprehensive enumeration of
available languages and tools without identifying the best options.
According to Tobias and Hofmann (2004), who evaluated four freely
available and JAVA based programming libraries, Repast is the most suitable
simulation framework for the applied modelling of social interventions
based on theories and data. In contrast, Terna (1998) focuses on Swarm,
which is the ground breaking and most dated tool of this type.
In economics, there remains a considerable gap between powerful general
purpose languages and packages easy learned (Tesfatsion 2002). On the one
hand, significant programming skills are needed in order to master general
purposes languages such as C++ and Java, where applications are built from
scratch. On the other hand there is a proliferation of ad-hoc packages, often
not powerful enough for many economic applications, which can't
communicate with each other and don't facilitate an easy sharing and
comparison of modelling features. Economic applications often opt for a
programming language or a generic but powerful software as NetLogo or
Swarm (e.g. Luna and Stefansson 2000).
Also environmental applications are more oriented towards ad-hoc
packages. Bousquet and Le Page (2004) survey some platforms developed
with OOP distinguishing between generic softwares (e.g. Swarm and
NetLogo), those dedicated to social and ecological simulation (e.g. Ecosim,

Repast and Cormas), and specific platforms for ad- hoc applications (e.g.
Manta, Arborscapes). According to Railsback et al. (2006) NetLogo is
highly recommended, compared to Mason, Repast and Swarm, even for
prototyping complex models. Cormas is a well tested software for
ecosystem management which supports participatory processes (Le Page et
al. 2000). Repast is well considered for its flexibility but requires higher
programming skills.
Many of the packages which have not been cited in this section can be
found in the appendices of Tobias and Hofmann (2004) and in Schut (2007).
3. ABM of socio-ecosystems and climate change
In the past 10 years there have been few studies that modelled socio-
ecosystems and included climate change elements related to mitigation or
adaptation issues. In this section we review those papers that, in our opinion,
are suitable to this purpose and are already published or close to publication.
However, we suggest to look at the following comparative analysis as a first
attempt to envision, in a comprehensive manner, the issue of climate change
through the lenses of ABM. Several research projects, which are currently
developing new relevant studies for this same issue, are expected in the near
future.
Janssen and de Vries (1998) are specifically concerned with the behavioural
aspects of ABM applied to climate change adaptation. Agents are groups of
decision makers who operate at the international level and have different
world-views and management styles towards climate change.
Dean et al. (1999), Werner and McNamara (2007), Entwisle et al. (2008)
and Filatova (2009) deal with ABM and land use. Dean et al. (1999) is an
early example of ABM of a local socio-ecosystems, which include climate
change elements in order to simulate human responses and the outcome of
adaptation. The model represents the behaviour of culturally relevant agents
on a defined landscape in order to test hypothesis of past agricultural
development and settlement patterns. Werner and McNamara (2007)
investigate how the economic, social and cultural factors surrounding the
human response to river floods, hurricanes and wetlands degradation affect
a city landscape. Entwisle et al. (2008) focus on the responses to floods and
drought at a regional level in terms of agricultural land use and migration,
explicitly taking into account social networks. Filatova (2009) incorporated
climate change related risks in an agent-based land market for coastal cities,
which simulates the emergence of urban land patterns and land prices as a
result of micro scale interactions between buyers and sellers.
Berman et al. (2004), Bharwani et al. (2005) and Ziervogel et al. (2005) are
the only published empirical field studies, which explicitly aim at exploring
local adaptation in the context of climate change and sustainable
development by means of ABM. As Bharwani et al. (2005) and Ziervogel et
al. (2005) refer to the same research project, we chose to review Bharwani
et al. (2005) for its more comprehensive model description. Grothmann and
Patt (2005) is a useful socio-cognitive model that can be used in ABM of

this kind where is important to capture the most significant behavioural
determinants of adaptation. It has been tested in similar studies but not
applied in the reviewed paper and therefore it is not present in table 2.
Berman et al. (2004) assess how scenarios associated with economic and
climate change might affect a local economy, resource harvest and the well-
being of an existing community. Bharwani et al. (2005) investigate whether
individuals, who adapt gradually to annual climate variability, are better
equipped to respond to longer-term climate variability and change in a
sustainable manner.
Barthel et al. (2008) developed an ABM framework for water demand and
supply future scenarios where the socio-ecosystem is enabled to react and to
adapt to climate change.
Hasselmann (2008), Beckenbach and Briegel (2009) and Mandel et al.
(2009) concern macroeconomic models which employ, more or less
explicitly, an agent-oriented framework in dealing with growth and climate
change at a regional to global level. Hasselmann (2008) introduces few
representative actors in a macroeconomic model of coupled climate-socio-
economic system conceptualized following a system dynamics approach.
The focus is on the evolution of this coupled system according to behaviour
of the agents pursuing different goals while jointly striving to limit global
warming to an acceptable level. Mandel et al. (2009) developed an agent-
based model of a growing economy where growth is triggered by the
increase of labour productivity proportionally to investments. Beckenbach
and Briegel (2009) investigate the relationship between innovations,
economic growth and carbon emissions.
Building on Parker et al. (2001) and Grimm et al. (2006) we review the
cited papers according to the following categories:
1. in stream of research we show how is possible to associate any of
them to one of the streams in table 1;
2. in system under study and climate issue we describe the object of the
model, it's physical boundaries and the climatic problem at stake;
3. in agents and environment we define how the respective concept are
applied in practice;
4. in emergence we identify which system-level phenomena truly
emerge from individual traits;
5. in interactions we depict how the complexity of interactions is
treated;
6. in heterogeneity we show how the diversity of the system elements is
captured;
7. in space and time we describe the spatial and temporal dimensions,
the process scheduling and the model initialization;
8. in behaviour we focus on how the model deals with behavioural
complexity;
9. in verification and validation we look at the strategies used to
understand the model performance and the ability to represent the
system under study;
10. finally, in technical aspects we identify the implemented

programming languages and tools and other technical issues.
The main results of this classification effort are reported in table 2.
Table 2 – Comparative analysis of agent-based models of SES with climate change
elements
3.1 Stream of research
This classification shows that the framework regarding the streams of
research proposed in table 1 remains valid in the climate change arena.
However, at this early stage there seems to prevail one distinct research
question. More than half of the studies we analysed are concerned about the
self-organization and co-evolution of the system. Not surprisingly, this is the
stream that paved the way to the application of ABM to social processes,
meaning that the first examples of ABM dealing with climate change are
following the most consolidated path of development.
Conversely, in Berman et al. (2004) the model purpose is to project how
local institutions shape human adaptation to hypothetical futures. In
Bharwani et al. (2005) the focus is on the emergence of strategies over time
as a part of a cultural process. In Barthel et al. (2008) the focal point is on
the implications for water management, given the system specific structure.
These are all examples of the stream of research on modelling organizations,
cooperation, and collective management.
Two very different studies are concerned about diffusion processes and
networks formation. Entwisle et al. (2008) considers social influence at a
local scale and at an empirical level , while Beckenbach and Briegel (2009)
is about the diffusion of innovation at a global scale and in abstract terms.
Janssen and de Vries (1998) and Beckenbach and Briegel (2009) are models
where the agent's architecture is a research question per se. Grothmann and
Patt (2005) could be added to this subset even though it is not exactly an
ABM model.
The programming phase is generally made to be case specific. Only in one
case (Mandel et al. 2009) a generic software is produced, which can be
applied to case studies other than the German economy.
Parallel experiments are not diffused but in three cases they are utilized to
substantially improve the credibility of the model. In Dean et al. (1999) and
Mandel et al. (2009) this is achieved a posteriori through statistical means.
In Bharwani et al. (2005) this is an iterative process based on the
participation of the stakeholders.
3.2 System under study and climate issue
ABM shows abilities to model local, regional and global systems both at a
very abstract or more realistic level. We can distinguish between two
typologies of ABM dealing with climate change: (a) the majority, that focus
on adaptation, analysing regional and local systems and (b) few global
models, that are concerned about mitigation (Janssen and de Vries 1998;
Hasselmann 2008; Mandel et al. 2009; Beckenbach and Briegel 2009). In
the first case the level of detail is at the community (or network of

communities) level. In the second case there is much more aggregation even
if a certain degree of heterogeneity is introduced by means of the agent-
based thinking. Notwithstanding the novelty of the methodology this
dichotomy appears quite conservative with respect to the climate change
literature. In no case adaptation and mitigation are treated together.
3.3 Agents and Environment
Agents can represent various human actors at different decisional levels.
Very surprisingly households emerge as the main category of agents in the
climate change arena, as if it was the basic unit of reference, independently
from the scope.
In general the number of agent's types is limited, in order to control
complexity. Most of the model employ 1 to 3 agent's classes. Werner and
McNamara (2007) is an exception with seven types of agents, which
exponentially increase the level of details and the heterogeneity complexity
of the model.
The notion of environment is treated in a variety of ways. Very often these
models rely on equations or indicators, which can be defined as sub-models
describing theoretical spaces of interaction. Most of the models employ
economic sub-models. In models dealing with land use and in Barthel et al.
(2008) there is a significant correspondence between the landscape of the
system under study and the environment of the ABM, however they also
employ non-spatial sub models. The best example is Filatova (2009) in
which the environment is constituted by the land market model where the
price negotiation process and transactions take place and the by the cellular
grid where the urban dynamics are represented.
3.4 Emergence
Emergence remains a central tenet of ABM dealing with climate change.
Most of the models identify the economic outcome as an emergent property
of the system. Other emergent properties are linked to demographic aspects
and, where the spatial dimension is explicit, to land use patterns, which can
be visualized on the grid. Those models that are concerned about mitigation
look at carbon emissions as emerging from the system behaviour. The
studies belonging to the stream on modelling organizations, cooperation,
and collective management see these outcomes as a consequence of
emerging behaviours.
3.5 Interactions
In the climate change arena, ABM is consistently employed in order to
capture the complexity of interactions. With the exceptions of Janssen and
de Vries (1998), Bharwani et al. (2005), Barthel et al. (2008) and
Hasselmann (2008) models investigate interactions both among agents and
between the agents and their environment. Most of the studies show
interdependencies across spatial and temporal scales. In Berman et al.

(2004), Hasselmann (2008) and Mandel et al. (2009) interdependencies are
particularly complex and can manifest with time lags and in form of
feedback loops. In Entwisle et al. (2008) and Beckenbach and Briegel
(2009) social influence is particularly crucial, given their main research
question.
3.6 Heterogeneity
In contrast with the mainstream literature on climate change economics,
with ABM the representative agent is avoided, even in those ground-
breaking macroeconomic applications (Hasselmann 2008; Mandel et al.
2009). Agents can vary for demographic characteristics, location, own
endowment, individual abilities, perception of the world, attitudes and
behaviour. Clearly, the level of diversity is linked to the level of detail of the
model and therefore this ABM ability can be more effectively employed in
the local dimension. However, some degree of aggregation is always
necessary. Heterogeneity can also concern the spatial attributes in those
cases in which the model is spatially explicit, as in Dean et al. (1999),
Werner and McNamara (2007), Barthel et al. (2008), Entwisle et al. (2008)
and Filatova (2009).
3.7 Space and time
Notwithstanding the suitability of the methodology, in the climate change
arena the spatial representation of the environment is not the prevailing
option. More than half of the model considered are not spatially explicit.
Not only those models which are dealing with the global system are aspatial
but also some dealing with local adaptation (Berman et al. 2004; Bharwani
et al. 2005). Instead, Dean et al. (1999), Werner and McNamara (2007),
Barthel et al. (2008), Entwisle et al. (2008) and Filatova (2009) are spatially
explicit and make use of cellular grids. However, in Filatova (2009) the
space represented by the grid remains abstract, while in the rest of these
spatially explicit models the space is based on a GIS capturing the real
geography of the system under analysis. Dean et al. (1999) and Filatova
(2009) implement CA, given the emphasis on neighbouring effects, as by
definition of MAS/LUCC.
Most of the models are run for a time period of approximately 100 years,
where every year is a time step. This is in average a time period of
significance in order to capture climate change effects both in adaptation
and mitigation terms. However, there can be exceptions in both directions.
Dean et al. (1999) consider a 1000 years time period, given the
archaeological value of their study. On the contrary Mandel et al. (2009)
investigate a period of 40 years. In Beckenbach and Briegel (2009) time
steps don't correspond to the years under consideration: a period of 30 years
is simulated in 120 steps, in order to capture more details about the
evolution of the system in the short to medium term. In Filatova (2009) time
is abstract and follows market cycles.
Process scheduling can be programmed in a quite simple way, by executing

the full repertoire of activities for all the agents each year (e.g. Dean et al.
1999), or in a more complex manner. For instance, in Berman et al. (2004)
the sequence of decisions to be taken by agents follows different time clocks
for different activities. Demographic change, household formation, seasonal
wage employment, and migration follow a five-year cycle. On the other
hand, the model recomputes hunting activities dynamically five times per
year.
Given the fact that the methodology shows a certain path dependency,
initialization is a prime object of testing. Initialization is strictly linked to
the model purpose. For example, in Filatova (2009) all land is assumed to be
under agricultural use and the city centre is exogenously set. Conversely,
Dean et al. (1999) is initialized with the available archaeological data while
Berman et al. (2004) with parameters obtained from field work and local
experts.
3.8 Behaviour
The prevailing options for modelling behaviour are: (1) goal oriented
heuristic rules drawn from field work expressed in form of statements and
(2) utility functions based on economic theory expressed in form of
equations. The first are preferred in the most empirical studies such as
Berman et al. (2004) and Bharwani et al. (2005) but there can be exceptions
mixing different options (e.g. Hasselmann 2008). The two studies that are
more concerned about the agent's architecture, Janssen and de Vries (1998)
and Beckenbach and Briegel (2009), employ respectively a genetic
algorithm (Holland 1992) and a satisficing rule (Simon 2000). Janssen and
de Vries (1998) simulated a learning process where agents may change their
mind when they are surprised by observations, and make adjustments in
their decisions according to their new perception of the problem. In
Beckenbach and Briegel (2009) the multiple-self nature of the economic
actor feeds different forces each of which is directed in favour of a possible
mode of action.
Other models insert elements of learning (e.g. Bharwani et al. 2005; Barthel
et al. 2008) and genetic evolution (e.g. Mandel et al. 2009). In Bharwani et
al. (2005) agents are endowed with the capacity of learning from previous
experience so that they can modify their decision trees. In Barthel et al.
(2008) each agent dispose of an history tracing successful and failed plan
execution of previous time steps providing them with learning capabilities.
In Mandel et al. (2009) agents update their belief according to information
from the previous time step. On the long term, technologies and prices
evolve genetically according to the profitability of firms. A genetic
algorithm regulates any economic sector entry and exit, imitation and
mutation.
On the behavioural side, it is worth noting that Berman et al. (2004) and
Bharwani et al. (2005) also admit forms of collective adaptation in order to
respond to harvest shortfalls.
3.9 Verification and Validation

ABM confirms it's main pitfall in validation and verification even in the
climate change arena. Almost half of the literature that we considered
simply don't treat the argument. This is mainly justified by the models’ level
of abstraction, which impose a serious limitation to achieve any form of
model testing. In contrast, those models that employed parallel experiments
(see section 3.1) definitely overcame this problem. In Dean et al. (1999)
verification and validation are extensively treated. Many iterations involving
altered initial conditions, parameters, and random number generators have
been performed in order to assess the model's robustness. Graphical output
of the model includes a map for each year of simulated household residence
and field locations, which runs simultaneously with a map of the
corresponding archaeological and environmental data. These paired maps
facilitate comparison of historical and simulated population dynamics and
residence locations in statistical terms. In Mandel et al. (2009) input-output
tables are used for validation, comparing real data and simulations results.
In Bharwani et al. (2005) the model is driven by data collected from the
field in a bottom-up process. Verification and validation, in accordance with
the “French school”, are achieved through the feedbacks deriving from the
iterative inclusion of stakeholders by means of interviews, questionnaires
and role games.
The remaining models are not fully satisfying from this point of view even
if some have produced significant efforts. In Barthel et al. (2008) the means
of verification and validation that have been applied are indirect and not of
numerical type, including expert knowledge and consumer experiences.
Filatova (2009) compare the model outcomes to the results deriving from
other theories. In particular the land market model has been able to replicate
qualitative properties of the standard equilibrium-based monocentric urban
market model. Berman et al. (2004) achieve statistical verification by means
of Monte Carlo simulations.
3.10 Technical aspects
Almost half of the models that we considered make use of an ABM
platform. Three of them used Repast, one Netlogo and one Vensim, which is
more appropriate for system dynamics but includes some agent-based
features. Four models are programmed from scratch making use of a all set
of different languages including Object Pascal, Visual Basic, UML and
JAVA. As expected OOP turns out to be a real mainstream with regards to
the implementation of ABM.
Quite surprisingly, Janssen and de Vries (1998) and Werner and McNamara
(2007) only rely on mathematical equations. This proves the ability of ABM
to be expressed in mathematical terms even if maths is not the ABM natural
environment.
4. Conclusions
This paper reviewed the state of the art in ABM. We were interested in

understanding how consolidated is this approach in dealing with the
complexity of the coupled human-environment systems. More specifically,
we wanted to investigate whether if ABM could be an innovative but sound
methodology to model the dynamics of SES exposed to climate change and
assess the related policies.
Our analysis suggests that ABM is today a quite consolidated
interdisciplinary approach. In particular we showed that, at the theoretical
level, the research questions are the same across social, economic and
environmental applications. We were able to identify six main research
purposes that we called the streams of research of ABM, as reported in
section 2. The resulting framework can be used to categorize any ABM
effort belonging to any sustainability dimension. This may support
Boulanger and Bréchet (2005) who concluded that ABM is the most
promising modelling approach for sustainability science. The intrinsic trans-
disciplinarity of the methodology certainly justifies its application to the
modelling of SES, where the human and the environmental systems co-
evolve and a significant integration of the knowledge belonging to different
domains is needed.
Past research often regarded ABM as bottom-up methodology alternative to
top-down equilibrium-based models but, to our knowledge, few publications
have some relevancy in the climate change arena. Therefore, we reviewed
those agent-based models of SES which included some kind of climate
change related issue in order to clarify how the main ABM elements are
applied, according to the systems under analysis. Our analysis, in section 3,
described how the notions of agent, environment, emergence, interaction,
heterogeneity, space and time, behaviour and validation are treated in each
study, including the computing languages, tools and platforms used for the
simulations. The results support the idea that ABM is an appropriate bottom-
up methodology for the exploration of climate policies.
ABM seems particularly well suited to the analysis of adaptation to climate
change of local systems. Applications of this type spawn across all the
streams of research of ABM composing the main body of work on agent-
based models dealing with climate change. Households are the most crucial
agents while the environment is the natural and economic landscape that can
be expressed in spatially explicit terms and/or in form of sub-models
describing theoretical spaces of interaction.
Surprisingly ABM also shows the possibility to be employed in more top-
down orientations where the main issue is mitigation at a global level. Few
ground-breaking studies are showing the way to insert agent-based thinking
into macroeconomic models overcoming some unrealistic aggregative
simplifications of traditional equilibrium models. One possible direction for
further development of ABM research on climate change is the joint
analysis of mitigation and adaptation.
In addition to the expected qualities of the methodology, i.e. the emergence
of outcomes at the macro-level from micro-interactions, some specific

strengths of ABM are particularly meaningful when dealing with climate
change. The main advantages of ABM applied to climate change related
issues are the abilities to take into account adaptive behaviour at the
individual or system level and to introduce a higher degree of heterogeneity
resulting into a more natural representation of the system, compared to
equilibrium-based models.
In the climate change arena adaptive behaviour means the possibility to
enable the SES to react, which is crucial in order to avoid unrealistic or
meaningless results. At this early stage, behavioural architectures are mainly
based on heuristic rules and on utility theory. More specific architectures
exist but are not often employed.
Heterogeneity is another particularly relevant aspect, because people have
different perceptions of the risk, environmental sensitivities, capacity to
cope with change and so forth. Neglecting this diversity may lead to missing
some crucial driver of change. ABM effectively shows the ability to
overcome this problem.
The main disadvantage of ABM, as in other domains of application, stays in
the challenges of testing the model which is not always very clear and often
neglected. Where feasible participatory approaches seem the most suitable
solution. For this reasons local applications may appear more robust.
Further research is needed to consolidate ABM applications to the global
system.
Two open issues should finally be highlighted and are related to
programming and documenting agent-based models.
While there already exist various ABM packages and tools that can be
employed in this field (e.g. Repast and NetLogo), it makes sense to think
about a dedicated platform, for the future, which could simplify the
modelling options into local and global systems and posses a library of
household type agents and of specific socio-cognitive models of adaptation.
This would certainly improve the accessibility of the methodology to those
who cannot spend too much time in learning a programming language.
Finally, a communication barrier remains evident. While our specification
effort in table 2 may be more appropriate to explain the models to a public
not trained on ABM, we also felt the need to find a common communication
standard of the models we were analysing. We therefore recommend to the
modellers to take into account a protocol such as in Parker et al. (2001)
and/or Grimm et al. (2006) for their future publications.
References
Albin, P., and D. K. Foley. 1992. Decentralized, dispersed exchange without
an auctioneer: A simulation study. Journal of Economic Behavior and
Organization 18, no. 1: 27-52.
Arthur, W. B. 1990. Positive Feedbacks in the Economy. Scientific American
262: 92-99.
———. 1993. On designing economic agents that behave like human

agents. Journal of Evolutionary Economics 3, no. 1: 1-22.
———. 1999. Complexity and the economy. Science 284, no. 5411: 107.
———. 2006. Out-of-equilibrium economics and agent-based modeling. In
Handbook of computational economics, ed. L. Tesfatsion and K. L.
Judd, 2:1551-1564. Elsevier, North-Holland. Vol. 2. Amsterdam.
Arthur, W. B., J. H. Holl, B. Lebaron, R. Palmer, and P. Tayler. 1996. Asset
pricing under endogenous expectations in an artificial stock market.
Santa Fe Institute Working Papers: 96-12-093.
Axelrod, R., and W. D. Hamilton. 1981. The evolution of cooperation.
Science 211, no. 4489: 1390-1396.
Axelrod, R. M. 1997. The complexity of cooperation: Agent-based models of
competition and collaboration. Princeton, NJ, USA: Princeton
University Press.
Baas, NA, and C. Emmeche. 1997. On Emergence and Explanation.
Intellectica 2, no. 25: 67-83.
Balmann, A. 1997. Farm-Based Modelling of Regional Structural Change: A
Cellular Automata Approach. European Review of Agricultural
Economics 24, no. 1: 85-108.
Barker, T. 2008. The economics of avoiding dangerous climate change.
Climatic Change, in press.
Barreteau, O., and F. Bousquet. 2000. Shadoc: a multi-agent model to tackle
viability of irrigated systems. Annals of Operations Research 94, no.
1: 139-162.
Barreteau, O., F. Bousquet, and J. M. Attonaty. 2001. Role-playing games
for opening the black box of multi-agent systems: method and
lessons of its application to Senegal River Valley irrigated systems.
Journal of Artificial Societies and Social Simulation 4, no. 2: 5.
Barthel, R., S. Janisch, N. Schwarz, A. Trifkovic, D. Nickel, C. Schulz, and
W. Mauser. 2008. An integrated modelling framework for simulating
regional-scale actor responses to global change in the water domain.
Environmental Modelling and Software 23: 1095-1121.
Batty, M. 2005. Cities and complexity. Cambridge, MA, USA: MIT Press.
Beckenbach, F., and R. Briegel. 2009. Multi-agent modelling of economic
innovation dynamics and its implication for analyzing emissions
impact. Working Paper, University of Kassel.
Bell, A. M. 2001. Reinforcement learning rules in a repeated game.
Computational Economics 18, no. 1: 89-110.
Berger, T. 2001. Agent-based spatial models applied to agriculture: a
simulation tool for technology diffusion, resource use changes and
policy analysis. Agricultural Economics 25, no. 2-3: 245-260.
Berman, M., C. Nicolson, G. Kofinas, J. Tetlichi, and S. Martin. 2004.
Adaptation and sustainability in a small arctic community: Results of
an agent-based simulation model. Arctic 57, no. 4: 401-414.
Bharwani, S., M. Bithell, T. E. Downing, M. New, R. Washington, and G.
Ziervogel. 2005. Multi-agent modelling of climate outlooks and food
security on a community garden scheme in Limpopo, South Africa.
Philosophical Transactions of the Royal Society B: Biological
Sciences 360, no. 1463: 2183.

Bonabeau, E. 2002. Agent-based modeling: Methods and techniques for
simulating human systems. Proceedings of the National Academy of
Sciences 99, no. 3: 7280-7287.
Booker, L. B., D. E. Goldberg, and J. H. Holland. 1989. Classifier systems
and genetic algorithms. Artificial Intelligence 40, no. 1-3: 235-282.
Boulanger, P. M., and T. Bréchet. 2005. Models for policy-making in
sustainable development: The state of the art and perspectives for
research. Ecological economics 55, no. 3: 337-350.
Bousquet, F., O. Barreteau, C. Le Page, C. Mullon, and J. Weber. 1999. An
environmental modelling approach: the use of multi-agent
simulations. In Advances in environmental modelling, ed. F. Blasco,
113-122. Elsevier. Paris.
Bousquet, F., C. Cambier, C. Mullon, P. Morand, J. Quensiere, and A. Pave.
1993. Simulating the interaction between a society and a renewable
resource. Journal of biological systems 1, no. 2: 199-214.
Bousquet, F., and C. Le Page. 2004. Multi-agent simulations and ecosystem
management: a review. Ecological Modelling 176, no. 3-4: 313-332.
Bower, J., and D. Bunn. 2001. Experimental analysis of the efficiency of
uniform-price versus discriminatory auctions in the England and
Wales electricity market. Journal of Economic Dynamics and
Control 25, no. 3-4: 561-592.
Bradbury, R. 2002. Futures, predictions, and other foolishness. In
Complexity and Ecosystem Management: The Theory and Practice
of Multi-Agent Approaches, ed. M. Janssen, 48–62. Edward Elgar.
Cheltenham.
Camerer, C. 2003. Behavioral game theory: Experiments in strategic
interaction. Princeton, NJ, USA: Princeton University Press.
Carley, Kathleen M. 1996. A comparison of artificial and human
organizations. Journal of Economic Behavior & Organization 31,
no. 2 (November): 175-191. doi:10.1016/S0167-2681(96)00896-7.
Cecconi, F., and D. Parisi. 1998. Individual versus social survival strategies.
Journal of Artificial Societies and Social Simulation 1, no. 2: 1–17.
Chan, N. T., B. LeBaron, A. W. Lo, and T. Poggio. 1999. Agent-based
models of financial markets: A comparison with experimental
markets. MIT Artificial Markets Project Working Paper 124.
Chattoe, E. 1998. Just how (un) realistic are evolutionary algorithms as
representations of social processes? Journal of Artificial Societies
and Social Simulation 1, no. 3.
Cohen, M. D., R. Riolo, and R. Axelrod. 2001. The role of social structure
in the maintenance of cooperative regimes. Rationality and Society
13, no. 1: 5-32.
Conte, R., B. Edmonds, S. Moss, and R. K. Sawyer. 2001. Sociology and
social theory in agent based social simulation: A symposium.
Computational & Mathematical Organization Theory 7, no. 3: 183-
205.
Cox, B. J. 1986. Object oriented programming: an evolutionary approach.
Boston, MA, USA: Addison-Wesley Longman Publishing Co.
Dawid, H. 1996. Adaptive learning by genetic algorithms: Analytical results

and applications to economic models. Revised Second Edition.
Berlin: Springer Verlag.
De Vany, A., and C. Lee. 2001. Quality signals in information cascades and
the dynamics of the distribution of motion picture box office
revenues. Journal of Economic Dynamics and Control 25, no. 3-4:
593-614.
Deadman, P., and R. H. Gimblett. 1994. A role for goal-oriented
autonomous agents in modeling people-environment interactions in
forest recreation. Mathematical and Computer Modelling 20, no. 8:
121-133.
Dean, J. S., G. J. Gumerman, J. M. Epstein, R. L. Axtell, A. C. Swedlund,
M. T. Parker, and S. McCarroll. 1999. Understanding Anasazi
culture change through agent-based modeling. In Dynamics in
human and primate societies: Agent-based modeling of social and
spatial processes, ed. T. A. Kohler and G. J. Gumerman, 179–205.
Oxford University Press.
Deffuant, G., S. Huet, J. P. Bousset, J. Henriot, G. Amon, and G. Weisbuch.
2002. Agent based simulation of organic farming conversion in
Allier département. In Complexity and Ecosystem Management: The
Theory and Practice of Multi-Agent Approaches, ed. M. Janssen,
158-187. Edward Elgar. Cheltenham.
Deffuant, G., D. Neau, F. Amblard, and G. Weisbuch. 2000. Mixing beliefs
among interacting agents. Advances in Complex Systems 3, no. 4:
87-98.
Drogoul, A., and J. Ferber. 1994. Multi-agent simulation as a tool for
modeling societies: Application to social differentiation in ant
colonies. Lecture Notes in Computer Science 830: 3-23.
Emonet, T., C. M. Macal, M. J. North, C. E. Wickersham, and P. Cluzel.
2005. AgentCell: a digital single-cell assay for bacterial chemotaxis.
Bioinformatics 21, no. 11: 2714-2721.
Entwisle, B., G. Malanson, R. R. Rindfuss, and S. J. Walsh. 2008. An agent-
based model of household dynamics and land use change. Journal of
Land Use Science 3, no. 1: 73-93.
Epstein, J. M., and R. Axtell. 1996. Growing artificial societies: Social
science from the bottom up. Cambridge, MA, USA: MIT Press.
Ferber, J. 1999. Multi-agent systems: an introduction to distributed artificial
intelligence. Addison Wesley Longman.
Feuillette, S., F. Bousquet, and P. Le Goulven. 2003. Sinuse: a multi-agent
model to negotiate water demand management on a free access
water table. Environmental Modelling and Software 18, no. 5: 413-
427.
Filatova, T. 2009. Land markets from the bottom up. Micro-macro links in
economics and implications for coastal risk management. University
of Twente, PhD Thesis.
Funtowicz, S., and J. Ravetz. 1995. Science for the post-normal age.
Perspectives on Ecological Integrity: 34-48.
Gilbert, N. 2004. Agent-based social simulation: dealing with complexity.
Working Paper, Centre for Research in Social Simulation, University

of Surrey.
Gilbert, N., and A. Abbott. 2005. Introduction to special issue: social science
computation. American Journal of Sociology 110, no. 4: 859-863.
Gilbert, N., and S. Bankes. 2002. Platforms and methods for agent-based
modeling. Proceedings of the National Academy of Sciences 99, no.
3: 7197-7198.
Gilbert, N., and K. G. Troitzsch. 1999. Simulation for the social scientist.
Buckingam, UK: Open University Press.
Gintis, H. 2000. Beyond homo economicus: evidence from experimental
economics. Ecological economics 35, no. 3: 311-322.
Goldspink, C. 2000. Modelling social systems as complex: Towards a social
simulation meta-model. Journal of Artificial Societies and Social
Simulation 3, no. 2: 1-23.
Grand, S., and D. Cliff. 1998. Creatures: Entertainment software agents with
artificial life. Autonomous Agents and Multi-Agent Systems 1, no. 1:
39-57.
Grimm, V. 1999. Ten years of individual-based modelling in ecology: what
have we learned and what could we learn in the future? Ecological
modelling 115, no. 2: 129-148.
Grimm, V., U. Berger, F. Bastiansen, S. Eliassen, V. Ginot, J. Giske, J. Goss-
Custard, T. Grand, S. K. Heinz, and G. Huse. 2006. A standard
protocol for describing individual-based and agent-based models.
Ecological Modelling 198, no. 1-2: 115-126.
Grimm, V., and S. F. Railsback. 2005. Individual-based modeling and
ecology. Princeton and Oxford: Princeton University Press.
Grothmann, T., and A. Patt. 2005. Adaptive capacity and human cognition:
the process of individual adaptation to climate change. Global
Environmental Change 15, no. 3: 199-213.
Gunderson, L. H., and C. S. Holling. 2002. Panarchy: understanding
transformations in human and natural systems. Washington, D.C.:
Island Press.
Guyot, P., and S. Honiden. 2006. Agent-based participatory simulations:
Merging multi-agent systems and role-playing games. Journal of
Artificial Societies and Social Simulation 9, no. 4.
Hare, M., and P. Deadman. 2004. Further towards a taxonomy of agent-
based simulation models in environmental management.
Mathematics and Computers in Simulation 64, no. 1: 25-40.
Hasselmann, K. 2008. Simulating human behaviour in macroeconomic
models applied to climate change. In Proceedings of the Heraeus
seminar on Energy and Climate, 25. Bad Hoeneff, Germany, May
26.
Holland, J. H. 1992. Adaptation in natural and artificial systems.
Cambridge, MA, USA: MIT Press.
Holling, C. S. 1978. Adaptive environmental assessment and management.
London: John Wiley & Sons.
Hommes, C. H. 2002. Modeling the stylized facts in finance through simple
nonlinear adaptive systems. Proceedings of the National Academy of
Sciences 99, no. 3: 7221-7228.

Huhns, M. N., and L. M. Stephens. 1999. Multiagent systems and society of
agents. In Multiagent Systems: A Modern Approach to Distributed
Artificial Intelligence, ed. G. Weiss, 79-122. Cambridge, MA, USA:
MIT Press.
Huston, M., D. DeAngelis, and W. Post. 1988. New computer models unify
ecological theory. BioScience, no. 38: 682-691.
Janssen, M. A. 2005. Agent-based modelling. In Modelling in ecological
economics., ed. J. L. R. Proops and P. Safonov, 155-172. Edward
Elgar Publishers. UK.
Janssen, M. A., and W. Jager. 2002. Stimulating diffusion of green products.
Journal of Evolutionary Economics 12, no. 3: 283-306.
Janssen, M. A., and B. de Vries. 1998. The battle of perspectives: a multi-
agent model with adaptive responses to climate change. Ecological
Economics 26, no. 1: 43-65.
Kohler, T. A., and E. Carr. 1996. Swarm-based modeling of prehistoric
settlement systems in southwestern North America. In
Archaeological applications of GIS: proceedings of Colloquium II,
ed. I. Johnson and N. MacLaren. Forli, Italy, September.
Kohler, T. A., and G. J. Gumerman. 1999. Dynamics in human and primate
societies: Agent-based modeling of social and spatial processes.
New York: Oxford University Press.
Krawczyk, K., W. Dzwinel, and D. A. Yuen. 2003. Nonlinear development
of bacterial colony modeled with cellular automata and agent
objects. International Journal of Modern Physics C-Physics and
Computer 14, no. 10: 1385-1404.
Langton, C. G. 1992. Life at the edge of chaos. In Articial Life II, vol. X of
SFI Studies in the Sciences of Complexity, ed. C. G. Langton, C.
Taylor, J. D. Farmer, and S. Rasmussen, 41-91. Redwood City, CA,
USA: Addison-Wesley.
Lansing, J. S., and J. N. Kremer. 1993. Emergent properties of Balinese
water temple networks: Coadaptation on a rugged fitness landscape.
American Anthropologist 95, no. 1: 97-114.
Latane, B. 1996. Dynamic social impact: The creation of culture by
communication. The Journal of Communication 46, no. 4: 13-25.
Le Page, C., F. Bousquet, I. Bakam, A. Bah, and C. Baron. 2000. CORMAS:
A multiagent simulation toolkit to model natural and social dynamics
at multiple scales. In Proceedings of the Workshop" The ecology of
scales". Wageningen, The Netherlands, June.
LeBaron, B. 2000. Agent-based computational finance: Suggested readings
and early research. Journal of Economic Dynamics and Control 24,
no. 5-7: 679-702.
Lempert, R. 2002. Agent-Based Modeling as Organizational and Public
Policy Simulators. Proceedings of the National Academy of Sciences
99, no. 3: 7195-7196.
Lomi, A., and E. R. Larsen. 1998. Density delay and organizational survival:
computational models and empirical comparisons. Computational &
Mathematical Organization Theory 3, no. 4: 219-247.
Luna, F. 2002. Computable learning, neural networks and institutions.

Studies in Fuzzines and Soft Computing 100: 211-232.
Luna, F., and B. Stefansson. 2000. Economic Simulations in Swarm: Agent-
based modelling and object oriented programming. Dordrecht, The
Netherlands: Kluwer Academic Publishers.
Macal, C. M., and M. J. North. 2005. Tutorial on agent-based modeling and
simulation. In Proceedings of the 37th Winter Simulation
Conference, 2-15. Orlando, FL, USA.
Macy, M. W., and R. Willer. 2002. From Factors to Actors: Computational
Sociology and Agent-Based Modeling. Annual review of sociology
28, no. 1: 143-166.
Mandel, A., S. Furst, L. Wiebke, F. Meissner, and C. Jaeger. 2009. Lagom
generiC: an agent-based model of growing economies. European
Climate Forum, Working Paper 1/2009. Potsdam. http://ecf.pik-
potsdam.de/Images/Lagom%20generiC.pdf.
Manson, S. M. 2001. Simplifying complexity: a review of complexity
theory. Geoforum 32, no. 3: 405-414.
———. 2005. Agent-based modeling and genetic programming for
modeling land change in the Southern Yucatan Peninsular Region of
Mexico. Agriculture, Ecosystems and Environment 111, no. 1-4: 47-
62.
Marks, R. E. 1992. Breeding hybrid strategies: Optimal behaviour for
oligopolists. Journal of Evolutionary Economics 2, no. 1: 17-38.
Martens, P. 2006. Sustainability: science or fiction? Sustainability: Science
Practice and Policy 2, no. 1: 36-41.
Mathevet, R., F. Bousquet, C. Le Page, and M. Antona. 2003. Agent-based
simulations of interactions between duck population, farming
decisions and leasing of hunting rights in the Camargue (Southern
France). Ecological modelling 165, no. 2-3: 107-126.
Matthews, R. B., N. G. Gilbert, A. Roach, J. G. Polhill, and N. M. Gotts.
2007. Agent-based land-use models: a review of applications.
Landscape Ecology 22, no. 10: 1447-1459.
Moss, S., C. Pahl-Wostl, and T. Downing. 2001. Agent-based integrated
assessment modelling: the example of climate change. Integrated
Assessment 2, no. 1: 17-30.
Nowak, A., R. R. Vallacher, S. J. Read, and L. C. Miller. 1998. Toward
computational social psychology: cellular automata and neural
network models of interpersonal dynamics. In Connectionist models
of social reasoning and social behavior, 277-311. Mahwah, NJ:
Lawrence Erlbaum Associates.
Pahl-Wostl, C. 2007. The implications of complexity for integrated
resources management. Environmental Modelling and Software 22,
no. 5: 561-569.
Parker, D. C., T. Berger, S. Manson, and W. J. McConnell. 2001. Agent-
Based Models of Land-Use and Land-Cover Change. LUCC Report
Series No. 6, Volume 1. http://www.globallandproject.org/
Documents/LUCC_No_6.pdf.
Parker, D. C., and V. Meretsky. 2004. Measuring pattern outcomes in an
agent-based model of edge-effect externalities using spatial metrics.

Agriculture, Ecosystems and Environment 101, no. 2-3: 233-250.
Parker, D.C., S. M. Manson, M. A. Janssen, M. J. Hoffmann, and P.
Deadman. 2003. Multi-agent systems for the simulation of land-use
and land-cover change: a review. Annals of the Association of
American Geographers 93, no. 2: 314-337.
Patt, A., and B. Siebenhüner. 2005. Agent Based Modeling and Adaptation
to Climate Change. Vierteljahrshefte zur Wirtschaftsforschung 2, no.
74: 310-320.
Perez, P., and D. F. Batten. 2006. Complex science for a complex world:
exploring human ecosystems with agents. Camberra: ANU E Press.
Prietula, M. J., K. M. Carley, and L. Gasser. 1998. Simulating organizations:
Computational models of institutions and groups. Cambridge, MA,
USA: MIT Press.
Railsback, S. F., Steven L. Lytinen, and Stephen K. Jackson. 2006. Agent-
based Simulation Platforms: Review and Development
Recommendations. SIMULATION 82, no. 9: 609-623.
Rammel, C., S. Stagl, and H. Wilfing. 2007. Managing complex adaptive
systems. A co-evolutionary perspective on natural resource
management. Ecological Economics 63, no. 1: 9-21.
Reynolds, C. W. 1987. Flocks, herds, and schools: A distributed behavior
model. Computer Graphics 21, no. 4: 25-34.
Rosenkopf, L., and E. Abrahamson. 1999. Modeling reputational and
informational influences in threshold models of bandwagon
innovation diffusion. Computational & Mathematical Organization
Theory 5, no. 4: 361-384.
Rouchier, J., F. Bousquet, M. Requier-Desjardins, and M. Antona. 2001. A
multi-agent model for describing transhumance in North Cameroon:
Comparison of different rationality to develop a routine. Journal of
Economic Dynamics and Control 25, no. 3-4: 527-559.
Rumelhart, D. E., and J. L. McClelland. 1986. Parallel distributed
processing: explorations in the microstructure of cognition, vol. 2:
psychological and biological models. MIT Press.
Savage, M., and M. Askenazi. 1998. Arborscapes: A swarm-based multi-
agent ecological disturbance model. Santa Fe Institute Working
Paper: 98-06-056.
Schelling, T. C. 1971. Dynamic models of segregation. Journal of
Mathematical Sociology 1: 143-186.
———. 1978. Micromotives and Macrobehavior. New York: Norton.
Schut, M. 2007. Scientific Handbook for Simulation of Collective
Intelligence. 2nd ed. Creative Commons, February.
http://www.scribd.com/doc/244626/Scientific-Handbook-for-the-
Simulation-of-Collective-Intelligence.
Simon, H. A. 1955. A behavioral model of rational choice. The Quarterly
Journal of Economics 69, no. 1: 99-118.
———. 2000. Bounded rationality in social science: Today and tomorrow.
Mind & Society 1, no. 1: 25-39.
Smith, V. L. 1989. Theory, experiment and economics. The Journal of
Economic Perspectives 3, no. 1: 151-169.

Takahashi, N. 2000. The Emergence of Generalized Exchange. American
Journal of Sociology 105, no. 4: 1105-1134.
Terna, P. 1998. Simulation tools for social scientists: Building agent based
models with swarm. Journal of Artificial Societies and Social
Simulation 1, no. 2: 1-12.
Tesfatsion, L. 2001. Structure, behavior, and market power in an
evolutionary labor market with adaptive search. Journal of
Economic Dynamics and Control 25, no. 3-4: 419-457.
———. 2002. Agent-based computational economics: Growing economies
from the bottom up. Artificial Life 8, no. 1: 55-82.
———. 2003. Agent-based computational economics: modeling economies
as complex adaptive systems. Information Sciences 149, no. 4: 262-
268.
Tesfatsion, L., and K. L. Judd. 2006. Handbook of Computational
Economics: Agent-Based Computational Economics. Vol. 2.
Amsterdam: Elsevier, North-Holland.
Tobias, R., and C. Hofmann. 2004. Evaluation of free Java-libraries for
social-scientific agent based simulation. Journal of Artificial
Societies and Social Simulation 7, no. 1.
Torrens, P. M. 2001. SprawlSim: modeling sprawling urban growth using
automata-based models. In Agent-Based Models of Land-Use and
Land-Cover Change. LUCC Report Series No. 6, Volume 1, ed. D.C.
Parker, T. Berger, S. M. Manson, and W. J. McConnell, 72-78.
Torrens, P. M., and D. O Sullivan. 2001. Cellular automata and urban
simulation: where do we go from here? Environment and Planning B
28, no. 2: 163-168.
Troisi, A., V. Wong, and M. A. Ratner. 2005. An agent-based approach for
modeling molecular self-organization. Proceedings of the National
Academy of Sciences 102, no. 2: 255-260.
Veldkamp, A., and P. H. Verburg. 2004. Modelling land use change and
environmental impact. Journal of Environmental Management 72,
no. 1-2: 1-3.
Voinov, Alexey. 2008. Systems Science and Modeling for Ecological
Economics. Elsevier, Academic Press.
Vriend, N. J. 2000. An illustration of the essential difference between
individual and social learning, and its consequences for
computational analyses. Journal of Economic Dynamics and Control
24, no. 1: 1-19.
Waldrop, M. M. 1992. Complexity: The emerging science at the edge of
order and chaos. Simon & Schuster Paperbacks. New York.
Weiss, G. 1999. Multiagent systems: a modern approach to distributed
artificial intelligence. Cambridge, MA, USA: MIT Press.
Weitzman, M. L. 2009. On modeling and interpreting the economics of
catastrophic climate change. The Review of Economics and Statistics
91, no. 1: 1-19.
Werner, B. T., and D. E. McNamara. 2007. Dynamics of coupled human-
landscape systems. Geomorphology 91, no. 3-4: 393-407.
Wooldridge, M., and N. R. Jennings. 1995. Intelligent agents: Theory and

practice. Knowledge engineering review 10, no. 2: 115-152.
Ziervogel, G., M. Bithell, R. Washington, and T. Downing. 2005. Agent-
based social simulation: a method for assessing the impact of
seasonal climate forecast applications among smallholder farmers.
Agricultural Systems 83, no. 1: 1-26.

Table 1 – Classification of references according to scientific domain and stream of research
ABM Streams \ Domains Social Economic Environmental
Self-organization and
co-evolution of the
system
(SOCES)
Schelling (1971)
Epstein & Axtell (1996)
Marks (1992)
Arthur et al. (1996)
Bower & Bunn (2001)
Hommes (2002)
Bousquet et al. (1993)
Deadman &
Gimblett (1994)
Kohler & Carr (1996)
Balmann (1997)
Torrens (2001)
Parker &
Meretsky (2004)
Diffusion processes and
networks formation
(DPNF)
Latane (1996)
Nowak &
Vallacher (1998)
Rosenkopf &
Abrahamson (1999)
Albin & Foley (1992)
De Vany & Lee (2001)
Tesfatsion (2001)
Janssen & Jager (2002)
Rouchier et al. (2001)
Berger (2001)
Deffuant et al. (2002)
ABM Main Research Questions
Modelling
organizations,
cooperation and
collective management
(MOCCM)
Axelrod &
Hamilton (1981)
Cohen et al. (2001)
Takahashi (2000)
Cecconi & Parisi (1998)
Prietula et al. (1998)
Lansing &
Kremer (1993)
Barreteau &
Bousquet (2000)
Becu et al. (2003)
Feuillette et al. (2003)
Mathevet et al. (2003)
Parallel experiments
(PE) Lomi & Larsen (1998)
Carley (1996) Arthur (1993)
Chan et al. (1999) Bousquet et al. (1999)
Barreteau et al. (2001)
Guyot & Honiden (2006)
Agent's architecture
(AA) Rumelhart &
McClelland (1986)
Holland (1992)
Booker et al. (1989)
Dawid (1996)
Chattoe-Brown (1998)
Gintis (2000)
Vriend (2000)
Bell (2001)
Luna (2002)
Reynolds (1987)
Drogoul & Ferber (1994)
Wooldridge &
Jennings (1995)
Grand & Cliff (1998)
Deffuant et al. (2000)
Manson (2005)
ABM Accessories and Tools
Programming
(P) Terna (1998)
Gilbert & Bankes (2002)
Tobias &
Hofmann (2004)
Luna &
Stefansson (2000)
Savage &
Askenazi (1998)
Le Page et al. (2000)
North et al. (2006)
Railsback et al. (2006)

Table 2 – Comparative analysis of agent-based models of SES with climate change elements
Reference Stream of
research
System
under
study1
Climate
issue2 Agents3 Environment4 Emergence5 Interactions6 Heterogeneity7 Space / Time8 Behaviour9 Verification
and
Validation
Technical
aspects
Janssen & de
Vries (1998) SOCES &
AA GL CE DM Economy-energy-
climate model EO, CE A-E Agent's cultural
perspectives Aspatial;
100 years GA Absent Mathematical
equations
Dean et al.
(1999) SOCES &
PE
LL
Arizona
(US) R HH SCG and production
model PLUC A-E, A-A Agent's age, location and
grain stocks;
SA
CA, GIS based;
1000 years HR Statistical Programmed in
Object Pascal
Berman et al.
(2004) MOCCM LL
Canada T, SP I, HH Environmental,
economic and social
indicators
EO and
demographic
change A-E, A-A Agent's age, HH type,
education, wage-work and
hunting-time capabilities
Aspatial;
40 years;
complex
scheduling
HR; CR Statistical
verification Programmed in
Visual Basic
Bharwani et al.
(2005) MOCCM &
PE
LL
South
Africa F, D Farming
HH Planting fields and
market place model Crop yields and
food security A-E Agent's wealth, crop type,
location and timing Aspatial;
100 years Decision tree
rules; LA; CR Participatory Repast
Werner &
McNamara
(2007) SOCES
LL
Georgia
(US)
F, H 7 sets of
economic
agents
Economic model and
landscape model PLUC A-E, A-A Agent's types, prediction
models and utility
functions; SA
GIS
re-sampled on a
100 x 100 grid;
~200 years
U functions Absent Matlab
Barthel et al.
(2008) MOCCM CRL
Germany T, R HH, C, F SCG Water supply and
consumption A-E Agents' type, location,
level, behaviour,
preferences and plans; SA
1x1 km cells,
GIS based;
100 years
U based
decision rules
and LA
Partial
Participatory
DeepActor
programmed in
UML 2.0
Entwisle et al.
(2008) DPNF CRL
Thailand R, F, D I, HH, C SCG and social
networks
Migration, social
connections and
PLUC A-E, A-A Agent's demography,
wealth, social ties; SA
GIS based grid;
time not
specified
Probability
rules Under
development Repast
Hasselmann
(2008) SOCES GL
T, CE F, HH,
Banks, DM Three levels
macroeconomic model EO, CE A-E Environment levels, agent's
objectives, physical units Aspatial;
100 years HR Absent Vensim
Beckenbach &
Briegel (2009) DPNF &
AA GL
economy CE F Sectoral demand model
and inter-sectoral
input/output tables EO, CE A-E, A-A Agents' prevailing force
among innovation imitation
routine
Aspatial;
120 time steps
equal to 30
years
Satisficing
rules balancing
different goals Absent Repast
Filatova (2009) SOCES LL
Holland F HH,
land owners SCG and Land market
model Land prices and
PLUC A-E, A-A Agents' location
preferences , individual
budget, risk perception; SA
CA, 35 x 63
cells; abstracts
space and time
U
maximization Structural NetLogo
Mandel et al.
(2009) SOCES &
PE & P
CRL
German
economy CE HH, F, DM,
Financial
system
Economic process
as schedule of
events
EO,
unemployment,
wages. A-E, A-A Agents' type, economic
activities, time steps Aspatial;
40 years HR, GA Statistical Lagom generiC
programmed in
Java
Notes to table 2:
1 Global Level (GL), Country or Regional Level (CRL), Local Level (LL). 2 Carbon Emissions (CE), Temperature (T), Rainfall (R), Snow Precipitations (SP), Floods (F),
Droughts (D), Hurricanes (H). 3 Households (HH), Individuals (I), Decision Makers (DM), Communities (C), Firms (F). 4 Spatial Cellular Grid (SPG). 5 Economic Output (EO),
Patterns of Land Use and Cover (PLUC). 6 Agent-Environment (A-E), Agent-Agent (A-A). 7 Spatial Attributes (SA). 8 Cellular Automata (CA), Geographical Information
Systems (GIS). 9 Heuristic Rules (HR), Utility (U), Learning Algorithm (LA), Genetic Algorithm (GA), Collective Response (CR).