Desktop version

Home arrow Sociology arrow Understanding Society and Natural Resources

Representation of Human-Environment Interactions in Land Change Models

A wide variety of land change models have been developed over the past two decades that have been reviewed numerous times (Agarwal et al. 2001; Priess and Schaldach 2008; Verburg et al. 2004b) based on different criteria. In this section we do not aim to provide an exhaustive review of these models, but rather will address the variation in ways that human-environment interactions are conceptualized in the different models. In contrast to a classification or representation based on the specific modeling technique used, e.g. cellular automata or agent-based modeling, the methods employed to represent human-environment interactions may be classified on a scale from deductive, theory-led approaches to fully empirical, inductive approaches to modeling. Overmars et al. (2007) provide such a scale from deductive to inductive reasoning and conclude that many of the existing models are neither fully deductive or inductive. But, still large differences exist in the role of theory and empirical data in conceptualizing the model. Especially the way in which decision making on land change is represented differs. In some approaches an almost completely deductive approach is taken by assuming rational agents that optimize income and tailor land change decisions towards that goal. Some of these models operate at the level of individual decision makers, e.g. farmers (Piorr et al. 2009) while others operate at the level of large world regions in which decision making is conceptualized for an aggregate (representative) agent (Havlík et al. 2011; Souty et al. 2012; Van Meijl et al. 2006). It may be questioned under what conditions the same behavioral assumptions are valid for both individual and highly aggregate agents. On the other end of the spectrum models that employ machine learning methods to relate land change to its determinants are found. Many machine learning techniques do not provide insight into the estimated relations and it is only the observed data that determine the relations employed in the model to simulate future land changes. Many other models fall somewhere in between these extremes. So called 'factor-led induction' (Overmars et al. 2007) employs theory to identify the factors driving land change decisions while the actual relations between these factors and land change are established using empirical estimation of statistical coefficients using observed data (Chomitz and Gray 1996; Nelson and Hellerstein 1997). Such a theory-based approach is important to explore for several reasons. It structures the model around the critical human-environment relationships identified within the theory, and focuses attention on the data required to explore those relationships. Similarly, many agent-based models of land change employ a range of empirical techniques to make a typology of different decision making types and parameterize the decision making rules in the model based on household survey results (Robinson et al. 2007; Smajgl et al. 2011; Valbuena et al. 2008). The latter group of models is of specific interest to the study of human-environment interactions. Multi-agent models simulate decision making by individual agents of land use change, explicitly addressing interactions among individuals. The explicit attention for interactions between agents makes it possible for this type of model to simulate emergent properties of systems. These are properties at the macro scale that are not predictable from observing the micro units in isolation. If the decision rules of the agents are set such that they sufficiently look like human decision making they can simulate behavior at the meso-level of social organization, i.e. the behavior of heterogeneous groups of actors. Multi-agent-based models of land change are particularly well suited to representing complex spatial interactions under heterogeneous conditions (Bousquet and Le Page 2004; Parker et al. 2003). Multi-agent systems are able to formalize decision-forming behavior of individual stakeholders, either based on theory (Happe et al. 2006), or based on observations and statistical analysis (Bousquet et al. 2001; Robinson et al. 2007; Valbuena et al. 2010a). In the initial years of application of agent-based models to land change, most multi-agent models focused on either hypothetical or simplified representations of the real world to explore interactions between agents and between agents and the environment. Especially the parameterization of agent behavior in models for real case-studies turned out to be very complex. However, more recently a larger number of applications of agent-based models to real case studies worldwide have been published, showing the potential of the approach to explore the land change dynamics in local to regional level case studies (Le et al. 2012; Robinson et al. 2012; Valbuena et al. 2010b). At larger spatial scales, ranging from the region to the global level the principles of agent-based modeling have not yet been applied in simulation models, leaving most models at that level with highly simplified representations of human-environment interactions (Rounsevell and Arneth 2011). The possibilities for either upscaling or outscaling agent-based models have been described by Rounsevell and colleagues (Rounsevell et al. 2012), but have not yet been applied in operational models.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >

Related topics