Agent-based modeling techniques simulate the microlevel behavior of individual actors within a larger system. These actor elements interact with one another as well as the economic and ecological systems that contain them. Unlike CA models, whose cells remain stationary in space and whose states evolve over time, agent-based models reflect the reality of actors moving across a landscape, making decisions, and interacting with other agents. These interactions result in phenomena such as migration, succession (land-use conversion), the development of infrastructure, and heterogeneous distribution of preferences. From microeconomic theory, it is assumed that the sum total of individual decisions reflects the market-level demand for such goods (e.g., single-family household).
The model agents can represent actors at one or more scales ranging from the individual or household to an agency or institution. In the case of the individual, social and economic characteristics can be allocated using an iterative proportional fit algorithm to yield a set ofsynthetic agents that represent the aggregate demographic characteristics of the population. Microsimulation approaches have gained in popularity as computer processing power and statistical methods for modeling choice behavior have improved. Agent interactions are modeled using discrete choice theory, random utility theory, or other probabilistic mechanisms and typically assume that the decision maker is seeking to maximize personal utility from the set of available alternatives. Individual decisions can be constrained by exogenous factors (e.g., population or employment growth) or allowed to freely operate within the model specification.
One example of an agent-based model that combines the framework of the gravity model with theories of individual choice is the Huff model. Originally conceived as a means to quantify consumer choice behavior, it has evolved to address the spatial implications of individual choices on land-use patterns. The model assumes the probability that an individual will select an alternative is proportional to the sum of the utilities of all of the considered choices. Adding a spatial component, such as the distance between the agent and its alternatives, extends the applicability of the model to a wide range of land-use management situations. An example of this is presented as follows:
where P¡j = the probability that consumer i selects option j Sj = the attraction factor of alternative j Dj = the distance between i and j and a and @ = empirically estimated parameters based on survey data.
In this case, the probability that consumer i will select alternative j depends on the attraction weight of the alternatives as well as the distance (or travel cost) between the consumer and the alternative.
Was this article helpful?