Traditional empirical land use change models generally assume one average land use decision-maker. Multi-Agent System (MAS) models, on the other hand, acknowledge existence of different types of agents, but their poor empirical embedding remains a serious handicap. This paper demonstrates how agent information can also be incorporated into empirical, biophysical land use models. Agent (farmer) information was captured in four farmer types by means of cluster analysis. The types were distinguished by age, education, property size, distance from residence, and the number of animals owned. This information was made spatially explicit as each field in the study area is related to a farmer, based on cadastral information. Statistical interaction terms between farmer type and landscape factors such as remoteness, soil quality, slope and aspect, were tested for significance in describing the observed occurrence of three land use changes: afforestation of arable land, abandonment of arable land, and restoration of the traditional Montado system. Results showed that each farmer type uses different criteria for selecting land for a certain land use change. For example, absentee farmers abandon the most remote areas while other farmer types do not use remoteness as a criterion for abandonment; active farmers select the most accessible fields for afforestation while other farmer types do not; absentee farmers select their best soils for restoration of the traditional Montado system, while active farmers tend to select poor soils. It is demonstrated that each farmer type shows a different relationship between landscape factors and land use changes. Hence, farmer-specific relationships between landscape and land use contribute significantly to the explanation of land use change.
- multiagent systems