Comparing Agent-Based Learning Models of Land-Use Decision Making

An agent-based model, incorporating a small set of primarily agent-based variables, was designed to explain private land-use decision making. Agents are landowners, who allocate their labor and land for different uses in regular time intervals. The goal is to understand what kind of spatial patterns emerge from different agent characteristics, and decision and learning mechanisms. Landscapes produced by two different learning models are compared to actual land-cover data. By calculating a set of spatial metrics from the simulated land-cover and comparing them to the metrics calculated from the actual land-cover data, the role of agent preferences for different land-uses is explored. The preliminary results suggest that the models capture relatively well the quantitative patterns of land-cover changes but they are poor in predicting the location of changes.