This thesis discusses the use of agent-based models for capturing land change in large frontier areas. Applying agent models in such areas is not straightforward, given the lack of data. To date, most agent based models of land frontiers study local areas using in-situ information. At regional scales, agent-based modellers need additional ways to describe collective decision-making. The work presents ideas to deal with the complexities of agent-based models at such scales: institutional arrangements and states. Institutional arrangements help to model multi-agent interaction by explaining why, although there are rules and norms for land use, these rules are not always followed. This formalism captures states and transitions of agents in a simulation and helps to build expressive models, where the agent strategies evolve depending of local and external factors. We validate our ideas by building a deforestation model in an area of 60,000 km2 in Amazonia. Results show that we need to set different arrangements to capture changes in agents’ behaviour, as they react to external conditions. Thus, combining the ideas of institutional arrangements and states improves the explanatory power of agent models for regional scales.
20 de julho de 2017