Exploring the Choice of Decision Making Method in an Agent Based Model of Land Use Change

In agent-based models of land use/cover change (ABM/LUCC), small changes in micro-level decision-making methods used by agents may significantly affect macro-level outcomes. Yet, the implications of choosing a specific decision making model are seldom explored in ABM/LUCC studies. This paper discusses an ABM/LUCC modelling study of smallholder farming households in the Amazonian varzea in Marajó Island, Brazil. These agents represent the 21 households within the community of Paricatuba. Farmers in this community cultivate acai as a primary source of income, in addition to other farming and economic activities such as offsite employment. In the model, agents make annual decisions to allocate scarce land, capital and labour resources to best provide revenue for the household. Household agents have the same overall goals, resources, information, and feasible actions available to them within the simulation environment. Alternative simulations are developed in which the household agents employ one of two primary decision-making methods, either based on linear programming or decision trees. A comparison of these methods in a Monte Carlo simulation indicates that in certain scenarios, alternative decision-making methods with otherwise common objectives and environments may lead to widely divergent outcomes. The evaluation of multiple decision making methods within a common model can be used to highlight the advantages and limitations of these methods and challenge assumptions.