About the Uncertainties in Model Design and Their Effects: An Illustration with a Land-Use Model

Although agent-based modeling is a strong modelling method in many aspects, its high degree of freedom in agent design can also be regarded as weakness. This freedom requires strong validation strategies during model design for empirical models, especially when models aim to be descriptive enough for policy support. Where theory or evidence does not support model design, assumptions are usually made. In these cases, arguments should be given for why the assumptions do not impair the validity of results. However, we believe that such justifications are sometimes weak in such kinds of models. In particular, we believe that the justification arguments are mostly plausible, but often not strong enough to overrule other plausible arguments leading to different designs. We believe that the reasons for this argumentative ambiguity are sometimes rooted in the type of underlying theory, framework, or validation strategy chosen. The point is that we suspect that simulation results can be sensitive to this ambiguity. To test this hypothesis, we selected a well-tried theory/framework/validation design strategy, and built alternative versions of a land-use change model in line with the underlying strategy. Results clearly show that levels and direction of simulated land-use change are significantly different among model versions.