GIS-Based Modeling of Archaeological Dynamics (GMAD): Weaknesses, Strengths, and the Utility of Sensitivity Analysis

Abstract

Simulative models of past behavior generally comprise many parameters or factors affecting model outcome. The interconnectedness and value settings of model parameters and factors form a source of potential uncertainty. The main question that has to be addressed in relation to the problem of uncertainty is: What extent do model connections and value settings have on the model's outcome, and how can outcomes be related to “real-world” observations? Few archaeological examples exist that seriously deal with this question. Additionally, the input data utilized in predictive and simulative models can amplify error as a result of archaeological survey bias and a lack of accuracy or applicability of environmental data, such as digital elevation models, soil maps, and groundwater tables. An “easy” response to the problem of uncertainty is to increase resolution, stemming from the uncritical assumption that more data and model parameters will produce more detailed and hence more reliable outcomes. In this paper, a preliminary methodology for developing sensitivity analyses for an archaeological model of hunter-gatherer behavior in the central Netherlands is presented that attempts to evaluate where, when, and how much error is introduced to the model at different stages of the multi-ordered, cascading framework. Consideration is also given to how this particular model, and archaeological models in general, may be refined or improved after the uncertainty analysis is complete.