Comparing two sensitivity analysis approaches for two scenarios with a spatially explicit rural agent-based model

• Published in: Environmental modelling & software : with environment data news 2014-04, Vol.54, p.196-210
• Main Authors: Schouten, Marleen, Verwaart, Tim, Heijman, Wim
• Format: Article
• Language: English
• Subjects: Agent-based models; agricultural policies; Animal, plant and microbial ecology; biodiversity; Biological and medical sciences; Complex systems; environmental-models; Fundamental and applied biological sciences. Psychology; General aspects. Techniques; land-use; landscapes; management; Methods and techniques (sampling, tagging, trapping, modelling...); protocol; schemes; Sensitivity analysis; simulation; Social simulation; uncertainty
• ISSN: 1364-8152; 1873-6726
• DOI: 10.1016/j.envsoft.2014.01.003

In this paper two sensitivity analysis approaches are applied for scenario analysis in a spatially explicit rural agent-based simulation. The simulation aims to assess the socioeconomic and ecological impacts of agricultural policy interventions, market dynamics and environmental change on a regional scale. Two different methods of sensitivity analysis are investigated: i) a one-at-a-time approach where each parameter is varied one after the other, while all other parameters are kept at their nominal values; and ii) a procedure based on Monte Carlo sampling where random sets of input parameter values are related to outputs of the simulation. The complementarity of both approaches and their contribution to the overall interpretation of the model is shown in two scenarios simulating alternative European policy instruments for biodiversity conservation. Results show that a mixed approach of sensitivity analysis leads to a better understanding of the model's behaviour, and further enhances the description of the simulation's response to changes in inputs and parameter settings. [Display omitted] •Use of complementary sensitivity analysis methods provides additional insight in sensitivities.•The one-at-a-time approach provides rapid insight into the importance of parameters.•The Monte Carlo analysis reveals parameter interactions and nonlinearities.•Parameter sensitivity of agent-based models can be different across different policy scenarios.•Sensitivity analysis can be used to identify opportunities for policy leverage.