Practical Points for the Software Development of an Agent-Based Model of a Coupled Human-Natural System

Modeling complex natural and human systems to support policy or management decision making is becoming increasingly common. The resulting models are often designed and implemented by researchers or domain experts with limited software engineering expertise. To help this important audience, we present our experience and share lessons learned from the design and implementation of an agent-based model of agricultural production systems in the Argentine Pampas, emphasizing the software engineering perspective. We discuss the model's design including the model classes; the activity diagram, and data flow; the package and folder layout; the use of design patterns; performance optimization; initialization approaches; the analysis of results; and model measurement, validation, and verification.