Validation of an Agricultural MAS for Southland, New Zealand

Authored by William Kaye-Blake, Chris Schilling, Elizabeth Post

Date Published: 2014-10-31

Sponsors: New Zealand Foundation for Research Science and Technology

Platforms: .NET

Model Documentation: ODD Flow charts

Model Code URLs: Model code not found

Abstract

This paper describes the process and results of validating a simulation model of agriculture for a region in New Zealand. Validation is treated as a process, in which simulation models are made useful for specific purposes by making them conform to observed historical trends and relationships. In this case, the model was calibrated to reproduce the year-by-year conversion to dairying from 1993 to 2012 in Southland, New Zealand. This was achieved by holding constant some elements of the simulation model, based on economic theory or data, and by running simulations on a range of values for two key parameters. The paper describes the model and process, and demonstrates that empirical validation is possible if approached pragmatically with a view to the intended use of the model. Important elements are: using stylised facts to limit the parameter space ex ante, establishing the range of model outcomes and focusing on the most likely parameter space, focusing the search for parameter values where there is the greatest uncertainty, and using historical data to calibrate models.

Tags

Agent-based model Agriculture Agent-based models Interdisciplinary Research Empirical Validation Validation Multi-agent simulation