Calibration of cellular automata based land use models: lessons learnt from practical experience

With an increasing interest and use of CA-based land use models in the planning and policy practice, calibration and validation gain in importance. Calibrating CA models, however, remains an ongoing challenge. Despite several attempts to develop (semi-)automatic approaches [1, 2], manual calibration, especially for calibrating land use change models with multiple dynamic land uses, remains the common practise [3]. Although there are numerous papers that discuss CA models, only few cover their (manual) calibration process, and a common knowledge base seems to be lacking. This presentation aims to start such a knowledge base by presenting a method for calibrating CA-based land use models, which has been developed over the past 25 years and builds on work by a number of people who have calibrated the Metronamica land use model [4, 5, 6] in its current or preliminary forms [7, 8, 9, 10]. In this methodology the calibration process is approached broader than the setting and fine-tuning of parameters, and includes all steps related to finding an appropriate parameter set and assessing its quality.