Empirically derived method and software for semi-automatic calibration of Cellular Automata land-use models

Land-use change models generally include neighbourhood rules to capture the spatial dynamics between different land-uses that drive land-use changes, introducing many parameters that require calibration. We present a process-specific semi-automatic method for calibrating neighbourhood rules that utilises discursive knowledge and empirical analysis to reduce the complexity of the calibration problem, and efficiently calibrates the remaining interactions with consideration of locational agreement and landscape pattern structure objectives. The approach and software for implementing it are tested on four case studies of major European cities with different physical characteristics and rates of urban growth, exploring preferences for different objectives. The approach outperformed benchmark models for both calibration and validation when a balanced objective preference was used. This research demonstrates the utility of process-specific calibration methods, and highlights how process knowledge can be integrated with automatic calibration to make it more efficient.