Multi-objective optimisation framework for calibration of Cellular Automata land-use models

Modelling of land-use change plays an important role in many areas of environmental planning. However, land-use change models remain challenging to calibrate, as they contain many sensitive parameters, making the calibration process time-consuming. We present a multi-objective optimisation framework for automatic calibration of Cellular Automata land-use models with multiple dynamic land-use classes. The framework considers objectives related to locational agreement and landscape pattern structure, as well as the inherent stochasticity of land-use models. The framework was tested on the Randstad region in the Netherlands, identifying 77 model parameter sets that generated a Pareto front of optimal trade-off solutions between the objectives. A selection of these parameter sets was assessed further based on heuristic knowledge, evaluating the simulated output maps and parameter values to determine a final calibrated model. This research demonstrates that heuristic knowledge complements the evaluation of land-use models calibrated using formal optimisation methods. •Generic framework for calibrating Cellular Automata land-use models.•Multi-objective optimisation used to trade-off predictive and process accuracy.•Rigorous consideration of model uncertainty during optimisation.•Final evaluation of multiple model parameter sets integrates heuristic knowledge.•Case-study application demonstrates framework utility.