An Innovative Proposal for Developing a Dynamic Urban Growth Model Through Adaptive Vector Cellular Automata

Monitoring urban growth through simulation models is becoming increasingly vital for the sustainable management of cities. Although various raster-based models have been developed over the past three decades, the irregular, fragmented, and heterogeneous geometric structure of urban areas poses significant challenges to effectively modeling complex land use and land cover (LULC) transitions. To address these limitations, this study proposes a novel urban growth simulation model based on vector cellular automata (VCA). In this model, dynamic neighborhood relationships are flexibly established using an algorithm called growth vectors (GVs). Open-access data from four time periods between 1990 and 2018 were utilized for three major European metropolitan areas: Istanbul, Berlin, and Madrid. During the calibration phase, the model was trained using three machine learning algorithms: Random forest, support vector machine, and multi-layer perceptron. For the simulation phase, an adaptive VCA-based urban growth model was developed to predict LULC changes through to 2040. The results demonstrate that the proposed algorithm can achieve a satisfactory level of accuracy in modeling urban growth.