Modeling Urban Growth using Hybrid Cellular Automata and Particle Swarm Optimization

Land and its cover has always been changing and evolving due to various human activities over time. Detecting, Identifying and predicting these changes could be very beneficial in sustainable monitoring and management of land use and cover both in urban and rural areas. Inthis field, to better understand the process of these spatio-temporal changes, cellular automata (CA) models are very common due to their dynamic simple structure and powerful spatial features. It is noteworthy, however, that the conventional cellular models confront many difficulties because of cell size sensitivity and also the inability to consider real spatial objects.On the other hand, novel object-based CA models which have managed to mitigate these problems are very complex and difficult to implement, and therefore, not very practical. This paper presents a hybrid cellular automata (HCA) model as a combination of cellular structure and vector concept. The space is still defined by an organized set of cells, but rasterized spatial objects are also utilized in the structure of transition rules. Particle swarm optimization (PSO) is also used to calculate the urbanization probability of cells based on their distance from development parameters such as road network or major city centers. The proposed model is applied separately to Landsat satellite imagery of the city of Tehran, Iran with 28.5, 57 and 114m spatial resolution, to simulate the urban growth from 1988 to 2010.Statistical comparison of the ground truth and the simulated image derived by the HCA model (for images with 28.5m resolution) shows an accuracy of 83/42 percent (Kappa coefficient),which is greater in comparison to the applied conventional CA model’s result with an 81/13percent accuracy. Moreover, decreasing the spatial resolution by a factor of one-fourth, has reduced the simulation accuracy of the HCA model by 1/19 percent which in comparison to conventional CA model’s 3/04 percent decrease, shows the lower scale sensitivity of the proposed model.