Multi Objective Optimization of Urban Land Use Allocation Using Meta-heuristic Algorithms and Spatial Metrics

Today, urban land use planning and management is an essential need for many developing countries. So far, lots of multi objective optimization models for land use allocation have been developed in the world. These models will provide set of non-dominated solutions, all of which are simultaneously optimizing conflicting social, economic and ecological objective functions, making it more difficult for urban planners to choose the best solution. An issue that is often left unnoticed is the application of spatial pattern and structures of urban growth on models. Clearly solutions that correspond with urban spatial patterns are of higher priority for planners. Quantifying spatial patterns and structures of the city requires the use of spatial metrics. Thus, the main objective of this study is to support decision-making using multi objective Meta-heuristic algorithms for land use optimization and sorting the solutions with respect to the spatial pattern of urban growth. In the first step in this study, we applied the non-dominated sorting genetic algorithm ΙΙ (NSGA_II) and multi objective particle swarm optimization (MOPSO) to optimize land use allocation in the case study. The four objective functions of the proposed model were maximizing compatibility of adjacent land uses, maximizing physical land suitability, maximizing accessibility of each land use to main roads, and minimizing the cost of land use change. In the next step, the two mentioned optimization models were compared and solutions were sorted with respect to the spatial patterns of the city acquired through the use of spatial metrics. A case study of Tehran, the largest city in Iran, was conducted. The six land use classes of industrial, residential, green areas, wetlands, Barren, and other uses were acquired through satellite imagery during the period of 2000 and 2012. Three scenarios were predicted for urban growth spatial structure in 2018; the continuation of the existing trend from 2000 to 2018, fragmented growth, and aggregated growth of the patches. Finally, the convergence and repeatability of the two algorithms were in acceptable levels and the results clearly show the ability of the selected set of spatial metrics in quantifying and forecasting the structure of urban growth in the case study. In the resulted arrangements of land uses, the value of the objective functions were improved in comparison with the present arrangement. In conclusion planners will be able to better sort outputs of the proposed algorithms using spatial metrics, allowing for more reliable decisions regarding the spatial structure of the city. This achievement also indicates the ability of the proposed model in simulation of different scenarios in urban land use planning.