One of the most important challenges of this era is the rapid growth of urbanization. According to the United Nations report, around 66% of the world’s population, equivalent to 6.4 billion will live in urban areas by the year 2050, while this number was only about 746 million people, equivalent to 30% of the world’s population in 1950. This increasing trend is followed by the issues such as providing citizens with proper housing, providing energy, health services, education, employment, transportation etc., which should be resolved by appropriate policies and solutions. In this regard, modeling urban growth might play a key role in guiding urban strategies. Planners have traditionally used various models to formulate urban growth. However, these models lack sufficient dynamism and mobility and do not take the social behaviors of individuals and their interactions into account. The shortcomings of the previous methods have led to an increase in the use of modern modeling methods like cellular automata and agent-based model, particularly when presenting a perspective of the future urban growth is expected. Although the use of agents is a common tool in the modeling of the earth systems, few studies have been carried out in this regard in Iran, and various existing foreign studies are not in full agreement with the existing situation yet, due to the complex nature of the land use change problem. Therefore, researchers are still trying to provide new models by focusing on available findings and different aspects of the problem. In this research, a multifunctional system has been developed for the simulation of the urban growth by integrating the irregular cellular automata and the agent.
Materials and Method
The study area in this research is NajiAbad in the city of Kashan that is one of the city’s new districts. The texture of this district has taken shape in a designed and regular manner. The average area of its parcels is 250 square meters whose formation is mainly towards the northwest-southeast direction. Land use map of the year 2006 (1385), slope map, soil type, access map, floodway and river map were used in this research. The data of the year 1385 are used for the simulation of urban growth in 1392 and the results are compared with real data of 1392 and thus the results resulted from the model are evaluated. We have presented forecast of urban growth for the year 1400 subsequently. In the presented model, cadastral polygons act as irregular automata which have their own status and properties. In this model, the changes are updated in each replication, and each time step is considered to be one year. The first step in the evaluation model is the overall proportion of the land parcels which is one of the effective parameters in the decision-making process of the agents.
The calculation of the spatial proportion of the parcels for development is carried out by irregular cellular automata and based on four criteria of neighborhood, physical proportion, accessibility and constraints. Twelve effective factors were classified in proportion with these criteria and were normalized before the combination. The overall proportion of each land parcel has been calculated based on weighted linear combination function. In the next steps, the activity of the agents starts in the model. Many actors play roles in the development of the urban environment.
In this research, the agents are classified into 3 general classes of urban planning agent, developing agents and family agents. The family agents were classified into 3 classes of families with high, medium and low income according to the income level of the families. The urban planning agent estimates land demands and issues the segmentation permits for a number of lands. The developing agent calculates the profitability of the parcels, and segments those having separation permits and high profitability. The family agents search the environment and choose suitable land for habitation based on their preferences. This process is followed up until all family agents are settled and the demands are achieved.
Results and Evaluation
In this research, the output of the model and urban growth map in the study area for the year 2013 (1392) is calculated based on the input data of 2006 (1385). In this research, each time step is considered to be one year. In order to evaluate the model results, real data of 1392 has been used. In this research, the error matrix was used to calculate the accuracy of the results and comparison criterion is Kappa index. The Kappa index is a value between 0 (nonconformance of the calculated and observed maps) and 1 (full match of calculated and observed maps). Although there is no global standard, the Kappa index greater than 0.80 is often considered as a criterion of the proper conformation of calculated and observed maps. The Kappa index in this research was calculated to be 71% based on the error matrix. In this calculation the area of the previous developed regions has been eliminated. Although the elimination of the area related to these regions relatively reduces the overall accuracy of the model, it leads to a more accurate evaluation of the results. The accuracies of the user and producer in the developed lands feature a higher overall accuracy of the model, which can be a reason for desirable design of the model and its adaptation to the reality.
In this research, cellular automata were used to simulate the variations in the status of each land parcel in comparison with different spatial factors. Although the conventional cellular environment in cellular automata methods facilitates the possibility of urban growth modeling, it was attempted in this research to conduct modeling on the scale of irregular polygons of lands and in the form of base parcel. Although the use of cellular automata on this scale makes calculations more complicated and difficult, the results of modeling can be evaluated more realistically. In this research, the agents were classified into 3 general classes of urban planning agent, developing agent and family agents. The family agents were classified into 3 classes of families with high, medium and low income according to the income level of the families. The evaluation of the results with real data showed that the accuracy of the model was 71%. This study has been conducted in a vector structure and local scale that can be extended to other areas. This modeling can be done on a regional scale in future works.