关键词: NetLogo, 城市扩张, 多智能体模拟, 广州市番禺区, 矢量数据

Abstract: Urbanization has great potential to stimulate China's domestic demand and is an important driving force of building a well-off society in an all-round way. However the limited land resource in the process of urbanization has caught great attention of the government. So, research on the mechanisms of urban land expansion can offer effective guidance for urbanization. Multi-agent modeling provides a new method to study urban land expansion. And it is of theoretical and practical significance to adopt multi-agent modeling, a bottom-up approach, to explore the characteristics and mechanisms of urban land expansion. However, the existing multi-agent models are mainly based on raster, where grid size has a major impact on the simulation results, and grid size, neighborhood shape and neighborhood size also lead to different simulation results. As a result, it is a big challenge to choose an appropriate size for better simulation results. In order to overcome the limitation of the traditional raster-based multi-agent simulation, a vector-based multi-agent system of urban land expansion dynamic simulation was developed, in which household agents selected their own optimal places of living based on evaluation indices of urban residential suitability in the human-environment science. The evaluation indices mainly include the following four aspects: traffic accessibility, living convenience, land price and landscaping. The indices of traffic accessibility are obtained by calculating the distance from road and subway station based on exponential distance decay function. The indices of living convenience are obtained by calculating the distance from the public facilities such as school, hospital and so on, based on the same function. The indices of land price depend on the average price of the land in the neighborhood. The indices of landscaping are obtained based on the distance to rivers or the green in the same way. Finally, the four indices are summed up, by a weighted factor, to indicate the living suitability. The model runs as follows: household agents select their ideal residential package to live based on their preferences of living environment, the four evaluation indices as mentioned above. The land type of this package will be converted to urban land when the total number of household agents that select this package exceeds a certain threshold. At this point household agents can choose to stop selecting and settle down, or continue to select until they find their ideal residential package. This vector-based model was used to simulate urban land expansion process from 2003 to 2008, taking the district of Panyu of Guangzhou City as a study area and using the GIS-Extension module of NetLogo platform. Eventually, the simulation results were compared to the actual land use situations, showing that the conversion accuracy is 63.09%, the non-conversion accuracy is 90.74%, and the total accuracy is 86.04%. The simulation shows that vector-based multi-agent simulation of urban land expansion not only can eliminate the impact of the scale on the simulation results, but also has high simulation accuracy. At the same time, to some extent the vector-based multi-agent simulation is useful for understanding and discussing the mechanisms of urban expansion, and offers effective support for making the decisions on the construction of new urbanization.

Key words: multi-agent simulation, NetLogo, Panyu District, Guangzhou City, urban expansion, vector data