Study On The Agent-Based Urban Land Use Planning Decision Support Model

China is at a stage of rapid urbanization.It is crucial to choose a scientific method for urbanization.Under the current background of New Urbanization strategy and Smart City construction,in order to advance new urbanization strategy,it is necessary to improve the scientificness of urban planning and improve the rationality and scientificness of decision-making for urban planners.In order to promote the development of smart cities,it is necessary to grasp the underlying laws of urban growth.Integrating various emerging technologies,and establishing a planning decision support system which includes high-quality geographic information extraction,spatio-temporal process mechanism analysis,and urban expansion simulation,could help to understand the historical process and support future planning.Therefore,building an urban land use planning decision support model that integrates "top down" and "bottom up" will help sort out the situation of construction land,understand the spatio-temperal evolution process of urban expansion,and predict the direction of urban expansion,so as to improve the rationality,scientificity of spatial decision-making in urban land use planning.,and to provide useful decision support for sustainable development of the city.This study takes the former municipal district of Changzhou City as an example.It combines the geographic object-oriented image classification,random forest algorithm,expansion indexes,and agent-based model to extract construction land information from 1991 to 2015,analyzes the spatio-temporal evolution process of urban space expansion,and simulates the scale and layout of construction land under different scenarios,in order to provide decision support for urban planning and land use planning.The main contents and conclusions are as follows:(1)Intergating random forest algorithm with the geographic object-oriented image classification,the classification rules can be explored based on the spectrum,geometry,texture and other features of the objects sample automatically,and the rule sets can be built automatically.And the overall accuracy and Kappa coefficient of 6 classification results of Landsat images in the study area from 1991 to 2015 exceeded 94.04%and 0.90,respectively,which proved that the classification quality was good,validating the effectiveness of the random forest object-based image classification method.(2)By applying the expansive intensity index and the landscape expansion index to the random forest object-oriented image classification results from 1991 to 2015,it was found that there was a spatio-temporal difference in the intensity and pattern of construction land expansion in the study area.During the study period,the scale of construction land increased significantly and grew rapidly in the study area.The construction land of the study area experienced high speed expansion(1991-1995),medium speed expansion(1995-2000),second high speed expansion(2000-2005)and low speed expansion(after 2005).Its main expansion pattern has been evolved from edge-expation and outlying from 1991 to 2000,to edge-expation from 2000 to 2005,edge-expation and infilling after 2005.The continuous increase of mean LEI shows that urban expansion of the study area evolved from diffusion to coalescence,during the study period.(3)The "combining up and down" agent-based simulation model constructed in this study can be used for urban spatial expansion simulation.This study designed "top-down" three-level agents designed,according to the actual process of land use planning.The first-level Agentt control the total amount of construction land and restrictions.The second-level Agenti determines the allowable construction area and the conditional construction area through the interaction with the third-level Agentu.The third-level three sub-classes of Agentu added new construction sites in the designated allowable and conditional construction areas.For the bounded rationality assumption of second-level agent,this study starts from the third-level agent selection preferences,builds different construction suitability evaluation systems,uses the AHP to determine the weights to calculate the objective suitability evaluation results,uses the learning function to complete the annual comprehensive suitability iterations,and uses the Monte Carlo method to select new construction land.It has been verified that both the Lee-Salle index and the Kappa coefficient are greater than 0.8,which validates the high accuracy of simulation results and the effectiveness of the model.(4)This study designed three types of different scenarios,namely trend continuation,increment reduction,and protection development,to realize simulations under different total area control of construction land and different constrained conditions,and effectively arrange the layout of construction land.Combining with the actual situation and growth trend of the population and construction land scale in the study area,it can be concluded that the "reduction" of construction land increment is the current trend.In this context,to guide the construction land layout more aggregated and orderly,delineation allowable and conditional construction areas can be helpful,taking the results of the protection development scenario in to consideration.