The Simulation Study Of LUCC In Harbin Based On CA And MAS Coupling Model

Along with the accelerating urbanization process,urban spatial pattern changes very fast,and its pattern transformation simulation study become a research hotspot.The distribution and evolution of urban land is a complex system,which is affected by many factors,such as regional economy,society,nature,politics and so on.The traditional urban expansion simulation method is very limited and cannot meet the precision requirements.In recent years,the extensive application of cellular automata(CA)model and multi agent(MAS)model has injected new energy into the research of geological model.Modeling by the idea of combining bottom-up and top-down,comprehensive utilization of CA model and MAS model,make full use of their advantages,build environment restrict guide body spontaneous adjust the urban land use dynamic simulation system of interaction,can play a variety of comprehensive utilization of research due to the evolution of city pattern effect,helps to improve the simulation accuracy.This paper took Harbin city as the research object,by constructing the object-oriented classification method based on multi features,combining the characteristics of resolution to interpret the Landsat remote sensing data of 2000,2010 and 2015,which were divided into residential land,wetland,water and land,farmland and unused five categories land use types individually.When modeling,utilizing multiple factors which were constructed such as slope,exposure,traffic,service facilities,water,population and so on,using Markov model to calculate rotation matrix and area conversion probability matrix,and using ACO(Ant Colony Optimization)combined with the rule of transforming type conversion probability matrix to construct the CA model of land use type conversion rules.Next,using land planning outline,integrated the CA-MAS model to modify CA model by the MATLAB platform.Then getting the simulation results,and the implementation model was coupled.Finally,based on the datum of 2015,simulating and analyzing the 2020 urban spatial pattern.The main conclusions are as follows:(1)Through the construction of NDVI(vegetation index),BI(brightness),GVI(green),WI(humidity),Slope,ICA(independent component analysis component)and MNDWI(Modified Normalized Difference Water Index)and many other factors,using the method of object-oriented classification,the overall interpretation accuracy of remote sensing data was 84.47%,which was increased by 9.69%,10.85%and 10.80%compared with the method of maximum likelihood,artificial neural network and support vector machine.The results showed that the method has better classification effect and meets the precision requirement of follow-up study.(2)The ACO(Ant Colony Optimization)combined with land conversion probability matrix,constructing land conversion rules,effective will DEM(digital elevation),slop(slope,aspect,slope direction),Dis river(distance)river waters,Dis service(facilities),Dis trif(road)and pop(population distribution)to integrate and use.At the same time,using MATLAB platform to write the implementation model,implemented the intelligent of the model operation,and thus improved the operation speed along with the efficiency of the model.(3)The overall accuracy of the coupling model was 90.88%,the kappa coefficient was 0.887,the consistency was effective,and the various land types also achieved a good simulation accuracy.Model overall accuracy,each user category accuracy,consistency and model specific parameters of the model on the simulation were prominent,which showed that the CA-MAS coupled model evolution simulation had good effect in city spatial pattern and a certain potential of application.