Spatial Heterogeneity Oriented Neighborhood Of Vector CA For Urban Growth Simulation

Land use change is a spatial change process dominated by human beings.While bringing about urban growth,it has a profound impact on the natural ecosystem.Cellular Automata(CA)is a common land use change model,which can simulate macro complex dynamic pattern based on micro simple cell interaction.Compared with the traditional raster CA based on homogeneous rules,vector CA can more truly express the real world.However,it does not eliminate the variability of the model results caused by different neighborhood configurations.Neighborhood computation is a difficult problem in the research of vector CA neighborhood.Geographical space is heterogeneous,where the distribution of land use is complex,different land use types interact with each other,and the interactive modes described by distance are spatially different.The existing CA model has some limitations,using consistent rules to define the cellular proximity and neighborhood areas,and defining neighborhood rules based on spatial autocorrelation of the same type of land use.Therefore,this paper explored the neighborhood definition and rules of CA from the perspective of spatial heterogeneity,considering the complexity of land-use spatial distribution and the difference of spatial interaction patterns.On this basis,combined with zoning simulation and automatic screening mechanism of land use driving factors,a vector CA model considering spatial heterogeneity was designed to make CA more "geographic" in the field of urban research.The main research contents and achievements are as follows:(1)Definition of vector CA neighborhood considering space complexityReferring to the spatial distribution and morphological differences such as uneven density and different shapes of land units,the land use units were abstractly expressed based on graph theory,and the cellular proximity was obtained by Delaunay triangulation spatial clustering algorithm under edge length constraint.Referring to the difference of land use interaction space,the neighborhood area was defined by Delaunay triangulation clustering algorithm with multiple constraints of space and attribute.Then,the neighborhood definition of vector CA considering the space complexity was realized by combining the two above.(2)Expression of neighborhood effect of vector CA considering spatial differenceReferring to complexity and spatial difference of land use interaction,we firstly combined spatial metrics and distance decay function to construct sequential data.Then,the clustering algorithm based on DTW(Dynamic Time Warping)similarity was used to obtain the law of land use from historical data.Finally,the transformation probability under the influence of neighborhood space scenario was expressed in the form of neighborhood effect pattern library.(3)Construction of vector CA model considering spatial heterogeneityLocal constraints were defined based on heterogeneous constrained multi-order neighborhood and differentiated neighborhood effect pattern,which combined the region constraints and global constraints to determine the comprehensive transition probability of cells.Logistic stepwise regression method was used to automatically screen the driving factors of land use,and Markov chain was used to predict the number of construction land.(4)Verification of urban growth simulation in Jiangyin CityThe urban growth simulation of Jiangyin in 2017 showed that the neighborhood effect model library based on heterogeneous neighborhood could reveal diversified landuse interaction patterns.In general,the Kappa coefficient and Fo M accuracy of this model were 84.10% and 22.23% respectively,which were 1.90% ? 2.30% and 1.72% ? 5.09%higher than the traditional vector CA model without considering spatial heterogeneity.In addition,compared with scheme(A)based on all non-construction land,scheme(B)based on non-construction land changed in model calibration stage had higher accuracy.The above work enriches the theory and method research of Geographical Cellular Automata,and to a certain extent,promotes the transition of CA research focus to heterogeneous neighborhood and differentiated land use interaction pattern.At the same time,it has positive analysis and application value for urban growth and sustainable development.