Measurement On Parameter Sensitivity And Uncertainty Of Cellular Automata Model Of Urban Growth

In recent years,the urbanization level of China has been continuously improved,while the urban scale is rapidly expanding,the land use type has also undergone earth-shaking changes,resulting in many environmental problems.Therefore,how to guide city construction scientifically and rationally and use land resources efficiently is very important.Cellular automata,with its powerful spatiotemporal computing power,can simulate the complex process of geographical change by relying on local interactions,and has been widely used in the simulation of land use change.When the cellular automata model is used for the simulation of urban growth,the results are restricted by the transform rules,cellular size,neighborhood shape,neighborhood window and time step,different parameters contribute different degrees to the simulation results.It is very important to measure the parameter sensitivity of cellular automata accurately to improve the accuracy of the model.At the same time,as the basic data source of the model,the interpretation results of remote sensing image will inevitably produce errors in the process of collection and processing,and these errors will be transmitted through the simulation of cellular automata,causing a certain degree of uncertainty to the simulation results.However,the existing research mainly focuses on the influence of a single parameter on the simulation results.The selection of parameter scale level is not representative,the sensitivity of multiple factors interaction is not considered,and it is difficult to accurately measure the uncertainty of the simulation results caused by error transmission.In view of the above problems,the following work is carried out:(1)The constitution and mechanism of cellular automata model and the dynamic mechanism of the evolution of urban land use types are systematically analyzed.On this basis,two CA transformation rules are obtained by using the multi-criteria evaluation method and Markov model method,and the UG-CA model is finally constructed.(2)The transformation rules,neighborhood shapes,cell sizes,neighborhood windows and their interaction terms are selected as the sensitivity parameters.Considering the problem of multi-parameter and multi-level comprehensiveness,the uniform design idea is introduced,and the uniform test scheme of mixing level is designed to analyze the influence of sensitivity parameters on the simulation accuracy.Finally,the quadratic polynomial stepwise regression equation is constructed.The order of sensitivity of these parameters and interaction terms is as follows: transformation rule * neighborhood shape > transformation rule > neighborhood shape * neighborhood window > transformation rule * neighborhood window > neighborhood shape * cellular size > transformation rule * cellular size.The optimal scale combination of parameters of UG-CA model is determined as follows: the transformation rule is Markov-CA,the neighborhood shape is Von Neumann,the cell size is 15m*15m,and the neighborhood window is 13*13.(3)Monte Carlo method is used to add 30% random error interference in the original data source,and the simulation is carried out on the basis of the optimal scale combination of UGCA model,and the results are compared with the simulation results without error interference,so as to analyze the error transmission characteristics.It is found that the clustering degree of ground objects and the neighborhood space of central cell will reduce the uncertainty caused by errors.In addition,the paper also sets the time step in annual,quarterly and monthly units to analyze the relationship between error transmission and model iteration time.The final results show that the dense time step can effectively reduce the uncertainty of data source error in the simulation transmission process.In summary,this paper makes an in-depth analysis of the parameter sensitivity and uncertainty of error transfer of the cellular automata model of urban growth,explored a sensitivity measurement method suitable for multi-factors and multi-levels and studied the transmission rules of data source errors,improved the simulation accuracy of the cellular automaton model for urban growth which is of great significance to accurately simulate urban development trend and rationally allocate land resources.