Land-use Change Modeling With Cellular Automata Using Land Natural Evolution Unit

Cellular Automata(CA)is an important model for simulating land use change,and the size and shape of the cellular units can affect the simulation results.Therefore,it is crucial to determine reasonable cellular units when using CA models for land use change modeling.However,the existing methods for defining cellular units are inadequate in describing the natural evolution characteristics of land use,leading to mismatches between simulated units and actual land use change units.To address this issue,this study proposes a method for determining cellular units based on Land Natural Evolution Units(LNU).This method is based on the principle of geographic similarity,using geographic environmental data based on existing natural meteorological data and social environment data as a foundation.The geographic similarity is calculated by comparing the geographic environmental similarities of different pixels,aggregating into the same LNU if the similarity exceeds the set threshold,and using it as a cellular unit for simulating land use change in cellular automata.Based on this,this study designs a LNU-CA model based on the coupling of geographic detectors for weight allocation and natural LNU for simulating land use change.Finally,a case study is conducted in the Chenggong district of Kunming City,Yunnan Province,from 2009 to2016 to evaluate the effectiveness of using LNU-defined cellular units in simulating land use change.During the experiment,a control group comprising 76 combinations of cellular automata for LNU defined by different weight allocation methods(average weight method,coefficient of variation method,entropy weight method)and different geographic similarity thresholds was set up for accuracy comparison.This was used to evaluate the sensitivity of cellular automata for natural LNU sensitivities to weight allocation and threshold settings.The experimental results and main conclusions of the case study are as follows:(1)The LNU-CA model,which uses the average weight method for weight allocation,showed the highest improvement compared to the regular grid-based cellular automata in terms of overall accuracy,Kappa,and FoM,with a maximum increase of3.22%,4.23%,and 118.88%,respectively,at different similarity thresholds.The LNUCA model,which uses the coefficient of variation method for weight allocation,showed the highest improvement compared to the regular grid-based cellular automata in terms of overall accuracy,Kappa,and FoM,with a maximum increase of 2.63%,3.37%,and116.48%,respectively,at different similarity thresholds.The LNU-CA model,which uses the entropy weight method for weight allocation,showed the highest improvement compared to the regular grid-based cellular automata in terms of overall accuracy,Kappa,and FoM,with a maximum increase of 3.19%,4.18%,and 122.77%,respectively,at different similarity thresholds.The LNU-CA model,which uses the Geodetector method for weight allocation,showed the highest improvement compared to the regular grid-based cellular automata in terms of overall accuracy,Kappa,and FoM,with a maximum increase of 3.24%,4.25%,and 125.47%,respectively,at different similarity thresholds.(2)Through the comparative analysis of simulation accuracy,it was found that the principle of Geodetector can accurately capture the key geographic environmental driving factors of land natural evolution processes from a geographic relationship perspective,determine their driving forces quantitatively,and use this quantitative method to assign reasonable weights to different geographic environmental driving factors when identifying natural landform units and constructing LNU-CA models.This approach can more accurately approach the natural evolution process of land and simulate land use change.Compared with traditional weight allocation methods that only calculate driving forces from a mathematical relationship perspective,geographic detectors take into account the spatial heterogeneity present in the environmental evolution process,and the scientificness of weight allocation was validated through the accuracy verification of the LNU-CA model for simulating land use change based on natural landform units.