LUCC Simulation And Prediction Of Chengdu Plain Based On RS And GIS

Land cover/change simulation and prediction is important to better understand the land transfer mechanism and rational planning for future land use.The paper is completed based on the scientific research platform of "Key Laboratory of Ministry of Education" and "Provincial Land and Resources Collaborative Innovation Center" and the National Natural Science Foundation.The land use data of Chengdu plain were extracted from the 2005 a and 2015 a images using the decision tree and the object-oriented classification method respectively,based on the RS and GIS techniques.Transfer process between the land use types in the study area were analyzed.The correlation coefficient was used to select the best variable to quantify the degree of LUCC expression.The correlation between spatial influence factors(Land use area,spatial auto correlation,dispersion degree)and LUCC under the factorize township boundary and grade 13 fishing net scale was analyzed and the optimal mesh scale is calculated by using the principal component comprehensive score.The regression equation between LUCC and spatial influence factor,topographic factor(elevation,slope,slope direction),distance factor(distance from the road,distance from the water,distance from the woodland,distance from the arable land,distance from the construction site),kernel density factor(road core density,water body nuclear density,forest core density,farmland nuclear density,construction land nuclear density)were used to construct using the principal component and linear regression.The probability diagram of the influence of each factor on LUCC is produced by using the regression equation.The LUCC during the 2015 and 2025 a were to predict using the Markov model,and the spatial position of LUCC in 2015 a was predicted combing with the probability map.The predicted accuracy of the results meets the requirements.The patterns of LUCC in 2025 a was finally predicted.The result showed that:(1)The residual space rate of change was most likely to express LUCC in all variables expressing the extent of LUCC.The score of the grid of 270 m x 270 m between the township bolder and grid with13 levels was the higher.So the optimal gridscale of the relationship between the spatial expression factor and LUCC was 270 m x270m.The regression equation of water and cultivated land was higher,respectively0.462 and 0.344,in all expression of principal component and linear regression equation.The linear regression of forest land,grassland and cultivated land was high,with R2was0.418,0.615,and 0.591.Therefore,the probability of quantitative expression of LUCC is better.(2)The fitting degree of the 6 order equation in the regression equation between the LUCC and elevation or slope was the highest.The fitting degree of altitude and water regression equation was 0.847,and forest land was 0.703,and cultivated land was0.448,and construction land is 0.574,respectively.The fitting degree of slope and water regression equation was 0.947,and forest land was 0.996,and cultivated land was 0.871,and cultivated land was 0.866,respectively.So the altitude and slope can be a good quantitative expression of LUCC probability.The fitting degree of aspect was low.The regression equation of the power function is the highest from the distance of the road,with R2=0.7.The regression equation is not the same as that of the land use type,which is included mainly six order and power function.For the whole study area,the regression equation of building kernel density and LUCC was low,but the degree of fitting of the individual building equation was very high,so the land use type was different in space,and the degree of LUCC was also different.(3)The predicted area of the construction land,water,unused land will continue to increase,while the area of forest land,arable land and grassland will decrease continuously in 2025 a.The predicted accuracy on construction land was higher with75.11%,followed by 71.23% of cultivated land,69.97% of forest land and 43.5% of water body based on RS and GIS.Therefore,the land use model was suitable for land with high degree of spatial clustering,while the discrete prediction accuracy was low.