Research On Population Spatialization Method Based On GTWR Model

Population distribution refers to the distribution of population in different regions within a certain period of time.Population distribution information is of great significance in many major issues such as studying regional planning,formulating social policies,and analyzing social and economic development.The traditional census data has some shortcomings such as low spatial resolution,long update period,unstable survey unit,and lack of fine data in the space unit.In response to these problems and practical application needs,researchers based on advanced spatial information technology,using real-time and accurate Earth observation information big data,combining social unit-based statistical data with ecological data based on natural environment units,and jointly converted to A unified basic geographic unit with high resolution,through demographic data and specific models to reverse the specific distribution of population within the study area,is called "spatialization." The main methods and models of the existing population spatialization are to use simple linear regression to estimate the relationship between population density and various indicators,and then invert the impact of space-time heterogeneity.Therefore,this paper uses spatial-temporal weighted regression method to spatialize the population.It is of great significance to improve the spatialization accuracy of population data.Taking Chengdu downtown area as an example,based on the data of land use type,elevation data,transportation,medical care,education and housing,based on the administrative division map of downtown Chengdu,the use of arcgis software for land use types and The elevation raster data is masked,the nuclear density analysis is performed on the vector data of each interest point,and the Geographically and Temporally Weighted Regression model is used to spatially model the population data,and the population distribution data of the 1KM*1KM in the city center of Chengdu is established.The inversion results of linear regression model,time-weighted regression model and geographically weighted regression model were compared and analyzed.The study found that:(1)Calculating the accuracy of the four population spatialization models,the accuracy of the Geographically and Temporally Weighted Regression model is basically less than 6%;the accuracy of the geographically weighted regression model is between 4%and 8%;and the accuracy of the time-weighted regression model is 7%.-14%between;while the linear regression model is more than 20%accurate.(2)In the four population spatialization models tested,the space-time geo-weighted regression model considering the spatial-temporal heterogeneity has the best fitting effect,and the spatial heterogeneity in the model is found to be greater than the temporal heterogeneity.(3)In addition to the error of the spatial population model prediction errors,but also from each element data extraction grid,the extraction grid size selection,the original resolution of the raster data and other factors,and therefore high-precision data and a reasonable lattice Net size is an effective way to increase the spatialization of population data.