Application Of Population Data Spatialization Method Based On Multi-Source Data-Taking Qingdao City As An Example

In the process of geographical statistical research,demographic data have problems such as low spatial and temporal resolution,difficulty in data acquisition,inconsistency between the boundary of the administrative region and the boundary of the study area,which brings great inconvenience to the study of population distribution law,urban geography,urban planning,natural resources development,utilization and protection,disaster risk assessment and other work closely related to population distribution.The spatialization of the statistical population data in the administrative region to a fine-scale grid can better display the spatial distribution characteristics and density differences of the population,and can provide more valuable population data in the relevant research combined with the spatial analysis of population,laying a foundation for carrying out more objective and practical spatial analysis research after multi-source data fusion.This article takes Qingdao City as the research area,collects and processes multi-source remote sensing data,and applies various modeling algorithms to achieve the spatialization of population data at the scale of 1km grid in Qingdao City and 100 m grid in the Shinan District of Qingdao City.The population prediction values at grid scale are obtained and the accuracy indicators are analyzed.The main content of the article is:(1)Based on existing population spatialization modeling algorithms,this thesis elaborates on the acquisition and preprocessing operations of nighttime lighting data,land use data,DEM data,building height and other data that can be used for population spatialization modeling in Qingdao.(2)The population spatial modeling was carried out on the 1km scale grid in the administrative district of Qingdao and the 100 m scale grid in Shinan District of Qingdao,respectively.The population value at the grid scale was obtained and relevant parameters were calculated to evaluate the modeling accuracy,and the advantages of multiple regression,geographically weighted regression,spatial lag model,random forest and other algorithms were analyzed and the optimal modeling algorithm in different population characteristics partition is obtained.