High-resolution Population Spatialization Based On Multi-source Data And Random Forests Model

The spatial distribution of population refers to the distribution of population in the region,which is the spatial manifestation of population,the result of the comprehensive action of various natural,social and economic factors,and the objective reflection of different regions of human activities.The study of population spatial distribution is of great significance for urban planning and sustainable development.Census data is an important indicator that reflects the spatial distribution of population,and is of great significance to the actual distribution of population in the study area.To study the spatial distribution of population,geospatial information contained in the population census data needs to be deeply mined,China's census data is generally published in administrative units,such as districts,counties,and towns.However,this form of census data is difficult to reflect the geographical and spatial heterogeneity of the population distribution within the region,and it is difficult to accurately grasp the population distribution pattern within the region.Therefore,it is necessary to estimate the actual distribution of population in the geographical world by some technical methods on the basis of the national public census data and with the support of massive and diverse geographical data that have important influences on population distribution,so as to depict the distribution of population in the real world.This process of re-expressing census data on a more accurate spatial scale is called population data spatialization.The geographic data used in the process of population spatialization that have important effects on population distribution is called population spatialization Supporting data.With the advent of geospatial big data era and the development of computer technology,on the one hand,population spatialization technologies and methods emerge endlessly,on the other hand,there is an explosive growth of population spatialization supporting data.The method of population spatialization is constantly enriched,especially machine learning has made significant progress in the method of population spatialization,and the nonlinear relationship between various supporting data in the process of population spatialization has been discussed,and the supporting data of population spatialization has exploded The rapid growth has made the data used for population spatialization increasingly massive and diverse.This paper takes Changsha as the research area,combines POI(Point of Interest),DEM(Digital Elevation Model),night light and land use data to build a random forest model to spatialize the population in Changsha,simulate the regional population distribution in the real world,and the Spatial autocorrelation analysis was used to analyze the population distribution pattern and regularity of Changsha City,which provided basic data for the study of population visualization in Hunan Province.The main research results are divided into the following three aspects:(1)Population distribution characteristics of Changsha: The spatial distribution pattern of population in Changsha City is in the form of "one middle-two pairs-multidirectional extension".The population is mainly concentrated in the central region,and the population density in the surrounding areas is relatively low.The biggest factor affecting the population of Changsha City is economic activity,which is represented by POI data and night lighting data.(2)Massive geospatial data in the application of population spatialization: With the continuous development and improvement of geographic information technology,at this stage,fully open source data is very rich,with open source data enough to support all the work of population spatialization.In addition,the combination of traditional geospatial data and massive,multi-source new geospatial data has made the process of population spatialization a new era of geospatial information research,making geospatial information research more abundant and scientific.Among the errors caused by different reasons and data in the process of population spatialization,the errors caused by attribute data have the greatest impact on the population estimation model and need to be paid attention to and controlled most.(3)The combination of geographic grids and random forests has a unique advantage in the process of population spatialization: using geographic information technology to establish a grid for the entire study area is convenient,efficient,and fast.All data are linked by unique coordinates through geographic coordinates.It is convenient to construct a population estimation model.The random forest model is used for population spatialization with good effect and fast calculation speed,which can achieve better dispersion effect,especially in areas with obvious feature differentiation.