Modeling And Application Of Population Spatial Distribution In Historical Periods With Multisource Data

The anthroposphere is the critical component of the earth surface system.Human activities deeply change the natural environment and atmosphere composition.Thus,there is a demand to obtain an explicitly spatial population distribution pattern which is of great significance to the research of human-induced historical land use and climatic environmental evolution.In this study,firstly,we used the GeoDetector method to select the environmental factors which significantly affected the population distribution.Furthermore,based on this,we established a population density forecast model by using the stepwise multivariable liner regression model(SMLRM)and random forest regression model(RFRM),and verified the accuracy and stability of the model.Based on prefecture-level census data derived from the historical documents,we reconstructed the spatially explicit population datasets for the traditionally agricultural area of China which contained six periods(1776,1820,1851,1880,1910 and 1953).Finally,this study analyzed the spatiotemporal variations of population from 1776 to 1953,and discussed the driving factors of population change.The main conclusion can be listed as follows:(1)On the county scale in eastern China,the altitude,slope,relief amplitude,the distance to the nearest provincial and prefectural cities,and climatic moisture index have great explanation power for the spatial differences of population,which are the dominant environmental factors affecting population distribution.Relatively,the distance to rivers has less influence on the spatial distribution of population.(2)The Stepwise Multivariable Linear Regression model and the random forest regression model based on the independent variables of altitude,slope,relief amplitude,the distance to the nearest provincial and prefectural cities and climatic moisture index can well simulate the spatial distribution of population density at the county scale in eastern China.The determination coefficient(R~2)of the RFRM is 0.73(P<0.01),which is higher than 0.14(P<0.01)of the SMLRM;the root mean square error(RMSE)of the SMLRM is 1419 persons per km~2,which is higher than that of the RFRM(802persons per km~2).It indicates that the RFRM with nonlinear relationship has a strong ability to simulate the spatial variations of population distribution in eastern China.(3)The spatial distribution changes of population density based on the reconstruction of the RFRM showed that the overall spatial distribution pattern of population in traditional agricultural area of China is relatively stable from 1776 to1953.The population is densely distributed in the middle and lower reaches of the Yangtze River Plain,the Huang-Huai-Hai Plain,Sichuan Basin and the Pearl River Delta.However,in different periods,there are obvious differences in the spatial distribution characteristics of population.From 1776 to 1851,the population in traditional agricultural area of China showed an increasing trend,the main population growth area and high population density agglomeration area was the Yangtze River Delta area.From 1851 to 1880,the characteristics of population change were polarized.On the one hand,affected by the war and drought,the population in the Yangtze River Delta area and Guanzhong Plain showed a downward trend,and the population agglomeration characteristic in those areas weakened;on the other hand,the population in Huang-Huai-Hai plain,Sichuan Basin and Pearl River Delta kept growing.The high population density agglomeration areas in Huang-Huai-Hai plain and Sichuan Basin were taking shape.From 1880 to 1953,the population in traditional agricultural area of China showed a wildly growth trend.The main population growth areas were Huang-Huai-Hai plain and Sichuan Basin,and the high population density agglomeration areas in those areas were finally formed.(4)Natural conditions and the development of agricultural determined the basic population spatial distribution pattern.War,natural disasters and migration are the dominant driving forces of population spatial and temporal distribution pattern changes in the traditional agricultural area of China.