| Population spatialization is that the population based on administrative unit is more accurately distributed in the regional space,which can show the spatial distribution information of population more comprehensively.Population raster data was of great significance to the implementation of fine population management and the coordination of the contradiction among population,resources,environment,and development.As an indicator of global warming,the Tibetan Plateau was experiencing rapid warming,almost 1.5 times the global average.Benefiting from climate warming,there were more and more areas suitable for human life and production in the Tibetan Plateau.The study of population spatialization on the Tibetan Plateau was not only conducive to clarify the characteristics of population distribution on the Tibetan Plateau,but also helped to balance the contradiction between ecological protection and population expansion.Meanwhile,this was more helpful to improve the characteristics of global population distribution,such as the highest altitude of human habitation.Secondly,the Tibetan Plateau was bordered by Myanmar,India,Bhutan,Nepal,and other countries,and its geographical location was of great strategic significance.President Xi Jinping had also stressed many times that"to govern the country must govern the border,and the border must first stabilize Tibet".The first important task of stability was to grasp the information of population distribution.Therefore,understanding the population distribution of the Tibetan Plateau was of great strategic significance for national measures such as stabilizing and assisting Tibet.In this paper,taking the Tibetan Plateau as the study area,altitude,slope,relief degree of land surface,normalized vegetation index,corridors such as rivers and roads,nighttime light data,point-of-interest data and human footprint data were selected as the influencing factors of population distribution from two aspects of natural conditions and social economy.On the basis of the township scale data of the sixth national population census in 2010,combined with the administrative division of Tibetan Plateau,the demographic data of the Tibetan Plateau were spatially transformed into 30×30 m raster data by using single model(the random forest model,the XGBooost model and the Cubist model)and two multi-model ensemble methods(weighted average ensemble model and stacking ensemble model).The fitting accuracy of each model was verified,and the reason for the difference of each model was discussed.At the same time,the fitting results were compared horizontally with World Pop population data set and the Google Earth images to further verify the accuracy of the results.In addition,based on the optimal fitting results,this paper used the spatial analysis method to study the characteristics of population distribution,population spatial differentiation and spatial autocorrelation in the Tibetan Plateau,as well as the coupling relationship with various index factors.Finally,the influence of various index factors on the population distribution of the Tibetan Plateau were discussed,and the advantages and adaptability of the single model and the integrated model were also discussed.The main conclusions were as follows:(1)the fitting accuracy of all models were better than that of the World Pop data set(RMSE=1143.81,R~2=0.814),and the best fitting accuracy of the Cubist model(RMSE=865.77,R~2=0.893)was obtained in a single model by piecewise linear regression method.The spatialization results of the ensemble models were better than those of the individual models.The results of the Stacking ensemble model(RMSE=799.79,R~2=0.908)by the method of secondary machine learning were more robust than those of the weighted ensemble model(RMSE=832.86,R~2=0.891),and the Stacking fitting accuracy was 30.08%higher than that of the World Pop data set.On the whole,the population spatialization method based on the Stacking ensemble model could maximize model accuracy and achieve good quality population fitting.The stacking ensemble methods could also be applied to other studies in complex terrain areas.(2)The resident population in the Tibetan plateau was 11.736 to 12.048 million in 2010,with an average population density of 4.56-4.68 person/km~2.The population distribution was dense in the southeast and sparse in the northwest in the horizontal dimension,depicting a tertiary distribution in the vertical dimension.The agricultural developed areas such as Hehuang Valley and the middle reaches of the Yarlung Zangbo River were the core areas of the population distribution in the Tibetan Plateau.Through the calculation of heterogeneity on the grid scale,it was found that the population distribution heterogeneity of the Tibetan Plateau was generally high,while the areas with low heterogeneity were mainly distributed within the city.Through spatial autocorrelation analysis,it was found that the global Moran index was 0.8241,and there was a significant positive spatial autocorrelation in population distribution,which was dominated by low-low cluster distribution,and high-high aggregation distribution mainly occurred in the eastern part of the Tibetan Plateau such as Xining,Lhasa,Golmud City and Xigaze City.Combined with the results of spatial autocorrelation analysis and heterogeneity,it was found that the areas with low heterogeneity of population distribution were distributed in high-high population aggregation areas,while the areas with high population heterogeneity were distributed in low-low population aggregation areas.(3)the upper limit of population distribution in the Tibetan Plateau was 5800 m.In addition,about 65%of the population of the Tibetan Plateau was mainly distributed in the range where the slope was less than 25°,the relief degree of land surface was less than 300 m,and the distance from the river was less than 8 km.It showed that the population of the Tibetan Plateau was mainly distributed in the areas with low altitude,small slope,flat topography and abundant water resources.In terms of social economy,the population was mainly distributed in areas with perfect infrastructure,developed economy and convenient transportation.Among the factors that affect the population distribution,altitude,distance from the built-up area and human settlement index were the three main factors that affected the population distribution.Generally,geographical factors determined whether there was population distribution in a certain area.and socio-economic factors determined the number of population distribution in this area. |
PDF Full Text Request