Spatial Simulation Of Population In Mountain Cities Based On GIS And Multi-source Data

Population data are widely used in many fields such as ecology,economy,environment,and land disasters.However,due to restrictions on administrative divisions and other restrictions,there are problems such as low temporal and spatial resolution,which greatly limit the research application.The spatialization of population data has important implications for the integration of disciplines,the spatial research on the coordination of ecological environment and population,and the loss assessment of natural disasters at specific scales.With the increasing progress of technologies such as Geographical Information System(GIS)and Remote Sensing Technology(RS),especially Night-time light data and Remote Sensing images acquired at multiple spatial scales,there are more and more data and models used to simulate population spatialization.However,the spatial distribution of the population in mountain cities is different,of which the simulation results are relatively poor,and the models still have room for optimization.For the spatialization of population in mountain cities,the suitability of the model should be taken into account at first.Because of the largely undulating topography of the mountain cities,the population is mostly concentrated near the mouth of the low-lying estuary,and there are sporadic distributions on other mountains,resulting in great differences of population densities.It is difficult to simulate the distribution of the population with a single data.Therefore,the data of multiple sources should be considered to participate in the population space model.In this paper,by comparing and analyzing the main population spatialization models,we selected models suitable for mountain cities to conduct trial simulations for Chongqing's population and examined the accuracy.In the end,the models which are best suited the spatialization of population in mountain cities were selected.Afterwards,we spatialized the statistical population of Chongqing in the years 2000,2005,2010 and 2015 and analyzed the temporal and spatial evolution.The main contents are as follows:(1)By comparing historical spatialization models,it is found that range attenuation model and spatial interpolation model are mostly used in plain city where the population distribution is decreasing from the central city to the periphery.The partition density model is mainly used in the district where the same type of internal population density is consistent and the intelligent model is mostly applied to fine-scale research.These models are not suitable for mountain cities where the spatial distribution of population is very different.Multi-variable regression model and multi-factor fusion model are more commonly used for population simulation in mountain cities due to the addition of correction factors and topographic factors.(2)Firstly,the multivariate regression model was used to spatialize the population of Chongqing in 2010.By constructing the Habitat Index and doing multiple regression analysis with population density,it was found that the average relative error of the population is 44.69% in carrying out accuracy verification at district and county scales without considering the elevation,and the larger errors are mostly distributed in districts and counties with an average elevation of 600 m or more;After adjusting for the Habitation Index by elevation,the average relative error of the population at the districts and counties scale was reduced to 27.17%,indicating that the elevation of the mountainous city has a great influence on the distribution of the population.(3)Then,the multi-factor fusion model was used to spatialize the population of Chongqing in 2010.Through the analysis of the correlation and the actual situation between factors and population,it was discovered that the overall correlation between DEM and population density is only-0.28,but for the districts and counties whose average elevation is above 355 m,the correlation of population density reach-0.855,and the correlation is relatively high;while the correlation coefficient between water area and population is 0.872,but there is no distribution of residential areas near a large proportion of water,it was not suitable as a factor of population simulation in Chongqing;The correlation coefficient of population and traffic factor is 0.804 and traffic factors are mostly distributed around residential areas,and the closer to the road,the more residential areas in the area.Therefore,after removing the water area factor and increasing the traffic factor,the four factors including terrain,traffic,DMSP-OLS and NDVI were selected as the impact factors of the simulated population.(4)Different land-use types have different effects on population distribution.This paper subdivides the settlements staying away from the rural into different land types,determines their respective weights with multiple regression method.The simulation for all kinds of population which is far from rural settlements is more detailed,and the methods for determining weights are more scientific.(5)The population results simulated by the multi-factor fusion model show that the population density in the urban residential areas is consistent because the population weights of urban residents' land uses are assigned the same value,and the urban residents' regional environment is the same or similar,which result that the population weights of the impact factors is similar and the population differences within the simulated urban areas are extremely small.Therefore,this paper introduces the kernel density of industrial sites as a factor to optimize the weights of residential populations to differentiate the population weights within the urban areas.The article verified the accuracy of the simulation result with the sixth census of villages and towns statistics,and found that the correlation between the optimized simulated population and the statistical population density reached 0.837.Compared with the multi-variable regression model of 0.6 and 0.799 for the pre-optimization multi-factor fusion model,the correlation increases more,and the simulation is better than the previous two.Therefore,this paper used the optimized multi-factor fusion model to simulate the spatial distribution of the population in 2000,2005,2010 and 2015 in Chongqing.(6)Through a comparative analysis of the simulated four-period population,it was found that the overall population of Chongqing shows an increasing trend,while the rural population shows a gradual decreasing trend.Compared with the first two time periods,the population growth of 2010-2015 is the largest.The population of Chongqing is mainly located near the urban area and gradually expands to the surrounding areas.Especially in Yubei District,the population rapidly expands from the southwest of the Yubei District to the northeast.In the northeastern and southeastern parts of Chongqing,there is less population distribution in the area around the basin,and the overall growth trend is not obvious.The distribution of population in the paralleled ridge-valley area in northern Chongqing is larger than that in southeastern and northeastern Chongqing,especially in Wanzhou District and Kaizhou District,where the population density is higher and growth is faster.The population density in hilly region of western Chongqing is widely distributed and grows rapidly.