Ecological Modelling 181 (2005) 461–478 Surface modelling of human population distribution in China Tian Xiang Yuea,∗, Ying An Wanga, Ji Yuan Liua, Shu Peng Chena, Dong Sheng Qiua, Xiang Zheng Denga, Ming Liang Liua, Yong Zhong Tiana, Bian Ping Sub a Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, 917 Building, Datun, Anwai, Beijing 100101, China b College of Science, Xi’an University of Architecture and Technology, Xi’an 710055, China Received 24 March 2003; received in revised form 23 April 2004; accepted 4 June 2004 Abstract On the basis of introducing major data layers corresponding to net primary productivity (NPP), elevation, city distribution and transport infrastructure distribution of China, surface modelling of population distribution (SMPD) is conducted by means of grid generation method. A search radius of 200 km is defined in the process of generating each grid cell. SMPD not only pays attention to the situation of relative elements at the site of generating grid cell itself but also calculates contributions of other grid cells by searching the surrounding environment of the generating grid cell. Human population distribution trend since 1930 in China is analysed. The results show that human population distribution in China has a slanting trend from the eastern region to the western and middle regions of China during the period from 1930 to 2000. Two scenarios in 2015 are developed under two kinds of assumptions. Both scenarios show that the trends of population floating from the western and middle regions to the eastern region of China are very outstanding with urbanization and transport development. © 2004 Elsevier B.V. All rights reserved. Keywords: Surface modelling; Population distribution; Grid generation; Geographical information system 1. Introduction et al., 2004). Surface modelling includes development of digital terrain models, spatial interpolation models, A surface model is the mathematical representa- area-based matching models and a multi-resolution ap- tion of a surface in such a form that it can be used proach. Three widely used principal ways of struc- in design calculations. Since the first digital terrain turing a digital terrain model are triangulated irregu- model for road design was produced by the Mas- lar networks, regular grid networks and contour-based sachusetts Institute of Technology in 1957, surface networks (Moore et al., 1992). Spatial interpolation modelling has begun to be developed (Stott, 1977; Yue models include interpolation by drawing boundaries, trend surface analysis, moving averages, Kriging in- terpolation, spline curves and finite element method ∗ Corresponding author. Tel.: +86 10 64889633; fax: +86 10 64889630. (Stein, 1999; Sabin, 1990; Shipley, 1990). Area-based E-mail address: [email protected] (T.X. Yue). matching models include radiometric model and least 0304-3800/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2004.06.042 462 T.X. Yue et al. / Ecological Modelling 181 (2005) 461–478 squares correlation (Mustaffar and Mitchell, 2001; Geography of Chinese Academy of Sciences, 1987); Heipke, 1997; Gruen, 1985; Foestner, 1982). The transforming population data from census to grid multi-resolution approach is an image-driven surface (Tobleret al., 1997), apportioning census counts to each estimation, which is characterized by three phases that grid cell based on probability coefficients (Dobson et are shape modelling, multi-resolution model construc- al., 2000), and estimating population using nighttime tion and variable resolution representations (Sarti and light data (Sutton, 1997; Sutton et al., 1997, 2001; Lo, Tubaro, 2002; Cignoni et al., 1998). 2001; Sutton et al., 2003). Surface modelling of population distribution (SMPD) that was developed on the basis of grid gen- eration method (Yue et al., 2003) is aimed at formulat- 2. Background of major data layers ing population in a regular grid system, in which each grid cell contains an estimate of total population that In addition to total population in every province, is representative for that particular location. Compiling the major data layers that are matched with their corre- population data in grid form is by no means a new ap- sponding SMPD variables include net primary produc- proach. For instance, Adams (1968) presented a com- tivity (NPP), elevation, city distribution and transport puter generated grid map of population density in west infrastructure distribution. The elevation is a natural Africa; Population Atlas of China presented grid pop- factor and has a slow change with time so that it is ulation data for several regions in China (Institute of spatial variable and could be regarded as a temporal Fig. 1. Spatial distribution of cities in 2000 in China (unit: thousand persons grouped by non-agricultural population in urban district). T.X. Yue et al. / Ecological Modelling 181 (2005) 461–478 463 constant within 100 years. Although NPP is based on ern China in modern history. During the period from climate and soil, it could be modified by human activ- 1843 to 1893, urban population proportion had slow ities so that it is a spatial and temporal variable. City growth, which was increased from 5.1% to 6.0% aver- distribution and transport infrastructure distribution are agely in China; in the area of lower reaches of Yangtze spatial and temporal variables, which are determined River the urban population proportion was increased by both natural factors and human activities, have a from 7.4% to 10.6%, in the coastal area of south rapid change with time in China during recent 100 China from 7.0% to 8.7%; in inland area the pro- years. portion paced up and down between 4.0% and 5.0%. From 1895 to 1931, in areas along coast and Yangtze 2.1. Spatial distribution of cities in China River, northeast China and north China cities were de- veloped rapidly, while in inland area cities were de- Urbanization is a process of the concentration of veloped very slowly, even at a standstill. In the early population in cities. Spatial distribution of cities and 1930s, urban population proportion was about 9.2% in proximity to cities are essential factors for human pop- China. From 1931 to 1949, the turbulent and unstable ulation distribution of China. The spatial distribution situation led to slow population growth in China and of cities in China has had the feature that city den- the urban population proportion increased to 10.6% sity is much higher in eastern China than in west- (Zhang, 1997). Cities in China spatially concentrates Fig. 2. Spatial distribution of railways in 1930 in China. 464 T.X. Yue et al. / Ecological Modelling 181 (2005) 461–478 in coastal area, especially Yangtze River Delta, Peal tion proportion would increase at the average rate in River Delta and Beijing–Tianjin–Tangshan area. In recent 5 years, annually 1.44%, it would be 57.82% 2000, 42.1% of the 667 major cities of China dis- in 2015. tributed in eastern China where area accounts for 9.5% of the whole area of China; 34% distributed in mid- 2.2. Spatial distribution of transport infrastructure dle China where area accounts for 17.4%; 23.8% dis- in China tributed in western China where area accounts for 70.4% (Urban Society and Economy Survey Team of Transport infrastructure is a primary indicator of National Bureau of Statistics of People’s Republic of human population distribution (Dobson et al., 2000). China, 2001). The distribution densities of cities in Roads and railways are especially indicative because eastern China and in middle China were respectively of their vital role in human well being. Construction 13.1 times and 5.8 times the one in western China of a piece of railroad in 1881, which was from Tang- (as seen in Fig. 1). The urban population proportion Shan city to Feng-Nan county about 9.7 km, initiated was 36.22% in 2000 (National Bureau of Statistics railroad development in China. There was 14,411 km of People’s Republic of China, 2001). According to of railroad in China in 1930 (Fig. 2) and 21,800 km National Report of China Urban Development (China in 1949 (Fig. 3). Length of railway in operation was Mayor Association, 2002), the urban population pro- 68,700 km in 2000 (Fig. 4)(Year Book House of China portion would be 46.9% in 2015. If the urban popula- Transportation and Communications, 2001). However, Fig. 3. Spatial distribution of railways in 1949 in China. T.X. Yue et al. / Ecological Modelling 181 (2005) 461–478 465 Fig. 4. Spatial distribution of railways in 2000 in China. relative research results showed that the appropriate cations, 2001). In recent 10 years, major projects of length of railroad should be 100,000 km in China, from highway construction include seven east–west main which current railroad length has a great gap (Chen and trunk roads and five south–north main trunk roads Zhang, 2000). To implement the western development as well as three important sections, of which total strategy, railroad building in China is paying attention length is about 35,000 km (Fig. 8). The seven east–west to strengthening the linkage between eastern region and main trunk roads include highways from Suifenhe western region of China, speeding up construction of to Manzhouli, from Dandong to Lasa, from Qing- accesses to central Asia and southeast Asia and improv- dao to Yinchuan, from Lianyungang to Huerguosi, ing connection within the western region of China. To- from Shanghai to Chengdu, from Shanghai to Ruili tal length of railroad in China would reach 81,653 km and from Hengyang to Kunming. The 5 south–north in 2015 (Fig. 5). main trunk roads include highways from Tongjiang of In 1902, the first automobile was imported in Heilongjiang province to Sanya of Hainan province, China and in 1906 the first piece of road was con- from Beijing to Fuzhou, from Beijing to Zhuhai, structed.