The Comparison and Modeling of the Driving Factors of Urban Expansion for Thirty-Five Big Cities in the Three Regions in China

1School of Government, Beijing Normal University, Xinwai Street, Beijing 100875, China 2State Key Laboratory of Resources and Environmental Information Systems, Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, No. A11, Datun Road, ChaoYang District, Beijing 100101, China 3State Key Laboratory of Water Environment Simulation, School of Environment, Beijing Normal University, Xinwai Street, Beijing 100875, China


Introduction
Across the world, rapid urban expansion has encroached on croplands, forests, water bodies, barren land, and even villages on a large scale [1][2][3][4][5][6][7][8][9][10].The wholesale transformation of natural ecosystems and agricultural lands has resulted in widespread environmental changes in the twenty-first century [1].The urbanization process also affects the water, carbon, aerosol, and nitrogen cycles in the climate system.Although urban land occupies only approximately 2.4% of global landmass, it is home to more than half of the world's people [11].In large cities, air, water, and noise pollution, traffic congestion, and high housing prices influence inhabitants' quality of life.
The Chinese urbanization process has attracted global attention [5][6][7][8].The country has experienced an unprecedented scale and rate of urban expansion over the last four decades.In 1978, China's urbanization level was only 17%, but by 2010, its urban dwellers accounted for 49.68% of the total population [12,13].There were 654 cities in China in 2010.The four municipalities and twenty-seven capital cities are the major cities of China.But Dalian, Qingdao, Ningbo, and Xiamen are as important as their provincial capitals.Hence, we had added the four big cities as the objects of study (Figure 1).In 2010, the thirty-five cities had a combined population of 248 million, accounting for 19.93% of the total population.Their urban population was 147 million, accounting for 38.08% of the total urban population of China [13].
Urban expansion is the spreading out of the cities and townships into the periphery areas with the urban development.A better understanding of urban expansion and its determinants is necessary to assess the impact of possible future developments on the environment, economy, and society at large [14].The urban construction land expanded dramatically and had encroached on a great amount of arable land and threatened the grain safety of China [6,10].At the same time, the urban expansion indicated the periodical difference and the regional difference in Eastern, Middle, and Western China [6,15].The land use had become more intensive with economic development in Eastern China while urban land expanded remarkably in middle and western regions [6,15].
Urban dynamic model can improve our understanding of the mechanisms of urban expansion.The CA (cellular automata) and ABM (agent-based model) have emerged from complexity, chaos, and cybernetics and have been widely used to simulate urban dynamics.Social and economic factors are the major factors of land use dynamics.The spatiotemporal dynamics and simulations of urban expansion of the individual cities and regions in China have been widely studied [1,3,[16][17][18][19][20]. Spatial regression analysis and panel analysis have been used to simulate the urban expansion [15,21].
The comparisons of urban expansion and its driving factors among the different regions in the different periods are very few.Because the economic system has transformed from the planning to market economy, the regional policies have impacted the economic growth and urban expansion.In order to speed up the economic development, the central government encouraged the coastal and Eastern China to be developed firstly after 1978.Then the big development of Western China was put forward in 2000.The comparisons of urban expansion in Eastern, Middle, and Western China and its driving factors will be beneficial to formulate the policies.The policies should promote the economic development and urban development while minimizing environmental impacts [1].
The decision-makers of the central government and local governments need to understand the factors that drive urban expansion for the different regions.However, the driving factors differ for different regions and periods.It will be beneficial, therefore, to distinguish these driving factors for policy makers.In this context, it is of great importance to build a national-scale and regional-scale urban expansion model and study its driving forces.The objectives of this study are (1) to quantify the relationships between urban expansion and its driving factors for thirty-five big cities in China and (2) to distinguish the driving factors for urban expansion among the three regions of China.The original data of the eight indices of driving factors are extracted from the Chinese Statistics Yearbook.The units of the original data of the eight indices had been listed in Table 1.All the data was processed into the increment rate.We had listed the mean, maximum, minimum, and standard deviation of the eight indices of the thirty-five cities between 1990 and 2000 in China (Table 1).

Data Source and Method
Urban expansion is defined as a largely continuous area covered by urban facilities but may include some isolated facilities [22].The data on urban expansion are generated by the Ministry of Housing and Urban-Rural Development (MOHURD).MOHURD is based on the actual land area that is granted the permission by the local governments to be developed in each city each year.The State Bureau of Surveying and Mapping investigated the actual increase The relative error between the statistical urban land and derived total urban land area at the national was less than 2% and is considered as the best proxy of the real urban land in China [23].
2.2.The National and Regional-Scale Urban Expansion Model and Simulation.Urban expansion is the result of complex interactions between humans and the environment [14].POPI, GDPI, IICC, SIPC, RI, SIPI, TIPI, and ECI are studied using the national-scale and regional-scale urban expansion model (NRUEM).Statistical techniques are well suited to investigate the relationships between these variables.Panel data analysis is performed to study the relationships between urban expansion and its driving factors [21].It is used to identify the major factors of urban expansion.Spatial heterogeneity and unobservable variables may cause the relation between socioeconomic variables and urban expansion.We study the relationships between urban expansion and its social and economic driving factors for each of the three regions.A model for the relationship between urban expansion and its social and economic determinants can be estimated using (1).
The panel data analysis is based on the several panels.The urban expansion model can be expressed as follows: where UE  is the urban expansion of city  at the period  and is the independent variable;   is the driving factor of urban expansion for city  at the period  and the  variable of dependent variable;   is a regression coefficient that is estimated for city ; and   is an error term.
Hausman test is used to choose the panel data model.For the model of  panels, we can use (1).The testing of the sample data will decide whether it is mixed regression model, variable intercept model, and variable coefficient model.There are two hypotheses [24].
0 : The coefficient of variables is the same for all the panel cities.But the intercepts are different.The model is the varying intercept model.
1 : The coefficient of variables and intercepts were all the same for the panel cities.The model was the mixed regression model.
There are two methods of  test statistics: where  is the number of panel cities;  is the number of observational periods of the samples;  is the number of the variables. 1 ,  2 , and  3 are the residual sum of squares of regression models of varying coefficient model, varying intercept model and the mixed regression model.We test the hypothesis  1 firstly.If  2 is less than the threshold value of  distribution, we cannot refuse the hypothesis  1 and do not need to check the hypothesis  0 .Then mixed regression model is used to simulate the sample data.Otherwise, we should refuse the hypothesis  1 and continue to test the hypothesis  0 .If  1 is less than the threshold value of  distribution, we cannot refuse the hypothesis  0 .Then the variable intercept model is used to simulate the sample data.Otherwise, we can refuse the hypothesis  0 .Then the variable intercept and variable coefficient model as (1) can be used to simulate the sample data.Therefore, we use the Hausman test to decide if we use the mixed regression model, variable intercept model, and variable coefficient and intercept model.
We used Eviews 6.0 to calculate the coefficient   .Compared with time-series data model and cross-sectional data model, panel data model can avoid fluctuations over time.Therefore, the regression results are more effective.
According to the gradient development theory, the economic development will transfer to the low technological region from the high technological region.In order to speed up the economic development and improve the economic efficiency, the central government of China had put forward the three regions division in Seventh Five-Year Plan (1986)(1987)(1988)(1989)(1990).By the combination of economic development and   2).We had calculated the coefficient   for urban expansion and driving factors for panel U 1 , panel U 2 , panel U 3 , panel U 4 , and panel U 5 in China (Table 3).Then we calculated the coefficient   for urban expansion and driving factors for panels U 1 , U 2 , U 3 , U 4 , and U 5 in Eastern China (Table 4), Middle China (Table 5), and Western China (Table 6).
At the same time,  1 ,  2 , and  3 are the residual sum of squares of regression models of varying coefficient model, varying intercept model, and the mixed regression model.We can get  1 ,  2 , and  3 values using the Eviews 6.0 software as follows: We can calculate  2 using (2) and  2 = 0.1019.
Because  2 is less than the threshold value, then we cannot refuse the hypothesis  0 .The model is the varying intercept model.The coefficient of variables is the same for all the panel cities.But the intercepts are different.The model is the varying intercept model.
We used Eviews 6.0 to calculate the coefficient   .Compared with time-series data model and cross-sectional data model, panel data model can avoid fluctuations over time.Therefore, the regression results are more effective.

The Urban Expansion of the Studied Cities.
Over twenty years, the urban land area of the thirty-five provincial capitals in China expanded by 8917 km 2 or 222.19%.Urban expansion was faster in Eastern China and Western China and comparatively slow in Middle China (Table 2).This regional pattern can be observed in other studies.For example, Eastern China and Western China expanded more dramatically than Middle China from 1989 to 2000 [15].For Eastern China, rapid economic development stimulated urban expansion.For Western China, the "Great Western Development Policy" stimulated investment and infrastructure construction.However, the primary grain-producing land is located in Middle China.The cropland protection policy constrains urban expansion in that region.Hence, urban expansion in Middle China was slow.
For the cities of Eastern China, urban land expanded the most dramatically from 2000 to 2005 (period U 4 ) (Figure 1).The urban land of the cities in Middle and Western China expanded dramatically in periods U 4 and U 5 (2000-2010).For period U 4 , urban land expanded most dramatically in Eastern China, then Middle and Western China.For period U 5 , urban land expanded dramatically in Middle, Western, and Eastern China (Table 2).
More than 60% of the studied cities (including Beijing, Fuzhou, Xiamen, Guangzhou, Guiyang, Haikou, Hangzhou, Hefei, Harbin, Hohhot, Jinan, Qingdao, Nanjing, Nanning, Shenyang, Dalian, Shijiazhuang, Taiyuan, Tianjin, Yinchuan, and Zhengzhou) expanded the most dramatically in period U 4 (2000)(2001)(2002)(2003)(2004)(2005).Approximately 68% of these cities are located in Eastern China and 22% in Middle China.In Western China, only Yinchuan and Guiyang had the highest expansion rates in period U 4 .During this period, the rapid urban development of China resulted in the wasteful use of land.After that, the central government implemented a strict land use policy in cities [7,8].In October 2008, the General Planning Outline of National Land Use (2006-2020) was issued.The measures in the planning outline slowed urban expansion in Eastern China, as urban expansion was  constrained by the control policy.However, for Western China, economic development was the first priority.The urban expansion of most cities in Western China was the most rapid between 2005 and 2010.During period U 5 (2005-2010), Changchun, Changsha, Kunming, Nanchang, Urumqi, and Xi'an expanded the most dramatically.The urban expansion rate of Kunming was 22.9% and that of Urumqi was 18.52%.The urban expansion of the other twenty-five cities was slower in period U 5 than in period U 4 .
City-level built-up areas sometimes increase abruptly due to the effect of a large built-up area project.For example, the urban land of Beijing expanded by 20.15%, 2.31%, 142.21%, and 14.21% during the four time periods (periods U 2-5 ).The urban expansion of Shanghai was also uneven across the four periods (56%, 41.03%, 49.09%, and 139.85%).Relatively speaking, the urban expansion of Guangzhou increased evenly by 42.31%, 66.41%, 70.53%, and 70.44% in the four periods, respectively.

The Driving Factors of National-Scale Urban Expansion.
Many studies argue that economic growth stimulates urban expansion in China through a positive correlation between the two indicators [15,25].In the Pearl River Delta, largescale investments in industrial development played the major role in urban land conversion [1].Urban land expansion is not  only the consequences of economic growth, but also drivers of such growth [22].We studied the relationship between urban expansion and eight social and economic indicators.
For period U 1 (1991-2010), national-scale urban expansion was positively correlated with the four driving factors (POPI, SIPC, SIPI, and RI).The regression coefficient between urban expansion and POPI was 0.2038, indicating the highest impact on urban expansion.Table 3 summarizes the model for urban expansion and the driving factors that were included.The relationship between urban expansion and the three driving factors (SIPC, SIPI, and RI) was statistically significant at a 5% level while the relationship between urban expansion and POPI was statistically significantly at a 1% level (Table 3).Salary increase per capita (SIPC) motivates farmers to migrate to the cities.Increases in secondary industrial production (SIPI) can impact urban expansion.For period U 1 , population increment (POPI) and salary increase per capita (SIPC) were the dominant driving factors for urban expansion, followed by SIPI and RI that were the other major driving factors (Table 3).
Urban expansion had different characteristics in the different time periods.In period U 2 (1991-1995), urban expansion was positively correlated with population growth (POPI), economic development (GDPI), and electricity consumption increment (ECI) at 1% level (Table 3).Urban expansion was significantly positively correlated with tertiary industry product increment (TIPI) at 5% level.The regression coefficient between urban expansion and POPI was the highest.Urban expansion was also strongly influenced by economic development and electricity consumption increment (GDPI and ECI).In period U 3 (1995)(1996)(1997)(1998)(1999)(2000), urban expansion was positively correlated with the three driving factors of GDPI, POPI, and ECI (Table 3).The regression coefficient with GDPI was the highest.For period U 4 (2000)(2001)(2002)(2003)(2004)(2005), urban expansion was positively correlated with four driving factors: SIPI, RI, IICC, and ECI (Table 3).The regression coefficient with SIPI was the highest.For this period, SIPI was the greatest driving factor for urban expansion.RI and IICC were also major driving factors for urban expansion.In period U 5 (2005-2010), urban expansion was positively correlated with IICC, POPI, and SIPI.IICC had the highest regression coefficient and was the most important driving factor of urban expansion.POPI stimulated urban expansion at a less high level.
From above analysis, the urban expansion was correlated with the social and economic factors.However, the driving factors were dramatically different for different periods.For the periods U 2 (1991-1995) and U 3 (2000)(2001)(2002)(2003)(2004)(2005), POPI and GDPI were the primary factors.For the period U 3 (1995)(1996)(1997)(1998)(1999)(2000), the secondary industry and road increment became the dominant driving factors while investment increment in capital construction and electricity consumption became the primary driving factors for period U 5 (2005-2010).

The Urban Expansion of Eastern, Middle, and Western
China and Its Driving Factors.To compare the differences between the three regions, we studied the relationships between urban expansion and its driving factors in Eastern, Middle, and Western China.
In Eastern China, urban expansion was positively correlated with POPI, RI, and ECI during period U 1 (Table 4).Population increment (POPI) was the strongest stimulant of urban expansion.Road increment was the secondly most important driving factor for urban expansion.In period U 2 , urban expansion in Eastern China was positively correlated with two driving factors (RI and GDPI) (Table 4).The regression coefficient was the highest with RI (0.1545).For this period, RI was the dominant driving factor for urban expansion.GDPI also stimulated urban expansion.In period U 3 , the urban expansion of Eastern China was correlated positively with the two driving factors (ECI and RI) (Table 4).The regression coefficient with ECI was the highest (0.0378).For this period, ECI was the major driving factor for urban expansion.For period U 4 , urban expansion in Eastern China was correlated positively with four driving factors (GDPI, ECI, TIPI, and SIPI).Urban expansion was most highly correlated with GDPI (0.2312, statistically significant at a 1% level).For this period, GDPI became the dominant driving factor.ECI and TIPI also stimulated urban expansion during this time.For period U 5 , urban expansion of Eastern China was correlated positively with SIPI (statistically significant at a 10% level).Urban expansion was most strongly correlated with secondary industry increment.For this period, secondary industry was the most important driving factor for urban expansion.For Eastern China, POPI and RI were the dominant factors for the 20 years.However, RI was the dominant driving factor during period U 2 .For the period U 3 , ECI was the major driving factor.GDPI was the major driving factor for the period U 4 .The secondary industry was the major driving factor of urban expansion for the period U 5 .In Eastern China, the correlation between urban expansion and IICC cannot pass the significance test.
In Middle China, urban expansion was positively correlated with SIPI and GDPI during period U 1 (Table 5).The regression coefficient with SIPI was 0.1287 (statistically significant at a 5% level).For the four periods (U 2 , U 3 , U 4 , and U 5 ), urban expansion was positively correlated with GDPI, based on the panel data analysis (Table 5).Urban expansion was most strongly correlated with SIPI in period U 2 , GDPI in period U 3 , TIPI in period U 4 , and GDPI in period U 5 .In period U 2 , SIPI and GDPI were the major driving factors for urban expansion.The secondary industry and economic development stimulated urban expansion in Middle China.For period U 3 , urban expansion was correlated positively with GDPI, SIPI, and TIPI.In period U 4 , urban expansion was correlated positively with TIPI, ECI, POPI, and GDPI.In this period, tertiary industry stimulated urban expansion.Finally, in period U 5 , urban expansion had the highest correlation with GDPI.It also exhibited positive correlations with TIPI and ECI.For Middle China, GDPI was the most important driving factor for urban expansion in periods U 3 and U 5 .SIPC and IICC were the most important driving factors of urban expansion for periods U 2 and U 4 .In Middle China, the correlation between urban expansion and IICC cannot pass the significance test.For the twenty years, secondary industry and economic development were the most important factors of urban expansion.
In Western China, urban expansion was correlated positively with GDPI and ECI in period U 1 (statistically significant at a 1% level) (Table 6).The regression coefficient between urban expansion and GDPI was 0.0902.GDPI had the highest correlations with urban expansion.Over the twenty-year study period, economic development stimulated urban expansion in Western China.Moreover, electricity consumption had stimulated urban expansion.In period U 2 , urban expansion was positively correlated with POPI (statistically significant at a 1% level), SIPI (statistically significant at a 5% level), and TIPI (statistically significant at a 10% level) (Table 6).It exhibited the highest correlation with population growth.For period U 3 , urban expansion was significantly positively correlated with GDPI, SIPI, and TIPI (Table 6).In period U 4 , urban expansion was positively correlated with POPI (statistically significant at a 5% level) (Table 6).In period U 5 , urban expansion was significantly positively correlated with GDPI (statistically significant at a 1% level).It exhibited the highest correlation with GDPI.SIPI and RI were correlated with urban expansion at 5% level.ECI was correlated with urban expansion at a 10% level.For Western China, economic development and population growth had stimulated the urban expansion.Economic development (GDPI) was the major driving factor of urban expansion for the period of U 1 , U 3 , and U 5 .Population growth (POPI) was the major driving factor for urban expansion during two of the studied five-year periods (U 2 and U 4 ).But the correlation between urban expansion and IICC cannot pass the significance test.

Figure 1 :
Figure 1: The spatial distribution of the thirty-five big cities in the three regions in China.

2. 1 .
Data Source.We utilized urban land area in the thirtyfive studied cities published in the Chinese Statistics Yearbook between 1990 and 2010.We studied five periods of urban expansion for the provincial cities in China (panel U 1 : 1990-2010; panel U 2 : 1990-1995; panel U 3 : 1995-2000; panel U 4 : 2000-2005; panel U 5 : 2005-2010).City-level data for eight socioeconomic variables, population increase (POPI), GDP (Gross Domestic Product) increase (GDPI), increase in investment in capital construction (IICC), salary increase per capita (SIPC), increase in roads (RI), increase in secondary industrial production (SIPI), increase in tertiary industrial production (TIPI), and increase in electricity consumption (ECI), were included.POPI indicates the impact of population growth on urban expansion.GDPI, IICC, SIPC, SIPI, and TIPI indicate the impact of economic growth.Investment in capital construction (IICC) is the investment of the new builtup reconstruction and built-up extension of the engineering of the enterprises, institutions, and administrative units.RI and ECI indicate the impact of road construction and electricity consumption.The electricity consumption indicates the energy consumption and supports the economic growth and urban expansion.

Table 1 :
Statistics summary of driving factors of urban expansion in 1990-2010 in China.

Table 2 :
Urban expansion rate of Eastern China, Middle China, and Western China (%).Figure1).Shanghai, Beijing, and Guangzhou are the key economic centers in China.Shanghai is the central city of the Yangtze River Delta urban clusters.Guangzhou is the central city of the Pearl River Delta urban clusters.Beijing and Tianjin are the central cities of the surrounding Bohai Bay urban clusters.There are apparent differences between the three regions.Comparison of the differences between the three regions will help the central government to formulate the sustainable development policies, especially the urbanization policies.We had calculated urban expansion rate of Eastern China, Middle China, and Western China for panel U 1

Table 3 :
Urban expansion and driving factors in 1990-2010 in China.

Table 4 :
Urban expansion and driving factors in Eastern China.

Table 5 :
Urban expansion and driving factors in Middle China.

Table 6 :
Urban expansion and driving factors in Western China.