Urban expansion in Asia evaluation spati
Acknowledgement:
We would like to acknowledge the funding support from the National Aeronautics and Space Administration (NASA)’s Land Cover and Land Use Program through the grant to Michigan State University (NNX09AI32G) and “Urbanization in Asia” Project at Asian Development Bank (ADB). We thank Prof. Anxin Mei at East China Normal University and Prof. Bin Zhao from Fudan University in Shanghai, Prof. Yaowen Xie and Yongchun Yang from Lanzhou University, Prof. Xi Cheng and his colleagues at Xinjiang Ecology and Geography Institute, China Academy of Science in Urumqi, and Prof. Yong Liu at Southwest University in Chongqing for their support of our research, data sharing, and insights on urban land development of their respective cities. We also appreciate the support of Nathan Moore and Jianjun Ge for our research of Urumqi and Shanghai. Any opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of NASA or ADB.
1. Introduction
Asian cities have experienced rapid urbanization in recent years. As of 2010, the Asia and Pacific region had 43% of the population living in urban areas. Notably over the last two decades, the Asia-Pacific urban proportion has risen by 29%, more than any other region in the world (UNPD, 2010 & 2009). A well-known example, China, dramatically increased its urbanization ratio, i.e., urban population as a percentage of total population, from 18% in 1978 to over 50% in 2011. The rapid urbanization process exerts tremendous pressures on social, economic and environmental sustainability (Pickett et al., 2001). Urban sprawl is becoming a common phenomenon across the Asian continent. Although most urban expansion tend to occur in coastal cities, such as Shanghai and Mumbai, where the economy experienced the fastest growth, in recent years, large-scale urban development also occurred in traditionally resource-limited and environmentally vulnerable regions, such as cities in inland China and within the dryland regions of East Asia including Chongqing, Urumqi, Ulaanbaatar.
Despite the widely acknowledged urbanization and its severe impact, most previous studies on Asian urbanization focused on individual cities, and there is a lack of systematic analysis to present a comprehensive view of Asian urbanization, largely due to difficulty in collecting land use data over long time periods.
This paper evaluates the urban land expansion of nine major cities in Asia by relying on data processed from satellite images, in combination with existing land use maps of different periods. Further, using Shanghai as a case, it identifies the driving forces for urban land change and simulates future urban land use change under different scenarios. This paper will significantly improve our understanding of urbanization in Asia and offer support for policy makers and urban planners in land management and decision-making.
The rest of the paper is organized as follows: Section 2 describes the study area, data, and methodology. Section 3 presents findings on urban land expansion of the selected cities, driving forces for urban expansion and simulation of land use change of Shanghai. Section 4 further discusses (1) characteristics of urbanization in different types of Asian cities, such as coastal mega cities, inland mega cities, and major cities in the dryland region of East Asia, socio-economic driving force for urban expansion, and the implication of simulation. Section 5 concludes and offers policy implications.
2. Study Area, Data, and Methodology
2.1 Study Area
We selected nine cities from three countries to assess the urban expansion of major cities in East and South Asia (Figure 1). These nine cities can be classified into three different types according to their geographic locations, including coastal mega cities, inland mega cities, and major cities in the dryland region of East Asia. Their socio-economic profiles illustrated in Figure 2 show that these cities share certain common characteristics. We We selected nine cities from three countries to assess the urban expansion of major cities in East and South Asia (Figure 1). These nine cities can be classified into three different types according to their geographic locations, including coastal mega cities, inland mega cities, and major cities in the dryland region of East Asia. Their socio-economic profiles illustrated in Figure 2 show that these cities share certain common characteristics. We
Figure 1. Geographic locations of selected cities
Profile of selected ci>es
G 91 176 Chongqing
Urban built‐up area (square kilometer)
Figure 2. Socio-economic profile of selected cities 1
Coastal Megacities- Shanghai, Mumbai, and Hangzhou
Coastal cities in China generally have experienced rapid economic growth and started their unprecedented urbanization as early as in the 1980s. Similar to Chinese coastal cities, coastal cities in India, such as Mumbai and Calcutta, are also loci of urbanization as the country started its economic reform in the 1990s. Here we selected Shanghai and Mumbai as our case cities as they are the largest and the most vibrant economic centers
1 Note: The bubbles and the figures inside the bubbles represent the sizes of the population (10,000 persons) of the selected cities.
Sources: The urban built-up areas of all cities in 2010 are obtained from our own assessment based on the satellite imageries. For Chinese cities, i.e., Shanghai, Hangzhou, Chongqing, Urumqi, Lanzhou, Yinchuan, and Hohhot, the GDP per capita data is calculated based on 2010’s data from China Statistical Yearbook 2011 (total city) and the population data (city proper) is from the statistical yearbook of respective province or city in 2011 (Shanghai Statistical Yearbook, 2011; Zhejiang Statistical Yearbook, 2011; Chongqing Statistical Yearbook, 2011; Urumqi Statistical Yearbook, 2011; Gansu Statistical Yearbook, 2011; Ningxia Statistical Yearbook, 2011; Inner Mongolia Statistical Yearbook, 2011). For Urumqi, the city proper also deduct Dabancheng district due to its rural orientation. For Ulaanbaatar, population data of 2010 is from http://www.citypopulation.de/Mongolia.html ; GDP per capita data is Mongolia Statistical Yearbook 2011. For Mumbai, the population data of 2010 is from the United Nations. (United Nations, Department of Economic and Social Affairs, Population Division. 2012. World Urbanization Prospects, the 2011 Revision. http://esa.un.org/unpd/wup/GIS-Files/gis_1.htm ), GDP per capita data of 2010 is from http://www.docin.com/p- 195719235.html .
of their respective countries, the scale and speed of their urbanization reflect the trend of first-tier cities in Asia’s emerging countries. Our selection of Hangzhou is based on consideration that the city represents second-tier cities in the coast area, the coast provincial capitals, which have experienced similar urbanization frenzy during the period but have their own features due to their special status in their respective provinces and competitions among second-tier cities in the coastal area in China and, to a larger extent, in the region.
Shanghai (31°12’N 121°30’E) is located at the confluence of the Yangtze River and the East China Sea. Situated on an alluvial plain with an average elevation of 4 meters above
sea level, the city proper has a total area of 6340.5 km 2 . Located in the subtropical monsoon climate zone, Shanghai has four distinct seasons with hot, humid summers and
cool, wet winters. After China signed the Nanjing Treaty with the United Kingdom in 1842, Shanghai was forced to open as a port city for foreign trade and quickly grew from
a small fishing village to an international metropolis as well as China’s largest economic center and has remained the position since then. Despite falling behind other coastal cities in terms of speed of economic development at the beginning of the economic reform initiated in 1978, Shanghai eventually caught up with other coastal cities and resumed its leading economic position in China after the central government established the Pudong New Area in the 1990s. Shanghai’s city core is divided into two parts by the Huangpu River, a tributary of Yangtze river: (1) the west of the river that has been the city center since the nineteenth century, and (2) the east of the river, the newly developed Pudong New Area since the 1990s. As the largest economic center in China, Shanghai had a population of 23 million in 2010, with 14 million officially registered and 9 million as a floating migrant population. Among the 14 million official residents, 12.6 million are urban population.
Hangzhou (30°15’N 120°10’E) is located 180 km southwest of Shanghai in the Yangtze River Delta. Similar to Shanghai, Hangzhou has a humid subtropical climate with four
2 distinctive seasons. It has a total administrative area of 16596 km 2 , with 3068 km in the city proper and a total population of 4.35 million, with 73% as urban population in 2010.
Regarded as one of the most beautiful and historic cities in China, Hangzhou is built around the West Lake and has numerous tourist spots, such as the West Lake, Qiantang River, and the recently preserved Xixi Wetland. Benefiting from its proximity to Shanghai, Hangzhou accelerated its economic growth after the economic reform, particularly after 1992. Its GDP per capita is RMB 86,329 (equivalent to $12,751) in
2010, listed as the seventh of China’s 35 major cities 2 . Similarly to other large coastal cities in Asia, Hangzhou experienced rapid urban expansion and transformed itself from a
medium city to a mega-city. However, to distinguish itself from other large cities in the Yangtze River Delta area, especially Shanghai, Hangzhou has promoted itself as a city with good life quality through various planning efforts. Part of the planning efforts involve Hangzhou, as a pilot city selected by the central government, to test various land policies such as land banking and using land trusts and land bonds to finance a land reserve system (Zhu, 2007).
2 The 35 major cities include all provincial capitals plus Dalian, Ningbo, Xiamen, Qingdao, and Shenzhen, but not include Lhasa due to the lack of data.
Mumbai (18°58’N 72°49’E), bounded by the Arabian Sea to the west, is located at the mouth of the Ulhas River on the western coast of India. With an average elevation of 14
meters above the sea level, it has a total area of 603.4 km 2 . Mumbai has a tropical wet and dry climate with the southwest monsoon season from June to September. With 12
million population in 2011 (India Census), Mumbai is the largest city and the most important economic center in India; it generated 6.16% of India’s GDP in 2008. In 2010, its GDP per capita is $2800 (Rs 125000), more than two times of the national average ($1211, Rs 54527). With textile production and seaport activities as initial economic engines, the city has diversified its economy into sectors such as information technology, engineering, and health care. Especially, the city experienced the boom of its economy since the economic liberalization reform in 1991.
Inland Megacities-Chongqing
Chongqing (29°33’N 106°34’E) is located at the southeast part of Sichuan Basin and the upper reach of Yangtze River. With a total area of 82,401 km 2 including the city proper
and the counties under its administration, it is the largest direct-controlled municipality, (the area of city proper is 2735 km 2) , 2.39 times of the total area of Beijing, Tianjin, and
Shanghai, the other three directly-controlled municipalities. With the surrounding mountains to its north, east, and south, Chongqing is known as the “Mountain City.” Chongqing has a monsoon-influenced humid subtropical climate, with most of the year very humid. Along with Nanjing and Wuhai, Chongqing is known as one of the “Three Furnaces” of the Yangtze River” because of its very long, hot, and humid summer. It is also known as the “Fog City,” as on average, it has over 100 days of fog per year. Chongqing has been an important economic center and transportation hub of Southwest China. During the second Sino-Japanese War (1937-1945), many universities and factories relocated from the coastal area to Chongqing, the relative safe inland city and the wartime capital at the time, to continue the education and manufacturing production against the Japanese army. Due to the massive relocation, Chongqing quickly developed itself from an inland port to a heavily industrialized city despite the heavy bombing of the Japanese army during the period. The city was demoted to a sub-provincial city within Sichuan province in 1954, but in 1997 it was separated from Sichuan Province and made
a direct-controlled municipality. Due to its inland and remote location, Chongqing traditionally has been an important industrial base for weapons research, development, and manufacturing. However, the city has diversified its industrial base, especially since 1997, into sectors such as food, autos, chemicals, textiles, machinery and electronics. Currently it is the third largest center for motor vehicle production and the largest for motorcycles. Although by 2010 the city had GDP of $117 billion, its GDP per capita was only $3544 far lower than the national average of $10464. With a total population of 6.12 million in its city proper, about 4.25 million are urban population.
Major cities in the dryland region of East Asia
Located in the arid region of Northwest China and Mongolia, Urumqi, Lanzhou, Yinchuan, Hohhot are provincial capitals of Xinjiang Uyghur Autonomous Region,
Gansu Province, Ningxia Hui Autonomous Region, and the Inner Mongolia Autonomous Region, respectively; Ulaanbaatar is the capital city of Mongolia. Most of the cities are developed along rivers and are surrounded by mountains, as typical river valley cities in arid regions. The cities feature semi-arid climates, with low annual temperature and precipitation and long cold winters; while Yinchuan and Hohhot have hot and humid summers, Urumqi has hot, dry summers and Ulaanbaatar has brief and warm summers. As the provincial or national capitals, they play important roles in economic development, and political and cultural activities of their respective regions/countries. Having a population range between one and three million (Figure 2), they all have experienced rapid urbanization and vast urban sprawl in the recent decades. Characteristic land use changes are the expansion of urban land at the cost of agricultural land and the conversion of other types of lands for agricultural usage.
Urumqi (43°48’ N, 87°35’ E), literally “the beautiful pasture land” in the ancient Mongolian, is located at the north slope of Tianshan Mountain and the south edge of the Junggar Basin in the central north part of Xinjiang. Lying between mountains to the southwest and northeast, the city is composed of the Chaiwobao-Dabancheng Valley in the south and the alluvial plains of the Urumqi and Toutun rivers in the north. 2,500 km (1,400 miles) from the nearest coastline, Urumqi is also the most remote city from any sea in the world. Urumqi has a semi-arid climate, with precipitation of 286 millimeters (11.4 in) and 2523 sunshine hours in an average year. The hot summer (average temperature around 30°C in July) is in direct contrast with its cold winter (average temperature -7.4°C in January), with summer slightly wetter than winter. With a population of 2.29 million in its city proper in 2010, Urumqi has 49 ethnic groups, and the non-Han ethnicity population is 25% of the total population. Multiethnic groups live in compact, mixed communities consisting of primarily Uyghur, Han, Hui, Kazak, Mongolian, Kirgiz, and Xibe ethnicities. The city was once an important town on the northern route of the Silk Road, essential to Sino-foreign economic and cultural exchanges.
Lanzhou (36°02’N 103°48’ E) is located in the upper course of the Yellow River and bounded by mountains on the south and north sides. Due to its unique geography, Lanzhou has developed in a “Dumb-bell” shape. The distance between the west and the east of the city is about 35 km whereas the distances from the north to the south vary dramatically from 2 km to 8 km. In 2010, the city proper’s total population reached 1.3 million and 0.9 million of them are urban. Belonging to the middle temperate zone and with the average altitude 1520 m, Lanzhou has a moderate climate with annual average temperature of 11.2°C, notably without freezing winters or hot summers. Like other cities in Northwest of China, Lanzhou is dry with an annual precipitation of 327 mm, primarily occurring from June to September. Although daily temperature varies widely, it has plenty of sunshine with sunlight hours of 2446 and more than 180 frost-free days per year. As one of the oldest industrial bases of China, Lanzhou is the largest industrial city in the upper stream of the Yellow River and an important base of the raw materials industry in China’s West. At the intersection of inland China, Northwest of China, and Tibet Plateau, Lanzhou is the transportation center of northwestern China with four of China’s main railway lines and six national highways converging here.
Yinchuan (38°28’ N 106°16’ E) is located in the middle of the Yinchuan Plain with the Helan Mountain to its west and the Yellow River running through the city from southwest to northeast. With a desert climate and an annual average temperature of 9.0°C, Yinchuan’s winters are cold, windy, and dry whereas summers are hot and humid. Yinchuan has a long history of development, especially agricultural and commercial activities. Irrigation systems in Yinchuan were developed during the Han Dynasty to improve wheat and rice production (Chen and Gao, 2007). Yinchuan’s urbanization accelerated post 1949 when the P.R. China was established: the total population of the city proper grew from 200,000 in 1949 to nearly 1.99 million by 2010, with a non- agriculture population increasing from about 30,000 to 1.29 million. The city is also a center for the Muslim (Hui) minority people that account for 25% to 30% of Yinchuan’s population.
Hohhot (40°49’ N 111° 39’ E), the “green city” in Mongolian, is located on the Tumuochuan Plain, the south central part of Inner Mongolia Autonomous Region, surrounded by Daqing Mountain to its north and the Yellow River and Hetao Plateau to its south. Hohhot has a population of 1.21 million (0.92 million urban population) in its city proper in 2010. It has a cold semi-arid climate with a low annual average temperature of 6.7°C and the annual precipitation of 400mm. The city has long, cold and very dry winters and hot, somewhat humid summers, with strong winds especially in spring. Although most residents are Han (87.2%), Hohhot has a significant presence of ethnic minorities, especially Mongolian (8.6%) and Hui (1.6%). Founded by Mongol ruler Altan Khan in the late 16th century, the city has a rich cultural background. Serving as the region's administrative, economic, and cultural centre, it is also called the "Dairy Capital of China" due to two giant dairy producers headquartered in the city – Mengniu and Yili.
Ulaanbaatar (47°55’ N 106°55’ E), meaning "Red Hero" in Mongolian, is located in the north central part of Mongolia in a valley on the Tuul River at the foot of the mountain Bogd Khan Uul. Ulaanbaatar is the coldest national capital in the world, due to its high elevation, relatively high latitude location hundreds of kilometers from any coast, and the effects of the Siberian anticyclone. It has a monsoon-influenced, cold semi-arid climate featuring long, cold and dry winters and brief, warm summers. As the capital of the country, Ulaanbaatar is the largest city in Mongolia. Hosting one third of the total population of Mongolia (Bolormaa, 2010), its population reaches 1.15 million in 2010. Apart from the political importance of the Ulaanbaatar, the city contributes 48% of industrial output, 52% of construction, 41% of trade, 75% of hotel and restaurant services as well as 56% of transportation and communications in Mongolia (Herro et al, 2003). Ulaanbaatar’s urban development is concentrated along the Tuul River valley with an east-west built-up area of approximately 24 km long.
2.2 Data & Methodology
Urban growth indicators
In this paper, we adopt four sets of indicators to measure various aspects of the urban land use and population dynamics based on previous work conducted by other researchers (e.g., Huang et al, 2007; Kasanko et al, 2006; Schneider and Woodcock, 2008; Schwartz, 2010; Tsai, 2005) (Table 1). The first and second sets of indicators measure size and growth of urban built-up areas, the most fundamental aspect of urban land use, and their relation to non built-up areas. The third and fourth sets of indicators link population with land use, measuring population density in urban built-up area and total area, as well as how population occupies the available built-up space by examining the ratio of population growth to the growth of built-up areas in different periods.
Table 1. Urban land use indicators used for the selected cities
Category Indicator
Unit
Description
1. Built-up area 1.1 Built-up area
Km 2 built-up area in 1990, 2000, 2010 new urban built-up area during the periods
1.2 Growth in built-up area Km 2 1990-2000, 2000-2010 1.3 Percentage change,
rate of built-up area expansion for the annual percentage change
periods of 1990-2000, 2000-2010 2. Density of built-up area urban land as a percentage of all land in
2.1 Urban land density
2.2 Percentage change in change of urban land density for the periods urban land density
of 1990-2000, 2000-2010 3. Population density
urban population density in built-up area in 3.1 Urban population density
1990, 2000, 2010, = urban population/built- in urban land
person/Km 2 up area
3.2 total population density total population density in 1990, 2000, in total land
person/Km 2 2010, = total population/total land 4. Urban density population growth in contrast with urban
growth for the periods of 1990-2000, 2000- 4.1 Population growth in
2010, = change in population /change in contrast with urban growth
person/Km 2 built-up land
Satellite image data and processing
In this study, we mainly use urban built-up land and associated characteristics to assess urban expansion. The urban built-up land of our selected cities was derived from multi- source satellite imagery including Landsat 7 Enhanced Thematic Mapper plus (ETM+), Landsat 5 Thematic Mapper (TM), and SPOT5. Detailed information on all satellite images used is listed in Table 2. All level 1G Landsat images, downloaded from the USGS website, have been geometric corrected to the UTM coordinate system. After radiometric enhancement, one year SPOT 5 images of Shanghai, purchased from SPOT Image Corporation, were registered to the Shanghai ETM+ images, and resampled (RMSE <0.5 pixels) using the nearest neighbor algorithm to a nominal pixel size of 5m×5m. The administrative boundary datasets of China, India, and Mongolia were downloaded from the geographical information center of China and the GADM organization (Available at http://gadm.org), respectively. The Asia map is obtained from ShareGeo Open (Available at http://www.sharegeo.ac.uk/handle/10672/22 ).
To extract urban built-up land of our selected cities, we used an integrated approach combining supervised classification, unsupervised classification, and visual interpretation. The classification and post-classification enhancements were completed using Erdas Imagine 9.3. Except for the urban built-up land data of Shanghai in 2010, derived from SPOT5, all other data were extracted from Landsat.
For the coastal cities with low average elevation, such as Shanghai, Hangzhou and Mumbai, we followed the steps below. First, a supervised classification with maximum likelihood algorithm was performed to derive the primary urban built-up land data. Second aided by comparing with Google Earth and other high-resolution images, a visual interpretation classification was conducted to enhance the above classification result. Third a serial of post-classification processes were employed to refine the classification. We followed this general process for the other aforementioned cities. For the major cities in dryland East Asia, the terrain complexity made the task of automatically extracting the urban information difficult except for Yinchuan. Therefore, visual interpretation, in addition to supervised and unsupervised classification, was particularly important for identifying urbanized areas, including large townships or villages in suburban areas and the concrete surfaces in the cities. This additional step was completed for Urumqi, Lanzhou, Hohhot and Ulaanbaatar. We found that a machine-classified result resulted in lower accuracies. We conducted an accuracy assessment for the 2010 urban land maps, based on independent literature and Google Earth high-resolution images. We consider our classification sufficiently accurate with kappa coefficients ranging from 0.80 to 0.93 and producer and user accuracies of all cities over 0.9 except for Hangzhou’s producer accuracy (0.83) and Chongqing’s user accuracy (0.87) (Table 3). Yinchuan, as expected, has slight lower values of kappa coefficient and producer and user accuracies than most of the other cities.
Table 2. Description of satellite images of selected cities
2010 Sensor Path
Path Row Date
Urumqi TM
ETM+
TM
07/18/ TM
131 35 Lanzhou
35 TM
35 TM
07/01/ Yinchuan
TM 129
TM
TM
07/12/ Hohhot
126 32 2010 Ulaanbaatar
TM 126
TM
TM
TM 131
TM
TM
Note: *Shanghai 2010’s data is derived from the following SPOT images: SCENE 5 295-286 09/04/22 02:29:29 2 J SCENE 5 295-287 09/04/22 02:29:38 2 J SCENE 5 295-288 09/06/03 02:23:09 2 J SCENE 5 296-287 10/03/16 02:27:58 2 J SCENE 5 297-288 10/03/16 02:28:07 2 J The numbers after SCENE5 indicate the SPOT Scene ID in the format of the SPOT grid reference system (GRS)
Table 3. Accuracy assessment for urban built-up area of 2010
Other data, particularly economic statistics, and population are from data sources such as the China Statistical Yearbook, Statistical Yearbook of the provinces and the cities, and planning and government policy documents (NBS, 1991, 2001, & 2011; Shanghai Bureau of Statistics, 1991, 2001, & 2011; Hangzhou Bureau of Statistics, 1991, 2001, & 2011; Chongqing Bureau of Statistics, 1991, 2001, & 2011; Urumqi Bureau of Statistics, 1991, 2001, & 2011; Gansu Bureau of Statistics, 1991, 2001 & 2011; Ninxia Bureau of Statistics, 1991, 2001, & 2011; Inner Mongolia Bureau of Statistics, 1991, 2001, & 2011; Mongolia Bureau of Statistics, 2011).
CLUE-s Model & Logistic Regression
CLUE-s Model
We used the CLUE-s model for our urban land simulations. The Conversion of Land Use and its Effect (CLUE) was designed for the dynamic land use change simulations at the national and continental scale using the empirical analysis between the land use and its driving factors (Veldkamp and Fresco, 1996; Verburg et al.; 1999). Because of its inability to work with high-resolution data, the Conversion of Land Use and its Effect at Small regional extent model (CLUE-S) was developed, which incorporates the competition of spatial and temporal land use dynamic at multi scales, from 20 to 1000 m in case studies (Verburg et al.; 2002; Verburg and Veldkamp, 2004; Overmars et al.; 2007). The CLUE-S model has been widely and successfully applied in many regional researches, such as Europe (Erdogan et al.; 2011) and Asia (Liao et al.; 2010; Luo et al.; 2010; Zhang et al.; 2003; Verburg et al.; 2002). The CLUE-s model includes two We used the CLUE-s model for our urban land simulations. The Conversion of Land Use and its Effect (CLUE) was designed for the dynamic land use change simulations at the national and continental scale using the empirical analysis between the land use and its driving factors (Veldkamp and Fresco, 1996; Verburg et al.; 1999). Because of its inability to work with high-resolution data, the Conversion of Land Use and its Effect at Small regional extent model (CLUE-S) was developed, which incorporates the competition of spatial and temporal land use dynamic at multi scales, from 20 to 1000 m in case studies (Verburg et al.; 2002; Verburg and Veldkamp, 2004; Overmars et al.; 2007). The CLUE-S model has been widely and successfully applied in many regional researches, such as Europe (Erdogan et al.; 2011) and Asia (Liao et al.; 2010; Luo et al.; 2010; Zhang et al.; 2003; Verburg et al.; 2002). The CLUE-s model includes two
In order to use the CLUE-s model to simulate Shanghai’s urban land till 2020, we first obtained historic urban land maps and identified spatial determinants for urban land use changes. We derived urban land maps of Shanghai of 2000, 2005 and 2010 from satellite images (including TM, ETM+ and SPOT images) by the process described earlier. These maps allowed us to characterize the spatial pattern of Shanghai’s urbanization from 2000-2010. To be able to simulate the future land use change, we needed to identify the underlying driving forces, determined by both the human and the nature systems (Liu et al., 2007); we achieved this through logistic regression.
Logistic regression
As urban landscapes are affected by both human and nature systems (Liu et al.; 2007), it is of great importance to examine the relationships between land use and the driving factors. Logistic regression is used in the CLUE-s model to calculate the suitability of any location for certain land use types. The probability for converting location i into land type k can be calculated through a binomial logit model. The logistic regression model is denoted as follows:
Log (Pi/(1−Pi)) =β 0 + β 1 X 1i + β 2 X 2i +……+ β n X ni (1)
Where Pi is the probability of location i for the occurrence of the considered land use type k. Log (Pi/(1−Pi)), the log transformation of the ratio of the probability that conversion occurred (Pi) to the probability that conversion does not occur (1-Pi), can be expressed as a linear combination function of the explanatory variables X ni , the location factors. Using the actual land use patterns as dependent variable, the coefficients ( β n ) are estimated through the logistic regression. The value of Relative Operating Characteristics
(ROC) is used to validate the model performance (Pontius and Schneider, 2001).
We used 12 variables (Table 4), including several socioeconomic variables, geographical attributes and land use variables, to identify the most influential variables in the urbanization process. These variables were chosen based on literatures (Zhang et al.; 2011; Han 2009; Deng et al.; 2009). With the 2000 land use map as an initial map, we used these variables in the Clue-s model to calculate the suitability of each grid for urban land conversion thus simulating the urbanization of Shanghai until 2020.
Table 4. Details of variables used in the logistic regression
Meaning Notation
Unit
Generation Method
Source
Calculated the Euclidean Distance to Main
Own Roads
DMR
Km
distance to main roads for calculation each cell in ArcGIS 10.
Calculated the Euclidean Distance to CBD
DCBD
Km
distance to CBD for each Own calculation
cell in ArcGIS 10. Calculated the time cost
distance for each cell based Own Time Cost to CBD
TCOST
Hour
on 2000's road network data calculation in ArcGIS 10.
Industrial Land Percentage of industrial Own Density
land in 1km * 1km grid calculation Percentage of
IDEN
Percentage of agricultural Own Potential Industrial
PPI
Land land in 1km * 1km grid calculation Dummy variable for
Land Planning of LP
whether or not the cell falls Gong, 2007 1999-2020
n.a.
in the 2000's industrial land plan
SBS,2001 Product per Capita
Gross Domestic GDP
RMB 10,000 per
Survey
person
Foreign Direct Investment per
FDI
US$ 10,000 per
Capita Income
Population Density 2 PDEN 10,000 persons/km Survey
SBS,2001 Calculated based on census
Higher Education data. The percentage of HEDUR
SBS,2001 Percentage
people with college or higher education
Immigrant Rate IMGR
Survey
SBS,2001
Second, to examine future land use pattern in Shanghai from 2000-2020, we designed three different scenarios for our CLUE simulation, considering different urban growth rates and whether or not green land protection policy (Shanghai Bureau of Urban Planning, Land, and Resources, 2001 and 2002) is implemented (Table 5). These three scenarios are: (1) a base scenario that continues the historic trend of urban land conversion without any green land protection policy; (2) a decelerating scenario where Second, to examine future land use pattern in Shanghai from 2000-2020, we designed three different scenarios for our CLUE simulation, considering different urban growth rates and whether or not green land protection policy (Shanghai Bureau of Urban Planning, Land, and Resources, 2001 and 2002) is implemented (Table 5). These three scenarios are: (1) a base scenario that continues the historic trend of urban land conversion without any green land protection policy; (2) a decelerating scenario where
Table 5. Three scenarios for CLUE-s simulation
Scenario Name Description
Urbanization occurs at a linear rate following the historic trend without any green land protection
Base policy (the scenario is designed for comparison
purpose) Urbanization occurs at a decelerating rate (urban
Deceleration land conversion follows a nonlinear trend) and without any green land protection policy Urbanization occurs at a decelerating rate (urban
Restricted land conversion follows a nonlinear trend) and with green land protection policy
3. Findings
3a. Urban growth indicator and urban expansion
Our urban growth indicators illustrate that these nine cities have experienced extensive urbanization since the 1990s. Overall, costal cities, including Shanghai, Hangzhou and Mumbai have much larger urban built-up areas than other cities. It is also worth noting that while Chinese coastal cities slowed down their expansion rates, all other cities increased their expansion rates from 2000-2010, enhancing their urbanization processes. At the individual city level, the annual growth ratios of urban built-up area varied from 2% (Ulaanbaatar, 1990-2000) to 27% (Yinchuan, 2000-2010).
Figure 3. Expansion of urban built-up area of selected cities, 1990-2010
Urban built‐up area (km2 )
UlaanBaatar Hohhot Yinchuan Lanzhou Urumuqi Chongqing Mumbai Hangzhou Shanghai
Figure 4. Urban built-up area of selected cities, 1990, 2000, and 2010
Urban land density
UlaanBaatar Hohhot Yinchuan Lanzhou Urumuqi Chongqing Mumbai Hangzhou Shanghai
Figure 5. Urban land density (the percentage urban built-up area to the total area),
1990, 2000, and 2010
Urban popula>on density (person/km2)
UlaanBaatar Hohhot Yinchuan Lanzhou Urumuqi Chongqing Mumbai Hangzhou Shanghai
Figure 6a. Urban population density, 1990, 2000, and 2010
Total popula>on density (person/km2)
UlaanBaatar Hohhot Yinchuan Lanzhou Urumuqi Chongqing Mumbai Hangzhou Shanghai
Figure 6b. Total population density, 1990, 2000, and 2010
Urban density change
UlaanBaatar Hohhot Yinchuan Lanzhou Urumuqi
2000‐2010 Chongqing
1990‐2000 Mumbai Hangzhou Shanghai
Figure 7. Urban density change, 1990, 2000, and 2010
Urban land density further confirmed the urbanizing trend of the Chinese coastal cities. While Shanghai and Hangzhou have 44% and 27% urban land density respectively, all other cities have less than 20% urban land density. Among them, Ulaanbaatar and Chongqing have the lowest urban land density, 3% and 5%, respectively.
Urban population density indicated that Mumbai, Chongqing, and Lanzhou have much higher urban population density than the other cities, over 11,000 person/Km 2 ,
significantly more than Shanghai or Lanzhou. For total population density, Mumbai had the greatest density with 4460 person/Km 2 and Shanghai held a close second with
2227 person/Km 2 .
Excepting Ulaanbaator in 1990-2000, urban density change, the ratio of urban population change in contrast of urban built-up land change, are below 100%, indicating that urban population did not increase as fast as urban built-up land, i.e., urbanization in our selected cities happened physically, reflected by land use change, whereas the urban population has followed the physical urbanization. It is interesting to note that some cities, such as Shanghai, Hangzhou, Chongqing, and Urumqi, have increased their urban density change in the 2000-2010 than that of 1990-2000, whereas others have the reverse trend, which indicates an accelerating sprawl pattern of urban expansion.
Coastal Mega cities
Among all cities, Shanghai has the largest urban built-up area and increased its urban built-up area more than five times from 1990 to 2010, reaching a total urban built-up area
of 2815 Km 2 , far above the other selected cities. For each of 1990-2000 and 2000-2010, the city added over 1000 Km 2 to its urban built-up land. While before 1990 the city of 2815 Km 2 , far above the other selected cities. For each of 1990-2000 and 2000-2010, the city added over 1000 Km 2 to its urban built-up land. While before 1990 the city
increased by 45.2% to 2461 km 2 from 2000 to 2005, it only increased 18.3% from 2005 to 2010 (Figure 8).
Figure 8. Expansion of urban built-up area, Shanghai, 2000-2010
Our investigation indicates that Hangzhou expanded urban built-up area in its city proper
2 over 7 times in 20 years, from 118 km 2 in 1991 to 841 km in 2010, with a much faster rate in the second decade. Although the newly added urban built-up area is mainly
located in the north of Qiantang River around the main city area, the area of south of Qiantang River also experienced fast growth. Hangzhou’s urban expansion reveals a close relationship between newly developed lands with the locations of original towns, such as Xiaoshan in the south of Qiantang River, and Linping in the northeast of main city.
The urban expansion in Mumbai has accelerated over the past two decades, as its urban built-up area expanded 2.5 times, mainly surround the Bombay Bay, from 263 km 2 in
1992 to 666 km 2 in 2011. During these two decades, the relative expansion rate was evidently higher in the east of Bombay Bay (New Mumbai and Nawa Sheva) than the
west side (Greater Mumbai), which has revealed that most new urban construction land in Mumbai was promoted by the development of the satellite city (New Mumbai), independent of the old urban district. As an island city surrounded by water, Mumbai suffers serious space restrictions, therefore continuous urban expansion in Mumbai is suppressed, in contrast, leapfrogging urban development dominated during these two decades, at significantly higher rates.
In-land Mega City – Chongqing
The built-up area in the main city of Chongqing has experienced rapid growth since
2 1988, expanding urban built-up area 4.3 times, from 67 km 2 in 1988 to 296 km in 2010, mainly towards north and south. At earlier periods, the core urban area was built at the
intersection of the internal valleys of the Yangtze and Jialing Rivers and near the edges of nearby mountains. However, Chongqing’s urban spatial pattern has undergone rapid transformation from concentration to diffusion during the period from 2001 to 2010 and has been dominated by polycentric development and outer area urban growth.
Dryland East Asia
Urumqi has experienced fast urban land expansion from 1990, its urban built-up area
expanded from 212 Km 2 to 289 Km in 2000, but sharply increased to 412 Km in 2010.
A slightly rotated (towards west) T shape features the current urban built-up of Urumqi, with wide area in the north and narrow stripe in the southern part, respectively. Urumqi’s urban expansion has been affected by natural factors, administrative boundaries, and transportation factors. As a river valley city, the initial population of the city settled along the old river bed (Hetan, in Chinese) that runs southeast to northwest but later development expanded the city to the northwest and southeast along the valley. Some of the oasis ecosystems, which were cultivated during and shortly after the civil war that ended in 1949 (officially), were converted into impervious urban lands. In addition to the natural factors, the administrative boundary of Urumqi city also constrained urban expansion as Miquan, and Changji are closely bordered Urumqi. Finally, the urban expansion in the late 1990s along the major transportation routes was believed to be a result of the rapid development of public transportation and increased use of private vehicles.
The urbanized area of Lanzhou grew over two times from 1990 to 2010, mostly occurring between the existing urban built-up areas and at the northern bank of the Yellow River. The city experienced fast expansion in the last decade. Similar to Urumqi, Lanzhou’s expansion is constrained by its geography. It is worth mentioning that in the recent decade, the city has leapfrogged its surrounding mountains and established new industrial zones as well as residential communities outside the original core.
Among the five cities of dryland East Asia, Yinchuan experienced the most dramatic
2 expansion. The urban area only occupied 47 km 2 in 1990, but grew to 65 km in 2000, then skyrocketed to 240 km 2 in 2010. Two of its major urban districts, Xixia and
Xinqing, separated by farmland in 1990, have become connected over time.
2 Hohhot increased its urbanized area from 89 Km 2 in 1990 to 117 km in 2000, then to 229 km 2 in 2010. The rate of urban expansion in Hohhot increased after 2000 with an annual area increase of close to 10 km 2 /year. While development occurred mostly along the
eastern urban fringe area from 1990 to 2000, urban built-up lands were added to the rest of the urban periphery area from 2000 to 2010.
Although Ulaanbaatar fell behind the other major cities in China’s Northwest in overall urbanized rates and the speed of urban expansion, it nevertheless expanded noticeably
2 from 82 Km 2 in 1990 to 142 Km in 2010, especially after 2000. Examining high- resolution imagery from the Google Earth, we found that the settlement areas in the 2 from 82 Km 2 in 1990 to 142 Km in 2010, especially after 2000. Examining high- resolution imagery from the Google Earth, we found that the settlement areas in the
3b. Spatial determinants and simulation of Shanghai
Through logistic regression, we identified four variables as influential variables (significance at the 0.05 level) (Figure 9). They are: the distance to main roads (DMR), industrial land density (IDEN), population density (PDEN), and land use planning (LP), which are also identified by others as significant factors for urban land changes in Shanghai (Deng et al. 2005; Han 2009; Zhang et al. 2011;). The four significant driving factors have different impacts on the presence of urban lands. For the distance to main roads, every 1 km further decreases the possibility of urban land presence by 7.01%. The other 3 variables make positive contributions to the urban lands’ presence. Every 1% of
increase in the industrial land density and every 1000 persons per km 2 increase in population density increased the presence of urban lands by 6.77% and 11.67%,
respectively. The influence of land planning is the most significant among the four factors. The possibilities of urban lands’ presence are 55% more in the places with urban land plans. With 2000’s land use map as an initial map, we then used these four variables in the Clue-s model to calculate the suitability of each grid for urban land conversion thus simulating the urbanization of Shanghai until 2020.
Figure 9. Four significant variables identified as spatial determinants for Shanghai’s
urban land conversion
Our Clue-s model simulation indicates a lasting trend of urbanization in Shanghai from 2000 to 2020 (Figure 10). Generally, the urbanization happens around the city core and existing urban areas in suburbs. Moreover, most of the urbanization takes place along the axis from northwest to southeast, modifying the original city’s shape that developed along an axis from northeast to southwest. Our simulation illustrates distinct spatial Our Clue-s model simulation indicates a lasting trend of urbanization in Shanghai from 2000 to 2020 (Figure 10). Generally, the urbanization happens around the city core and existing urban areas in suburbs. Moreover, most of the urbanization takes place along the axis from northwest to southeast, modifying the original city’s shape that developed along an axis from northeast to southwest. Our simulation illustrates distinct spatial
Figure 10. Clue-s simulation of three scenarios
4. Discussion
4.1 Characteristics of urbanization in Asia Coastal mega cities
Our findings first confirm that coastal mega cities have led urbanization over other types of cities of this study, measured by indicators related to urban built-up land and urban population. Coastal cities, including Shanghai, Hangzhou and Mumbai, have much larger urban built-up areas than other cities, but they have decreased their expansion rates in comparison with the other cities. Further, the large urban density of Chinese coastal cities indicates a limited potential for further urbanization in comparison to the other cities. All coastal cities measured have very large base urban populations. Conversely, current urban population density is not as high, especially as compared with Chongqing or Lanzhou where the urban built-up potential area geographiclly limited leading to extremely high urban population density. It is also interesting that Chinese coastal cities have significantly higher urban density change in the 2000-2010 period than 1990-2000 period, indicating that the increase of urban population is catching up to the past production of converted urban built-up area.
Another distinct characteristic of urbanization of coastal mega cities is the polycentric urban development pattern. Past research on Hangzhou (Yue et al, 2010) and others (Wu, 1998; Han, 2005) have indicated that urban land conversion mainly occurred near the city core or sub-centers, leading to a polycentric pattern of development, rather than Another distinct characteristic of urbanization of coastal mega cities is the polycentric urban development pattern. Past research on Hangzhou (Yue et al, 2010) and others (Wu, 1998; Han, 2005) have indicated that urban land conversion mainly occurred near the city core or sub-centers, leading to a polycentric pattern of development, rather than
Inland mega cities (case of Chongqing)
Chongqing’s urban development is severely constrained by its geographic location as a peninsular city formed by two rivers and surrounded by mountains. For a long time, Chongqing had a small and concentrated urban built-up area built on limited hilly areas of the intersected valleys of the Yangtze and Jialing Rivers. Therefore, to counter its quite large urban population base, Chongqing’s urban population density was extremely high, especially in the 1990, comparable to Mumbai’s level. Nevertheless, the recent decade witnessed Chongqing’s development beyond the surrounding mountains, with bridges and tunnels making land beyond the Yuzhong Peninsular available for urbanization. The relatively low urban land density also indicates that unlike Shanghai, Chongqing can still convert a large amount of non-urban land to urban built-up land, thus facilitating urbanization far beyond its existing urban core.
Chongqing’s urbanization has been closely associated with its strategy of transport- oriented urban development. In recent years, Chongqing has undergone remarkable construction of its infrastructure, as its hilly topography requires higher unit investment. An impressive project initiated in 2003, named the eight-hour-Chongqing traffic project, has set up a transport network that allows one commuting from the urban core to the rural counties by expressways within eight hours. Since 2008, a new traffic project has been under construction, targeting at connecting multiple urban clusters with commuting time less than half an hour, using inner and outer ring expressways, light rails and subways, bridges over the rivers, and tunnels across the mountains.
Major cities in dryland East Asia
Through literature review on major cities in dryland East Asia, we identified three common trends of urbanization and land conversion: (1) the increase of urban land at the cost of other land, (2) a typical land conversion pattern of “grass land => agricultural land => urban land,” and (3) the reverse trend to increase forest land mainly through government intervention.