Scholarly article on topic 'The Analysis of Space Use around Shanghai Metro Stations Using Dynamic Data from Mobile Applications'

The Analysis of Space Use around Shanghai Metro Stations Using Dynamic Data from Mobile Applications Academic research paper on "Social and economic geography"

Share paper
Academic journal
Transportation Research Procedia
OECD Field of science
{"Shanghai metro stations" / "data from mobile applications" / "space use" / "transit-oriented development" / "Baidu Heat Map" / POI}

Abstract of research paper on Social and economic geography, author of scientific article — Zhongnan Ye, Yihui Chen, Li Zhang

Abstract The present study employs dynamic data from mobile applications such as Baidu Heat Map and POI to quantify the space use situation around metro stations in central city of Shanghai. A model is established on this basis to describe the relationship between space use situation and other characteristics of station areas. The results indicate that the intensity and diversity around the metro stations are not always in accordance with high floor area ratio and mixed land use, they also affected by other characteristics of the station areas such as location, bus stop density and urban morphology.

Academic research paper on topic "The Analysis of Space Use around Shanghai Metro Stations Using Dynamic Data from Mobile Applications"

Available online at



Transportation Research Procedía 25C (2017) 3151-3164 ■ ■ w «J «J

World Conference on Transport Research - WCTR 2016 Shanghai. 10-15 July 2016

The Analysis of Space Use around Shanghai Metro Stations Using Dynamic Data from Mobile Applications

Zhongnan Ye a,b*, Yihui Chen c, Li Zhangd

aTongji University, 1239 Siping Rd., Shanghai 200092, China bEast China Architecture Design & Research Institute, 151 Hankou Rd., Shanghai 200002, China cTongji Urban Planning & Design Institute, 1111 North Zhongshan Rd., Shanghai 200092, China dTianhua Urban Planning & Design Ltd., North Caoxi Road, Shanghai 200030, China


The present study employs dynamic data from mobile applications such as Baidu Heat Map and POI to quantify the space use situation around metro stations in central city of Shanghai. A model is established on this basis to describe the relationship between space use situation and other characteristics of station areas. The results indicate that the intensity and diversity around the metro stations are not always in accordance with high floor area ratio and mixed land use, they also affected by other characteristics of the station areas such as location, bus stop density and urban morphology.

© 220177 The Author's. Published by Elsevier B.V.


Keywords: Shanghai metro stations; data from moMe appHcations; space use; transit-oriented ^ve^ment; Baidu Heat Map; pO1

1. Background

ín recent; years, urban raü transh system has been an m^rtant; strate^c sotown to soWe the growmg proMems of traffic congestion and a useful tool to acMeve sustemaMe development m urban fransptortation due to ite mass-frans^ fos^s^ed sale ptunctaah envhonmenteh energy-an^an^sav^ featores. Also, h has a s^mficart impact

on urban land along the railway. Therefore, it is of great value to build up a coordinative relationship between urban land and transit infrastructure and to improve sustainable urban development through research on features along the

* Corresponding author. Ye Zhongnan; Tel.:+86-139-1805-7139; fax:+86-021-63330213. E-mail address: 179459°

2352-1465 © 2017 The Authors. Published by Elsevier B.V.

Peer-review under responsibility of WORLD CONFERENCE ON TRANSPORT RESEARCH SOCIETY. 10.1016/j.trpro.2017.05.353

railway, such as land use, spatial layout and so on. Within such context, tracking and analyzing the development and operation status of surrounding areas of metro stations with effective quantification methods will help us to learn and handle the pattern and law of land development and spatial form around metro stations. It will also help to establish respective performance evaluation system and provide guidance for development around new metro stations (Ye, 2014).

Research on the development around metro stations started in 1960s internationally, most of which came from Britain, the United States, Japan and Singapore. A great deal of research proved that the development of public transit and the use of urban land space are inseparate. TOD (transit-oriented development) is the most commonly recognized and accepted development form nowadays (Arefeh Nasri et al.,2014). Brought up in the United States since 1990s, TOD has always been a leading theory to guide sustainable land use development. With many years of research and practice of TOD theory, the 3D principles are formed including high density in development (Density), diversity in land use (Diversity) and good design (Design) (Calthorpe, P., 1993). The three principles actually reflect the goal of TOD, which is to establish favorable environment, high intensity of land use and multi forms of industry around transit stations.

Quantitative analysis of exiting research mainly focused on density and diversity. Although analytical methods and mathematical models were different in these studies, development intensity, usually presented by floor area ratio (FAR), was used as the leading indicator of density, and land use types the leading indicator of diversity.

However, floor area ratio and land use types are two static indicators of the planning and construction status of areas around transit stations. They can affect how people use the space, but cannot indicate the actual use of space. The density and diversity that TOD promotes are for high intensity of activities and busy businesses, rather than unoccupied office buildings, empty residential blocks and dull commercial facilities. Therefore, this study tried to better understand the spatial characteristics and patterns of surroundings of metro stations in a metropolis using more dynamic and accurate data from mobile applications (APP) to describe density and diversity.

2. Research methodology

With the arrival of the era of mobile internet and big data, new forms of data has been widely exploited and applied in urban research, providing a new perspective for researchers either in time or in space dimensions. This study tried to analyze the space use in surrounding areas of 340 metro stations within Shanghai central city based on two types of data from mobile applications, Baidu Heat Map and POIs.

Data from Baidu Heat Map was vectorized, geocoded, and assigned to demonstrate the space use intensity of metro station areas. POI data went through the process of standardization, main component analysis and geocoding to describe the diversity level of business forms. Based on the two types of data, this study designed TOD index to directly reflect the space use characters of metro station areas.

To further explore the factors that affect the space use around metro stations, this study followed the mixed effect regression modeling approach. The main independent variables of the model included conditions of the metro station itself and its surrounding built environment (Formula 1). In this model, floor area ratio and land use composition were also considered as influencing factors, thus their real impact on the space use in metro station areas would be explored.

C = aX + pu + £ (Formula 1)

• C is a TOD index of surrounding areas of a metro station;

• X is the matrix composed of variables that indicate the property of the station or its surrounding environment;

• U is interactive variables matrix;

• a and ¡3 are matrices composed of coefficients of X and U

• e is the residual error.

3. Data description

3.1. Heat Map data

Baidu Heat Map is an attached function to the mobile application Baidu Map, which offers visual display of urban population aggregation using different color blocks. The original intention was to help people to avoid busy scenic areas while travelling. In the mean time, the big data it presents is a great convenience to conduct research on temporal and spatial distribution of urban population.

The data of Baidu Heat Map is based on the real time positions collected from the mobile application of Baidu Company. Judging from its coverage, it is not a full sample data so that it is unable to reflect the actual number of population distribution. Yet the data is gathered from over two hundred million users, which makes it quite effective when indicating the level of urban population aggregation.

Fig. 1. interface of Baidu Heat Map.

As required by the research, raw data went through the process of vectorization and geocoding (Fig.2). Also, heat degrees 1 to 7 were assigned to different color zones. The higher the degree is, the higher the population density is, and vice versa. It should be noted that certain extent of inaccuracy may exist when replacing population distribution data with any kind of big data (Chaogui Kang et al., 2012). Baidu Heat Map can only approximately show the trend of population distribution geographically even though it is based on position data acquired from hundreds of millions of Baidu mobile users.

Fig. 2. vectorized and variable-assigned Baidu Heat Map data.

Considering the fact that citizen activity usually presents periodic variation on a weekly basis, and difference shows between population distribution on weekends and that on weekdays, the study utilized a self-developed program to intercept data at hourly intervals from Baidu Heat Map of Shanghai central city for a week (from May 22nd to May 28th 2014). 108 heat maps were snapped in total and heat data of areas around metro stations was extracted (Fig.3). Data of weekdays (from Monday to Friday) and that of weekend (Saturday and Sunday) were respectively investigated during the process.

Fig. 3. part of the heat map data around stations of Line 10.

According to Peter Calthorpe's definition of TOD (1993), the radius of the influenced area from the transit station should be about 400 meters (1/4 mile). In recent years, many scholars believed this range should be far more than 400 meters since rail transit has the largest capacity in public transport. For instance, in the analysis of the impact of New Shuttle (Ina Line) on the land price along the route in Saitama, Japan, the impact was within 2000 meters from a transit station. In addition, Taojun Feng (2009) established a mathematical model to estimate the influencing range of transit station and Shijing Zhou (2012) conducted a theoretical calculation and a case study of Hong Kong, and concluded that the core development is within 300 to 350 meters from a transit station and secondary development should be within 660 to 730 meters.

The study considered how the data listed above adapted to the case of Shanghai, and extracted data within 200 meters, 500 meters, and 800 meters from each metro station. 57120 pieces of basic data were acquired in total (Fig.4) mainly including names of the metro stations, time of interception, radius, areas of all levels of heat zones, and etc.

Fig. 4. part of the data collected from stations of Line 10.

3.2. POI data

A POI(Point of Interest) is a point of information on a digital map. Each POI contains basic information like name, address, latitude and longitude, and extended information required by the map user. A large quantity of POI lays the key foundation of a digital map. POI data this study used came from the mobile application Tencent Map (Fig. 5).

, - 1 & ' «aiiiB

Fig. 5. points of interest on Tencent Map.

The research utilized a self-developed program to obtain POI data of over 40 types of facilities within 200 meters, 500 meters and 800 meters from each metro station in Shanghai central city (Tab.1). And 293908 pieces of data were cleansed and categorized by their labels and point properties. The "park" and "green space" categories were merged, "hotels" and "inns" were merged; unqualified points like "Gate 1, Heping Park", "automotive parts

mart" were deleted. Facilities were further divided into four categories, commercial and recreational, transportation, medical and education, parks and open space. Correlation analysis was conducted among the facilities that provide similar services and only one item in each category with high degree of correlation was retained. For instance, restaurants and bars, or movie theatres and KTVs, offer services alike and have high level of correlation statistically (Tab.2), thus restaurants and movie theatres were chosen as the representatives of dining and recreational facilities. In the end, 11 types of facilities including restaurant, hotel, supermarket, bank, movie theatre, bus stop, parking lot, primary school, high school, hospital, park were selected as the main objects of observation and investigation (Tab.3).

Table 1. Part of the digital map POI data.

No. Name of POI Category Latitude Longitude

SH-CN-0114 Shanghai China Travel International Ltd. life service-travel agency 121.4242 31.1947

SH-CN-0321 Shengkai Stationery Shilong Store shopping-stationery 121.4433 31.1568

SH-CN-0394 Xingsheng Fruit Mart shopping-market-fruit market 121.4439 31.1568

SH-CN-0914 Zhenghua Self-service Store shopping-franchised store-liquor and cigarette 121.4439 31.1569

SH-CN-1011 Honghuantian Hunan Restaurant catering service-Chinese restaurant- Hunan cuisine 121.4440 31.1568

SH-CN-1205 Xiangdong Graphic Production life service 121.4437 31.1570

SH-CN-1768 Henghui Model Shop shopping-franchised store 121.4432 31.1576

Table 2. Part of the correlation coefficients of facilities with similar services.

Dining Financial Recreational Life

Restaurant Bar Cafe Bank Credit cooperative Movie theatre KTV Billiards room Supermarket Convenience store Food market

Restaurant 1.00 0.79 0.83 0.24 0.19 0.56 0.50 0.41 0.61 0.59 0.47

Bar 0.79 1.00 0.89 0.33 0.29 0.67 0.61 0.52 0.55 0.54 0.36

Cafe 0.83 0.89 1.00 0.34 0.23 0.66 0.59 0.55 0.49 0.49 0.42

Bank 0.24 0.33 0.34 1.00 0.85 0.40 0.52 0.49 0.50 0.50 0.43

Credit cooperative 0.19 0.29 0.23 0.85 1.00 0.41 0.48 0.46 0.47 0.50 0.39

Movie theatre 0.56 0.67 0.66 0.40 0.41 1.00 0.91 0.83 0.60 0.76 0.29

KTV 0.50 0.61 0.59 0.52 0.48 0.91 1.00 0.84 0.60 0.77 0.27

Billiards room 0.41 0.52 0.55 0.49 0.46 0.83 0.84 1.00 0.55 0.63 0.40

Supermarket 0.61 0.55 0.49 0.50 0.47 0.60 0.60 0.55 1.00 0.87 0.79

Convenience store 0.59 0.54 0.49 0.50 0.50 0.76 0.77 0.63 0.87 1.00 0.84

Food market 0.47 0.36 0.42 0.43 0.39 0.29 0.27 0.40 0.79 0.84 1.00

Table 3. Data selection POI facilities.

Category of facilities Name of facilities

Commercial and Recreational Restaurant, hotel, supermarket, bank, movie theatre

Transportation Bus stop, parking lot

Medical and Education Primary school, high school, hospital

Parks and Open Space Park

4. Results and discussion

4.1. Space use intensity

To more intuitively and quantitatively demonstrate the population aggregation status in a certain range from the metro station, the study designed a space use intensity coefficient Q based on the need of research and the data's own characteristics, the calculation of which is shown in formula 2 as below. Space use intensity coefficient Q indicates a weighted result of areas of zones of different heat degrees ranging from 1 to 7 with higher number representing the higher use intensity.

(Formula 2)

Qrt is the space use intensity coefficient of the area within the radius r from the station at time t; n is the heat degree from Baidu Heat Map with n=1-7; Sn is the area of the zone of heat degree n within the scope of study; Sr is the area of the land within the radius r from the station.

A general result of how space is used around metro stations in Shanghai was obtained after averaging the data from workdays and weekend (Fig.6).

••VVrV'-'v y

; v. >r*


Fig. 6. space use intensity around metro stations in Shanghai.

4.2. Diversity of facilities

Two factors affected the diversity of facilities, the absolute quantity of facilities, and the proportion of their combination of the complexity and quantity. From POI statistics of 11 types of facilities, substantial deviation appears in absolute quantity of different facilities, among which restaurant significantly outnumbers other categories. To avoid the interference caused by variation in quantity due to the facility's own feature, the study used language R to standardize the data and results were scored as the evaluation basis of diversity. Principal component of the data was analyzed and normalized, thus diversity of facilities was scored within 500 meters from each metro station in Shanghai central city (Tab.5, Fig.7). Same results could be obtained within 600 meters, 700 meters, 800 meters from the station.

Table 4. Part of the standardized scores of facilities within 500 meters from each station.

Restau- Hotel Super- Bank Movie Bus Parking Primary High Hospital Park

rant market theatre stop lot school school

Xinjiangwancheng 0.65 0.71 0.78 0.72 0.71 0.84 0.82 0.77 0.74 0.74 0.72

East Yingao Road 0.75 0.71 0.78 0.72 0.71 0.77 0.79 0.77 0.74 0.74 0.88

Yanchang Road 0.99 0.73 1.26 0.79 0.72 0.94 1.31 0.92 0.84 0.94 0.88

Jingan Temple 3.01 0.92 1.80 1.72 1.91 1.94 2.80 0.97 0.74 0.79 0.62

South Shaanxi Road 2.89 0.84 2.78 1.01 2.12 1.79 2.39 0.91 0.86 0.74 0.73

Century Avenue 1.88 1.42 1.53 1.28 1.23 2.43 2.29 0.90 0.74 0.76 0.72

Yangsi 0.70 0.65 0.87 0.73 0.71 1.09 0.87 1.11 0.77 0.79 0.71

Tongji University 0.84 0.79 0.87 0.80 0.76 0.87 1.03 1.10 0.84 0.74 0.72

Table 5. Part of the diversity scores of facilities within 500 meters from each station. Name of the station Diversity Score

Shenjiang Road 0.01

South Yanggao Road 0.10

North Yanggao Road 0.11

Shanghai University 0.13

Changzhong Road 0.13

Shanghai Zoo 0.14

Deping road 0.17

Huamu Road 0.18

Yuanshen Sports Center 0.30

Jiangpu Road 0.36

Fengqiao road 0.36

South Xizang Road 0.45

Chifeng Road 0.47

Dalian Road 0.58

South Shaanxi Road 0.65

Xintiandi 0.65

North Sichuan Road 0.77

South Huangpi Road 0.94

East Nanjing Road 0.98

Fig.7. diversity level of facilities around each station.

4.3. TOD index

To better show if the space use around metro stations corresponded with the density and diversity principles of TOD, the study established TOD index C (formula 3) based on space use intensity Q and diversity of facilities D. and TOD index around each station was calculated (figure 8, table 6).

C = a%Qr +ß%Dr


C is the TOD index around each station

Qr is the space use intensity within the radius r from the station. Dr is the diversity of facilities within the radius r from the station. R is the radius of the scope of study.

a and p are index weights, determined through Analytic Hierarchy Process.

Fig.8. TOD index around each station.

Table 6. Part of the TOD indexes around each station.

Line Name of the station TOD index

1 South Huangpi Road 8.40

2 East Nanjing Road 8.08

9 Xujiahui 7.82

2 Jingan Temple 7.75

8 Dashijie 7.57

9 Shangcheng Road 7.50

10 Jiangwan Stadium 7.46

2 West Nanjing Road 7.41

7 Changshou Road 7.05

10 Yuyuan Garden 7.04

2 Lujiazui 7.02

1 Shanghai Stadium 6.99

1 South Shaanxi Road 6.90

3 Zhanghuabang 1.11

3 Songfa Road 1.10

12 Fuxing Island 1.09

10 Sanmen Road 1.09

7 Dachangzhen 1.07

11 Longyao Road 1.07

7 Nanchen Road 1.07

10 East Yingao Road 1.05

10 Xinj iangwancheng 1.05

12 Shenjiang Road 1.04

6 Wuzhou Avenue 1.03

7 Qihua road 1.01

6 Oriental Sports Center 1.00

7 Houtan 1.00

4.4. Regression analysis of influencing factors

The results show the relationship between the TOD index of the station and conditions and surroundings of the station itself (table 7). Three main variables that indicate the feature of the station are number of lines at the station, weighted distance from the station to city center and sub-center and regional metro station density. Four indexes reflecting the built environment are floor area ratio, land use mix (entropy), average block size and bus stop density.

Table 7. Result for the mixed-effect regression model.

Variables Coeffcient Standard error p-value

Constant 1.59 0.025 0.000

Station property variables

Number of lines transferring at the station 0.35 0.0011 0.000

Weighted distance from the station to city center and sub-center -0.27 0.006 0.000

Metro station density in 2km -0.19 0.002 0.000

Built environment variables

Floor area ratio 0.28 0.024 0.077

Land use mix (entropy) 0.12 0.003 0.014

Average block size -0.28 0.000 0.000

Bus Stop density 0.33 0.017 0.009

As it is shown in table 7 , number of lines at the station is positively correlated to TOD index, meaning that it would be easier to form TOD around transfer stations. Yet weighted distance from the station to city center and subcenter is negatively correlated, which means TOD tends to be formed around stations with better locations. For remote stations, very few of them can be surrounded by community environment of high density and diversity. Regional metro station density is also negatively correlated, which indicates that eminent TOD can hardly be formed in regions with dense stations due to the competition among facilities around them. In four indexes reflecting the built environment, floor area ratio, land use mix (entropy), and bus stop density are positively correlated to TOD index, among which floor area ratio and land use mix has less influence on the density and diversity of areas around stations than expected while in fact bus stop density is more influential. In addition, average block size is negatively correlated which means that the higher the road network density is, the higher the accessibility is, and the better for TOD to be formed.

From the results above, it can be seen that proposed high floor area ratio and land use mix were unable to guarantee the transit-oriented development that stands for the realization of high intensity of activities and diversity of facilities around metro stations. On the contrary, without research of location, traffic environment and spatial features of the station itself, blindly plannings of high floor area ratio and land use mix are inconsistent with the development needs of the station areas and would cause a waste in resources.

5. Conclusion

With traffic problems getting worse day by day, rail transit is more and more accepted and promoted in major cities in the world and urban land development around rail transit stations is getting rising attention. The study believes whether the development around stations is reasonable or not depends on how people actually use the space, thus principles of TOD, like density and diversity, should be represented by the intensity of actual use of space and the variety of its facilities. Based on such knowledge, the study took full advantage of big data resources, traditional data like floor area ratio and land use composition were replaced by Baidu Heat Map and POI information provided by two mobile applications. TOD index was established under such ground to truly describe the development status of land around stations by investigating the actual conditions of how space is used. In the end, the study established a mixed effect regression model to further discuss factors that affect the space use around rail transit stations.

The results showed that space use around rail transit stations were affected by various factors besides floor area ratio and land use composition. Noticeable influences include the location of the station, the number of lines being transferred at the stop, block sizes, density of stations, etc. Therefore, when planning areas around rail transit stations, the characters of the station itself should be investigated to guarantee its potential of high-density and diversified development. With proposed TOD area, high level of floor area ratio and land use complexity should be applied while density of road network and public transit system should also be taken into consideration.

In addition, as an attempt to conduct urban research using big data, this study explored analytical methods of internet big data from mobile applications and showed its merits comparing to traditional data resources, which revealed great potential in research of urban land and space.


This research is supported by China Intelligent Urbanization Co-creation Centre for High Density Region dominated by Tongji University. The author would like to thank Professor Zhiqiang Wu from Tongji University for his guidance and advice. The author would also like to thank Fengyuan Zhu, Jiaqi Ni and Liang Zou for their support and valuable comments. The authors are solely responsible for all statements in this paper.


Arefeh Nasri, Lei Zhang, 2014. The analysis of transit-oriented development (TOD) in Washington,D.C. and Baltimore metropolitan areas. Transport Policy 32, 172-179.

Baoxing Qiu, 2012. Compactness and diversity: two core elements of sustainable urban development in China. City Planning Review 10, 11-18. Boarnet, M. G, et al., 1999. Transit-oriented development in San Diego County. Journal of the American Association of Planners 65, 80-95. Calthorpe, P., 1993.The Next American Metropolis: Ecology, Community, and the American Dream. Princeton Architectural Press, Princeton. Cervero,R., 1994. Transit-based housing in California: evidence on ridership impacts. Transport Policy 3, 174-183.

Cervero, R., 2007. Transit-oriented development's ridership bonus: a product of self-selection and public policies. Environment and Planning A 39, 2068-2085.

Chaogui Kang, et al., 2012. Towards Estimating Urban Population Distributions from Mobile Call Data. Journal of Urban Technology 11, 3-21. Floriano Scioscia, et al., 2014. A framework and a tool for semantic annotation of POIs in Open Street Map. Procedia - Social and Behavioral Sciences 111, 1092-1101.

Glen Searle, et al., 2014. Positive and negative factors for transit oriented development: case studies from Brisbane, Melbourne and Sydney.

Urban Policy and Research 32, 437-457. Sung H, et al., 2011. Transit-oriented development in a high density city: Identifying its association with transit ridership in Seoul, Korea. Cities 1, 70-82.

Taojun Feng, 2009. The scope of urban rail transit station's effect on the value of land. Southwest Jiaotong University, Chengdu, pp. 142. Shijing Zhou, 2012. Research on land reasonable development intensity around urban rail transit stations. Beijing Jiaotong University, Beijing, pp.99.

Zhongnan Ye, 2015. Research on urban facilities around metro stations based on POI data—the case of Shanghai metro Line 10. Urban Development Studies 22, 1-6.