Land Use Policy 82 (2019) 269–282

Contents lists available at ScienceDirect

Land Use Policy

Transit oriented development among metro station areas in Shanghai, T

a
School of Resource and Environmental Sciences, Wuhan University, Wuhan, China
b
Shanghai Surveying and Mapping Institute, Shanghai, China
c
Collaborative Innovation Center of Geospatial Technology, Wuhan University, Wuhan, China

ARTICLE INFO ABSTRACT

Keywords: Transit Oriented Development (TOD), the integration of transport system with land use, has gained considerate
TOD priorities in planning strategies towards urban sustainability. To accrue a better overall leverage of the benefits
Node-place model arising from TOD practice, it is necessary to explore the variations and typology among TODs. This paper extends
Accessibility the classic ‘node (transport) – place (land use)’ model by incorporating the oriented characteristics that represent
Land use and transport integration
the morphological and functional ties between transport and land use. The model is applied to the case of
Land use planning
Shanghai, China. Fuzzy AHP (Analytic Hierarchy Process) is used to construct the indicators system of each
dimension (node, tie and place) in the model and typology among TODs is divided by SOM (self-organizing
map). We find that the TOD index value declines from the urban center to the outskirts of the city obviously.
Four typologies are identified among TODs, including the Integrated (all high node, tie and place index value),
the Functionally place-developed (low node index value and high tie and place index value), the
Morphologically node-developed (high node index value and low tie and place index value) and the Dispersed
(all low node, tie and place index value). Based on the evaluation, we put forward an optimization plan for the
areas with low TOD index values, which are consistent with the planned lines under construction. All the
methods demonstrated in this study are easy to perform and can be widely applied to the assessment of TOD
typologies worldwide. This study produces some generalized knowledge that are useful for implementing TOD
practice within land use planning.

1. Introduction 2015; Wey et al., 2016). TOD advocates the maximization of urban
developments around public transport nodes and the creation of com-
1.1. Background munities with diverse land use, moderate to high density, and walkable
environment (Bertolini and Spit, 1998; Dittmar and Ohland, 2004). It
During the recent past, academic and professional circles have combats the uncontrolled urban sprawl and is therefore closely linked
shown considerate interest towards the planning strategies that can with the New Urbanism and Smart Growth policies (Bertolini et al.,
help mitigate the adverse consequences of accelerating urban growth 2012; Dittmar and Ohland, 2004; Givoni and Banister, 2010; Sung and
and mobility (Lyu et al., 2016; Wey et al., 2016). One common re- Chang, 2017). In this regard, the TOD has become the dominant
commendation is the integration of transport system with land use paradigm for land use planning.
(Banister, 2008; Cervero, 2004; Pan et al., 2011; Suzuki et al., 2013), TOD, under favorable conditions, delivers a variety of benefits in-
with Transit Oriented Development (TOD) as a typical term (Cervero, cluding increased non-motorized access to public transport systems,
1998a,b; Nasri and Zhang, 2014; Schlossberg and Brown, 2004). Whilst reduced traffic pollution and energy consumption, popularized public
scholars have proposed a diversity of definitions (Boarnet and Crane, transport ridership, boosted healthier lifestyles (e.g., walking and cy-
1997; Cervero, 2004; Parker et al., 2002), the planning for TOD in cling), creating attractive destinations, shaping urban and regional
nature attempts to achieve cohesive territories and to promote walking, polycentrality, mitigating urban sprawl, and accommodating economic
cycling, and using public transit instead of driving private cars (Vale, growth (Huang et al., 2017; Langlois et al., 2016; Lindau et al., 2010;

⁎
Corresponding author at: No.129 Luoyu Rd, Wuhan, Hubei Province, China.
E-mail addresses: [email protected], [email protected] (S. Su).

https://doi.org/10.1016/j.landusepol.2018.12.003
Received 10 August 2018; Received in revised form 15 October 2018; Accepted 2 December 2018
0264-8377/ © 2018 Elsevier Ltd. All rights reserved.
Z. Li et al. Land Use Policy 82 (2019) 269–282

Lyu et al., 2016; Papa and Bertolini, 2015; Singh et al., 2017). In North New Jersey, Washington, and New York. Many European cities later on
America, approximately 200 TODs have been established and 400 sites implemented TOD as their primary land use planning approach such as
are providing potential favorable forms of TOD practice (America, R. Amsterdam, Copenhagen, Helsinki, Karlsruhe, Lisbon, Munich, Naples,
2007). Planners have discovered a variety of TOD forms as well as Rome, Stockholm and Zurich (Bernick and Cervero, 1997; Cervero,
varying morphological and functional characteristics among individual 1998a,b; Curtis et al., 2009; Dittmar and Ohland, 2004; Papa and
TODs within a system (Atkinson-Palombo and Kuby, 2011; Belzer and Bertolini, 2015; Vale, 2015). Recent decade has seen the TOD adoption
Autler, 2002). To accrue a better overall leverage of the benefits arising in Asian governments such as China, South Korea, Japan, and Singapore
from TOD practice, it is necessary to explore the variations and typol- (Cervero and Day, 2010; Loo et al., 2010; Sung and Oh, 2011). Al-
ogies among TODs (Center for Transit-Oriented Development, 2011; though TOD adoption followed one basic philosophy in different cities,
Kamruzzaman et al., 2014). the specific focuses varied greatly with cases. American planners stress
The essence of TOD variations is the division of transport nodes the re-concentration of urban developments around transit nodes
(station areas) into hierarchical organization according to the in- (Hemsley, 2009), while European governors underline the redevelop-
dividual-specific attributes and services as well as surrounding land use ment of existing transit station areas (Bertolini and Spit, 1998). TOD is
mix, densities, activities, and facilities (Singh et al., 2017). On the regarded as a means of formulating mass transit corridors for mega-
contrary, the nature of TOD typology is the clustering of transport region growth in Asia (Zhang, 2007; Yang and Lew, 2009), which is
nodes that share common functional and morphological characteristics considered to be a strategy of re-connecting the existing dense urban
(Lyu et al., 2016). Based on the analysis, planners can not only develop developments around transit (Lindau et al., 2010).
remedial actions (e.g., investment in certain type of TODs) to improve
the existing situation, but also formulate more targeted strategies (e.g.,
location selection of new TODs) to promote TOD (Arrington and 1.2.2. Methodologies of TOD variations and typology
Cervero, 2008; Kamruzzaman et al., 2014; Lyu et al., 2016; Renne, Since TOD takes a diversity of forms, it requires a practical tool that
2007; Reusser et al., 2008). Researchers have demonstrated different captures the case-specific characteristics separately and allows for the
methodologies to measure TOD variations and typology (Lyu et al., comparisons among individuals, so that unique recommendation can be
2016; Nasri and Zhang, 2014; Papa and Bertolini, 2015; Singh et al., proposed to achieve the desired form. The node-place model (Fig.1) is
2014, 2017; Vale, 2015). However, the complementary two issues have the most popular approach in the literature for TOD description
not been frequently examined in one case. Most importantly, seldom (Bertolini, 1999). Under the analytical framework, the ‘node’ represents
studies have further showcased the usefulness of the quantitative out- the transit station (transport) and ‘place’ refers to the surrounding area
comes in land use planning (e.g., solutions to TOD promotion and im- (land use). The basic idea is a feedback cycle between transport and
plications for optimizing the transit systems). In addition, former stu- land use: (1) improving transport services and supply (the node value)
dies were overwhelmingly conducted in Western developed nations, will benefit the diversification and intensification of land use in the
and the non-Western countries have received relatively less attention. It surrounding area; (2) diversifying and intensifying the land use (the
thus requires efforts to address such important knowledge gap. place value) will benefit the infrastructure further development. Thus,
TOD has a two-fold meaning: a transit station for people to catch the
1.2. Literature review public transport services and a place for people to socialize, live, re-
create, and shop (Bertolini, 1999). The node-place distinguishes five
1.2.1. Conceptualizations archetypal TOD forms (Fig.1). The first category is the ‘balanced’ form:
Calthorpe, an American architect, first introduced the TOD concept transport and land use are coordinated developed and integrated to-
(Calthorpe, 1993) and defined it as “mixed-use communities within gether; they provide the most favorable conditions for each other. The
10 min walking distance along the regional transit systems and these second category is the ‘dependent’ situation: transport and land use are
communities act as strategic development points with complementary both less developed; they have low probability of tensions but auton-
jobs, public services, uses, and retail in the neighborhood”. Boarnet and omous development dynamics are difficult to be generated. On the
Crane (1997) argued that TOD referred to the development or re- contrary, the third category is the ‘under stress’ form: transport and
development of land near rail transit stations in order to leverage the land use are both highly developed; they have a strong competition for
public investment and promote usage of rail transit systems. Parker space and high probability of tensions. The last categories are the
et al. (2002) conceptualized TOD as a transportation related land use
strategy that coordinated communities with public transit systems.
Schlossberg and Brown (2004) considered TOD as an integrated ap-
proach where land use planning and transportation that increased the
pedestrian access to transit stop. Nasri and Zhang (2014) regarded TOD
as a planning approach that provided transit service along with mixed-
use and high density development so as to encourage transit patronage.
Transit Agency also proposed a diversity of TOD definitions. Even
though scientific and political communities have no universally ac-
cepted TOD definition, the common goal of TOD remains to promote
public transit through integrating land use with urban traffic, which is
typically featured by moderate to high density developments around
transit nodes with walkable and mixed use communities. With respect
to the land use characteristics of TOD, residential land accounts for
20%–60%, central office/commercial land accounts for 30%–70%, and
public space accounts for 5%–15%. However, the TOD does not mean
that motorized trips or private automobile would be largely replaced by
transit. Rather, it represents that public transit acts as an essential al-
ternative to private automobile and a preferred choice in spite of con-
tinuous automobile dominance.
The TOD principles were first adopted by American planners and
progressively applied to California, Georgia, Massachusetts, Florida, Fig. 1. Five archetypal TOD forms (Source: Bertolini, 1999).

270
Z. Li et al. Land Use Policy 82 (2019) 269–282

‘unsustained’ or ‘unbalanced’ forms: the ‘unbalanced node’ (transport urban sustainable development. Under the circumstances, how to use
far exceeds land use development) and the ‘unbalanced place’ (land use reasonable planning strategies to solve these problems has attracted the
far exceeds transport development). attention of many scholars (Lyu et al., 2016; Wey et al., 2016). Among
Following the conceptual basis of node-place model, scholars have them, the integration of public transport and urban land use is regarded
developed case-specific variables to measure the TOD around transit as an effective means. Urban planners and managers have clearly pro-
nodes (metro stations in most cases). Typical application examples in- posed the adoption of the TOD model, in an attempt to promote the
clude Amsterdam in the Netherlands (Bertolini, 1999), Brisbane in development of the TOD model in China from the perspective of policy
Australia (Kamruzzaman et al., 2014), Lisbon in Portugal (Vale, 2015), and construction. Therefore, many local governments (e.g., Shanghai,
Ostrava in the Czech Republic (Ivan et al., 2012), Switzerland (Reusser Beijing, Wuhan, Hangzhou, and Shenzhen) have gone for the TOD
et al., 2008; Zemp et al., 2011), Tehran in Iran (Monajem and concept which is a typical development strategy of integrating land use
Nosratian, 2015), and Tokyo in Japan (Chorus and Bertolini, 2011). planning and public transit and constructed a number of rapid transit
Facilitated by cluster analysis or principal component analysis, ad- projects to ease the transporting pressures. However, it faces challenges
ditionally, these metro station areas are categorized into different to implement the TOD within a Chinese context, since theoretical and
typologies that have a common set of node or place characteristics practical progress are relatively lagged. For example, stakeholders have
(Atkinson-Palombo and Kuby, 2011; Balz and Schrijnen, 2009; different expectations on TOD practices, and most of them place too
Bertolini, 1999; Calthorpe, 1993; Chorus and Bertolini, 2011; Center for much emphasis on the concept. Due to the insufficient understanding of
Transit-Oriented Development, 2010, 2013; Delta Metropolis the TOD concept, during the implementation of TOD in many cities,
Association, 2014; Dittmar and Poticha, 2004; Evans et al., 2007; overwhelming attention is paid to the characteristics of public transit
Kamruzzaman et al., 2014; Nasri and Zhang, 2014; Renne, 2009; stations, which makes the disintegration of land use and transport
Reusser et al., 2008; Schlossberg and Brown, 2004; Zemp et al., 2011). (Zhao et al., 2018). Moreover, some pioneering work has reported the
It is argued that classification of TOD typologies can enable the per- governance strategies or described the TOD phenomenon in China
formances evaluation and comparison across and within metro station (Cervero and Day, 2010; Mu and Jong, 2012, 2016; Yang et al., 2016).
areas and further enhance their design, planning, and operational ac- Nevertheless, quantitative studies are rarely conducted to examine the
tivities. In particular, successful benchmarks can be identified for jud- TOD variations, typology, and the subsequent optimization for planning
ging the TOD quality and further highlighting the priorities for further implications. This paper aims to the fill in the knowledge gap and select
action (Center for Transit-Oriented Development, 2010). the Shanghai as a typical example.
The node-place model contributes to the understanding of the Shanghai, located in the eastern coast (Fig.2), is the largest me-
transport and land use development of the metro station areas; how- tropolis in China. Metro acts as the most essential component of rapid
ever, it cannot reveal the morphological and functional interrelation transit system in Shanghai and is connected closely with other public
between transport and land use. For example, non-walkable pedestrian transport systems. The Shanghai Metro is China’s third oldest rapid
environments should disconnect the node with the place and further transit system as well as the world’s second largest one by station
reduce the morphological interrelation. In addition, if the station node number and the largest one by route length. At present, 14 lines with
does not have a radiation effect on the surrounding land use develop- 333 stations are in operation and the total length amounts to 548 km. It
ment, the functional interrelation should be largely weakened. It is is observed that the daily average ridership reaches 10 million on
referred to as TAD (Transit Adjacent Development) for the situation workdays. According to the Master Plan, the total route length is pro-
that transport and land use development are consistent but fail to be jected to reach 800 km and the number of metro stations would exceed
morphological and functional interrelated (Renne, 2009). Thus, scho- 500, forming an extensive network covering the whole city. In parti-
lars suggest to incorporate the ‘oriented’ dimension into the node-place cular, the local governors have introduced the TOD concept into land
to evaluate the TOD and classify the typologies among metro station use planning and adopted the TOD as critical strategy to sustainable
areas (Vale, 2015). The oriented characteristics are believed to create urban development. However, the traditional node-place model is in-
favorable local conditions to strengthen the feedback cycle between capable of clearly classifying the TOD areas into different typologies.
transport and land use (Wegener, 2004). Singh et al. (2014) proposed a Some neighboring stations have similar transport and land use char-
spatial multi-criteria assessment approach to TOD evaluation. Papa and acteristics, but present varying morphological and functional inter-
Bertolini (2015) developed a TOD-degree index to measure how urban relations between transport and land use. In this regard, it requires a
development was concentrated along rail corridors. Vale (2015) used new model to quantify the TOD variations and typology among the
the pedestrian shed ratio as an oriented indicator to extend the node- metro station areas. The findings should be of general values for cities
place model. Lyu et al. (2016) introduced the proximity indicators into both in China and beyond.
the node-place model. Based on the principles of sustainable transpor- Focusing on the metro station areas in Shanghai, our specific ob-
tation, Wey et al. (2016) established the TOD evaluation criteria. Prior jectives are to: (1) develop a practical tool by extending the node-place
studies have greatly advanced the knowledge about TOD, but three model to describe TOD variations; (2) demonstrate an alternative ap-
issues should be further deepened. First, functional attributes (e.g., proach to TOD classifications and characterize the TOD typology; and
transit service capacity) are not frequently considered for node mea- (3) optimize the metro system as solutions to TOD promotion and
surement. Second, practical alternative tools should be developed to provide implications for land use planning.
extend the node-place model and further describe TOD variations and
typology. Last, there lacks of quantitative studies for the non-Western 2. Materials and method
developing countries, since overwhelming majority of cases focused on
the Western developed nations. 2.1. Overall methodological framework

1.3. The present study 2.1.1. Extended ‘node-place’ model: ‘node-tie-place’ model
In reference to the term of ‘Transit (transport) – Oriented –
China has been experiencing rapid urbanization during the past Development (land use)’ and the model of ‘node (transport) – place
decades, and over 70% of the total population would be urban by 2035 (land use)’, we argue that the oriented characteristics should be in-
(Feng et al., 2018; Heilig, 2012; Zhang and Su, 2016). A variety of corporated to extend the ‘node-place’ model. The oriented character-
urban problems (e.g., traffic congestion) have emerged and posed great istics represent the morphological and functional ties between transport
threat to sustainability (Wan and Su, 2016, 2017 Xiao et al., 2019; Su and land use and thus play a key role in strengthening their feedback
et al., 2019). Transportation is one of the core problems that restrict cycle. It has been argued that the transit usage and subsequent land use

271
Z. Li et al. Land Use Policy 82 (2019) 269–282

Fig. 2. Location of Shanghai city (China) and the metro stations and lines within it.

diversification and intensification can both be enhanced by their ties in incorporating the ‘tie’ dimension as a third axis orthogonal with the
morphology and function (Lyu et al., 2016; Wegener and Fuerst, 2009; node dimension and place dimension (Fig.3). In theory, the node-tie-
Wegener, 2004). We therefore extend the node-place model through place identifies eight archetypal TOD forms:

272
Z. Li et al. Land Use Policy 82 (2019) 269–282

Fig. 3. Eight archetypal TOD forms under the ‘node-tie-place’ model.

(1) The ‘integrated’ form: transport and land use are both highly de- typologies; (4) employ a tabu-search algorithm to optimize the TOD
veloped and integrated together with close morphological and networks; and (5) discuss some implications for urban planning. The
functional ties; overall methodological framework is shown in Fig.4.
(2) The ‘morphologically double-developed’ form: transport and land
use are both highly developed but separate with each other with
loose morphological and functional ties; 2.2. TOD sites delineation
(3) The ‘double undeveloped’ form: transport and land use are both less
developed but integrated together with close morphological and The Master Plan of Shanghai emphasizes the clustering of urban
functional ties; developments around metro stations. We therefore limit our samples to
(4) The ‘dispersed’ form: transport and land use are both less developed TODs around metro stations. Then, the areas which are delineated are
with weak morphological and functional ties; TOD sites that are within a walking distance from the metro stations, in
(5) The ‘morphologically node-developed’ form: transport has mor- accordance with previous TOD literature (America, R, 2007; Guerra
phological link but no functional link with land use; transport is et al., 2011; Hale and Charles, 2006; ITDP, 2014; Lyu et al., 2016;
highly developed and land use is less developed. Parker et al., 2002; Vale, 2015). No fixed criteria that determine the
(6) The ‘functionally node-developed’ form: transport has functional specific value for the walking distance. Depending on the demographic
link but no morphological link with land use; transport is highly and geographical characteristics, prior literature has used a diversity of
developed and land use is less developed. case-specific distances. For example, American researchers used the
(7) The ‘morphologically place-developed’ form: land use has mor- distance ranging from 400 m (0.25 mile) to 800 m (0.5 mile) (Atkinson-
phological link but no functional link with transport; land use is Palombo and Kuby, 2011; Austin et al., 2010; Schlossberg and Brown,
highly developed and transport is less developed. 2004), while European scholars determined the TOD scope as a 700 m
(8) The ‘functionally place-developed’ form: land use has functional distance from transit stops (Bertolini, 1999; Reusser et al., 2008; Zemp
link but no morphological link with transport; land use is highly et al., 2011; Vale, 2015). Lyu et al. (2016) also selected the 700 m
developed and transport is less developed. threshold to delineate the TOD sites in Beijing. A 10 min walking dis-
tance was demonstrated in other studies (Molster and Schuit, 2013;
Singh et al., 2017). In this study, we use a 15 min walking distance
2.1.2. Main steps along the routes to define the TOD sites, and delineate the areas around
To apply our node-tie-place model into the Case of Shanghai City, the metro stations as TOD areas. This specific threshold value is derived
we: (1) geographically delineate TOD precincts for analysis; (2) estab- from the Master Plan of Shanghai, which attempts to establish a 15 min
lish the indicators for measuring ‘node (transit)’, ‘tie (oriented)’, and walkable neighborhood for residents. Within the 15 min walking dis-
‘place (development)’; (3) aggregate the indicators into integrated TOD tance to community, residents can access to all the basic public services
index using Fuzzy AHP (analytic hierarchy process); (4) apply the self- via walking.
organizing maps (SOM) to classify the TOD areas into different

Fig. 4. Methodological framework.

273
Z. Li et al. Land Use Policy 82 (2019) 269–282

2.3. TOD indicators and aggregation performed for each goal gi and generates m values Mjgi in triangular
fuzzy numbers. The main steps for the AHP are described as follows:
Scholars have demonstrated or proposed a great variety of TOD (1) Computing the fuzzy synthetic extent value Si for xi using Eq.
indicators across different contexts (Atkinson-Palombo and Kuby, 2011; (1).
Bertolini, 1999; Center for Transit-Oriented Development, 2013; 1
m n m
Cervero and Jin, 2009; Chorus and Bertolini, 2011; Dittmar and Si = Mgii Mgii
Ohland, 2004; Ivan et al., 2012; Kamruzzaman et al., 2014; Monajem j=1 i=1 j=1 (1)
and Nosratian, 2015; Papa and Bertolini, 2015; Pollack et al., 2014;
Reusser et al., 2008; Singh et al., 2017; Song and Deguchi, 2013; Vale, where
2015; Zemp et al., 2011). Lyu et al. (2016) conducted a systematic m m m m
review of prior studies and divided these indicators into three cate- Mgii = lj, mj , uj
gories, namely transit, oriented and development. In reference to the j=1 j=1 j=1 j=1 (2)
extended ‘node-tie-place’ model, earlier literature and data availability,
1
we first list 60 indicators from three dimensions: node, tie and place. n m
1 1 1
More specifically, the node dimension indicates the service capacity Mgii = n m , n m , n m
i=1 j=1 i=1
u
j=1 i i=1
m
j=1 i i=1
l
j=1 i (3)
and priority of stations within the metro network. The tie dimension
represents the morphological and functional link and integration be- (2) Computing the degree of possibility (P) between two fuzzy
tween transport and land use development. The place dimension re- synthetic extents (Eq. (4)).
veals the land use development around the 15 min walking distance
around the metro stations. Then, Pearson’s correlation analysis is ap-
P (S2 S1) = hgt (S1 S2) = µ S2 (d ) (4)
plied to calculate the correlations among the 60 indicators. One in- (3) Computing P for a convex fuzzy number greater than k convex
dicator is discarded when it presents strong correlation (|r| > 0.75) fuzzy numbers Si (i = 1, …, k) following Eq. (5).
with two or more indicators. The number of indicators is reduced to 42.
P (S S1, S2, ..., Sk ) = min P (S Si ) (5)
Expert panel evaluation is arranged to justify the remaining 42 in-
dicators for evaluating the TOD variations in Shanghai. Thirty-six ex- (4) Computing the vector W’ based on Eq. (6).
perts from Shanghai Urban Planning and Land Management Bureau,
W = (d (A1 ), d (A2 ), ..., d (Ak ))T (6)
Shanghai Transport Committee, and Shanghai Institute of Surveying
and Mapping as well as fifteen urban scientists participated in the where
evaluation process and all of them (51 in total) expressed strong will- d (Ai ) = min P (Si Sj ) (7)
ingness to participate in the TOD measurement. In particular, expert
judgment was guided by five criteria: measurability, discriminating (5) Normalizing W’ into a non-fuzzy number for each comparison
ability, scientific validation, suitability, and scale appropriateness (Su matrix.
et al., 2012, 2019). Experts assigned a score to each item between 1
(very unfit) and 5 (very fit) by judging its relevance to the five criteria. 2.4. Classification of TOD typology
The average score for each item was then calculated by assigning equal
weight to all criteria and experts. Equal weight was assigned because no The SOM algorithm is employed to classify the TOD typology.
significant differences in scores were identified among the experts Compared to traditional cluster approaches such as k-means and hier-
(P = 0.16) and the criteria (P = 0.09). When one or more experts as- archical clustering, SOM is more advantageous in three aspects
signed very low values to certain items, or pointed out severe unsuit- (Delgado et al., 2017; Su et al., 2011): (1) no requirements of priori
ability, these items were discarded. For each criterion, items with the knowledge; (2) ability to handle nonlinearity; and (3) suitability for
top 5 scored items were kept. We further classified the kept items into visualization. As a type of unsupervised computational neural net-
principle components and asked the experts to judge the classification. works, SOM can deal with highly nonlinear and dimensional data and
After three-round feedbacks, all the experts reached consensus on the reduce the input layer into regular array of low dimension (Kohonen,
classification and the components that constitute the indicator system 2001). Both the input layer and output layer are constituted of a
for evaluating TOD variations in Shanghai. Detailed description of the number of neurons. Neurons in these two layers are connected and the
indicators is presented in Table 1. Original data in 2015 for these in- strength of each connection is referred to as connection intensity or
dicators are obtained from the Shanghai Institute of Surveying and weight (Kohonen, 2001). The working principle of SOM can be sum-
Mapping. marized into three main steps: (1) original vector is put into the SOM;
The final indicators do not share equal importance for the evalua- (2) the dissimilarity between the input vector and weight vector; (3)
tion of TOD variations. Before aggregating these indicators into in- update the weight vector until feature mapping becomes stable. The
tegrated indices (node index, tie index, place index and the integrated SOM keeps learning through changing the computational units and the
TOD index), we should standardize, normalize, and weight them ac- weight vectors of their neighbors until it meets the stopping criteria
cording to the perception of the involved professionals. Based on the (Park et al., 2014; You and Yang, 2017). For example, the set number of
rankings assigned by each expert to each criterion, we employ the fuzzy iterations has been completed or the weight vectors have become
AHP method to aggregate the indicators into one integrated index (TOD stable. The computational units, which are represented as the output
index) and three sub-indices (node index, tie index and place index). neurons, follow a two-dimensional hexagonal arrange. No agreement
Fuzzy AHP combines fuzzy comprehensive evaluation on the basis of has been reached on how to determine the number of output neurons
AHP. It still uses AHP to determine the weight of each index, and then (Arribas-Bel et al., 2007). In our study, a series of SOM with different
uses the fuzzy comprehensive evaluation method to evaluate the fuzzy number of output neurons have been trained and the number is set to be
index. It can solve the inconsistency of evaluation thinking when there 9 (3 × 3), under which one neuron includes very similar metro stations.
are many indexes of the evaluation system, and it is applicable to the In addition, several trials are also conducted to determine the initial
index system of our study. radius and cycle number according to the model performances in-
Under the fuzzy AHP methodology, linguistic variables are used to dicated by quantization error for resolution and topographic error for
indicate the comparative judgments from the decision makers (Chang, typology preservation (Kohonen, 2001; Park et al., 2014; You and Yang,
1996). Let: an object set is represented by X={x1, x2,…, xn} and a goal 2017). Finally, the clusters are delineated and labelled in reference to
set is represented by G={g1, g2,…, gn}. For object xi, extent analysis is the weight vectors for different input variables among the

274
Z. Li et al. Land Use Policy 82 (2019) 269–282

Table 1
The indicator system for evaluating TOD variations in Shanghai.
Dimension Component Indicator Explanation of indicators/ Calculation method

Node Metro station bearing Number of entrances and exits Get the number of entrances and exits of various metro stations from Shanghai Metro
capacity official website
Number of service orientations Get the transfer status of each metro station from Shanghai Metro official website (two-
way service on common lines, four-way service on two-way transfer stations, and so
on)
Average distance of station(meter) The average distance from the current metro station to each service direction to the
next station
Maximum passenger flow (person) The maximum passenger flow currently available at the metro station
The number of compartments Number of compartments passing through current metro station
Metro station service Starting interval (seconds) The interval between departures within the metro station, if there are multiple lines,
capability take the shortest interval between departures
Operating time (minutes) Daily operating hours of this station
Metro station Reachable station within 20 minutes Use Baidu traffic big data to climb stations that can be reached by rail transit within
connectivity 20 minutes and count their number
Rail transit network integration Calculate the degree of integration of road traffic network in Shanghai based on space
syntax theory (The higher the level of integration, the greater the flow of people, the
road is in a more convenient position in the entire road network).
And the distance-weighted average of the integration indicators within 15 minutes’
walk of the Shanghai metro station is used to measure the traffic pressure around the
metro station.
Tie Accessibility Average distance to the workplace Select the points of interest in the area within 15 minutes around the metro station, and
calculate the average distance from the metro station to the points of interest in each
workplace according to Shanghai road network.
Average distance to medical facilities Select the points of interest in the area within 15 minutes around the metro station, and
calculate the average distance from the metro station to the points of interest in each
medical facility according to Shanghai road network.
Average distance to educational venues Select the points of interest in the area within 15 minutes around the metro station, and
calculate the average distance from the metro station to the points of interest in each
educational venue according to Shanghai road network.
Average distance to nursing homes Select the points of interest in the area within 15 minutes around the metro station, and
calculate the average distance from the metro station to the points of interest in each
nursing home according to Shanghai road network.
Average distance to recreation areas Select the points of interest in the area within 15 minutes around the metro station, and
calculate the average distance from the metro station to the points of interest in each
recreation station according to Shanghai road network.
Average distance to the residential quarters Select the points of interest in the area within 15 minutes around the metro station, and
calculate the average distance from the metro station to the points of interest in each
residential quarter according to Shanghai Road Network.
Average distance to bus stations Select the bus stop within 15 minutes around the metro station, and calculate the
average distance from the metro station to the bus stops according to Shanghai road
network.
Walkability Proportion of walkable blocks within 15 minutes Calculate the proportion of 15-minute walkable blocks in the service area of metro
walking distance from the station stations.
Average walkability of residential quarters Calculate the average walkability of residential quarters according to the revised Walk
Score metric (Su et al., 2017)
Road walkability Ratio between pedestrian width and attached green spaces width
Place Land use diversity Residential area within 15 minutes’ walking Taking area within 15 minutes’ walk as a threshold, count the area of residential land
distance from metro station (Dr) with threshold range.
Retail, hotel, restaurant within 15 minutes’ Taking area within 15 minutes’ walk as a threshold, count the area of commercial land
walking distance metro station (Dc) with threshold range.
Educational, health, and cultural facilities within Taking area within 15 minutes’ walk as a threshold, count the area of public
15 minutes’ walking distance from metro station management and public service land (Df) with threshold range.
Public transport and service facilities within Taking area within 15 minutes’ walk as a threshold, count the area of municipal public
15 minutes’ walking distance from the metro facilities land (Da) with threshold range.
station
Land use mixture Land function mix a = max(Dr,Dc,Df,Da)
b = min(Dr,Dc,Df,Da)
c=(Dr+Dc+Df+Da)/4
d = Dr+Dc+Df+Da
F = 1-((a-b)/d)-(a-c)/d))/2
The land function mix characterizes the degree of land use mixture in the areas
surrounding the metro station.

neighborhoods. The R software is used to complete the SOM learning. metro stations in populous places should promote the integration of
transport and land use. Therefore, the new metro stations should be
2.5. TOD optimization chosen in existing population agglomerations to serve more people.
Combined with urban land development and the near future and long-
Two issues should be given priorities in TOD optimization: (1) en- term road network construction planning, optimization should meet the
hancing the integration between transport and land use for metro sta- existing demand at the same time for subsequent urban development
tions with lower TOD index value; (2) selecting potential sites for metro space, and also meet the future urban transportation, environmental
stations in the blind area. Generally speaking, the intensification of development and the urban planning layout to the long-term develop-
human activities will lead to diverse forms of land use. Establishment of ment of the requirements.

275
Z. Li et al. Land Use Policy 82 (2019) 269–282

Fig. 5. Index variations among the metro stations within Shanghai.

DBSCAN clustering is used to determine existing population ag- With the above algorithm, the cluster number 9999 is considered to
glomeration centers and potential population agglomeration centers be noise, and other sample points have been classified according to the
based on population density, defining areas that must be connected to density. In the study, we take Eps as 1.5 km, N as 1, and after executing
existing metro networks and areas that need to be connected to the the DBSCAN algorithm, the natural discontinuity points are classified
network. Then use the heuristic path finding algorithm to plan the for the number of points owned by each cluster, and the discontinuous
metro line. points when the classification number is 3 are 7 and 20 respectively.
Thus, the area where the cluster size exceeds 20 is defined as the area
(1) Selecting a candidate point for metro stations that must be connected to the metro network, and the area with the
scale larger than 7 is defined as the area that needs to be connected to
Firstly, the census block with the population density of 25% in the existing network.
Shanghai is extracted, then the closure of the existing metro stations is
demarcated, and the clustered areas outside the closure are clustered. In (2) Carrying out route planning for metro stations
the end, it is expected to establish a new metro station and line to
connect larger clusters into the existing metro network. Eventually it is Using the heuristic path search algorithm to search for the path from
expected that large clusters will be connected to the existing metro the point which must be connected to the existing metro network, the
network through the establishment of new metro stations and lines. algorithm has three constraints: Firstly, the angle between the last
Using the DBSCAN method for spatial clustering, the algorithm is station connection of the proposed metro station and the target metro
described as follows: line on the route and the extending direction of the route should not
Input: sample point, distance threshold Eps and point number exceed 90°; Secondly, the deviation in the direction of path growth
threshold N; must not exceed ± 30°, which will ensure that the construction cost of
Output: all the generated clusters meet the density requirements; the new line is at a low level; Thirdly, it is necessary to maintain a
S1: Set the cluster number of each point to 0 and the temporary relatively small change in the total length of the route when entering
cluster number tempID to 1; the intermediate station in series (as many lines as possible need to be
S2: It is determined whether there is a point where the cluster connected to the metro station). The algorithm of the metro line
number is 0. If it does not exist, the algorithm ends; otherwise, S3 is planning is as follows:
reached; S1: starting from the point where each metro line must be con-
S3: The point where the first cluster number in the set is 0 is sear- nected, connect the end of each existing metro line;
ched in its Eps neighborhood. If the number of searched points is not S2: For each line segment generated in S1, search within its ± 30°
less than N, then the cluster number at that point is tempID. All the range to find the nearest required join point within the range, the point
searched points numbered 0 are added to the set M. must be connected, or the end point of the existing metro line. Replace
S4: It is determined whether the set M is empty. If it is empty, let the line with the line connecting the nearest point of the line at the end
tempID = tempID + 1 to S2. Otherwise go to S5; of the line associated with the line, and repeat S2 to reach the nearest
S5: For the first point 0 in the set M, the point q is searched in the point as the target metro station;
Eps-neighborhood. If the number of searched points is not less than N, S3: For each route generated in S2, find out the starting point (the
then the cluster number of the point q is tempID, and the searched non- point that must be connected to the existing metro network) and the
zero points are added to the set M. Otherwise, the cluster number for ending point (the location of the target metro station), connect every
point q is 9999. Let M = M – q, go to S4. point and end point on the path. If the angle between any connection

276
Z. Li et al. Land Use Policy 82 (2019) 269–282

direction and the extension direction of the line where the termination Table 2
point is located is greater than 90°, the line is removed; Explanation for the TOD typologies among the metro stations within Shanghai.
S4: For the remaining lines, the line with the shortest Euclidean Typology Evaluation features
distance between the starting point and the ending point is selected as a
preliminary selection. After the space visualization, the overlapped Integrated High node index value
High tie index value
lines should be properly filtered according to the actual situation, and
High place index value
use the second shortest instead of the shortest. Functionally place-developed Medium node index value
High tie index value
High place index value
3. Results Morphologically node-developed Medium node index value
Medium tie index value
Medium place index value
3.1. TOD variations Dispersed Low node index value
Low tie index value
As shown in Fig.5, higher TOD index values are observed for the Low place index value
stations within the core urban metro network formed by the Lines 1, 2,
3, 4, 7, 8, 9, 10 and 11. Moreover, the TOD index value for the stations
gradually decreases from the central city to the suburb. Patterns of the 3.2. TOD typology
three sub-indices are similar to that of the TOD index. Specifically, the
stations with higher node index value are mainly those along the metro As aforementioned, the extended model identifies eight archetypal
Line 2 connecting Pudong and Puxi, as well as those connecting the TOD forms. In the case of Shanghai, four TOD categories are identified
north and south metro Lines 1, 3 and 7. The stations with medium node for all the metro stations which are the Integrated (all high node, tie
index values are particularly concentrated in Huangpu, Jing'an, and and place index value), the Functionally place-developed (low node
Xuhui District on the Metro Line 4 loop. The stations with higher tie index value and high tie and place index value), the Morphologically
index value are mainly distributed in the central area, while those with node-developed (high node index value and low tie and place index
higher place index values are mainly distributed on the west and north value) and the Dispersed (all low node, tie and place index value)
sides of the central city. However, majority of the stations within the (Table 2). In terms of their distributions (Fig. 7), the transfer stations in
core areas of the central area (Jingan District, Huangpu District, Pu- the central urban area mainly belong to the Integrated type. The node
dong New Area near the Huangpu River) present relatively low values index value, the tie index value and the place index value are all very
of the place index. TOD variations for the metro stations located in the high and the coordinated development between transport and land use
central urban area of Shanghai are shown in Fig. 6. Metro stations is realized. The Functionally place-developed TOD stations are mainly
closer to the central urban areas (Jing'an District, Huangpu District, located in some non-transfer stations in the core urban areas and on
Xuhui District) present higher the TOD index value. In addition, the various metro lines in the central city. These stations have medium
TOD index value of the transfer stations is fairly good, and overall it node index value but high tie index value and place index value. It
presents a declining rating-circle structure. indicates that the service capabilities of these stations should be

Fig. 6. Index variations among the metro stations in the central area of Shanghai.

277
Z. Li et al. Land Use Policy 82 (2019) 269–282

Fig. 7. TOD typologies among the metro stations within Shanghai.

improved. The Morphologically node-developed TOD stations are gen- Line 5 or Line 8. Based on Metro Line 3, it extends north to Baoshan
erally located outside the central area. They have medium node index District, and extends southwards to the west of Dishui based on Metro
value, tie index value and place index value. The Dispersed TOD sta- Line 16 (Fig. 8). Comparing the construction plan of the Shanghai
tions, with low node index value, tie index value and place index value, Metro in recent years, we can find that the optimized metro line 2 ex-
are mostly located at the ends of the metro routes leading to each tends westward to the Qingpu District and is fully consistent with the
district. Metro Line 17 under construction in 2017. The metro line 9 extends
southwards to the Jinshan District, which is in line with the Jinshan
3.3. TOD optimization Railway which serves the intercity express metro service. Based on
Metro Line 5 or Line 8, it extends south to Fengxian District and is in
The proposed TOD metro stations and connecting lines are pre- line with the south extension of Line 5. The above comparison results
sented in Fig. 8. The recommended new metro stations are mainly give evidence to the reliability of the optimization method, and show
distributed in Qingpu, Songjiang, Jinshan, Fengxian and Baoshan. that the existing rail transit planning takes into account the needs of the
Therefore, it is proposed to extend westward to Qingpu District based population reasonably, and can connect Shanghai's urban areas and the
on Metro Line 2 and to southwards to Jinshan District based on Metro outer suburbs more economically and conveniently.
Line 9 and to extend southwards to Fengxian District based on Metro

278
Z. Li et al. Land Use Policy 82 (2019) 269–282

Fig. 8. The proposed new metro stations and lines.

4. Discussion and conclusions encourage and guide residents to choose more frequently metro for a
trip. Roughly speaking, stations on the metro lines from the urban
4.1. TOD variations and typology center to the urban areas show medium tie index values. The average
distance from each station to the workplace and place of residence is
In this paper, we propose an extended ‘node-place’ model to eval- similar. Such metro stations meet the daily needs of residents based on
uate the TOD. We find that the stations with high TOD index value are work and life. The tie index value of metro stations in the suburbs is
mainly distributed in the central urban area, and the TOD index value also medium, as the supporting public facilities around the station are
of the stations on each metro line from the central city to the suburbs relatively incomplete and the overall walkability is low. It indicates that
gradually decreases. It indicates that the TOD construction in the core the coverage of convenience facilities of these station areas need to be
districts in central area of Shanghai (such as Jing'an District, Xuhui improved, towards pedestrian-friendly goals.
District, Huangpu District, and the west side of Pudong New District) is In terms of place index, the stations with high place index value are
very effective. In addition, for the central urban area, the TOD devel- mainly located on the west and north sides of the central urban area.
opment level of the transfer stations is better. For the node index, The land use in the station area is very diverse. The core function of
higher values are observed for metro stations closer to the central urban these stations is to serve residents and they can serve as a demonstra-
area. Stations with large traffic flow have high node index value, be- tion area for the development of the TOD mode. However, the place
cause these stations all have good capacity, service capability and index value of metro stations in the core areas of the central area
connectivity. The node index value of stations with transfer capabilities (Jing'an District, Huangpu District, and Pudong New Area near the
are generally higher than those of ordinary stations, such as the People's Huangpu River) is relatively low. The reason is that the proportion of
Square Station and Yishan Road Station. The node index value of metro commercial facility land in these areas exceeds 10%, and the proportion
station gradually decreases from the center of the city to the periphery of residential land is relatively low. They are stations with commercial
of the city, which is in line with the demand for metro. services as the mainstay. In the central city, its place index value varies
In terms of the tie index, stations with high tie index value are also from the center to the peripheral area. To some extent, it reflects the
distributed in the central urban area. Rich and diverse commercial fa- functional distribution in central area of Shanghai. In the core area of
cilities along the street and well-arranged public service facilities are the metro network, the station mainly serves as a business center, and
important factors in inducing residents' walking activities. From such gradually transits to the peripheral circle to serve as residential area
stations, it is easy to access various public facilities, which would and then spreads to the surrounding districts.

279
Z. Li et al. Land Use Policy 82 (2019) 269–282

TOD variations in Shanghai, which present obvious disparities be- insufficient, and it is difficult to promote the diversification of land use
tween the central urban area and the suburban areas, are consistent in the station area.
with the TOD circle division model proposed by Calthorpe (1993). It
was pointed that TOD would be planned following a layout pattern and (3) Outskirts station: functions vary, advance slowly
that the shape of the circle planned in the TOD could be region-specific
(Calthorpe, 1993). For example, the circle in Dallas of USA is divided The outskirts stations mainly play the role of connecting Shanghai's
into three parts: a core circle with a radius of 200 m (Core), a center major external transportation nodes and population centers of cities
circle of 400 m (Center) and a circle of 800 m (Edge) (Dallas Area Rapid and towns. However, overall TOD construction progress is relatively
Transit, 2008). The intensity of development around the Arlington Rail slow. For some stations that have important service functions for re-
Transit site in the United States gradually decreases away from the sidents, they need to plan ahead and focus on development of these
station (Dittmar and Ohland, 2004). Cities such as Florida (Group, RP, stations.
2018), Phoenix (Atkinson-Palombo and Kuby, 2011) and Denver In order to advance the TOD mode construction in Shanghai and
(Ratner and Goetz, 2013) also clearly adopted the 800 m circle solve the above problems, the following ideas can be adopted:
boundary radius in the TOD planning model and 1.6 km of secondary
zone circle radius. (1) Central city station: Optimize status, improve steadily
There are a number of metro stations within cities, with different
locations, grades and functions, which makes the TOD planning rela- Encrypt the metro line in the core area of the central area, optimize
tively complex. Therefore, classification of TOD typology should pro- the land use pattern in the service area of the metro station with the
vide critical planning implications. Calthorpe (1993) proposed two main purpose of meeting residents' needs, and build a slow-moving
basic TOD types: city-level TOD and neighborhood-level TOD. In the urban space in the center of TOD transportation node to further im-
planning practice of Vancouver in Canada (TransLink, 2012), Sacra- prove the supply of public facilities within walking distance of the
mento in USA (Gleave et al., 2009), the TOD areas were also divided station area.
into different types. In our study, four TOD categories are identified
among the metro stations within Shanghai. The Integrated TOD sta- (2) City loop outer station: Interconnection, rational expansion
tions, which fulfill the coordinated development between transport and
land use, are concentrated in the core urban areas. In particular, these Near-central urban stations outside the city loop have a good basis
stations are generally the transfer stations which are featured by high for TOD mode development and need to be improved. It is re-
capacities to serve large flows of people, good accessibility to be commended to add new urban loop lines to include these stations in the
reached and diverse surrounding land to generate vitality. The Func- core network area. Divert the pressure from the central city station,
tionally place-developed TOD stations are the non-transfer stations in plan to enhance the diversity of land use, optimize the walkability of
the core urban areas and those on various metro lines in the central city. the road network within the metro station, improve the level of TOD
Due to the convenience of geographical location, the land use diversity construction, and form a new boundary for urban expansion.
and accessibility of such stations are relatively good. However, because
the passenger flow and departure frequency of these stations are rela- (3) Outskirts station: Emphasis planning, implement development
tively low, the station service capability is medium. The Morphologi-
cally node-developed TOD stations are generally located outside the The outskirts stations are located at the end of the metro routes
core area of the central city. These stations have strong service cap- leading to all districts. Efforts should be made to build metro station
abilities and they still serve a lot of passengers. However, since these complexes and guide the development of high-intensity and high-den-
stations are not in the core development area of the city, the land use is sity land, take advantage of the commuting of metro stations to attract
not diverse, and the accessibility of the station to other places is also and gather people.
low. The Dispersed TOD stations are mostly located at the ends of the For the central area, an encryption strategy should be adopted on
metro lines leading to all districts. Due to the inconvenient location, the existing metro network to form a loop based on existing station
such stations are in the suburb and are not performing well in all as- connections. For the surrounding districts, the analysis shows that po-
pects. pulation distribution in Jinshan District, Fengxian District, Qingpu
District, Songjiang District, northern Baoshan District and Pudong New
4.2. Planning implications Area is relatively concentrated, and the demand for metro stations is
more urgent. According to the results of the algorithm, it is proposed to
At present, the TOD development level of the Shanghai metro sta- extend westward to Qingpu District on the basis of Metro Line 2 and
tion presents a “high-medium-low” distribution from the urban center extend southwards to Jinshan District on the basis of Metro Line 9 and
to the outskirts of the city. The main issues in the current TOD con- southwards to Fengxian District on the basis of metro line 5 or 8. Based
struction in Shanghai are summarized as follows: on Metro Line 3, it extends north to Baoshan District and Metro Line 16
to the south to the west of Dishui Lake. These proposed options can thus
(1) Center city station: Land intensive, limited development serve as the facilitating references for TOD construction planning.

The central urban area mainly hosts the core functions of Shanghai. 4.3. Strengths, limits and prospects
The degree of land intensive utilization is high, and the road resources
in the central urban area are relatively stable. The further development In this study, the original ‘node-place’ model is extended, and the
of the TOD construction of the station and the upgrading of the metro oriented feature is incorporated to form the ‘Transit (Node) - Oriented
network is bound to be limited by land resources. (Tie) - Development (Place)’ model. The oriented representation of the
morphological and functional ties between the transport system and
(2) City loop outer station: High potential, insufficient attractive land use, which makes the tie between station service capability and
land use diversity more intimate and the model system more complete.
The station outside the city loop is the main direction of future Therefore, it plays a key role in strengthening its feedback cycle so that
urban development in Shanghai. The land plasticity is relatively good. the model can identify different TOD typology better. In addition, this
However, the interconnection between stations needs to transfer extended ‘node-tie-place’ model framework is universal, the strength of
through the central city station. The attraction of the station is our methodology lies in its high transferability over different

280
Z. Li et al. Land Use Policy 82 (2019) 269–282

geographies and the simplicity that makes it repeatable and easy to Curtis, C., Renne, J.L., Bertolini, L., 2009. Transit oriented development: making it
implement. Some cities in China are single-centered and have a rela- happen. Geojournal 16, 1190–1195.
Dallas Area Rapid Transit, 2008. Dallas Transit-oriented Development (TOD) Guidelines:
tively dense population. The advantages in urban density provide huge Promoting TOD Around DART Transit Facilities.
potential for public transport development. From this point of view, Delgado, S., Higuera, C., Calle-Espinosa, J., Morán, F., Montero, F., 2017. A SOM pro-
China is in favor of using this extended model to implement the TOD totype-based cluster analysis methodology. Expert Syst. Appl. 88, 14–28.
Delta Metropolis Association, 2014. SprintCity Update 6. Delta Metropolis Association.
concept. The proposed index system of the model can be further opti- Dittmar, H., Ohland, G. (Eds.), 2004. The New Transit Town: Best Practices in Transit-
mized and improved in the future dynamic monitoring, and can be Oriented Development. Island Press, pp. 19–39.
updated in conjunction with the conditions of different regions, such as Dittmar, H., Poticha, S., 2004. Defining Transit-oriented Development: the New Regional
Building Block.
"station layer number". However, it is worthwhile to mention that the Evans, J.E., Pratt, R.H., Kuzmyak, J.R., Stryker, A., 2007. Tcrp Report 95 Chapter 17:
established indicators system is not universal. Geo-big data such as cell Transit-oriented Development – Traveler Response to Transportation System
phone signal and check-in data can be used as indicators for the tie Changes.
Feng, Y., Wu, S., Wu, P., Su, S., Weng, M., Bian, M., 2018. Spatiotemporal character-
dimension. Moreover, independent evaluation of the special TOD con-
ization of megaregional poly-centrality: evidence for new urban hypotheses and
struction area can also be considered, and a comprehensive coverage implications for polycentric policies. Land Use Policy 77, 712–731.
and multi-scale comprehensive evaluation system can be constructed. Givoni, M., Banister, D., 2010. Integrated Transport: From Policy to Practice.
Further consideration can also be made for the bus stations. In parti- Gleave, S.D., Jackson, G., Anglin, K., 2009. Sacramento Regional Transit: A Guide to
Transit Oriented Development (TOD).
cular, how to design more robust models for optimizing the public Group, RP. A framework for transit oriented development in Florida.
transport system and land use should be given priorities in following Guerra, E., Cervero, R., Tischler, D., 2011. The half-mile circle: does it represent transit
studies. station catchments? University of California Transportation Center Working Papers
2276, 101–109.
Hale, C., Charles, P., 2006. Making the Most of Transit Oriented Development
Acknowledgement Opportunities. 29th Australasian Transport Research Forum (pp.1-19). The
Australasian Transport Research Forum.
Heilig, G.K., 2012. World Urbanization Prospects: the 2011 Revision. United Nations.
We thank Editor J.A. Zevenbergen and two anonymous reviewers Hemsley, W., 2009. The Commercial Reality of TOD in Australia.
for providing useful comments. This study is supported by the National Huang, R., Moudon, A.V., Zhou, C., Stewart, O.T., Saelens, B.E., 2017. Light rail leads to
Key Research and Development Plan (No.2018YFF0215006). more walking around station areas. J. Transp. Health.
ITDP (Institute for Transportation & Development Policy) (2014). The TOD Standard v2.
1. Retrieved from https://www.itdp.org/tod-standard/.
References Ivan, I., Boruta, T., Horák, J., 2012. Evaluation of railway surrounding areas: the case of
ostrava city. Urban Transport XVIII 128, 141–152.
Kamruzzaman, M., Baker, D., Washington, S., Turrell, G., 2014. Advance transit oriented
America, R, 2007. Why Transit-oriented Development and Why Now? Reconnecting
development typology: case study in Brisbane, Australia. J. Transp. Geogr. 34, 54–70.
America. Okland, CA, USA. .
Kohonen, T., 2001. Self-organizing maps, ser. Information Sciences. Springer, Berlin,
Arribas-Bel, Daniel, Nijkamp, P., Scholten, H., 2007. Multidimensional urban sprawl in
pp. 30.
Europe: a self-organizing map approach. Comput. Environ. Urban Syst. 35, 263–275.
Langlois, M., Wasfi, R.A., Ross, N.A., El-Geneidy, A.M., 2016. Can transit-oriented de-
Arrington, G.B., Cervero, R., 2008. Effects of TOD on Housing, Parking, and Travel. Tcrp
velopments help achieve the recommended weekly level of physical activity? J.
Report.
Transp. Health 3, 181–190.
Atkinson-Palombo, C., Kuby, M.J., 2011. The geography of advance transit-oriented de-
Lindau, L.A., Hidalgo, D., Facchini, D., 2010. Curitiba, the cradle of bus rapid transit.
velopment in metropolitan phoenix, arizona, 2000–2007. J. Transp. Geogr. 19,
Built Environ. 36, 274–282.
189–199.
Loo, B.P.Y., Chen, C., Chan, E.T.H., 2010. Rail-Based Transit-Oriented Development:
Austin, M., Belzer, D., Benedict, A., Esling, P., Haas, P., Miknaitis, G., et al., 2010.
Lessons from New York City and Hong Kong 97. Social Science Electronic Publishing,
Performance-based Transit-oriented Development Typology Guidebook. Guidelines.
pp. 202–212.
Balz, V., Schrijnen, J., 2009. From Concept to Projects: Stedenbaan. The Netherlands. .
Lyu, G., Bertolini, L., Pfeffer, K., 2016. Developing a TOD typology for Beijing metro
Banister, D., 2008. The sustainable mobility paradigm. Transp. Policy (Oxf) 15, 73–80.
station areas. J. Transp. Geogr. 55, 40–50.
Belzer, D., Autler, G., 2002. Transit Oriented Development: Moving From Rhetoric to
Molster, A., Schuit, S., 2013. Voetsporen Rond Het Station.
Reality. Communities.
Monajem, S., Nosratian, F.E., 2015. The evaluation of the spatial integration of station
Bernick, M., Cervero, R., 1997. Transit Villages in the 21st Century. McGraw-Hill, New
areas via the node place model; an application to metro station areas in Tehran.
York.
Transp. Res. D Transp. Environ. 40, 14–27.
Bertolini, L., 1999. Spatial development patterns and public transport: the application of
Mu, R., Jong, M.D., 2012. Establishing the conditions for effective transit-oriented de-
an analytical model in the Netherlands. Plan. Pract. Res. 14, 199–210.
velopment in china: the case of Dalian. J. Transp. Geogr. 24, 234–249.
Bertolini, L., Spit, T., 1998. Cities on Rails: The Redevelopment of Railway Station Areas.
Mu, R., Jong, M.D., 2016. A network governance approach to transit-oriented develop-
E & FN Spon, London and New York.
ment: integrating urban transport and land use policies in Urumqi, china. Transp.
Bertolini, L., Curtis, C., Renne, J., 2012. Station area projects in Europe and beyond:
Policy (Oxf) 52, 55–63.
towards transit oriented development? Built Environ. 38, 31–50.
Nasri, A., Zhang, L., 2014. The analysis of transit-oriented development (TOD) in
Boarnet, Marlon, Crane, Randall, 1997. L.A. story: a reality check for transit-based
Washington, D.C. And Baltimore metropolitan areas. Transp. Policy (Oxf) 32,
housing. J. Am. Plan. Assoc. 63, 189–204.
172–179.
Calthorpe, P., 1993. The Next American Metropolis: Ecology, Community, and the
Pan, H., Shen, Q., Liu, C., 2011. Transit oriented development at urban periphery: insights
American Dream. Princeton Architectural Press, Princeton.
from a case study in Shanghai. China. Transp. Res. Rec. 224, 95–102.
Center for Transit-oriented Development, 2010. Performance-based Transit Oriented
Papa, E., Bertolini, L., 2015. Accessibility and transit-oriented development in European
Development Typology Guidebook. Center for Transit-oriented Development.
metropolitan areas. J. Transp. Geogr. 47, 70–83.
Center for Transit-Oriented Development, 2011. Portland Metro TOD Program and TOD
Park, S.H., Hosoishi, S., Ogata, K., Kuboki, Y., 2014. Clustering of ant communities and
Strategic Plan Case Study.
indicator species analysis using self-organizing maps. Comptes Rendus - Biologies
Center for Transit-Oriented Development, 2013. Transit-Oriented Development Typology
337, 545–552.
Strategy for Allegheny County.
Parker, T., McKeever, M., Arrington, G., Smith-Heimer, J., 2002. Statewide Transit-
Cervero, R., 1998a. The Transit Metropolis: a Global Inquiry. Island Press,
Oriented Development Study: Factors for Success in California.
Washington, DC.
Pollack, S., Gartsman, A., Boston, M., Benedict, A., Wood, J., 2014. Rating the
Cervero, R., 1998b. Transit Villages in California: Progress, Prospects, and Policy
Performance of Station Areas for Effective and Equitable Transit Oriented
Reforms.Institute of Urban and Regional Development. University of California,
Development. Transportation Research Board 93rd Annual Meeting.
Berkeley.
Ratner, K.A., Goetz, A.R., 2013. The reshaping of land use and urban form in Denver
Cervero, R., 2004. Transit-Oriented Development in the United States:
through transit-oriented development. Cities 30, 31–46.
Experiences,Challenges, and Prospects. Transportation Research Board of the
Renne, J.L., 2007. Measuring the Performance of Transit Oriented Development in
National Academies.
Western Australia. Planning and Transport Research Centre and Institute for
Cervero, R., Day, J., 2010. Suburbanization and transit-oriented development in china.
Sustainability and Technology Policy.
Shanghai Urban Plann. Rev. 15, 315–323.
Renne, J.L., 2009. From transit-adjacent to transit-oriented development. Local Environ.
Cervero, R., Jin, M., 2009. Rail and property development in Hong Kong: experiences and
14, 1–15.
extensions. Urban Stud. 46, 2019–2043.
Reusser, D.E., Loukopoulos, P., Stauffacher, M., Scholz, R.W., 2008. Classifying railway
Chang, D.Y., 1996. Applications of the extent analysis method on fuzzy AHP. Eur. J. Oper.
stations for sustainable transitions – balancing node and place functions. J. Transp.
Res. 95, 649–655.
Geogr. 16, 191–202.
Chorus, P., Bertolini, L., 2011. An application of the node-place model to explore the
Schlossberg, M., Brown, N., 2004. Comparing transit-oriented development sites by
spatial development dynamics of station areas in Tokyo. J. Transp. Land Use 4,
walkability indicators. Transp. Res. Record J. Transp. Res. Board 1887, 34–42.
45–58.
Singh, Y.J., Fard, P., Zuidgeest, M., Brussel, M., Maarseveen, M.V., 2014. Measuring

281
Z. Li et al. Land Use Policy 82 (2019) 269–282

transit oriented development: a spatial multi criteria assessment approach for the city Wan, C., Su, S., 2016. Neighborhood housing deprivation and public health: theoretical
region Arnhem and Nijmegen. J. Transp. Geogr. 35, 130–143. linkage, empirical evidence, and implications for urban planning. Habitat Int. 57,
Singh, Y.J., Lukman, A., Flacke, J., Zuidgeest, M., Van Maarseveen, M.F.A.M., 2017. 11–23.
Measuring TOD around transit nodes-Towards TOD policy. Transp. Policy (Oxf) 56, Wan, C., Su, S., 2017. China’s social deprivation: measurement, spatiotemporal pattern
96–111. and urban applications. Habitat Int. 62, 22–42.
Song, J., Deguchi, A., 2013. Evaluation and typology of railway station areas in a 30km Wegener, M., 2004. Overview of land-use transport models. Handbook of Transport
circumference surrounding central Tokyo from view of transit-oriented development. Geography and Apatial Systems 5. pp. 127–146.
J. Archit. Plan. 78, 413–420. Wegener, M., Fuerst, F., 2009. Land-use Transport Interaction: State of the Art. Urban/
Su, S., Zhi, J., Lou, L., Huang, F., Chen, X., Wu, J., 2011. Spatio-temporal patterns and regional.
source apportionment of pollution in Qiantang River (China) using neural-based Wey, W.M., Zhang, H., Chang, Y.J., 2016. Alternative transit-oriented development
modeling and multivariate statistical techniques. Phys. Chem. Earth 36, 379–386. evaluation in sustainable built environment planning. Habitat Int. 55, 109–123.
Su, S., Zhang, Z., Xiao, R., Jiang, Z., Chen, T., Zhang, L., Wu, J., 2012. Geospatial as- Yang, P.P.J., Lew, S.H., 2009. An Asian model of TOD: the planning integration in
sessment of agroecosystem health: development of an integrated index based on Singapore. In: Curtis, C., Renne, J.L., Bertolini, L. (Eds.), Transit Oriented
catastrophe theory. Stochastic Environ. Res. Risk Assess. 2012 (26), 321–334. Development: Making It Happen. Ashgate Publishing, Ltd, pp. 91–106.
Su, S., Pi, J., Xie, H., Cai, Z., Weng, M., 2017. Community deprivation, walkability, and Yang, J., Quan, J., Yan, B., He, C., 2016. Urban rail investment and transit-oriented de-
public health: highlighting the social inequalities in land use planning for health velopment in beijing: can it reach a higher potential? Transp. Res. Part A Policy Pract.
promotion. Land Use Policy 67, 315–326. 89, 140–150.
Su, S., Zhou, H., Xu, M., Ru, H., Wang, W., Weng, M., 2019. Auditing street walkability You, H., Yang, X., 2017. Urban expansion in 30 megacities of China: categorizing the
and associated social inequalities for planning implications. J. Transp. Geogr. 74, driving force profiles to inform the urbanization policy. Land Use Policy 68, 531–551.
62–76. Xiao, Y., Wang, D., Fang, J., 2019. Exploring the disparities in park access through mobile
Sung, H., Chang, G.C., 2017. The link between metropolitan planning and transit-oriented phone data: evidence from Shanghai, China. Landsc. Urban Plann. 181, 80–91.
development: an examination of the Rosario plan in 1980 for Seoul, south Korea. Zemp, S., Stauffacher, M., Lang, D.J., Scholz, R.W., 2011. Classifying railway stations for
Land Use Policy 63, 514–522. strategic transport and land use planning: context matters!. J. Transp. Geogr. 19,
Sung, H., Oh, J.T., 2011. Transit-oriented development in a high-density city: identifying 670–679.
its association with transit ridership in Seoul, Korea. Cities 28, 70–82. Zhang, M., 2007. Chinese edition of transit-oriented development. Transp. Res. Record J.
Suzuki, H., Cervero, R., Iuchi, K., 2013. Transforming Cities With Transit. Transp. Res. Board 2038, 120–127.
TransLink, 2012. Transit-oriented Communities Design Guidelines: Creating More Livable Zhang, Q., Su, S., 2016. Determinants of urban expansion and their relative importance: a
Places Around Transit in Metro Vancouver. comparative analysis of 30 major metropolitans in China. Habitat Int. 58, 89–107.
Vale, D.S., 2015. Transit-oriented development, integration of land use and transport, and Zhao, P., Yang, H., Kong, L., Liu, Y., Liu, D., 2018. Disintegration of metro and land
pedestrian accessibility: combining node-place model with pedestrian shed ratio to development in transition China: a dynamic analysis in Beijing. Transp. Res. Part A
evaluate and classify station areas in lisbon. J. Transp. Geogr. 45, 70–80. 116, 290–307.

282