Relationship between transit modal split and intra-city trip ratio by car for compact city planning of municipalities in the Seoul Metropolitan Area

Cities 70 (2017) 11–21

Contents lists available at ScienceDirect

Cities journal homepage: www.elsevier.com/locate/cities

Relationship between transit modal split and intra-city trip ratio by car for compact city planning of municipalities in the Seoul Metropolitan Area

MARK

Seungil Lee⁎, Youngsoo An, Kijung Kim Department of Urban Planning and Design, The University of Seoul, Siripdaero 163, Dongdaemun-gu, Seoul 02504, Republic of Korea

A R T I C L E I N F O

A B S T R A C T

Keywords: Compact city Urban structure Transit modal split Intra-city trip ratio by car Seoul Metropolitan Area

Compact-city planning factors are commonly applicable even to metropolitan areas. In most cases, however, planning policies based on theses factors fail to consider that travel patterns are not uniform in each metropolitan area. Furthermore, the travel pattern of inter- and intra-municipality that results from spatial interaction between a central city and its various sub-centres and suburbs in a metropolitan area has not been fully explored. A consideration of the specific urban system could therefore provide an answer to the question of why certain factors have different effects on the transit modal split and car travel distance between municipalities of a metropolitan area. The aim of this study was therefore to find an effective way to establish compact-city planning policies in municipalities of the Seoul Metropolitan Area (SMA). An investigation of the changed travel pattern in each municipality based on the changed relationship between transit modal split (TMS) and intra-city trip ratio by car (ITR) between 2006 and 2010 found that the SMA became more car-dependent: TMS and ITR of the municipalities declined together, and ITR decreased much more in the outskirts. Based on the relationship between the two factors, the effects of changes in land use and transportation were estimated using a combination of cluster and regression analysis. This revealed that, in municipalities of Seoul and its adjacent subcentres, there is a need to promote transit-oriented development (TOD) by creating high-density areas within close proximity to city railroad stations. In contrast, it is necessary, in municipalities on the outskirts of the SMA, to restrict large-scale developments, such as large retail centres, and instead promote a mixture of self-sufficient land uses. In the intermediate municipalities that lie between these two, TMS and ITR can be increased through TOD near railroad stations, or ITR alone can be increased through a greater mix of land use. These findings could assist in implementing effective compact-city planning policies in each municipality of the SMA, and could also be applied to the other metropolitan areas in Korea or elsewhere in the world.

1. Introduction There has been growing public awareness over the past few decades of the damaging impact that cars have on cities (Newman & Kenworthy, 2006) in terms of their noise, discomfort, psychological and physical insecurity, loss of amenity and social space, and atmospheric pollution (Council of Europe, 2002). Transport energy consumption and greenhouse gas emissions have also been linked to a dependency on cars in many cities, which is often the result of urban sprawl. The universal conviction that this sprawl increases travel distances, and therefore transport energy consumption (da Silva, Costa, & Brondino, 2007), provides a solid base to accept a more compact city as an appropriate solution. Indeed, Newman and Kenworthy's (1989) study of the relationship between density and energy use in an international sample of cities has been widely cited in support of compact city policies (Handy, 1996). However, despite this, the feasibility of compact cities, especially for metropolitan areas, has been called into question by the

⁎

Corresponding author. E-mail address: [email protected] (S. Lee).

http://dx.doi.org/10.1016/j.cities.2017.06.004 Received 22 August 2016; Received in revised form 11 May 2017; Accepted 4 June 2017 0264-2751/ © 2017 Elsevier Ltd. All rights reserved.

dominant trend of expanding urban development (Gordon & Richardson, 1989; Breheny, 1992, 1995; Wegener, 1996; Lee, 2001; OECD, 2012). Most metropolitan areas in the world have become decentralized (Anas, Arnott, & Small, 1998; Aguilera, 2005; OECD, 2012), leading to a shift in the debate over the feasibility of compact cities to the relationship between decentralization and travel distance. Some studies (Gordon, Richardson, & Jun, 1991; Schwanen, Dieleman, & Dijst, 2001) have suggested that a decentralized distribution of employment and people shortens commuting distances. Most other studies (Aguilera, 2005; Casello, 2007), however, have found decentralization to be associated with increased commuting distances and a decrease in transit modal split. Despite these contradictory results, the compact city model has tended to be the core issue in studies relating to urban structure policies for reducing car dependency in metropolitan areas (Handy, 1996; Wegener, 1996; Cervero, 1996; Lee, 2001; Schwanen et al., 2001; Aguilera, 2005; Casello, 2007; Sung & Oh, 2011; OECD, 2012; Lee,

Cities 70 (2017) 11–21

S. Lee et al.

expansion of urban development, which is characterised by low density, segregated land use, and an insufficient provision of infrastructure (OECD, 2012). This generally results in an increase in travel distance and a greater dependence on cars, thereby increasing transport energy consumption (da Silva et al., 2007) and providing a solid basis for establishing a compact city model like that suggested by Newman and Kenworthy (1989). This consists of a dense city centre and a commitment to mass transit (particularly rail) and other non-automotive means of transport, which has provided the motivation for empirical analysis of the relationship between urban structure and travel patterns in many international cities (Handy, 1996). Going against physical planning policies aimed at reducing car travel distances and increasing transit modal split, Gordon and Richardson (1989) argued, however, against the adoption of monocentric compact cities in the U.S. from an econometric point of view. In addition to Gordon and Richardson's (1989) arguments, international studies relating to changes in the urban structure of metropolitan areas have found the monocentric compact city model to be infeasible owing to the dominant trend of expanding urban development (Breheny, 1992, 1995; Wegener, 1996; Lee, 2001; OECD, 2012). In light of the current trend towards polycentric decentralization, there have been a number of empirical studies of the relationship between urban structure and travel patterns in metropolitan areas. Gordon et al. (1991) and Schwanen et al. (2001) have suggested that a polycentric distribution of employment and people shortens commuting distances, whereas Aguilera (2005) and Casello (2007) have found polycentricity to be associated with increased commuting distances and a decrease in transit modal split. Despite these contradictory results, the compact city model has tended to be the core issue in studies relating to urban structure policies for reducing car dependency in metropolitan areas (Handy, 1996; Wegener, 1996; Cervero, 1996; Lee, 2001; Schwanen et al., 2001; Aguilera, 2005; Casello, 2007; Sung & Oh, 2011; OECD, 2012; Lee et al., 2013), which have been based on the idea that a compact city is also applicable to the decentralized urban structure that is more common in today's urban context (OECD, 2012). In other words, when viewed on a metropolitan scale, a compact city need not presume any specific urban form (i.e., monocentric or polycentric). Consequently, the compact-city planning factors suggested by Newman and Kenworthy (1989) have been the focus of academic and political interest to find a solution to the high car dependency in metropolitan areas.

Yi, & Hong, 2013), which have been based on the idea that a compact city is also applicable to the decentralized urban structure that is more common in today's urban context (OECD, 2012). For many metropolitan areas throughout the world, the compactcity planning factors for increasing transit modal split and decreasing car travel distance have been examined in those studies, which confirmed the general findings of Newman and Kenworthy (1989). Indeed, findings from the empirical studies have been used to establish policies to reduce car use in many metropolitan areas (OECD, 2012). In most cases, it has, however, failed to consider that travel patterns and behaviour are not uniform in each metropolitan area. This means that different metropolitan areas should pursue different compact-city planning policies. Furthermore, the travel pattern of inter- and intramunicipality that results from spatial interaction between a central city and its various sub-centres and suburbs in a metropolitan area has not been fully explored. Considering a specific urban system (Parr, 2012) is therefore necessary to determine why certain planning factors have different effects on transit modal split and car travel distance not only between metropolitan areas but also between municipalities of a metropolitan area. This study focused on municipalities of a metropolitan area. The total urban population grew universally during the 2000s, and in most countries, this increase followed a pattern of spatial dispersion of the urban population (Veneri, 2015). Urban development also occurred in the metropolitan areas of Asia (Anas et al., 1998; Newman & Kenworthy, 2006), including the rapidly changing Seoul Metropolitan Area (SMA) (Jun, 2012; Go & Choi, 2013; Yi & Lee, 2014), which represents one of the best examples of decentralized metropolitan areas in Asia. When compared to metropolitan areas in the US and Europe, the SMA has a central city with a much stronger centrality (Veneri, 2015: 8), a rapidly growing suburb owing to urban development (Jun, 2012; Go & Choi, 2013; Yi & Lee, 2014), and a public transportation network that operates mostly within the central city and limitedly to some major employment sub-centres (Lee et al., 2013). Because of these urban structural features, which accordingly affect the travel pattern across the set of municipalities that form Seoul's urban system (Parr, 2012), compact city policies should be based on an understanding of how such features influence the relationship between transit modal split and car travel distance in each municipality. This approach, which has not been pursued yet in other empirical studies, could lead to finding an effective combination of compact-city planning factors for each municipality of the SMA. Based on this background, the purpose of this study is to suggest an effective approach for establishing compact-city planning policies in municipalities of the SMA. To this end, the changed travel patterns of the municipalities between two time periods (i.e. 2006 and 2010) were first investigated with regards to the relationship between transit modal split and car travel distance. The influence of the land use pattern and transportation system on these two indexes of the relationship was then assessed through cluster analysis and regression analysis. Finally, any variables identified as having an apparent influence on both indexes were suggested as factors influential to any policies aimed at compact cities in each cluster of the municipalities. The findings could inform the implementing of effective compact-city planning policies in each municipality of the SMA by taking into account their different urban features. This was intended to be of importance not only to the municipalities of the SMA, but also any one of the many newly rising nations that are experiencing issues with car-focused-traffic systems in their metropolitan areas, and which are seeking to expand city railroads to promote TOD.

2.2. Compact-city planning factors Newman and Kenworthy (1989) found that, in a number of large cities around the world, there is a clear negative relationship between the gasoline consumption per capita and the population and job densities. Further investigation of this negative correlation suggested that the urban structure within a city is fundamental to its overall gasoline consumption. Detailed land use policy factors used in this study included the strength of the city centre and the proportion of the population living in the inner city. In the case of the former, more jobs in the city centre generally mean that mass transit systems were more viable, whereas a significant negative correlation with the latter indicates that a greater mixture of residences and businesses exist in the core of the city (Anas et al., 1998). Transportation system factors such as modal split, provision of private cars, traffic speed, and the availability of roads and parking infrastructure also exhibit a strong correlation with gasoline use if they make car use preferable to other options (Newman & Kenworthy, 1989). Breheny (1995) introduced a British study (ECOTEC, 1993) carried out for the Department of the Environment that provided evidence of a relationship between density and energy consumption, and which confirmed the general findings of Newman and Kenworthy (1989). It also demonstrated that there is an increase in average distance travelled with decreasing urban size; i.e. the lowest level of travel is in metropolitan areas owing to their higher urban densities, shorter travel

2. Literature review 2.1. Compact city and metropolitan areas Urban sprawl is a term commonly used to describe the uncontrolled 12

Cities 70 (2017) 11–21

S. Lee et al.

countries. Cities can therefore no longer be identified by very dense settlements alone, as their economic and spatial extent includes an important part of peri-urban and rural territory (Veneri, 2015). For example, although Korea, Mexico, and Chile have cities with core populations > 80%, the population in peri-urban and rural territories grew faster than in the cores over the last decade. The greatest trend towards decentral spatial structures have been observed in the UK, Slovenia, Korea, US, and Japan (Veneri, 2015). The Seoul Metropolitan Area (SMA) in Korea is different from many other cities in having a strong centrality (Veneri, 2015: 8), rapidly growing suburb owing to urban development (Jun, 2012; Go & Choi, 2013; Yi & Lee, 2014), and a public transportation network that operates mostly within the central city and limitedly to some major employment sub-centres (Lee et al., 2013). Because of these urban structure features compact city policies can only be established through an understanding of how they influence the relationship between transit modal split and car travel distance in each municipality. This approach to urban system planning, which has not been pursued yet in other empirical studies, could lead to obtaining an effective combination of compact-city planning factors for each municipality of the SMA. Unlike the relevant studies, an effective approach to establishing compact-city planning policies in municipalities of the SMA is suggested in this study to provide an answer to the question of why certain factors have different effects on the transit modal split and car travel distance to each municipality. To this end, the changed travel patterns of the municipalities between the two time periods (i.e. 2006 and 2010) are investigated with regards to the relationships between transit modal split and car travel distance. The influence of land use pattern and transportation system on these two indexes of the relationship is then assessed to find effective combinations of compact-city planning factors for each municipality.

distances, and mass-transit facilities (ECOTEC, 1993). Marique and Reiter (2012), on the other hand, found, through a comparison of four suburban districts in Belgium, that a good mix of workplaces, schools, shops, and dwellings in each neighbourhood allows for reduced travel distances. This, therefore, seems to be the best strategy for reducing transport energy consumption in suburban areas, as the means of transport used has only minimal impact. Casello (2007) analysed the potential for increasing the transit modal split in the polycentric metropolitan area of Philadelphia, and the results suggest that a coordinated policy of improved transit service and some disincentives for private vehicle use are necessary to achieve greater modal split and an improved transit system. The original transit modal split was linked to land use in terms of the density and mixture around existing and new transit nodes, and this has led to the idea that the time spent inside private vehicles is considered far less uncomfortable than the time spent walking to, waiting for, and transferring between mass transit systems by a factor of up to three or four times (Webber, 1976). The concept of transit-oriented development (TOD) that was derived from this idea has recently gained renewed attention (Sung & Oh, 2011; Lee et al., 2013) in relation to high-density, mixed land use developments that prioritize pedestrians by being located within easy walking distance of a major transit station (Evans, Pratt, Stryker, & Kuzmyak, 2007). Through an investigation of the literature relating to empirical studies of TOD planning factors in American cities, Ewing and Cervero (2001) identified that the application of these factors tends to reduce the total number of trips and distances travelled by vehicles by 3–5%. In summary, the land use policy factors (Newman & Kenworthy, 1989) for car travel distance and transit modal split exhibited a high interrelationship in terms of their effect on reducing car dependency. Although both were equally affected by the density and mixture of land uses, they had completely different spatial depths, i.e. overall density and mixture distribution in a city were related to car travel distance, whereas those only around transit nodes were related to transit modal split. Furthermore, the effect was also different according to the size and structure of the urban area. Moreover, the transportation system factors (Newman & Kenworthy, 1989) affected the travel patterns more clearly than land use policy factors.

3. Study area and travel pattern 3.1. The Seoul metropolitan area The spatial range of this study encompasses the Seoul Metropolitan Area (SMA) in South Korea, which is a large region with a concentration of small municipalities that serves as a centre for politics, economy, culture, and the other important functions for neighbouring provinces and the rest of the country (Yi & Lee, 2014). The SMA consists of the capital city of Seoul, the city of Incheon, and the province of Gyeonggi (see Fig. 1). Statistics reveal that about 24.9 million people reside in the 11,808 km2 area that is the SMA, which represents almost half of country's population (Statistics Korea, 2014). Of these, 12 million people (24.1% of the total population) live in Gyeonggi, and as of 2010, 10.5 million people resided in the 603.3 km2 area that is the central city of Seoul. This gives Seoul a population density of 17,404 people/km2, making it the most densely populated city in the world (Sung & Oh, 2011) and ensuring strong centrality (Veneri, 2015: 8). As shown in Fig. 1 and Table 1, the SMA includes two metropolitan cities (Seoul & Incheon) and one province (Gyeonggi) that contains 79 municipalities, i.e. 20 cities (Si), 6 counties (Gun), and 53 boroughs (Gu). The temporal scope of this empirical study is the year 2010, and only 78 municipalities were included as spatial units (The island of Wando-gun was excluded.). All variables used for analysis were recalculated according to the spatial units. The urbanized areas1 of the SMA shown in Fig. 2 were developed along a southerly axis from Seoul along with their neighbouring municipalities. The city railroad network shown in Fig. 2 is the main public transportation network serving these developed areas. In

2.3. Relationship between transit modal split and car travel distance As already mentioned, the compact-city planning factors have been empirically verified and/or implemented in many metropolitan areas regardless of their urban structure (Handy, 1996; Wegener, 1996; Cervero, 1996; Lee, 2001; Schwanen et al., 2001; Aguilera, 2005; Casello, 2007; Sung & Oh, 2011; OECD, 2012; Lee et al., 2013). The results for each metropolitan area, however, have varied according to the travel pattern that is created by the spatial structure. In other words, travel patterns are not the same in each metropolitan area. Based on this, it is often argued that different metropolitan areas should pursue different compact city policies (OECD, 2012). Furthermore, the role of municipalities is embedded in a spatially wider organization of socio-economic activities, which affects travel between and within municipalities by car and transit systems (METREX, 2010). Parr (2012) suggested that the spatial structure of a metropolitan area needs to be considered as an urban system that extends across a set of municipalities. In an urban system spread across municipalities with increasing population, the overall effect of heightening competition for space is an expansion in the city area and an increase in average density (Parr, 2012). The travel patterns of municipalities in a metropolitan area are therefore best defined through the relationship between car travel distance and transit modal split, from which each municipality can determine effective compact-city planning policies to increase either or both indexes in response to the expansion of the urban system. Veneri (2015) empirically verified that the nature of population distribution throughout an urban system differs across OECD cities and

1 The urbanized areas of the SMA were defined in this study according to Korean urban planning law (‘Act on Comprehensive Plans for Construction in the National Territory’): i.e., residential area, commercial area, industrial area, green-space area, and greenbelt.

13

Cities 70 (2017) 11–21

S. Lee et al.

Fig. 1. Municipalities of the Seoul Metropolitan Area (SMA). (Source: Yi and Lee (2014).)

Table 1 Details of the SMA (unit: 1,000,000 people, 1,000,000 households, km2). Item

Total

Seoul

Incheon

Gyeonggi

Population Household Area City (Si) County (Gun) Borough (Gu)

23.836 8.415 11,808 20 6 53

9.784 3.578 605 – – 25

2.663 0.929 1032 – 2 8

11.379 3.908 10,171 20 4 20

Source: 2010 Census in Korea, Statistics of Urban Planning in 2010, Yi and Lee (2014).

1974, the first city railroad line was completed, and there are now 19 lines with 981.5 km of expansion operating mostly between the central city and its some limited employment sub-centres (Go & Lee, 2012; Lee et al., 2013). 3.2. Travel patterns The travel patterns of 78 municipalities in the SMA were analysed using the transit modal split (TMS) and intra-city trip ratio by car (ITR) of each municipality. Note that only trips made during the morning peak hour from 7 to 9 were investigated, and all non-motorized means of transport (e.g. walking and use of bicycles) were excluded from the analysis. The TMS was calculated as the ratio of the number of trips made by mass transit relative to the total number of trips occurring in a municipality. The mass transit systems in the study were urban rail and bus, which were considered separately (see Eq. (1)):

TMS =

Fig. 2. Urbanized areas and city railroad of the SMA.

Tijm1 Number of trips from municipality i to j by rail. Tijm2 Number of trips from municipality i to j by bus. Tijmt Number of trips from municipality i to j by all modes. mt Mode of transport: rail (t = 1), bus (t = 2), and car (t = 3). The TMS ratios calculated for each municipality using Eq. (1), as shown in Fig. 3, demonstrate that those municipalities that are connected to the central city by urban rail have a higher TMS, whereas

∑ Tijm1 + ∑ Tijm2 ∑ Tijmt

(1) 14

Cities 70 (2017) 11–21

S. Lee et al.

Fig. 3. Transit modal split of municipalities in the SMA.

from the inter-zonal trip distance by car and intra-zonal trip distance within a municipality using the following equation:

those located further away have a relatively low transit modal split. The municipality of Pyeongtaek-si circled in Fig. 3, for example, has a very low transit modal split despite being connected by urban rail. A curve drawn on the basis of distribution (see Fig. 3) reveals that, the TMS gets smaller as the distance from the centre of the central city to the centre of each municipality increases. The intra-city trip ratio by car (ITR) was adopted as a simple index for determining the distance travelled by car, and represents the proportion of trip distance within a municipality (as intra-zonal trips in Fig. 4 according to the concept of the compact city by Newman and Kenworthy (1989)) to the total trip distance to all municipalities (as inter- and intra-zonal trips in Fig. 4). The limitation of this index is that the average distance by car can in fact be much shorter than the index depending on the size of the municipality. As with TMS, only trips generated in the morning peak hour from 7 to 9 were considered for ITR. The ITR for each municipality was calculated

ITR =

∑ Tiim3 ∑ Tijm3

mt

(2)

Tij Distance of trips from municipality i to j only by car (t = 3). i Origin municipality. j Destination municipality. The ITR values calculated for each municipality using Eq. (2) presented in Fig. 5 show that those municipalities in and immediately adjacent to the central city Seoul, which are connected by city railroad, have a low ITR. In contrast, the municipalities located on the outskirts of the SMA that are not connected to Seoul by the city railroad have a very high ITR. As shown by the curve in Fig. 5, this creates an opposite trend to that seen with TMS. 15

Cities 70 (2017) 11–21

S. Lee et al.

Fig. 4. Concept for calculation of intra-city trip ratio by car (ITR).

the relationship is not only consistent with the change in travel pattern in each municipality, but also the trend towards decentralization in the SMA (Jun, 2012; Go & Choi, 2013; Yi & Lee, 2014). The fact that the municipalities of Cluster 1 had the lowest ITR and highest TMS indicates that the majority of people living in the subcentres of Seoul and its adjacent municipalities work in the CBD or a sub-centre, much like in the three largest metropolitan areas in France (Aguilera, 2005). These municipalities are linked by a city railroad with an average distance of 12.5 km to the centroid of Seoul. From this, it can be deduced that the presence of a rail system is responsible for the high TMS; however, the decrease in TMS from 86.03% in 2006 to 71.48% in 2010 shows Cluster 1 follows a pattern of decentralization that is similar to that of North American urban development (Casello, 2007). The municipalities in Cluster 2 surround those of Cluster 1 (see Fig. 8), and therefore, their travel distances to the centroid of Seoul are inevitably longer. At the same time, the provision of a city railroad in these municipalities is much less than those of the municipalities in Cluster 1 (see Fig. 2). This gives the municipalities in Cluster 2 an intermediate ITR and TMS among the three clusters. Almost half of the residents in Cluster 2 commute to Seoul by car over an average distance of 27.7 km; however, the TMS increased from 42.79 to 46.03% owing to the provision of a new city railroad during the time period, which can explain the increased number of municipalities in Cluster 2. In the municipalities of Cluster 3, almost half of all trips occur within the municipality. This means that half of the residents work in the same municipality, while the other half commute to the sub-centres in Cluster 2 or even Cluster 1. The change in ITR from 76.2 to 47.57% means that trips to sub-centres outside of the municipality increased, while the decrease in TMS from 31.8 to 23.41% indicates a greater reliance on cars rather than mass transit. The fact that the mean values of TMS and ITR for the whole SMA decreased from 57.5 to 52.7% and 14.52 to 13.75%, respectively, indicates an increasing tendency towards decentral development that promotes a dependency on cars, which is consistent with global trends (Anas et al., 1998; Aguilera, 2005; OECD, 2012).

Using the common denominator of distance to the central city, a relationship between TMS and ITR can be defined as the travel pattern of the SMA. As shown in Fig. 6, this creates an inverse exponential curve (Y = 65.324e− 0.022X) when TMS is on the y-axis and ITR on the x-axis, which has a high correlation of 68.76%. This curve means that the municipalities of Seoul and those sub-centres near the central city have a well-fed public transportation system, which results in a high TMS and a low ITR for inter- and intra-zonal trips. Those municipalities on the outskirts, however, have poor transit infrastructure, which creates a completely opposite travel pattern. In comparison with the curve for 2006 (Y = 66.463e− 0.014X with a correlation of 0.5179%), the curve for 2010 as mentioned above became lower and steeper. This change can be explained by the fact -that the SMA became more decentralized (Jun, 2012; Go & Choi, 2013; Yi & Lee, 2014). According to the curves in Fig. 6, those municipalities closer to the central city exhibit a similar change between TMS and ITR. On the other hand, the municipalities on the outskirts have a contrasting change in that ITR decreased much further than TMS. Therefore, this change can be explained by the fact that the SMA became more car-dependent. 4. Empirical analysis of compact-city planning factors 4.1. Cluster analysis From the difference in travel pattern, it could be argued that the relationship between TMS and ITR represents the spatial structure of the SMA. This is empirically investigated here through cluster analysis to classify the municipalities in terms of their spatial structure and understand the changes in travel pattern caused by differences in land use and/or transportation. K-means clustering was used for this analysis as this was able to classify the municipalities more simply than a hierarchical cluster analysis (Lee et al., 2013). Seoul and its adjacent municipalities, the municipalities in the outskirt of the SMA, and intermediate municipalities (see Fig. 8) provided the three clusters for analysis using the data from 2006 and 2010 shown in Table 2. For the year 2010, the 32 municipalities belonging to Cluster 1 exhibited the lowest ITR and the highest TMS, whereas the 13 municipalities in Cluster 3 displayed the opposite behaviour. The 33 municipalities in Cluster 2 displayed a behaviour that was an intermediate between the two extremes (see Fig. 7 (left)). Between 2006 and 2010, there was only a slight change in the number of municipalities belonging to each cluster (see Table 2) and in the index value of each (see Fig. 7 (right), in which each arrow indicates the change in the two indexes of each municipality between 2006 and 2010). This change in

4.2. Regression analysis The cluster analysis identified that the municipalities within each cluster of the SMA have their own unique spatial structure features at a time period, which create a difference in travel pattern. Thus, in order to find the most effective compact-city planning factors for each municipality, these features were subjected to regression analysis to determine which factors most significantly affect the TMS and ITR of a 16

Cities 70 (2017) 11–21

S. Lee et al.

Fig. 5. ITR of municipalities in SMA.

located within easy walking distance of a major transit station (Ewing & Cervero, 2001; Evans et al., 2007; Sung & Oh, 2011; Lee et al., 2013). However, as this study is based on a spatial level of municipalities, it cannot differentiate the diversity within a station area from that of the whole city. Finally, ‘housing price’ as a proxy variable, which is not a planning factor by itself, represents the relative urban size of a municipality and is related to the increase in average distance travelled with decreasing urban size (ECOTEC, 1993). Variables belonging to transportation planning factors were selected with respect to the physical conditions influencing the choice of transport from three modes: car, rail, and bus. Economic factors such as price, income, and vehicle efficiency (Newman & Kenworthy, 1989; Gordon & Richardson, 1989) are excluded for reasons outlined in the ‘Literature review’ chapter. However, the socio-demographic factors, which may affect modal choice and travel length, were considered in

municipality. This study focused on the compact-city planning factors suggested by Newman and Kenworthy (1989), i.e. physical planning factors related to land use and transportation (see Table 3). The population and job density in a municipality are among the most representative variables for a compact city, as these indicate the spatial intensity of residents and workers. The ratio of residential and commercial floor area of buildings to the total floor area of a municipality can also provide an indication of land use in a municipality; i.e., a ‘ratio of commercial area’ that is much larger than that of other municipalities suggests a centre or sub-centre in the metropolitan area, whereas a high ‘ratio of residential area’ likely represents an origin point for trips to the central city area. The term ‘diversity’ is used to indicate a mixture of land uses in a municipality and is also related to the use of transit in terms of TOD in that it refers to high-density, mixed land use developments that place a priority on pedestrians and are 17

Cities 70 (2017) 11–21

S. Lee et al.

Fig. 6. Relationship between transit modal split (TMS) and ITR in the SMA.

and its adjacent sub-centres. These same municipalities also had the highest ‘housing price’, indicating that high prices are caused by a strong centrality and an increasing urban size of municipalities in the central city area of the SMA. The independent variables of the second group also show a relation to the central city and sub-centres of the SMA, but the extent of this is not as great as with the first group. Instead, these variables are more likely to affect the decentralization of employment and people within municipalities of the SMA.

Table 2 Clustering of municipalities for 2006 and 2010. Year

Cluster

2006

Mean of clusters

2010

Number of cases Mean of Clusters Number of cases

TMS ITR TMS ITR

Total

1

2

3

86.03 4.21 34 71.48 4.87 32

42.79 2.44 29 46.03 9.04 33

31.8 76.2 15 23.41 47.57 13

57.5 14.52 78 52.7 13.75 78

4.3. Results and interpretation Two multiple regression models for TMS and ITR were assumed for the 78 municipalities of the SMA, and the results of these model estimations are summarized in Table 5. The F-tests for each model indicate an appropriate goodness of fit, while the fact that the R-squares of both models exceed 0.7 implies that physical planning factors have a more profound impact on TMS and ITR than economic factors. It is these physical planning factors that should therefore be used to establish policies aimed at reducing car dependency in the SMA (Newman & Kenworthy, 1989; Gordon & Richardson, 1989). In the TMS model, the ‘population density’, ‘housing price’, ‘ratio of highway’, and ‘city railroad station density’, all achieve significance levels of 0.1, while the ‘ratio of high school graduates’ and ‘ratio of white-collar workers’ achieve significance levels of 0.01; in the ITR model, the ‘ratio of high school graduates’ and the ‘ratio of white-collar workers’ achieve significance levels of 0.01, both 'ratio of commercial area' and 'distance to nearest IC' achieve significance levels of 0.1, and 'city railroad' and 'diversity' achieve a significance level of 0.5 (see Table 5). The empirical fact that the municipalities in and around the central city have a higher TMS (see Fig. 3) indicates that the significant variables for TMS affect the polycentricity of urban functions in a few municipalities of Seoul and some of its adjacent sub-centres. Indeed, the decentralization variables of ‘city railroad’ and ‘city railroad station density’ play a key role in defining the transit-oriented travel pattern in the municipalities. However, although the diversity of an area near a

this regression analysis. To take into account the infrastructure features of each mode, the ‘ratio of road area’, ‘ratio of highway’, ‘distance to nearest IC’, ‘city railroad’, ‘city railroad station density’, and ‘bus stop density’ were used in the analysis. The reason for dividing the public transportation mode into rail and bus is the difference in the spatial range of the service area, i.e. the provision of city railroad is just limited to Seoul and some of its adjacent municipalities (Lee et al., 2013), whereas bus lines are located even in municipalities on the outskirts of the SMA. The term ‘city railroad station density’ is used to define the access to the service in municipalities located between the central city area and outskirts of the SMA. The ‘ratio of high school graduates’ and ‘ratio of white-collar workers’ can represent the travel behaviour of commuters as socio-demographic factors of transportation (Kenworthy & Laube, 1999; Go & Choi, 2013). The descriptive statistics for the regression analysis presented in Table 4 show that the independent variables can be classified into two simple groups: those with a significant difference between municipalities and those with a small or moderate difference. The variables belonging to the first group are the ‘ratio of commercial area’, ‘population density’, ‘job density’, and ‘housing price’. The apparent difference in these means that residences, workplaces, and commercial areas are heavily concentrated within just a few municipalities, namely Seoul 18

Cities 70 (2017) 11–21

S. Lee et al.

Fig. 7. Clusters for 2010 (left) and changes between 2006 and 2010 (right). Table 3 Variables used for the regression analysis. Category

Variable

Unit

Data source

Land use planning factors

Ratio of residential area Ratio of commercial area Diversitya Population density Job density Housing price

%

KOSIS

%

KOSIS

Transportation planning factors

Socio-demographic factors

Ratio of road area Ratio of highway Distance to nearest ICb City railroad City railroad station density Bus stop density Ratio of high school graduates Ratio of white-collar workers

– Persons/km2 Jobs/km2 10,000 Won/ m2 % % m

KOSIS KOSIS MOLIT MOI KTDB KTDB

Dummy Number/km2

KTDB KTDB

Number/km2 %

KTDB KOSIS

%

MOLIT

Sources: Korean Statistical Information Service (KOSIS) for 2010; Korea Transport Database (KTDB) for 2010; Ministry of Land, Infrastructure and Transport (MOLIT) for 2010; Ministry of the Interior (MOI) for 2010 a Diversity is calculated in this study using the Herschman-Herfindal Index, HHI (Henderson, 1997; Song, Merlin, & Rodriguez, 2013): n HHI (Herschman − Herfindahl Index) = ∑i = 1 P 2i , where Pi indicates the ratio of the floor space for building use i. The maximal value of HHI = 1 occurs when the land in a city is completely concentrated with a building use. For diversity, therefore, HHI is to be modified to 1-HHI. b The distance is calculated in this study from the centroid of municipality, i.e. from the location of municipality office to the nearest highway interchange.

Fig. 8. Clusters of municipalities for 2010.

station is able to increase TMS, the overall ‘diversity’ of the municipality often remains unchanged. Meanwhile, the statistical fact that the municipalities in the outskirts of the SMA have a higher ITR (see Fig. 5) explains why the polycentric variable of ‘ratio of commercial area’ has a negative effect on ITR. That is, the travel pattern of higher ITR in outskirt municipalities can be explained by a greater ‘diversity’ and an absence of city railroad. However, a mixture of different land uses does at least result in reduced car travel distances (Marique & Reiter, 2012). The socio-demographic variables of travel behaviour strongly affect both TMS and ITR, but in the contrasting ways; TMS is positively influenced by the ‘ratio of high school graduates’ and ‘ratio of whitecollar workers’, whereas the two ratios have a negative effect on ITR. The residents and workers in the municipalities of Seoul and its adjacent sub-centres, where city rail is available and ‘housing price’ is therefore higher, use more transit, whereas those in the outskirts

depend more on cars (Go & Choi, 2013). In order to identify the compact-city planning factors for each municipality through regression analysis and develop policies for reducing car dependency, the features of each cluster need to be considered through an integration of both analysis results. In other words, the exponential relationship between TMS and ITR (see Fig. 6) that results from the polycentric structure of the SMA has to be accounted for by examining the variables for each model by cluster.

19

Cities 70 (2017) 11–21

S. Lee et al.

Table 4 Descriptive statistics for each variable.

Dependent variables Independent variables

Variable

N

Min value

Max value

Mean

SD

Transit mode share (TMS) Intra-city trip ratio by car (ITR) Ratio of residential area Ratio of commercial area Diversity Population density Job density Housing price Ratio of road area Ratio of highway Distance to nearest IC City railroad City railroad station density Bus stop density Ratio of high school graduates Ratio of white-collar workers

78 78 78 78 78 78 78 78 78 78 78 78 78 78 78 78

14.413 1.724 5.847 0.192 0.302 47.928 13.891 135.773 1.123 0.0000 159.621 0 0.0000 3.796 13.084 18.741

81.574 80.026 93.923 36.312 0.809 26,143.078 36,963.724 1086.117 46.005 38.793 14,063.400 1 1.592 178.439 33.747 43.475

52.700 13.750 31.133 3.416 0.594 8923.034 3593.115 391.068 12.259 10.560 2003.766 0.81 0.207 47.022 26.025 31.068

18.917 16.853 21.255 4.612 0.120 7379.468 5190.990 193.446 8.602 8.059 2347.275 0.397 0.270 31.136 4.652 5.314

Table 5 Regression analysis results. Independent variable

TMS

ITR

Adj. R2 F-value Significant probability Variable

0.789 21.548 0.000 Std. coefficient

0.727 15.630 0.000 Std. coefficient

Land use planning factors

Transportation planning factors

Socio-demographic factors

⁎ ⁎⁎

(Constant) Ratio of residential area Ratio of commercial area Diversity Population density Job density Housing price Ratio of road area Ratio of highway Distance to nearest IC City railroad City railroad station density Bus stop density Ratio of high school graduates Ratio of white-collar workers

0.024 0.011 − 0.037 0.279 − 0.168 0.203 − 0.118 − 0.113 − 0.074 0.042 0.294 0.036 0.430 0.606

Sig. 0.001 0.840 0.920 0.598 0.080⁎ 0.279 0.095⁎ 0.236 0.082⁎ 0.252 0.527 0.058⁎ 0.758 0.000⁎⁎⁎ 0.000⁎⁎⁎

0.133 − 0.228 0.202 − 0.136 0.242 − 0.123 − 0.064 − 0.035 0.147 − 0.176 0.147 − 0.181 − 0.359 − 0.624

Sig. 0.000 0.324 0.063⁎ 0.014⁎⁎ 0.447 0.171 0.371 0.572 0.636 0.050⁎ 0.023⁎⁎ 0.398 0.173 0.001⁎⁎⁎ 0.000⁎⁎⁎

Sig. < 0.1. Sig. < 0.5. Sig. < 0.01.

⁎⁎⁎

the boundaries of Cluster 1 or 3 (see Table 2). Policies to increase both TMS and ITR are therefore needed in these municipalities to reduce car dependency, i.e. the policy for Cluster 1 is needed to increase TMS, whereas the policy for Cluster 3 is required to enhance ITR. The TOD planning factors are particularly important for enhancing both indexes at the same time in municipalities provided with city railroads, but care is needed to avoid implementing contrary policies that might increase ‘population density’ outside of the area of influence of the city railroad station.

For the municipalities in Cluster 1, for which increasing TMS is more important than ITR in order to reduce car dependency, ‘population density’ needs to be increased through urban development. It is, however, more important for these municipalities to promote TOD through high-density development in the areas influenced by city railroad stations. The decrease in TMS that occurred in these municipalities from 2006 to 2010 (see Table 2) can therefore be interpreted as a failure of the TOD policy. In the municipalities of Cluster 3, it is necessary to restrict any development such as large retail centres (Larkham & Pompa, 1989; Gaussier, Lacour, & Puissant, 2003) that may increase the commercial floor area or urban size, in order to maintain the current ITR even with a strong tendency for the ITR to decrease in these municipalities (see Table 2). An extension of the city railroad from the central city and subcentres to these municipalities should result in a decrease in ITR, and at the same time, might increase the TMS. The development of an increased mixture of land uses could also result in an enhanced ITR through a greater jobs-housing balance (Cervero, 1989). The relationship between TMS and ITR, and the probability of its change over time (see Figs. 6 and 7), are the most important considerations for the municipalities of Cluster 2, which are likely to cross

5. Conclusion The goal of this study was to determine why certain compact-city planning factors have different effects on municipalities of a metropolitan area. This study investigated the changed travel pattern in each municipality of the SMA based on the changed relationship between TMS and ITR. The changes in the spatial structure of the SMA between 2006 and 2010 were compared in this study using the inverse exponential curves of the relationship. As a result, the TMS and ITR of the municipalities declined together, and the ITR decreased much more in the outskirts. It was observed that the SMA became more car20

Cities 70 (2017) 11–21

S. Lee et al.

Breheny, M. (1992). The contradictions of the compact city: A review. In M. Breheny (Ed.), Sustainable development and urban form (pp. 138–159). London: Pion. Breheny, M. (1995). The compact city and transport energy consumption. Transactions of the Institute of British Geographers, 20(1), 81–101 New Series. Casello, J. M. (2007). Transit competitiveness in polycentric metropolitan regions. Transportation Research Part A, 41, 19–40. Cervero, R. (1989). Jobs-housing balancing and regional mobility. Journal of the American Planning Association, 55, 136–150. Cervero, R. (1996). Mixed land-uses and commuting: Evidence from the American housing survey. Transportation Research Part A: Policy and Practice, 30, 361–377. Council of Europe. European urban charter, congress of local and regional authorities. (2002). http://sustainable-cities.eu/upload/pdf-files/URBAN_CHARTER_EN.pdf (Accessed 08.03.16). ECOTEC (1993). Reducing transport emissions through planning. London: HMSO. Evans, J. E., IV, Pratt, R. H., Stryker, A., & Kuzmyak, J. R. (2007). Transit-oriented evelopment report 95. Washington, DC: Transportation Research Board. Ewing, R., & Cervero, R. (2001). Travel and the built environment: A synthesis. Transportation Research Record, 1780, 87–114. Gaussier, N., Lacour, C., & Puissant, S. (2003). Metropolitanization and territorial scales. Cities, 20(4), 253–263. Go, D. H., & Choi, C.-G. (2013). Commuting distance and mode choice of residents in the new developments and the existing urban areas in Gyeonggi-do - Comparing the firststage new towns, the ordinary housing land developments(OHLD), and the existing urban area. Journal of Korea Planning Association, 48(2), 83–106. Go, J. Y., & Lee, S. (2012). An appraisal of the urban scheme for sustainable urban transport. International Journal of Urban Sciences, 16(3), 261–278. Gordon, P., & Richardson, H. W. (1989). Gasoline consumption and cities: A reply. Journal of the American Planning Association, 55, 342–346. Gordon, P., Richardson, H. W., & Jun, M.-J. (1991). The commuting paradox: Evidence from the top twenty. Journal of the American Planning Association, 57(4), 416–420. Handy, S. (1996). Methodologies for exploring the link between urban form and travel behavior. Transportation Research Part D, 1(2), 151–165. Henderson, V. (1997). Medium size cities. Regional Science and Urban Economics, 27, 583–612. Jun, M. J. (2012). The effects of Seoul's new-town development on suburbanization and mobility: A counterfactual approach. Environment and Planning A, 44(9), 2171–2190. Kenworthy, J. R., & Laube, F. B. (1999). Patterns of automobile dependence in cities: An international overview of key physical and economic dimensions with some implications for urban policy. Transportation Research Part A, 33, 691–723. Larkham, P. J., & Pompa, N. D. (1989). Planning problems of large retail centres: The West Midlands County, 1987. Cities, 6(4), 309–316. Lee, S. (2001). Umweltverträgliche Stadtentwicklung für Kwangju: Alternative Strukturen der Nachhaltigkeit für eine Stadtregion in Südkorea. Dortmunder Beiträge zur Raumplanung 94Institut für Raumplanung, Universität Dortmund. Lee, S., Yi, C., & Hong, S.-P. (2013). Urban structural hierarchy and the relationship between the ridership of the Seoul Metropolitan Subway and the land-use pattern of the station areas. Cities, 35, 69–77. Marique, A.-F., & Reiter, S. (2012). A method for evaluating transport energy consumption in suburban areas. Environmental Impact Assessment Review, 33, 1–6. METREX (2010). Intra-metropolitan polycentricity in practice: Reflections, challenges and conclusions from 12 European metropolitan area. Glasgow, United Kingdom: Nordregio. Newman, P. W. G., & Kenworthy, J. R. (1989). Gasoline consumption and cities: A comparison of U.S. cities with a global survey. Journal of the American Planning Association, 55, 24–37. Newman, P. W. G., & Kenworthy, J. R. (2006). Urban design to reduce automobile dependence. Opolis, 2(1), 35–52. OECD (2012). Compact city policies: A comparative assessment. OECD Green Growth Studies: OECD Publishing. Parr, J. B. (2012). Spatial-structure differences between urban and regional systems Special issue paper. The Annals of Regional Science, 49, 293–303. Schwanen, T., Dieleman, F. M., & Dijst, M. (2001). Travel behavior in Dutch monocentric and polycentric urban systems. Journal of Transport Geography, 9, 173–186. da Silva, A. N. R., Costa, G. C. F., & Brondino, N. C. M. (2007). Urban sprawl and energy use for transportation in the largest Brazilian cities. Energy for Sustainable Development, 11(3), 44–50. Song, Y., Merlin, L., & Rodriguez, D. (2013). Comparing measures of urban land use mix. Computers, Environment and Urban Systems, 42, 1–13. Statistics Korea. http://kostat.go.kr/portal/korea/index.action (Accessed 10.07.14). Sung, H., & Oh, J. T. (2011). Transit-oriented development in a high-density city: Identifying its association with transit ridership in Seoul, Korea. Cities, 28, 70–82. Veneri, P. (2015). Urban spatial structure in OECD cities: Is urban population decentralising or clustering? OECD regional development working papers, 2015/01. Paris: OECD Publishing. Webber, M. (1976). The BART experience: What have we learned? The Public Interest, 45, 79–108. Wegener, M. (1996). Reduction of CO2 emissions of transport by reorganisation of urban activities. In Y. Hayashi, & J. Roy (Eds.), Transport, land-use and the environment (pp. 103–124). Dordrecht, The Netherlands: Kluwer Academie Publishers. Yi, C., & Lee, S. (2014). An empirical analysis of the characteristics of residential location choice in the rapidly changing Korean housing market. Cities, 39, 156–163.

dependent. On the basis of this, a cluster analysis of land use patterns and transportation systems confirmed that the municipalities of each cluster of the SMA have their own unique spatial structure features, which create different travel patterns. A regression analysis of the variables for each model identified the following factors as being key to any policy aimed at reducing car dependency in each municipality by cluster: For the municipalities of Seoul and its adjacent sub-centres (Cluster 1), population density is to be increased through urban development in order to reduce car dependency by enhancing TMS. Although the change in its travel pattern indicated that both TMS and ITR decreased similarly, it will be more effective for the municipalities to promote TOD via high-density developments near station areas of the city railroad. However, it should be clarified that this suggestion is not directly supported by the analysis results of this study based on the spatial unit of municipality as TOD is a micro-level land development type within walking distance of the transit station. In municipalities on the outskirts of the SMA (Cluster 3), which had an urban structural change resulting in a decrease in ITR that was much greater than that of TMS, there is a need to restrict large-scale development that might increase the ratio of commercial floor area or the urban size, in order to maintain a high ITR. However, ITR can also be increased through development with a mixture of self-sufficient land uses. Intermediate municipalities (Cluster 2), which are subject to the fact that the urban structural change affected the decentralized concentration of urban functions in a few municipalities of Seoul and some its adjacent sub-centres, have two policy options with the similar weightings for TMS and ITR: concentrated or decentralized municipalities. The policy of increasing TMS for the concentrated municipalities is similar to that of Cluster 1, whereas the policy of ITR for the decentralized municipalities is the same as that of Cluster 3. The criteria for selecting which policy option to use is the provision of the city railroad, as TOD planning factors can also enhance ITR and TMS. These results could be applied to the establishment of polices for reducing car dependency in municipalities of the SMA that take into consideration the different features of each cluster. The relationship between ITR and TMS indexes are also applicable to identifying changes in land use pattern and transportation systems within each municipality over time, thereby allowing policies to be adjusted to suit changes in these conditions. For such a purpose, this application requires further empirical study using recent travel data that will be released in the near future. Moreover, the method used in this study could also be further improved theoretically as well as technically. Finally, the exponential relationship identified between these indexes has the potential to be applied in policy development for other metropolitan areas in Korea or other countries. Acknowledgements This work was supported by the 2015 Research Fund of the University of Seoul, by the Mid-career Researcher Program (NRF2015R1A2A2A04005886) funded by ministry of Science, and by the Architecture & Urban Development Research Program (AUDP) funded by the Ministry of Land, Infrastructure and Transport of the Korean government (17AUDP-B102406-03). References Aguilera, A. (2005). Growth in commuting distances in French polycentric metropolitan areas: Paris, Lyon and Marseille. Urban Studies, 42(9), 1537–1547. Anas, A., Arnott, R., & Small, K. A. (1998). Urban spatial structure. Journal of Economic Literature, 36(3), 1426–1464.

21