TOD has been adopted as an effective sustainable planning strategy across the world to achieve the integration of transport systems and land use. Even though the basic philosophy of TOD seems to be similar in all contexts, literature has shown a great deal of variation in its application for cases (Phani Kumar et al., 2020). For example, in America, TOD focuses mainly on re-centering suburban sprawl around transit nodes (Ewing et al., 2017). In Netherlands, TOD underlines the redevelopment of existing transit neighbourhoods (Singh et al., 2017). In India, TOD is regarded as a means of channelling metropolitan growth along transit corridors (Phani Kumar et al., 2018). Recent studies have developed typologies based on the characteristics of the existing urban structure, in order to reveal the spatial variation in TOD application. TOD typology is beneficial to urban planners and decision-makers to develop more targeted strategies based on the similar characteristics of each type.

There are two main methodologies in recent literature for developing TOD typologies: qualitative description and quantitative analysis. The qualitative description involves classifications based on the geographical and functional characteristics of neighbourhoods, whereas the quantitative analysis involves classifications by the quantitative segmentation of neighbourhoods (Phani Kumar et al., 2020). With the advantages of being able to test, analyze and compare, the development of quantitative analysis becomes more and more of significance for quantifying spatial diversity of TODs (Yang et al., 2021). The best-known approach for quantitatively segmenting neighbourhoods into TOD typologies is the node-place model. Developed by (Bertolini, 1999), the model integrates environmental characteristics into two dimensions: node and place (Figure 1). The node dimension refers to transport characteristics, and the place dimension represents urban development characteristics. The basic principle of the node-place model is balancing these two dimensions to achieve a state of integration of transport and urban development (Zhang et al., 2019).

Over the past 2 decades, the node-place model has been expanded in methodology and applications. One major modification is the expansion of indicators while maintaining the two original node and place indexes. Reusser et al. (2008) developed six indicators in addition to those proposed in the original model based on expert questionnaires. Monajem and Nosratian (2015) added spatial indices which might determine the structure of station area development. In addition to adding different indicators, some scholars introduced a third dimension to extend the node place model, considering factors such as the urban design features (Vale, 2015; Vale et al., 2018; Zheng and Austwick, 2023), the functional and morphological interrelation between transportation and urban conditions (Lyu et al., 2016; Su et al., 2021; Cummings and Mahmassani, 2022), traveller-related attributes (Cao et al., 2020; Dou et al., 2021; Zhou et al., 2023), and ecological environment (Yang et al., 2023).

Although several improvements have been made to the node-place model, they still have limitations. Using indexes to aggregate information about the two or three dimensions, some important types of detailed TOD features are missed. Therefore, one possible enhancement of the node-place model is to propose a multi-axial model to perform a detailed evaluation. Caset et al. (2018) has done a pioneer study along this line and introduced the butterfly model. This model is similar to the shape of a butterfly because it has two “wings”: the left-wing contains node-related indicators, while the right-wing contains place-related indicators (Figure 2). Each wing can also be divided into different parts instead of using a simple aggregated index. In this theoretical idea, each site area consists of a node value and a place value. The node value represents the traffic supply of a site area and the place value represents the spatial development of a site area. The model increases the node value of a location by improving accessibility, thereby creating favourable conditions for further development of the location. In turn, a place develops because of its growing transport demand, which favours the promotion of further development of the transport system. It assumes that a site area functions optimally when the node and location values are balanced. The butterfly model further elaborates this point. It is based on three salient features of node value and three salient features of place value. The determinants of node value are its status in the public transport, road and cycle networks. The determinants of place value are the density of residents, workers and tourists, the mix of land uses and their centrality. The latter refers to the extent to which the station itself functions as the centre of its surroundings. The butterfly model locates the interrelationships of these six features, i.e., the node values represent the positional values of the left wing and the place values representing the right wing. The station area functions best when the two wings of the butterfly are in balance with each other. The minimum requirement for a station area is that its location in the public transport network should match the density of residents, workers and tourists.

Compared to previous approaches, the butterfly model is able to evaluate TOD characteristics in more detail. And this more disaggregate way to visualize the components of each dimension allows urban planners to better evaluate and understand the development status of metro stations.

Although the butterfly model can be applied to all major cities around the world, the indexes must be modified to reflect the characteristics of TODs in China. We optimize the butterfly model based on the characteristics of TODs in China. We have constructed three first-class indexes at the node dimension and place dimension respectively. The node dimension is linked to the accessibility of the station and can be divided into three facets: accessibility of metro mode, accessibility of bus feeder mode, and accessibility of bicycle feeder mode. In terms of the place dimension, we add environmental features and spatial parameters into it and recategorize the place dimension following the ‘three-Ds’ principles (density, diversity, and design) discerned by (Cervero and Kockelman, 1997). Finally, the butterfly model with six indexes’ system is established, which is constituted by two levels: 6 first-class indexes and 21 second-class indicators. Below, a more thorough discussion of the indicators’ operationalization will be provided.

Figure 3 illustrates the butterfly model we developed for the context of China. According to the TOD theory, the Chinese 10-min life circle (Peng et al., 2023), and observations of walking distance in Tianjin (Wu and Yan, 2020), this study’s analysis threshold of distance is defined as 600 m. Eleven second-class indicators of node dimension are presented in Table 1. Table 1 also includes the ten second-class indicators of place dimension.

X i j ′ = X i − ⁡ min ⁡ X i j max ⁡ X i j − ⁡ min ⁡ X i j  positive  max ⁡ X i − X i j max ⁡ X i j − ⁡ min ⁡ X i j  negative  (2)

Cluster analysis is then applied to identify the classes which have common TOD features. Two-step cluster analysis is employed in this study in line with (Phani Kumar et al., 2020). Compared with k-means clustering (Zhang et al., 2019) and hierarchical clustering (Rodríguez and Kang, 2020), this method has two advantages. First, it can deal with categorical and continuous variables efficiently. More importantly, two-step cluster analysis can automatically determine the optimal cluster numbers. Before cluster analysis, we use the Min-Max standard method to normalize and standardize all the indicators between 0 and 1. In Eq. 2, Xij is the value of indicator i for station j, max Xij is the maximum value and min Xij is the minimum value of indicator i for all stations. Positives represent a greater value with a positive contribution and negatives denote a higher value with a negative contribution.

Almost all indicators are positively correlated with the logic of the butterfly model, with the exception of A3. Then, each first-class index is calculated by the average of all second-class indicators referring to them.

The twenty-three stations have been classified into five types by the two-step cluster analysis, namely, urban TODs, balanced TODs, specific TODs, future TODs, and unbalanced TODs. Each of the five categories has their own unique characteristics. Their spatial distributions are illustrated in Figure 5, while their corresponding numerical distributions and realistic circumstance of the representative stations of each cluster are displayed in Figure 6. Table 2 reports the levels of TOD indexes of each cluster.

TOD profoundly influences urban development, enhances transportation efficiency, boosts economic vitality by attracting businesses and creating jobs, optimizes land use, fosters community cohesion, improves environmental quality, and increases property values. Overall, TOD serves as a catalyst for sustainable urban growth by addressing transportation, economic, social, and environmental aspects. The TOD typology of metro stations provides valuable insights into TOD planning, equipping city managers and planners with a useful tool for developing more targeted strategies and avoiding the ‘one-size-fits-all’ solution. On the whole, the TOD development of Tianjin Metro Line 1 shows a “high-medium-low” spatial disparity from the city centre to the suburbs, presenting an obvious urban transect. Relative to its transportation supply, the degree of development of Cluster 3 is exceptionally low and it might be given priority by urban planners to maximize the benefits of investment. This may include increasing public service facilities, improving transportation connectivity, and attracting commercial investments. Elevating the development in this area could effectively balance urban development disparities and maximize investment returns. Cluster 1 has the highest indexes and the development potential has been fully exploited. For this cluster, it can be regarded as an organism integrating the metro and land use planning. “Urban Acupuncture”, which is a strategy of urban renewal focusing on the key urban area, could be used to stimulate urban vitality and emphasis should shift towards sustainable development measures, emphasizing the integration of urban planning and environmental conservation to ensure long-term sustainability. Although exhibiting a balanced butterfly visualization, cluster 2 can achieve a more ideal development pattern by improving pedestrian-friendly environment. Improving the quality and convenience of sidewalks would further optimize its development pattern, enhancing the region’s attractiveness and residents’ quality of life. Regarding Cluster 5’s relatively low density, exploring high-density development strategies is suggested. Advocating for moderate high-density development may enhance accessibility and attractiveness, promoting TOD’s advancement in the area.

In this paper, we use the revised and optimized butterfly model to assess the TOD typologies in the metropolitan context of China. Although the original node-place model is very simple and powerful as a framework for investigating the interaction between transport and land use, details of some important characteristics are disregarded by using indexes to aggregate node and place values. Using the multi-axial approach to extend the node-place model, the butterfly model can perform a more detailed evaluation and generate a better understanding of TOD characteristics. An empirical case study of Tianjin Metro Line 1 demonstrates the effectiveness of the approach.

This study obtains three major findings. First, the analysis reveals that stations in Tianjin Metro Line 1 have good integration between the metro system and land use, indicating that TOD principles have already been implemented. However, some stations need more attention because of the mismatch between the node and place value. Second, we identify five TOD typologies, which reflect different degrees of metro network accessibility, bus service, bicycle service, density, diversity and design, and they present significant spatial variations. The number of TOD typologies in this study is in line with the previous study (Wu and Yan, 2020), but more details and TOD characteristics are found under the modified butterfly model. Cluster 1 is located in the centre of the city with the highest scores on all indicators, whereas cluster 4 is located at the ends of the metro lines and the suburb areas with the low scores on all indicators. Cluster 2 is similar to cluster 1 with a balanced development status, but the scores are much lower. Cluster 3 includes only one station, namely, Tianjin West Railway Station. As an urban transportation hub, it has not developed too many urban functions except the transportation function. As for cluster 5, the values of node indexed are much higher than that of place indexes, exhibiting an unbalanced situation. Third, our focus on the metro corridor finds that stations with high TOD index values are located in the urban centre, and the TOD index values show high-medium-low spatial disparity from city centre to suburb areas. It means that TOD constructions in Tianjin’s central urban area are very effective.

This study presents several limitations. First, the differences in the weight of the indicators are not considered in the construction of the indicator system, and methods such as Analytic Hierarchy Process and Entropy Weighting (Yang et al., 2023) can be used in future research to identify the weight of each indicator. Second, focusing on one corridor Tianjin Metro Line 1, we do not consider the differences between corridors, and comparative analysis among different corridors can be conducted in the future. Finally, the aspect of the population, such as resident and employee also have not been taken into consideration in this study and this aspect should not be ignored in the future study.