Exploring the relationship of transport network and land use patterns: an approach through weighted centrality assessment

ABSTRACT

This study investigates the relationship between transport networks and land use patterns in the Delhi Metropolitan Area by employing a weighted centrality assessment approach. Recognizing the limitations of traditional monocentric and accessibility-based urban models, this study introduces a multimodal perspective that integrates both road and rail systems. Land use patterns are scrutinized in terms of spatial distribution of Residential, commercial, Public semipublic (PSP)/institutional, industrial and other landuse. Weighted centrality measures are assessed in terms of closeness, betweenness, and straightness, applied within a weighted framework to reflect the multimodal transport characteristics. This study introduces a path evaluation function (PEF) that incorporates transport network capacity and travel time and is calibrated with a turning parameter (α). Through multinomial logistic regression, we explore the complex relationship between weighted centrality of transport network and land use patterns. Results reveal a significant relationship, underscoring that higher centrality tends to favor commercial and public-semi-public/institutional uses, whereas centrality at lower spectrum favors the development of residential areas. This research not only advances the theoretical understanding of urban planning but also offers practical insights for policymakers, emphasizing the importance of multimodal transport considerations in urban development.

KEYWORDS:

• Weighted centrality
• land use patterns
• transport systems
• accessibility
• path evaluation function

1. Introduction

The relationship between transport and land use patterns is a key topic in urban studies, dating back to early 20th-century theories. The concentric zones theory (Burgess, 1925) and sector theory (Hoyt, 1939) are foundational theories in understanding urban development. These models, while recognizing the importance of transport characteristics in urban land use, often depicted a simplified, monocentric city form where they assumed fixed transport costs and did not adequately consider the evolving growth of public and private transport systems. Since mid-1900’s, various researchers recognized the polycentric evolution of cities (Erickson, 1986; Harris & Ullman, 1945; Heikkila et al., 1989). As urban areas continue to develop and become more complex, these early models fail to capture the dynamic and polycentric nature of modern cities, where transport and other evolving factors play a crucial role in shaping urban land use and structure (Berry & Kim, 1993; Ladd & Wheaton, 1991). Constantly growing and changing urban landscapes necessitate a thorough understanding of this interdependence. Central to this understanding is the notion of accessibility, a principle profoundly shaped by transport networks (Giuliano & Agarwal, 2017). As noted by Geurs and van Wee (2004), transport infrastructure significantly affects spatial distribution and land-use patterns by influencing accessibility. The conventional understanding is that areas with better accessibility of transport networks often experience greater development intensity, whereas those with less accessibility exhibit slower rates of development (El-Geneidy & Levinson, 2006). Thus, transport networks play a pivotal role in shaping the urban form by impacting accessibility. In assessing the impact of transport on land use, accessibility measures have become a widespread tool among urban planners and researchers. These metrics provide a quantitative method for assessing the degree to which different areas within a city or region have access to the transport network and, by extension, the various destinations and opportunities to which it connects (Handy & Niemeier, 1997). There are various measures of accessibility that range from simple cumulative opportunity measures (Black & Conroy, 1977; Breheny, 1978) to gravity measures (Geurs & van Wee, 2004; Hansen, 1959; Siddiq & Taylor, 2021).

However, the application of these measures is not without limitations. Hewko et al. (2002),argued that accessibility measures often inadequately represent the nuanced differences in accessibility within an urban area. This is because the measures often rely on aggregated data, which can obscure important variations in accessibility at more granular levels, such as neighbourhoods or individual properties. This is a serious concern because it can lead to misinterpretations about the actual accessibility of specific areas. To address this limitation, an alternative approach to aggregated accessibility measures has emerged: the application of network centrality measures. Originating from the field of graph theory, centrality measures offer a way to quantify the importance or influence of particular nodes within a network. In the context of transport networks, these nodes can represent anything from individual transit stops to entire neighbourhoods or districts (Freeman, 1978). The potential of network centrality measures in this field has begun to be realized recently. Several studies have shown that these measures can provide a more nuanced and comprehensive assessment of a transport network’s impact on land use. For instance, Wang et al. (2011) proved that street centrality captures location advantage in Baton Rouge, Louisiana, and plays a crucial role in shaping land use intensity in the city. Similarly, Liu et al. (2016), Rui and Ban (2014) and Song et al. (2023) showed that street centrality plays a crucial role in shaping urban form and land use intensity in Wuhan, Stockholm and Jinan, respectively. However, one of the key issues in the above studies is the often-singular focus on urban streets/roads, purely focusing on the configuration of nodes in a network based on topology and distance (Cheng et al., 2015). This method tends to neglect crucial supply aspects such as the network’s capacity and speed. As cities become more complex, multimodal transport systems become more prevalent. By focusing solely on one mode of transport, these measures risk ignoring the interplay between different modes and their cumulative effect on centrality.

This study endeavours to bridge this methodological gap by embracing a weighted centrality assessment that holistically considers the multimodal nature of modern transport networks. Through the use of path evaluation functions, this research probes the relationship between a multimodal transport system, encompassing both road and rail, and the spatial dynamics of land use within the burgeoning Delhi Metropolitan area.

2. Materials and methods

2.1. Study area

The National Capital Region (NCR) of Delhi, encompassing 35 districts across four states with the National Capital Territory (NCT) of Delhi at its heart, is home to more than 58.1 million residents (Census of India, 2011). This figure is projected to nearly double to 11 crores by 2041, with urbanization rates rising from the current 54% to about 67% (NCRPB, 2021). This study zeroes in on the Delhi Metropolitan Area within the NCR, covering both the NCT of Delhi and the Central National Capital Region, defined by the overlap of Delhi Metropolitan Area and administrative boundaries. Figure 1, shows the location of the study area.

Figure 1. Location of Delhi Metropolitan area in the National Capital Region of India.

Despite occupying only 13% of the National Capital Region’s (NCR) land, the study area is home to about 48% of its population. This substantial portion contributes significantly to the NCR’s economic and urban development. The area presents a unique blend of historical and contemporary features, rendering it a key location for analysing the overall urban growth.

2.2. Data sources and preparation

The study primarily relies on two datasets: 1. Transport network and 2. Existing landuse data.

2.2.1. Transport network

Transport network data were sourced from multiple databases, including Open Street Maps, the National Capital Region Planning Board (NCRPB) Master Plan (NCRPB, 2021), and the Delhi Development Authority (DDA) Master Plan (DDA, 2021). The integration of data from these sources was crucial in capturing the complete spectrum of the transport network within the geographical scope of the study.

The study employed Geographic Information Systems (GIS) to construct a detailed transport network. This process involved digitally mapping the layout of roads, railways, and other transport infrastructure based on the collected data.

Next, the following steps were undertaken to enhance the data quality:

• Connection of Unlinked Segments: Unconnected links within the network were identified and connected to ensure continuity and functionality.

• Removal of Redundant Elements: Shooting and dangling links, which represent errors or unnecessary fragments in GIS data, were identified and removed to streamline the network.

• Duplicate Link Elimination: Duplicate entries in the network data were identified and eliminated to avoid data redundancy and potential skewing of analysis results.

Once the basic structure of the transport network was completed in the GIS, speed and capacity data were fed into the network. This data was sourced from Bing API and Google Maps (for lane information), providing critical information about the traffic conditions on different network links.

This study adopts a primal approach for the transport network representation. The primal graph treats intersections/stations as nodes and roads/railway line as links, enabling the integration of critical parameters such as traffic density, speed, and actual metric distances (Oberoi et al., 2018). This approach contrasts with the dual graph method, where distances are considered homogenous, failing to reflect the true variability found in a transport network (Porta et al., 2006). The primal graph’s more intuitive representation, focusing on geographic dimensions and real distances, makes it a preferred choice for the study to assess network accessibility. Figure 2, shows the network representation (links and nodes) for the study area.

Figure 2. Transport network (nodes and links) in the study area.

To streamline the network structure and reduce computational time, the study excludes local roads from consideration. This approach aligns with Indian city planning standards, where local roads are typically not a central focus in master plans (MOUD-Govt of India, 2014). The resulting network has 6,721 nodes and 8,989 links with attributes such as Link ID, length, travel time, end node ID and their corresponding coordinates.

2.2.2. Existing landuse

Similar to transport networks, existing land use information was compiled from a diverse mix of sources. This included inputs from development authorities and various master plans, supplemented by data extracted from Google Maps and insights gained through primary surveys. The collected information encompasses various aspects of land usage, reflecting the current state and ongoing developments in the study area. This comprehensive dataset was then mapped and organized using Geographic Information Systems (GIS), allowing for detailed analysis and visualization of land use patterns, distributions, and potential development trajectories.

2.3. Weighted centrality assessment

Weighted centrality assessment is an analytical approach in network analysis that determines the importance of each node by considering both the node’s connectivity and the assigned weights of the edges (Singh et al., 2020). Traditionally, centrality measures were formalized for binary networks (Freeman, 1978), which are networks where connections are either present or absent, without any gradation of actual distance or strengths. In such binary networks, centrality measures purely rely on the structural arrangement of the nodes and their connections which fail to capture any of the important variability in networks (Alves et al., 2022). However, in weighted centrality assessment, the weights on the network are considered, leading to measures such as degree centrality, closeness centrality, betweenness centrality, straightness centrality, Page rank centrality etc. This approach allows for a more accurate analysis of fully weighted networks, as it captures the characteristics of nodes in highly connected weighted networks

In the array of weighted centrality measures identified by Porta et al. (2006), three stand out as particularly significant for this study, each focusing on a distinct aspect of network connectivity and accessibility. These are: closeness centrality (C^C), which gauges how near a location is to all others in the network; betweenness centrality (C^B), which assesses a location’s role as an intermediary among others; and straightness centrality (C^S), which measures the directness of routes from a location to all other points in the network. In centrality measures, the relative importance of a node is based on its relative positioning in the network (Disney, 2020). Access to all nodes is valued at each node represents an equal potential opportunity.

2.3.1. Closeness centrality

Closeness centrality is a measure of how close a node is to all other nodes along the shortest path in the network. Closeness centrality as defined by (Freeman, 1977):

(1) $\mathit{C}\mathrm{C}=\frac{\mathit{N}-1}{\sum _{\mathit{j}\in \mathit{N};\mathit{j}\ne \mathit{i}}\mathit{d}\mathrm{ij}}$(1)

Where, $\mathit{N}$ is the number of nodes and $\mathit{d}\mathrm{ij}$ is the smallest sum of link lengths (Distance/time) throughout all possible paths in a graph from i to j.

2.3.2. Betweenness centrality

Betweenness is a feature that accounts for a bypass node that neither acts as the origin nor the destination node. The interaction of two non-adjacent nodes depends on intermediate nodes that have strategic control and influence on them. The betweenness centrality of node measures how often that node is traversed by the shortest path between the given nodal pair (Freeman, 1977). The betweenness centrality is defined as (Porta et al., 2006):

(2) $\mathit{C}\mathrm{B}=\frac{2}{\left(\mathit{N}-1\right)\left(\mathit{N}-2\right)}\sum _{\mathit{j}=1;\mathit{j}\ne \mathit{i}}\frac{\mathit{n}\mathrm{jk}(\mathrm{i})}{\mathit{n}\mathrm{jk}}$(2)

Where $\mathit{N}$ is the number of nodes, $\mathit{n}\mathrm{jk}$ is the number of shortest paths between node j and k and $\mathit{n}\mathrm{jk}(\mathrm{i})$ is the number of shortest paths between j and k that contains node i. One of the key advantages of Betweenness centrality is that it can be used to assess vertex betweenness/link betweenness (Brandes, 2008; Yoon et al., 2006) which further can be used to assess traffic volumes on the links (Wang et al., 2011).

2.3.3. Straightness centrality

Straightness centrality is the measure of deviation from the virtual straight line along the shortest distance between a pair of nodes. In other words, straightness can be used synonymously to the detour index. The straightness index measures the extent to which a place can be reached directly and efficiently, like on a straight path from all nodes present in a graph. The straightness centrality can be measured using the given formula by Porta et al. (2006):

(3) $\mathrm{CS}=\frac{1}{\mathit{N}-1}\sum _{\mathit{j}=1;\mathit{j}\ne \mathit{i}}\frac{\mathrm{dijEucl}}{\mathrm{dij}}$(3)

where N is the number of nodes, $\mathrm{dijEucl}$ ^l is the Euclidean distance between node i and j.

All these centrality assessments can be undertaken for both globally and locally (Porta et al., 2012). For global centrality, the analysis encompasses all nodes within the network, providing a holistic view of the network’s interconnectedness and the relative importance of each node within this broad context. In contrast, the local centrality approach adopts a more focused perspective. For this study, the local centrality calculations are confined to nodes within specific geographical areas. For instance, when assessing local centrality in Gurgaon, only the nodes located within Gurgaon are considered. This method is replicated for other regions, allowing for a detailed and localized understanding of centrality that highlights regional variations and characteristics within the broader network.

As mentioned in section 1, centrality measures have predominantly been applied with a focus on the spatial positioning of nodes, primarily based on their distances. This approach, however, tends to treat all network features as uniform entities, thereby neglecting crucial physical attributes such as the intensity, strength, and capacity of both links and nodes. For example, in this framework, the centrality (C^c) of a node in a Mass Rapid Transit System is considered equivalent to that of a node on a local street, provided their distances to other nodes are comparable. This oversimplification poses a significant limitation to the study. It fails to account for the distinct characteristics of different transport systems, which can markedly influence land use (Kasraian et al., 2016) and offer varying levels of accessibility to their surrounding areas. Therefore, a more nuanced approach is needed, one that incorporates these diverse system characteristics to provide a more accurate and comprehensive understanding of network dynamics and their impact on adjacent land uses. The development of path evaluation function has been a key breakthrough in this case.

2.3.4. Path evaluation function

The architecture of the transport network poses a separate array of interconnected systems that are characterized not only by their topological complexity but also by their capacity, travel time, cost etc. Path evaluation function is defined by Sosnowska and Skibski (2018), as the function that asses the path between two nodes based on two classes i.e. sum of weight of edges and number of intermediaries. For each node the borderline parameters are number of edges and sum of weights. And each class is quantified using a constant ‘α’. Opsahl et al. (2010), named this constant as turning parameter. The benefit of the turning parameter ‘α’ is the fact that it can be used for adjustment or iteration in order to find a better fit to a given application.

The measures of closeness, betweenness and straightness centrality depend on recognizing and measuring the shortest paths connecting nodes within a network. To adapt these measures for networks with varying and multiple factor weights, the initial task involved redefining the process of identifying shortest paths and calculating their distances in the context of weighted networks. The study aims to integrate two critical factors – capacity and travel time – to enhance the accuracy and relevance of the path evaluation function. To achieve this, we define a function that simultaneously considers both these elements when determining the value of a path between two nodes in the network. Herein, let’s assume that $C_{\mathit{ij}}$ and $T_{\mathit{ij}}$ be the capacity and time taken to traverse the link between Node i and Node j, then the Path Evaluation Function (PEF) can be denoted as:

(4) $\mathit{PEF}=\mathit{\alpha }\times T_{\mathit{ij}}+\left(1-\mathit{\alpha }\right)\times \frac{1}{{\mathit{C}}_{\mathit{ij}}}$(4)

In this formula, α is a turning parameter that balances the importance of travel time and capacity. By adjusting α, we can alter the emphasis placed on each factor. Herein, it is important to highlight that the path evaluation function is not a weight in itself; rather, it serves as a guiding mechanism to determine the most appropriate or efficient path between two nodes within a network. Once a path is chosen using this function, the actual distance, travel time, or other relevant metrics of this path can then be calculated.

2.3.4.1. Calibrating the turning parameter (α) in the path evaluation function

The turning parameter, denoted as ‘α’, in the path evaluation function plays a crucial role in network analysis. Proper calibration of α is essential for ensuring that the path evaluation function aligns with specific network analysis objectives and real-world conditions. The study has used the framework used in Figure 3 to derive the ‘α’.

Figure 3. Framework to assess path evaluation function and centrality measures.

The initial step involved the detailed preparation of the transport network. Each link within the network was assigned two critical attributes:

• Capacity (in passengers/hour): This attribute represents the maximum passenger capacity of a link per hour. Given the mix of road and rail systems in the study area, standardizing these units was crucial. This has been achieved by converting PCU per hour (capacity measure for road) to passenger per hour.

• Travel Time: This measures the time needed to traverse a specific link.

Using these attributes, the Path Evaluation Function (PEF) of each network link was calculated as per Equation 2(2) $\mathit{C}\mathrm{B}=\frac{2}{\left(\mathit{N}-1\right)\left(\mathit{N}-2\right)}\sum _{\mathit{j}=1;\mathit{j}\ne \mathit{i}}\frac{\mathit{n}\mathrm{jk}(\mathrm{i})}{\mathit{n}\mathrm{jk}}$(2) . This method effectively captures both the connectivity and intensity aspects of the transport network. We then examined a variety of scenarios by adjusting the turning parameter (α).

As highlighted in section 2.3.2, there is a close relationship between betweenness centrality and traffic volumes on network links. Following this, the edge betweenness for each scenario of the varying PEF (varying α) was calculated. These betweenness centrality values were correlated with actual traffic flow data on the links to identify the most optimum ‘α’. The configuration with the highest correlation was selected for further analysis.

In the final stage, this optimized α value was employed to assess local closeness, global closeness, local straightness and global straightness centrality, offering a comprehensive view of the network’s dynamics under varying conditions.

2.4. Multinomial logistic regression analysis of land use patterns in relation to transport network centrality

In addition to the centrality measures of the transport network, this study extends its analysis to examine the influence of these measures on land use patterns using multinomial logistic regression (MNL). This approach is particularly pertinent for evaluating the relationships between network characteristics (local closeness, global closeness, local straightness and global straightness centrality) and the categorical nature of land use types.

Initially, the land use data was prepared and categorized to fit the multinomial framework of the logistic regression model. This involved classifying land use into distinct categories such as residential, commercial, and industrial, among others. The centrality measures derived from the network analysis were then introduced as independent variables in the MNL model.

Moreover, to refine our understanding of each node’s influence on surrounding land use, the Thiessen polygon approach was employed. This technique allowed for the delineation of influence areas for each node within the transport network, providing a spatially explicit method to link centrality measures with specific land use types.

3. Results

3.1. Network characteristics

The study area’s transport network extends over 3,842 kilometers, offering a blend of urban and non-urban/interurban roads alongside a metro system. Figure 4 shows the transport network characteristics in the study area. Urban roads account for 70.3% of this network, with traffic flowing at an average of 31.2 km/h. In contrast, the non-urban and interurban roads, which are more open and less congested, make up 19.5% of the network and support a faster average speed of 54.1 km/h. The metro rail, critical for the daily commute, represents 10.2% of the network and operates at an average speed of 36.7 km/h. For a detailed breakdown of the transport network’s composition and speeds, refer to Table 1 below:

Figure 4. Transport network characteristics for the study area. a) represents the typology of transport network in the study area; b) represents the peak hour speeds on the identified transport network i.E roads and Metro system.

Table 1. Transport network characteristics in the study area.

Classification	Transport network	Total length (in km)	Average speed (kmph)
Urban roads	Arterial roads	706.7	35.34
	Sub-arterial roads	645.6	34.22
	Collector roads	1350.0	27.59
Non-Urban roads	Expressway	208.90	63.70
	National highway	151.51	53.30
	State highway	121.33	51.64
	Other Non-Urban roads	266.22	48.21
Metro system	Metro system	392.21	36.70

3.2. Land use characteristics

The study encompasses an area of 6,492 square kilometers. Within this expanse, approximately 25% (1,623 square kilometers) is fully developed. The remaining area comprises agricultural land, fallow fields, forests, and water bodies. In terms of developed land, residential usage predominates, covering 19% of the total area. This is followed by industrial areas (2%), institutional/public and semi-public spaces (2%), and commercial zones (1%). Additionally, transport infrastructure occupies 2% of the total study area. Figure 5 shows the existing Land use distribution in the study area.

Figure 5. Existing landuse distribution for the study area-2021.

The study area has developed in a more pronounced multinuclear development due to development of a lot of employment nodes. Established metropolitan city centres in the study area includes Connaught Place and its Extension, commercial areas in Walled City and its Extension, Karol Bagh and Gurugram city centre. Figure 6 shows the existing landuse map of the study area.

Figure 6. Existing landuse of the study area-2021.

3.3. Weighted centrality assessment

3.3.1. Derivation of path evaluation function

The analysis determines the optimal turning parameter (α) for the transport network, based on correlation assessment between calculated link betweenness and actual traffic volumes. The Figure 7 illustrates these findings, showcasing the relationship between the varying betweenness centrality’s (with varying Path Evaluation Function (PEF)) and traffic flow, and highlighting the network dynamics under different scenarios.

Figure 7. Correlation assessment for varying betweenness centrality (with varying path evaluation function (PEF)) and traffic flow.

The following figure shows the variation of correlation coefficient of different betweenness centralities with different α values and traffic volume.

Figure 8. Correlation coefficients for separate α values.

Based on Figure 8, the path evaluation function for the study area was chosen to be:

(5) $\mathit{PEF}=0.80\times T_{\mathit{ij}}+\left(0.20\right)\times \frac{1}{{\mathit{C}}_{\mathit{ij}}}$(5)

Where, $T_{\mathit{ij}}$is the time travel to traverse a link and $C_{\mathit{ij}}$ is its capacity. This suggests that while capacity is a relevant factor in evaluating paths, it is considered less critical than travel time. The inverse relationship ($\frac{1}{{\mathit{C}}_{\mathit{ij}}}$) implies that higher capacity links will result in a lower PEF value, which can be interpreted as more desirable.

3.3.2. Centrality measures

The study computed closeness and straightness centrality metrics at both global and local scales, utilizing the derived Path Evaluation Function (PEF). Analysis of Global Closeness Centrality revealed a pronounced concentration in areas such as Connaught Place and the Old City in Delhi, indicating their significant centrality within the network. In contrast, the Local Closeness Centrality exhibited a more distributed, polycentric pattern. Examination of Global Straightness Centrality highlighted prominent linearity along the nodes encompassing Connaught Place, Old City area of Delhi, and the city centres of Ghaziabad, Faridabad, and Manesar. However, the Local Straightness Centrality presented a more uniform distribution, with a notable emphasis on nodes located in the eastern part of the study area, though without a discernible uniform pattern. Figure 9, shows the maps of each of the centrality measures.

Figure 9. Centrality maps for the study area; a) shows global closeness centrality, b) shows local closeness centrality, c)shows global straightness centrality and d) local straightness centrality for the study area.

3.4. Multinomial logistic regression

Following the computation of straightness and closeness centrality measures at both local and global scales, the study progressed to a more nuanced phase of analysis. Multinomial logistic regression (MNL) was employed to discern the impact of these centrality measures on land use patterns as shown in Figure 10. For the purposes of analytical simplicity, we have restricted our analysis till Residential, Commercial, Institutional/Public Semi-Public (PSP), and Industrial landuse.

The preliminary assessment offered insightful revelations. It was observed that local closeness centrality, global closeness centrality, and local straightness centrality were significant predictors of land use outcomes. The following figure shows the snapshot of the results from Multinomial logistic regression.

Figure 10. Results of multinomial logistics regression for various centrality measures and landuse (here this represents direction of relationship with respect to commercial); landuse 1 is residential, landuse 2 is industrial and landuse 3 is PSP/institutional.

The findings underscore the intricate relationship between transport network characteristics and urban development patterns. They reveal that not only the physical presence of transport infrastructure but also its configurational properties play a pivotal role in shaping the urban landscape.

3.5. Relationship between landuse and centrality measure

Investigation into the determinants of land use patterns within the study area has yielded compelling insights. It was established that global closeness, local closeness, and global straightness are significant indicators of land use types. To synthesize these insights, a composite centrality index was constructed. This index represents a normalized average of the three centrality measures, providing a singular, comprehensive metric of network connectivity.

Figure 11 illustrates the distribution of composite centrality across the study area. After the development of this composite index, a detailed mapping of land use against the composite centrality was undertaken. This spatial analysis unravelled the prevalence of specific land use categories in relation to varying levels of centrality.

In the lower composite centrality spectrum (below 0.6), residential areas predominate, with only a fringe of other zones (here, Cantonment area and special area of New Delhi Municipal Council). As centrality increases (0.6–0.75), residential landuse maintain dominance but begin to share space with industrial land uses. Moving towards the central ranges (0.75–0.85), commercial activities emerge alongside residential, hinting at a shift towards mixed-use development as centrality rises.

Figure 11. Composite centrality.

A significant transition appears as we approach higher centrality values (0.85–0.9), where commercial zones become prevalent along with PSP/institutional landuse. At the upper echelons of centrality (0.9–1), the trend culminates with commercial land uses, particularly in district centres, with PSP/institutional zones as a common secondary use. Figure 12, shows the relationship of composite centrality and landuse in the study area.

Figure 12. Relationship between composite centrality and landuse.

This gradient of land use against centrality underscores a transition from predominantly residential areas to commercial and public-centric zones as centrality increases. To enhance the precision of our analysis, we further examined the interplay between land use and composite centrality within the confines of Thiessen polygons around various transport system nodes. By this way the study explores the distribution of landuse along the immediate confine of various transport system nodes.

3.5.1. Land use vs composite centrality along road-based systems

3.5.1.1. Urban roads

Analysing the relationship between land use and composite centrality along Arterial, Sub-Arterial, and Collector roads reveals complex interaction, where residential zones thrive at lower centrality points, and commercial and public semi-public (PSP)/institutional (Inst.) land uses peak notably at certain higher centrality junctures, indicating specific access levels conducive to commercial and public services. Figure 13 and Table 2 shows the gradient of land use against centrality along urban roads.

Table 2. Key landuse across centrality ranges along urban roads.

Arterial Roads			Sub-arterial roads			Collector roads
Centrality	Primary landuse	Secondary landuse	Centrality	Primary landuse	Secondary landuse	Centrality	Primary landuse	Secondary landuse
Less than 0.67	Residential	PSP/Inst.	Less than 0.74	Residential	Industrial	Less than 0.81	Residential	Commercial
0.67–0.83	Residential	Commercial	0.74–0.83	Residential	Commercial	0.81–1	Commercial	PSP/Inst.
0.83–0.88	Commercial	PSP/Inst.	0.83–0.92	Commercial	PSP/Inst.
0.88–0.92	PSP/Inst.	Commercial	0.92–0.98	PSP/Inst.	Commercial
0.92–1	Commercial	PSP/Inst.	0.98–1	Commercial	PSP/Inst.

Figure 13. Relationship between composite centrality and landuse along: a) arterial roads, b) Sub arterial roads and c) Collector roads.

3.5.1.2. Non-urban/interurban highways

The examination of land use in relation to centrality presents a stark contrast between urban and non-urban/interurban highways. While urban roads demonstrate a clear transition from residential to commercial and other land uses with increasing centrality, non-urban or interurban highways display a markedly different pattern. Residential land use maintains a dominant presence along these highways, regardless of centrality measures. Figure 14, shows the distribution of landuse along non-urban/interurban highways.

Figure 14. Relationship between composite centrality and landuse along non-urban/interurban highways.

3.5.2. Land use vs composite centrality along Metro systems

Examining the composite centrality and land use trends along metro systems within the study area, we notice patterns akin to those observed in urban roads, albeit with minor differences. The metro-based system presents a significant share of commercial land use across all levels of centrality, with the prevalence of commercial areas remaining consistently high beyond centrality of 0.65 and peaking beyond the centrality value of 0.81. This trend suggests that metro stations, which are typically central to urban areas, may inherently bolster commercial activities.

Conversely, residential land use shows similar results wherein it is dominant along lower composite centrality spectrum (less than 0.81). PSP land use appears less frequently and does not show the expected increase (as with urban roads) with higher accessibility that one might expect. Figure 15, shows the distribution of landuse along metro system in the study area.

Figure 15. Relationship between composite centrality and landuse along Metro system.

4. Discussions and conclusion

The findings of this study offer significant implications for urban planners and decision makers seeking to navigate the complexities of transport network impacts on land use. By using weighted centrality assessment and path evaluation functions, we have provided a more granular understanding of how multimodal transport systems relate to land use in the Delhi Metropolitan area, a rapidly urbanizing region. This study confirms that centrality of transport network is a significant driver of land use patterns, with areas of high centrality more inclined toward commercial and public-semi-public (PSP)/institutional uses. The methodological framework employed in this study, particularly the weighted centrality assessment approach, can be adapted to other urban contexts, both in developing and developed countries. Cities facing rapid urbanization and those undergoing transportation infrastructure transformations can apply similar analytical approaches to understand how these changes might influence land use patterns and urban form.

This study underscores the importance of incorporating various modes of transport into urban planning and policy-making. The use of a composite centrality index to reflect the interconnectedness of different transport systems is crucial in identifying the areas with the highest potential for development. This approach equips urban planners with a nuanced tool for forecasting and guiding urban growth, ensuring efficient use of land and transport infrastructure. The findings advocate for a land-use planning approach that aligns with sustainable transport principles. The positive association between transport centrality and commercial, institutional, and residential developments aligns with the concept of Transit-Oriented Development (TOD), which emphasizes the development of compact, walkable communities with a rich mix of landuse near quality public transportation. Implementing such models can help minimize urban sprawl, reduce carbon footprints, and enhance the quality of urban life by creating more livable, accessible, and vibrant urban spaces.

Nonetheless, the methodology employed here is subject to certain constraints. It warrants significant adaptation for applicability to intangible network systems such as airports/ports. These systems present a unique challenge because they lack the tangible, link-based structure typical of road and rail networks. Airports and ports function as critical nodes in transport networks; however, their operational dynamics differ significantly. Unlike roads or railways, where links can be physically traced and measured, the connections in air and sea transport networks are less visible and more complex. They involve a variety of factors such as air routes, flight frequencies, maritime channels, and shipping routes, which are not as straightforward to quantify in spatial terms. Adapting our methodology to these systems would involve developing new parameters and metrics that capture the essence of these networks. Second, the turning parameter ‘α’ in the path evaluation function, while effective, may require generalization to apply to different scales of cities. This limitation suggests that what works for the Delhi metropolitan area may need adjustment for other urban contexts, especially when considering cities of varying sizes and developmental stages. In addition, future studies could enhance the path evaluation function by incorporating cost considerations, providing a more comprehensive economic perspective on transport network efficiency.

In conclusion, this study provides a substantive foundation for urban planners and policymakers. It emphasizes the role of transport centrality in the multifaceted landscape of urban development and highlights the need for multimodal considerations in future urban planning endeavours. As cities continue to grow and evolve, the methodologies and insights from this research can serve as valuable guides in the quest for sustainable and efficient urban development.

Author contributions

Dr. Sewa Ram led the design of the study, provided overarching guidance, and authored key sections of the manuscript. Jagannath Das was instrumental in collecting data, analysing it, and contributing to the manuscript’s writing. All authors have reviewed and approved the final version of the manuscript for publication.

Disclosure statement

No potential conflict of interest was reported by the author(s).