A Data-Driven Framework for Sustainability and Ergonomic Design of Urban Cycling Networks in the Métropole Européenne de Lille

Abstract

Sustainable urban mobility is gaining importance as cities seek to address congestion and environmental concerns, with cycling infrastructure being an essential component of urban transportation systems. This study proposes a novel integrated, data-driven modeling framework that uniquely combines sustainability and ergonomic design to evaluate and optimize urban cycling networks. A computational model incorporating graph-based analysis, isochrone mapping, and network discontinuity identification was used to assess cycling safety and accessibility within MEL. The findings highlight significant accessibility shortcomings caused by network discontinuities, unsafe segments, and missing links—issues frequently overlooked in conventional cycling network planning. Key employment centers in MEL were found to have limited cycling access, highlighting the need for cross-regional connectivity. The study suggests that targeted micro-interventions and improved connectivity can improve the sustainability and ergonomics of urban cycling networks. The methodological framework developed is scalable and adaptable, making it applicable to other metropolitan regions. This study offers actionable insights for urban planners, advocating for data-driven decision-making and micro-scale network improvements to create a more connected, efficient, and inclusive cycling network.

1. Introduction

Sustainable urban mobility is becoming increasingly important as cities aim to reduce congestion, improve air quality, and promote healthier lifestyles. Sustainable urban mobility refers to transportation systems in cities that are environmentally friendly and do not harm the environment or negatively impact social well-being [1]. This includes providing transportation options that reduce pollution and improve air quality while ensuring accessibility and efficiency for all users.

Building on this foundation, policymakers and researchers seek to identify and analyze the key factors that facilitate or hinder the successful adoption and implementation of sustainable urban mobility solutions [2]. Achieving a comprehensive understanding of the enablers and barriers is essential for cities to develop, maintain, and evaluate environmentally sustainable and efficient transportation systems. Multi-scale evaluations of cycling networks offer valuable insights into user-friendliness and data integration methods [3,4]. Methodological frameworks have been proposed to assess the cost-effectiveness of Sustainable Urban Mobility Plans and to evaluate mobility sustainability patterns across various urban contexts [5,6,7]. Furthermore, network structure and topological analyses provide critical perspectives on urban mobility readiness and sustainability [8,9]. Reviews of emerging technologies, such as electric and autonomous vehicles, highlight their influence on new paradigms of sustainable urban mobility [10], while case studies emphasize discrepancies between policy formulation and the practical application of sustainability indicators in urban settings [11].

To understand the multifaceted impacts of urban mobility, it is essential to consider sustainability’s three interconnected pillars—environmental, social, and economic—which are closely linked to transport systems [12]. Car transport infrastructure is a key determinant of the economic pillar, influencing economic performance and productivity through the facilitation of efficient mobility and the movement of goods and people. It also impacts the social pillar by enabling people to perform essential daily activities. However, car transport produces negative externalities, such as congestion, pollution, noise, and greenhouse gas emissions, which pose significant social, environmental, and health risks [13,14,15].

Cycling, along with walkability and green space access, is identified as a key indicator of urban design that supports public health [16]. Cycling network is a key component of urban transportation networks, providing an efficient alternative to car use [17]. The availability of bicycle lanes encourages more people to cycle, which has positive impacts on public health, such as increased physical activity [18]. Cycling offers valuable daily physical activity that increases caloric expenditure and falls within the moderate-intensity range, providing health benefits and potentially influencing overall health and longevity [19]. Implementing a cycling and walking network is associated with modest reductions in transport-related carbon dioxide (CO₂) emissions over a short period [20]. Building and maintaining bicycle infrastructure is less expensive than building and maintaining car infrastructure. As a result, investing in cycling networks offers significant benefits to urban areas and represents a cost-effective strategy for improving transportation options, public health, and sustainability [21].

Given these concerns, safety emerges as a critical factor influencing the viability of cycling as a transport mode [22]. Cyclists frequently share the roadway with motor vehicles, which poses significant safety risks—particularly during overtaking maneuvers when adequate lateral clearance is not always ensured [23]. Segregated cycleways have been shown to increase cycling volumes without raising accident rates, underscoring the critical role of a protected network in ensuring cyclist safety [24]. Cycling is often perceived as unsafe due to several factors, including inadequate infrastructure, insufficient separation from motor vehicles, long distances between origins and destinations, unfavorable weather conditions, and a lack of end-of-trip facilities, such as secure bicycle parking [3,22,25]. A cyclist typology classifying adults into four categories based on their comfort with cycling, interest in using it for transportation, and physical ability has shown that 56% of respondents in a survey (n = 908) fell into the “interested but concerned” group, identified as the key target for increasing cycling adoption [26]. Findings suggest that reducing traffic speed and increasing physical separation from motor vehicles could enhance comfort and cycling rates. Strengthening cyclist safety through protected infrastructure remains a key priority for improving cycling conditions in urban areas [27].

However, the lack of cycling infrastructure raises concerns about urban planning priorities and their alignment with global sustainable development goals. Despite the widely recognized benefits of cycling as a sustainable, non-polluting, and health-promoting mode of transportation, its integration into urban environments is often overlooked. This gap in planning and implementation underscores the need for targeted infrastructure development. Therefore, designing cycling networks that prioritize connectivity and safety is essential for a sustainable urban mobility system [3,15,21,28]. By improving connectivity, cycling networks enable efficient travel between key destinations, such as residential areas and employment centers. Furthermore, improved access to employment centers supports economic mobility and reduces commuting barriers.

Today, there is limited knowledge about the specific challenges within existing cycling network topologies in metropolitan cities. While cycling networks appear to have potential, discontinuities in connectivity and disparities in accessibility may exist, but they have not been comprehensively assessed. These potential issues indicate that cycling accessibility, such as access to employment centers, may be influenced by factors beyond distance, including network continuity and safety, which are often overlooked in traditional network planning.

Additionally, there is a lack of a comprehensive modeling framework that integrates both sustainability and ergonomic design while considering human factors in cycling networks. At its core, ergonomic design involves continuously incorporating the human factor into system design to ensure usability, comfort, and safety. Ergonomic design enables the design of systems that are human-centered, functional, and sustainable [29]. For urban cycling networks, ergonomic design refers to the development of urban cycling networks that are aligned with human needs, prioritizing usability and safety. It acts as a bridge between network planning and user experience by explicitly considering how cyclists interact with the physical network. This perspective goes beyond traditional urban cycling network design by embedding human-centric criteria—such as safety and accessibility to employment centers—into the computational modeling that informs the design process. Such a framework would enable a more data-driven, detailed evaluation of cycling conditions, helping to identify the extent of accessibility shortcomings and areas in need of improvement, as well as guiding efforts toward the development of a more sustainable and efficient cycling network.

This study aims to address these gaps by proposing a novel, integrated modeling framework for the ergonomic design of urban cycling networks through a data-driven approach. Unlike previous approaches, we propose a scalable framework that integrates conceptual, functional, physical, and process domains to optimize the cycling network in the MEL region (Métropole Européenne de Lille), using advanced Geographic Information Systems (GIS) techniques, graph-based modeling, and isochrone analysis. By focusing on computational modeling, it aims to support the ergonomic design of an urban cycling network using real-world data. This framework specifically examines how network discontinuities affect accessibility and investigates the interaction between the distribution of employment centers and the cycling network’s connectivity. By integrating these elements, the study provides a comprehensive, data-driven analysis intended to inform targeted improvements in the ergonomic design of sustainable urban cycling networks. With a population of 1.15 million, the selection of MEL is justified by its strategic significance in European sustainable mobility planning. It serves as a suitable case study for developing and testing models aimed at addressing cycling network challenges common to many metropolitan areas. The availability of detailed spatial and infrastructural data supports the rigorous application of such frameworks.

Cycling network computing and analysis involves related but distinct concepts such as connectivity, reachability, and accessibility. To ensure clarity, these terms are used as follows in this study:

• Connectivity is a topological property of the network structure that describes the extent to which nodes are physically linked by paths or edges. It characterizes the global structure of the network graph—typically modeled as an undirected or directed graph—and reflects its redundancy, resilience, and potential for multiple routing options. Connectivity does not depend on origin-destination pairs or user-specific attributes. High connectivity implies a greater number of route options and possible connections between points, regardless of practical constraints like distance or time.
• Reachability is a binary property defined for a pair of nodes in the network. A destination node is said to be reachable from an origin node if there exists at least one continuous path in the network graph from the origin to the destination. This property considers only the existence of a path and does not incorporate factors such as cycling time, safety, or comfort. It describes whether a particular location or set of locations can be reached from a given origin via the existing network, typically considering only the presence or absence of a continuous path (not its quality or efficiency).
• Accessibility is a concept that captures the ease with which a user can reach destinations from a given origin within the network graph. It combines the notion of reachability with user-relevant factors such as cycling time, network quality, safety, comfort, and suitability of the network. It reflects how effectively the network enables practical access to spatial opportunities (e.g., jobs, services, amenities). Accessibility thus reflects the actual ability of users to access opportunities and services, not just the existence of a path.

Table 1. Comparison of connectivity, reachability, and accessibility.

Term	Type	Focus	User Dependency
Connectivity	Structural property	Global network topology	No
Reachability	Binary origin-destination property	Existence of paths	No
Accessibility	User-centered measure	Practical ease of reaching destinations	Yes

provides a comparative classification of these concepts, specifying their type, focus, and whether they incorporate user-centered attributes.

For urban planners, the proposed modeling framework provides a clear, data-supported foundation for decision-making, enabling them to prioritize interventions that address critical gaps in the network. Policymakers can benefit from gaining a thorough understanding of the cycling network, including the areas of the network where investment is needed to develop and improve the infrastructure. Additionally, the methodology is scalable and applicable to other metropolitan regions, offering a replicable model for advancing sustainable urban mobility.

The remainder of this paper is structured as follows: Section 2 describes the proposed framework; Section 3 presents the computational methods; Section 4 discusses the results, including principal findings, critical analysis, spatial recommendations for the MEL area, and implications for the sustainable ergodesign of the cycling network; and Section 5 provides the conclusion, highlighting the overall contributions of the study and outlining directions for future research.

2. Proposed Modeling Framework

The proposed framework follows a structured engineering design process, integrating multiple domains to model, analyze, and optimize cycling network s, with a particular focus on sustainable urban mobility (Figure 1).

This framework is organized into four key domains [30]:

User Domain (Conceptual Model): Identifies user needs and sustainability objectives. These broad objectives are transformed into more specific, quantitative performance requirements that can be measured and evaluated in the following stages of the cycling network design.

Functional Domain (Conceptual and Mathematical Models): Translates user needs into measurable functional requirements. In doing so, the framework ensures that the cycling network design aligns with user goals and expectations.

Physical Domain (Graph-Based Mathematical Model and Computational Model): Represents the cycling network mathematically and processes it computationally.

Process Domain (Experimental and Computational Models): Applies results to improve network design and inform decision-making.

Each domain is elaborated in the following sections.

2.1. User Domain (Conceptual Model)

The User Domain answers the question: “What do the users of the cycling network need to accomplish?”. It represents the fundamental starting point of the framework, focusing on the desires, needs, and expectations of the end-users of the urban cycling network. In this context, the user is considered an active participant in the design and evaluation of urban mobility infrastructure.

One of the primary considerations in the User Domain is sustainability. The shift towards encouraging active mobility—such as cycling and walking—is not only a response to climate change but also a critical step towards reducing the environmental impact of transportation. Active mobility contributes to lower carbon emissions compared to private car use, helping metropolitan areas such as the Métropole Européenne de Lille (MEL) to meet their sustainability targets. This is especially important in densely populated urban areas where the carbon footprint of transportation is disproportionately high. Consequently, policies that promote cycling and walking over driving align with global sustainability goals, including reducing air pollution and reliance on fossil fuels.

In addition to sustainability, ensuring that the cycling network is accessible for cyclists is a cornerstone of the User Domain. An essential aspect of accessibility is the concept of ‘safe accessibility’, ensuring that cyclists can safely access key destinations, such as workplaces, educational institutions, parks, or public transport stations, within an acceptable amount of time. For instance, the design of the cycling network should consider the proximity of bicycle lanes to these destinations and minimize detours or obstacles. Ensuring that the network supports a wide range of users, including both seasoned cyclists and beginners, is essential for maximizing the potential of active transportation.

Therefore, a fundamental requirement in the User Domain is the creation of a cycling network that prioritizes the safety and comfort of cyclists. This includes designing bicycle lanes that are separated from high-traffic zones to minimize the risk of accidents. Cyclists should feel safe using the cycling network, which in turn encourages more people to opt for cycling as a regular mode of transportation. Comfort, too, plays a pivotal role—smooth, well-maintained roads and well-designed intersections improve the cycling experience and contribute to the attractiveness of cycling as a viable mode of transport.

As an illustration, a key user requirement derived from these broader considerations in the context of the Métropole Européenne de Lille (MEL) is: “A cyclist should be able to safely reach their destination within 30 min to consider cycling a viable option”. This time threshold, such as the 30 min cycling limit, serves as a critical benchmark for the design and optimization of the urban cycling network. If the network does not allow for reasonable cycling times to essential destinations, cyclists may be discouraged from using it, undermining the adoption of cycling as a sustainable and efficient mode of transport.

For the Métropole Européenne de Lille (MEL), these qualitative needs—safety and accessibility—form the basis for the Functional Domain.

2.2. Functional Domain (Conceptual and Mathematical Models)

The Functional Domain answers the question: “What must the cycling network do?”. It serves as the bridge between the qualitative user needs defined in the User Domain and the formal mathematical metrics used to model and design the network. This domain translates abstract user requirements into precisely defined, mathematically formal performance criteria that guide the development of the cycling network. While the model is formal and structured, it remains abstract, as it is not yet instantiated with real-world data. It ensures that the cycling network satisfies user expectations for safety and accessibility, which are essential for supporting cycling as a sustainable and practical mode of transportation.

2.2.1. Functional Requirements (Conceptual Model)

The Functional Domain translates key user needs into specific, measurable performance requirements, ensuring the network remains user-centered.

Some of the most critical functional requirements include maximum cycling time, cycling speed, connectivity threshold, and network density. Maximum cycling time ensures that cycling to a destination remains a viable option. A cyclist should be able to reach their destination within a reasonable time. Based on user preferences identified in the User Domain, the goal is to ensure that cycling routes do not exceed a maximum cycling time of 30 min. By establishing this time threshold, the design of the network ensures cycling remains competitive with other transport options, making it more likely that users will choose bicycles for their commutes.

Benchmark ranges for cycling speeds have been reported, but these values are derived from aggregated international studies and should not be interpreted as universal constants. For instance, for adult bicyclists riding on level roadways, a typical range of 19 to 26 km/h (12–16 mph) has been suggested [31]. However, average bicycling speeds vary widely depending on the environment, country, and individual factors. In France, for example, average cycling speeds differ between cities: Paris and Lyon, which are very dense, show annual average speeds between 13 and 13.5 km/h, whereas in Nantes, Toulouse, and other less dense cities, average speeds are higher, around 15.5 to 16.5 km/h [32]. This variability makes context-specific calibration essential when applying speed benchmarks in network modeling or accessibility analyses.

In our study, for the MEL region, the assumed cycling speed is 15 km per hour (km/h), reflecting the city’s intermediate urban density. This value represents a contextually reasonable estimate based on the available French city data. This average speed takes into account urban conditions, such as stop signs, traffic, weather, and terrain. It reflects the typical speed at which cyclists can comfortably travel, ensuring the network design accommodates a realistic pace for urban cycling. By using this speed and the maximum cycling time, we can estimate the distance cyclists can cover.

The connectivity threshold is a crucial requirement for a cycling network because it must be fully connected. A connected network allows cyclists to travel between any two points in the city without encountering dead ends or disconnected segments. High connectivity also improves accessibility by making it easier and more practical for individual cyclists to reach key destinations via direct, efficient, and safe routes. Without sufficient connectivity, cyclists may be forced onto longer or inconvenient paths, reducing both the functionality and attractiveness of the network.

Network density is a crucial factor in determining accessibility. A higher density means more nodes (destinations) are available within a given area, making cycling a more practical and attractive transportation option. By analyzing node distribution, we can evaluate whether the network provides sufficient destinations within a reasonable distance and multiple route options, improving the overall usability for cyclists.

2.2.2. Mathematical Model of the Functional Domain

The Functional Domain represents the cycling network abstractly as a mathematical object, without reference to real-world datasets. The mathematical model ensures that the network meets the established performance criteria. It provides a systematic way to analyze connectivity, reachability, accessibility, kinematic parameters (cycling times, speed, and distance), and network density.

The kinematic parameters—average cycling speed and maximum cycling time—represent functional requirements that define user-centered problems in the network. In this study, as previously argued, the average cycling speed is assumed to be $v=15\mathrm{k}\mathrm{m}/\mathrm{h}$ and the maximum acceptable cycling time is ${\tau }_{max}=0.5\mathrm{h}$.

Mathematically, the cycling network can be represented as a directed attributed multigraph:

$$G=(V,E,s,t,{\varphi }_V,{\varphi }_E)$$

(1)

where

-

$V$ is the set of nodes corresponding to intersections or destinations.

-

$E$ is the set of edges representing cycling segments.

-

$s,t:E\to V$ assign the source and target of each edge (directionality).

-

${\varphi }_V:V\to A_V$ assigns attributes to nodes (e.g., type).

-

${\varphi }_E:E\to A_E$ assigns attributes to edges (e.g., cycleway type, travel time).

-

$A_V$ is the attribute space for nodes, e.g., intersection type, node importance, coordinates, etc.

-

$A_E$ is the attribute space for edges, e.g., cycleway type, edge length, travel time, slope, surface type, one-way flag, etc.

Let note $G_{und}$ undirected graph of $G$ defined as follows:

$$G=(V,E_{und})$$

(2)

where $E_{und}=\left\{\right\{s\left(e\right),t\left(e\right)\}:e\in E\}$.

Connectivity, a structural property of the network, is satisfied if for every pair of nodes $\left(u,v\right)\in V$, there exists at least one path $P_{uv}$ consisting of edges in $E$. Formally, it can be expressed as follows:

$$\forall u,v\in V,\exists P_{uv}=\left\{e_1,e_2,\dots ,e_k\right\}\subseteq E,k\ge 1$$

(3)

such that

$$s\left(e_1\right)=u,t\left(e_k\right)=v,t\left(e_i\right)=s\left(e_{i-1}\right),\forall i=1,\dots ,k-1$$

(4)

Connectivity is independent of user parameters and ensures that the network is fully continuous.

Reachability is a binary origin-destination property. A destination $v\in V$ is considered reachable from an origin $u\in V$ if there exists a directed path $P_{uv}$ such that

$$s\left(e_1\right)=u,t\left(e_k\right)=v,t\left(e_i\right)=s\left(e_{i-1}\right),\forall i=1,\dots ,k-1$$

(5)

This can be expressed as a binary function:

$$R(u,v)=\left\{\begin{array}{cc}1,& if{P}_{uv}exists\hfill \\ 0,& otherwise\hfill \end{array}\right.$$

(6)

Reachability depends on connectivity but is independent of user parameters such as travel speed or travel time.

Accessibility is a user-centered measure that incorporates both network structure and cyclist parameters. For instance, a destination $v\in V$ is considered accessible from an origin $u\in V$ if there exists a directed path $P_{uv}$ whose total cycling time does not exceed the maximum allowable time ${\tau }_{max}$:

$$A(u,v)=\left\{\begin{array}{cc}1,& if{\sum }_{e\in E}{\tau }_e\le {\tau }_{max}\hfill \\ 0,& otherwise\hfill \end{array}\right.$$

(7)

where ${\tau }_e$ the cycling time along each edge.

Accessibility explicitly incorporates cyclist-specific parameters, such as speed and maximum cycling time, thereby providing a practical and user-centered measure of network performance.

Finally, the set of all accessible destinations from a given origin $u\in V$ is determined as follows:

$$V_{accesible}\left(u\right)=\left\{v\in V\left|A\right(u,v)=1\right\}$$

(8)

By evaluating $V_{accesible}\left(u\right)$ for all nodes $u\in V$, we obtain the spatial distribution of reachable destinations across the network. This information forms the basis for isochrone analysis, which quantifies the density of reachable nodes within specific time thresholds, thereby defining network density. Isochrone-based evaluation provides a functional requirement for ensuring that cyclists have access to a sufficient number of destinations within practical travel limits.

A key functional attribute introduced in this study is cycling safety, which represents the relative level of protection offered to cyclists along each edge. Safety is first expressed as a linguistic variable $S_l\left(e\right)$ for each edge $e\in E$ with values:

$$S_l\left(e\right)=\left\{VeryLow,Low,Moderate,High\right\}$$

(9)

These correspond, respectively, to shared roadways without dedicated infrastructure, painted bicycle lanes or shoulder-level accommodations, segregated bicycle lanes adjacent to roads, and fully separated, exclusive cycling infrastructure.

To enable quantitative analysis, these linguistic values are then transformed into a numerical safety degree $S_d\left(e\right)$ on a discrete scale from 0 to 3. Finally, for visualization and intuitive interpretation, each safety level is associated with a distinct color code along the network edges, for instance: yellow for 0, light green for 1, blue for 2, and black for 3. This mapping allows both analytical evaluation and spatial representation of cycling safety across the network.

2.3. Physical Domain (Mathematical Model and Computational Model)

The Physical Domain answers the question: “How does the system operate physically?” It translates the functional requirements into a graph-based representation of the actual cycling network, where the network’s physical properties are measurable and computable.

2.3.1. Mathematical Model

The Physical Domain concerns the real-world instantiation of the cycling network, derived from geospatial datasets such as OpenStreetMap (OSM). It provides a data-grounded graph representation of the transport infrastructure of a given city or region, specifically the Métropole Européenne de Lille (MEL). This representation captures the actual nodes (intersections, junctions, access points) and edges (road or cycle segments), along with their attributes such as length, slope, surface type, and bicycle accessibility, enabling a concrete and measurable basis for analyzing and optimizing the cycling network.

Unlike the Functional Domain, which represents the cycling network as an abstract directed attributed multigraph encoding connectivity, reachability, accessibility, and density conceptually, the Physical Domain represents the actual, measurable network as observed in the real world. Thus, the Physical Domain represents the real-world instantiation of the cycling network for the MEL by mapping the functional mathematical model onto geospatial data. The directed attributed multigraph (Equation (1)) defined in the Functional Domain, is instantiated using OSM data as:

$$G_{road}^{MEL}=(V_{road}^{MEL},E_{road}^{MEL},s,t,{\varphi }_V,{\varphi }_E)$$

(10)

where $V_{road}^{MEL}$ is the set of nodes corresponding to intersections, way endpoints, or access points extracted from OSM in MEL, and $E_{road}^{MEL}$ is the set of directed edges representing navigable road segments between them. The functions $s,t$ assign the source and target of each edge, thereby capturing the directionality of one-way streets and parallel lanes. Each node and edge is equipped with attributes through the mapping ${\varphi }_V$ and ${\varphi }_E$. The edge attribute space $A_E$ contains physical properties obtained from OSM. In addition, we derive a new attribute, the cycling time along each edge ${\tau }_e$. Since OSM encodes one-way streets, service lanes, and dedicated cycle tracks, $G_{road}^{MEL}$ is a directed multigraph rather than a simple graph.

From this road network, the cycling network of MEL is extracted as a filtered subgraph:

$$G_{cycle}^{MEL}=(V_{cycle}^{MEL},E_{cycle}^{MEL},s,t,{\varphi }_V,{\varphi }_E)$$

(11)

with:

-

$E_{cycle}^{MEL}\subseteq E_{road}^{MEL}$ retaining only edges that are bicycle-compatible (according to OSM attributes such as highway, cycleway, bicycle = yes, (see

Table A1. Network dataset’s structure.

Edges Attributes	Data Type	Description
u	Integer	Start node of the road segment.
v	Integer	End node of the road segment.
path	Object (Categorical)	Type of road (e.g., street, cycleway, alley).
osm_id	Object (Categorical)	Unique OpenStreetMap identifier for the road segment.
highway	Object (Categorical)	Classification of the road (e.g., residential, primary, cycleway).
level	Integer	Hierarchical level of the road within the network.
lanes	Integer	Number of lanes on the road segment.
width	Float	Width of the road segment (in meters).
bicycle	Integer (Binary)	Indicator for bicycle access (1 = allowed, 0 = not allowed).
bicycle_safety	Integer	Bicycle safety rating (0–3), where 3 represents a dedicated cycleway.
foot	Integer (Binary)	Indicator for pedestrian access (1 = allowed, 0 = not allowed).
foot_safety	Integer	Pedestrian safety rating (0–3).
max_speed	Integer	Maximum speed allowed on the road (in km/h).
motorcar	Integer (Binary)	Indicator for motor vehicle access (1 = allowed, 0 = not allowed).
geometry	Geometry	Spatial geometry of the road segment, represented as a line.

))

-

$V_{cycle}^{MEL}=\left\{v{\in V}_{road}^{MEL}|\exists e\in E_{cycle}^{MEL},v\in \left\{s\left(e\right),t\left(e\right)\right\}\right\}$

This filtering ensures that $G_{cycle}^{MEL}$ faithfully represents the cycling infrastructure available to users in MEL.

Connectivity, reachability, accessibility, network density, and safety are now evaluated on the real-world cycling network $G_{cycle}^{MEL}$, instantiating the abstract concepts from the Functional Domain (see Equations (1)–(9)).

2.3.2. Computational Model (Extraction and Network Analysis)

The computational model processes the graph representation of the cycling network to evaluate its performance against the functional requirements. This step involves applying algorithms to evaluate the functional requirements of the MEL cycling network, such as connectivity, reachability, accessibility, network density, and safety.

The computational model calculates the shortest path from a given starting point (e.g., a cyclist’s origin) to all other reachable points within the network. This step allows us to determine which parts of the network can be accessed within the defined 30 min threshold, a critical user requirement. Once the shortest paths are determined, the network is analyzed to create an isochrone, which represents all locations reachable within the 30 min cycling window. This process involves cutting the graph at the 30 min threshold and buffering each road segment to represent the accessible area. The resulting polygons, which show the accessibility zones, are then merged to form the final isochrone. This step provides a visual representation of the network’s performance, highlighting areas that are well-connected and those that might need improvements in accessibility.

The computational model also enables the identification of poorly connected areas within the network. By comparing the generated isochrones, it becomes possible to detect regions that do not meet the accessibility requirements. These areas may be underserved in terms of cycling infrastructure or may require additional routes to improve connectivity. The analysis can suggest potential solutions, such as new bicycle lanes or alternative paths, to ensure that the network is optimized for accessibility, efficiency, and consequently for sustainability. This computational model will be developed in detail in the next section.

These computational steps form the basis for decision-making in the Process Domain, where the outcomes of the analysis are used to inform further refinements and optimizations of the cycling network.

2.4. Process Domain (Experimental and Computational Models for Network Design)

The Process Domain answers the question: “How can the cycling network be improved based on analysis?”. This domain operates as the link between data-driven simulations and real-world decision-making. It serves as the stage where both experimental validation and computational testing are used to test new configurations, ultimately providing the necessary evidence and insights to inform urban planning decisions. The focus of this domain is to continuously improve the network’s safety and accessibility of the MEL by iterating on different scenarios and validating proposed changes.

2.4.1. Experimental Model (Validation of Network Changes)

While computational models offer a robust theoretical framework for assessing network performance, the real-world effectiveness of proposed changes must be validated. The experimental model incorporates field observations and user feedback to ensure that the computational outcomes align with actual cycling conditions, making the network design both realistic and effective.

The first step in the experimental validation is to compare the computed isochrones—areas representing all locations accessible within a 30 min cycling window—with real-world cycling conditions. This comparison helps assess whether the assumptions made in the computational model (such as the average cycling speed and road conditions) align with actual cyclist experiences. If discrepancies are identified, it can suggest that the initial model may need adjustments, particularly in terms of speed assumptions or network connectivity. Field observations bridge the gap between theoretical modeling and practical realities, allowing planners to refine the design based on observable, real-life data.

User feedback is another essential aspect of the validation process. Cyclists provide valuable insights into the accessibility and usability of the proposed network design. These insights can highlight issues that might not be captured by computational models alone, such as specific barriers to accessibility (e.g., poorly maintained paths, unmarked bicycle lanes) or subjective comfort concerns (e.g., road safety, traffic interactions). Incorporating user feedback ensures that the cycling network not only meets performance requirements but also caters to the actual needs and preferences of cyclists, ultimately creating a more cyclist-friendly urban environment.

This combination of field observations and user feedback ensures that network changes are thoroughly validated in the real world before full-scale implementation, thereby minimizing the risks of unforeseen issues arising after the network changes are made.

2.4.2. Computational Model (Simulating Network Improvements)

Once the initial network configuration has been validated, the next step is to use the computational model to simulate potential network improvements. This is where “what-if” scenarios and network reconfigurations are applied to predict the impact of proposed changes on accessibility and network efficiency. By adjusting the existing graph-based model, this computational approach provides planners and policymakers with insights into the potential outcomes of different interventions, allowing for evidence-based decision-making.

The “what-if” scenarios simulate the impact of specific changes to the cycling network, helping urban planners understand the potential effects on connectivity, reachability, and accessibility. For example, one scenario might involve adding a new bicycle lane along a busy road to improve accessibility to key destinations. The computational model can predict how this change will affect the 30 min isochrone, identifying new areas that become reachable by bicycle and whether the overall network efficiency improves.

Another scenario might prioritize certain roads for cycling by allocating more resources to bicycle infrastructure along selected routes. By adjusting the graph model to reflect these changes, planners can evaluate how such a priority shift impacts the overall connectivity and accessibility of the network.

Beyond isolated improvements, network reconfiguration allows for a broader set of modifications, such as the addition of new roads, the re-design of intersections, or the inclusion of bicycle-specific infrastructure like bicycle lanes or bicycle-sharing stations. The computational model can adjust the weights of network edges to simulate these infrastructure changes. For example, adding a bicycle lane to a road segment reduces the cycling time of that segment for cyclists, which can be captured by modifying the corresponding graph edge cycling time. These reconfigurations can be tested for their impact on the accessibility of the network, including the ability of cyclists to reach destinations within the 30 min threshold.

Through these computational analyses, planners can explore how multiple network changes might interact, allowing them to optimize network performance without needing to implement each change physically. By utilizing the experimental model to confirm real-world feasibility and by simulating various configurations through computational methods, the Process Domain ensures that the cycling network evolves in a way that is both theoretically sound and practically effective.

Through iterative testing and validation, the cycling network is adapted to meet user needs, as defined in the User Domain, while optimizing the performance requirements specified in the Functional Domain and computed in the Physical Domain.

3. Development of a Computational Model for MEL

3.1. Definition of the Geographic Boundaries of MEL

The first and most critical step in any geospatial analysis is to clearly define the geographic scope of the study. For our computational model of the Métropole Européenne de Lille (MEL), we begin by loading a shapefile containing the boundaries of the individual communes that compose the MEL region (Figure 2a). This shapefile provides an accurate spatial representation and serves as the foundation for all subsequent spatial data processing and analysis.

Because MEL consists of multiple administrative communes, these polygons are aggregated into a single, unified boundary representing the entire metropolitan area (Figure 2b). This consolidated boundary ensures that all spatial operations—such as road network extraction and infrastructure analysis—are geographically constrained within the defined MEL region.

Establishing a precise and unified geographic boundary prevents the inclusion of data from outside the study area, which could introduce errors and reduce the accuracy of downstream analyses. This step underpins the spatial integrity of the computational model and provides a consistent basis for integrating diverse geospatial datasets.

Supplementary technical information is provided in Appendix A.1.

3.2. Generation of the Geospatial Boundary Polygon for MEL

The primary goal in this step is to convert the MEL boundary into a standardized format suitable for further spatial analysis and data extraction. This format should ensure compatibility with a broad range of Geographic Information System (GIS) tools, including web applications, data visualization platforms, and other geospatial software. It should also facilitate data exchange and reproducibility across scientific and technical communities.

The standardized format should enable smooth integration into external tools for downstream operations, such as spatial querying, intersection analysis, and interactive map visualization. It also supports collaboration by simplifying the sharing of consistent spatial data with researchers and stakeholders.

To achieve this, the previously computed boundary polygon is first stored in a GeoDataFrame—a format widely supported in geospatial libraries such as GeoPandas—and then exported in the GeoJSON format.

Further technical implementation details are provided in Appendix A.2.

3.3. Extraction and Optimization of OpenStreetMap Data for MEL

The goal of this step is to extract OpenStreetMap (OSM) data specific to the Métropole Européenne de Lille (MEL) using its polygon boundary and to convert the result into a format optimized for efficient spatial processing and analysis. This ensures that only data within MEL is retained, eliminating irrelevant information from the wider Nord-Pas-de-Calais region and improving the performance of subsequent workflows.

This operation involves several key requirements. The extracted data must be geographically limited to MEL, relying on the predefined polygon boundary to enforce spatial focus and exclude unrelated features. Beyond spatial relevance, efficiency is essential: working with the full regional dataset would incur excessive processing time and memory consumption. Targeted extraction is therefore necessary to reduce computational overhead.

Equally important is the choice of tool. The extraction process must rely on software capable of handling large OSM datasets, applying spatial filters based on polygon geometry, and performing with minimal resource usage. In parallel, the data format must meet technical expectations. The output should be compact, easy to store, quick to load, and suitable for large-scale spatial queries. It must also integrate well with standard geospatial tools used in routing, infrastructure analysis, and visualization.

These technical and functional considerations must be supported by a workflow that is both reproducible and user-friendly. The resulting dataset should be structured for easy reuse and distribution, while also being immediately applicable to tasks such as extracting MEL’s cycling network and analyzing transportation infrastructure.

To address these requirements, the MEL boundary is applied using the osmium extract utility, part of the Osmium toolkit. This tool is well-suited to processing large OSM files and supports efficient polygon-based filtering. After extraction, the dataset is converted into the Protocolbuffer Binary Format (.pbf), which offers major advantages over the XML-based OSM format. The .pbf format significantly reduces file size and improves I/O performance, while also being widely supported by geospatial tools. As a result, the data is ready for high-performance querying, visualization, and further spatial analysis relevant to MEL’s mobility systems.

Appendix A.3 contains further technical details.

3.4. Road Network Extraction from OpenStreetMap for the MEL

The goal of this step is to extract, process, and structure the road network specific to the MEL from the previously generated OpenStreetMap dataset. This refines the dataset to include only road infrastructure relevant for mobility and transportation studies, preparing it for accurate spatial analysis and modeling.

This process requires several conditions to be met. The road network must be filtered from the comprehensive dataset to exclude unrelated geographic features such as buildings and waterways. It is necessary to project the extracted road network into a metric coordinate system that enables precise distance calculations essential for routing and network analysis, as the default GPS coordinate system (EPSG:4326) lacks this capability. The road network geometry must be simplified to reduce redundant nodes and edges while maintaining overall connectivity and structural integrity. Verification through visualization is needed to ensure that roads are correctly classified by hierarchy and that extraction and projection are accurate. Finally, the processed network must be stored in a format optimized for efficient storage, rapid retrieval, and preservation of spatial and topological information.

To address these requirements, the Policosm library is employed to extract road-related features from the MEL .pbf file, removing non-road elements. The extracted network is then reprojected from EPSG:4326 to EPSG:3950 (CC50 projection), a metric coordinate system that supports precise spatial measurements. A geometric simplification is applied to reduce complexity while preserving network connectivity. The network contains 278,813 edges, each representing a road segment. The resulting network is visually validated by plotting road hierarchies using a color-coded scheme to confirm the correct classification of highways, secondary roads, and local streets.

Further technical details are provided in Appendix A.4.

3.5. Structure of the Road Network Dataset for Bicycle Infrastructure Assessment

With the extracted road network data now prepared, this stage focuses on evaluating its applicability for cycling network analysis. A thorough understanding of the dataset’s composition provides the basis for quantitative assessments such as identifying high-risk segments, optimizing bicycle routing, and evaluating network connectivity.

The dataset presents a comprehensive and structured representation of the MEL road network, consisting of discrete road segments as edges. Each segment is described by 15 attributes encompassing numerical, categorical, and spatial data types (Table A1 in Appendix A.5), establishing a robust foundation for assessing network suitability across different transportation modalities.

Key attributes include integer-valued variables like the number of lanes and maximum speed, continuous attributes such as road width, and categorical attributes including road classification and permitted traffic modes. The inclusion of a geometry attribute enables precise spatial encoding of segments.

Each attribute conveys specific information pertinent to the characterization and categorization of road segments. For example, certain edges are classified as cycleways with elevated safety rankings (dedicated lanes), whereas others correspond to residential streets with comparatively lower safety indices for cyclists and pedestrians. Attributes such as width, lane count, and speed limit contribute to a detailed infrastructural profile.

A critical attribute is the binary bicycle accessibility attribute, denoting whether a segment permits bicycle traffic (1) or imposes restrictions (0). Complementing this is a bicycle safety attribute, ordinally scaled from 0 (shared roadway with motor vehicles) to 3 (segregated cycleway), derived through the application of OpenStreetMap (OSM) urban mobility tagging schemas. This attribute provides a systematic basis to quantify cycling safety levels throughout the network.

The spatial geometry attribute situates each segment within a geospatial context, supporting analyses of topological connectivity and network functionality when integrated with supplementary datasets.

3.6. Mapping Bicycle-Accessible Roads and Safety Levels in MEL

This step involves visualizing the road network in the MEL region that is authorized for bicycle access and categorizing these roads based on their safety levels for cyclists. The primary objective of this analysis is to generate a map that highlights roads where bicycles are permitted and to visually represent the varying levels of safety across these roads. This information is essential for understanding the current state of cycling infrastructure in MEL, and it aids in identifying areas that may require improvements.

To achieve this, the dataset is filtered to include only road segments where bicycles are allowed. This subset represents the bicycle-accessible roads in the region. The bicycle safety attribute is then used to classify these roads according to their safety levels. Typically, this safety rating ranges from 0 (indicating a road shared with motor vehicles and deemed unsafe for cycling) to 3 (indicating a dedicated cycleway that provides the highest level of safety for cyclists).

By plotting these bicycle-accessible roads with the bicycle safety attribute as the variable of interest, the map visually distinguishes the safety levels of different roads using a categorical color map (Figure 3). This method provides an intuitive and clear way to display areas with safer cycling conditions and areas where improvements are necessary. The color gradient allows for easy identification of roads with varying safety conditions, helping urban planners and cyclists quickly assess which routes are the safest or most challenging for cycling.

The creation of such a map is important for urban planning and transportation policy. It highlights the current gaps in safe cycling infrastructure, enabling stakeholders to focus on areas that require intervention. For instance, roads with lower safety ratings, such as those with shared lanes or higher traffic, may be targeted for upgrades like dedicated bicycle lanes or other safety measures.

In addition to mapping, the resulting figure can be saved in formats such as PDF, enabling the dissemination of this information for reports, presentations, or decision-making processes.

3.7. Incorporating Cycling Time Estimations into the Cycling Network

In this step, the goal is to integrate cycling time estimations into the cycling network, which is a critical parameter for assessing the efficiency and practicality of bicycle routes in urban areas. Estimating cycling times across the network is important for route planning, understanding cycling conditions, and informing infrastructure decisions.

To achieve this, the length of each road segment is extracted from the dataset, which is already in the CC50 projection, ensuring accurate measurements in meters. The program then calculates the time required for a cyclist to travel each segment by assuming an average cycling speed of 15 km per hour (km/h). This functional requirement, as specified in the conceptual model, relies on typical urban cycling speeds to provide a realistic estimation of cycling time.

The calculated cycling time is stored as a new attribute in the dataset, representing the time (in seconds) it takes for a cyclist to traverse each road segment. This allows the dataset to now reflect both the distance and the time it would take to travel along each segment.

The integration of cycling time into the dataset is significant as it provides a more complete understanding of the cycling network. Rather than merely considering distance, this step enables comparisons between different routes based on how long they would take to travel, which is a more practical metric for cyclists. It also opens the door for further analysis, such as optimizing bicycle routes for efficiency, evaluating the accessibility of various parts of the city, and identifying areas that could benefit from improved infrastructure to reduce cycling times.

By incorporating these cycling time estimates, the network becomes more useful for transportation planning, urban mobility analysis, and decision-making related to the improvement of cycling infrastructure. It allows urban planners and policymakers to identify and prioritize improvements to the cycling network that could make cycling in the city faster, safer, and more convenient for residents.

3.8. Geospatial Processing of IRIS Units Within MEL

This step focuses on identifying and preparing the IRIS (Ilots Regroupés pour l’Information Statistique) zones located within the boundaries of MEL for the subsequent analysis of bicycle accessibility. IRIS zones are small statistical units used in France for demographic and socio-economic analysis, making them a suitable reference for studying urban mobility patterns. By focusing on the 512 IRIS zones within the MEL boundary, we establish the study area needed to analyze which locations can be reached within a 30 min bicycle ride starting from a randomly selected IRIS. This is achieved through a series of geospatial operations, which are applied to both the IRIS data and the cycling network data.

The process begins by reading the IRIS shapefiles and converting them into a compatible projection, CC50 (EPSG:3950), which ensures consistency with the cycling network data. The IRIS dataset, which contains geographic boundaries for the entire Nord-Pas-de-Calais region (Department 59), is filtered using the MEL boundary to isolate the IRIS data that falls within the MEL region.

To further refine the area of interest and avoid including unnecessary IRIS data, the MEL boundary is buffered by 200 m, ensuring that only relevant IRIS boundaries are captured. The 200 m value was selected as a reasonable distance to exclude partial or marginally intersecting IRIS units while retaining those fully within the metropolitan area.

Once the appropriate IRIS data for MEL is selected, a spatial join operation is performed to retain only those IRIS units that intersect with the MEL boundary. The resulting dataset contains the IRIS units within the MEL area, which will be used for the subsequent analysis of reachable zones.

To visually assess the result, a plot is generated that superimposes the selected IRIS boundaries with the MEL region. In the plot, the IRIS boundaries are shown in their original configuration, and the MEL boundary is overlaid in transparent red (Figure 4). This visualization allows us to confirm the spatial overlap between the selected IRIS units and the MEL region, providing an intuitive understanding of the area under consideration for further analysis.

By performing these operations, we ensure that the IRIS data used for the reachable zone analysis is geographically constrained to the MEL area and that only the relevant IRIS units are included in the subsequent steps. This refined dataset serves as the basis for analyzing accessibility within the MEL region.

3.9. Identifying Accessible Zones Within 30 Min by Bicycle

This section examines the spatial accessibility of the cycling network by identifying areas that can be reached within 30 min from a randomly selected IRIS. The process begins with the selection of an IRIS as the starting point. Its boundaries are extracted from the IRIS dataset and converted into a compatible projection system to ensure accurate spatial computations. Since IRIS zones can have highly detailed boundaries, a simplification step is applied to reduce the number of points while maintaining the overall shape. This optimization minimizes computational complexity in subsequent steps without significantly affecting spatial accuracy.

Once the IRIS boundary is processed, its points are mapped to the nearest edges in the cycling network. The cycling network is then represented as a graph, where road segments serve as edges with an associated cycling time attribute.

To determine the 30 min travel zone, we applied Dijkstra’s shortest path algorithm, commonly used in transport network analysis for its computational efficiency in identifying optimal routes on weighted graphs [33]. The algorithm computes all road segments that can be accessed within the given time limit, progressively expanding outward from the starting IRIS. The search halts once the threshold is reached, ensuring that only realistic travel areas are considered.

The final output is a visual representation of the isochrone, depicting the extent of the 30 min travel zone (Figure 5). This visualization offers valuable insights into the accessibility of the cycling network within MEL, helping to identify well-connected areas and potential gaps in infrastructure.

By analyzing accessible zones within a fixed time frame, this method provides an effective approach for evaluating urban cycling accessibility. The results support data-driven decision-making for urban mobility planning, infrastructure improvements, and sustainable transportation policies.

3.10. Modeling Cycling Accessibility by Constructing Isochrone Polygons for Cycling Network Analysis

After identifying the accessible zones within the given time threshold, the next step is to generate a polygon that represents the actual area accessible by bicycle and short walking distances. This polygon provides a more precise spatial representation of accessibility, considering both the structure of the cycling network and human mobility constraints.

To construct the isochrone polygon, each accessible road segment (linestring) is first divided into smaller segments. Since movement is not limited strictly to the road itself, a buffer is applied around each segment to reflect a more realistic travel area. At both the start and end of each segment, circular buffers are created to account for additional walking accessibility, assuming a pedestrian speed of 0.8 m per second. This accounts for small deviations in movement, such as minor detours or transitions from cycling to walking.

Once all segments have been processed, the buffered areas are merged into a single continuous polygon. The resulting shape is somewhat asymmetric, as some areas are more accessible than others based on available time and road connectivity. This method ensures a more accurate representation of the real-world travel area rather than a simplistic radius-based approach.

The final isochrone polygon is then saved as a shapefile for further analysis and is plotted alongside the cycling network. The visualization highlights different cycling times within the isochrone using a color gradient, making it easier to interpret the results. The plotted map includes key features such as the selected IRIS, the cycling network, and the computed isochrone polygon, ensuring clarity in understanding the spatial extent of cycling accessibility.

This step is important for urban planning and sustainable mobility studies, as it provides a realistic estimation of how far a cyclist can travel within a set time while considering small walking distances. By incorporating real network constraints and travel behaviors, the isochrone polygon offers valuable insights for improving cycling infrastructure and accessibility in urban areas.

3.11. Superposition of Isochrone Polygons on IRIS Units for Localized Accessibility Analysis

The superposition of all isochrone polygons onto the IRIS units within MEL serves several key analytical and decision-making purposes. By overlaying the computed isochrones on these predefined units, the study gains a more localized and detailed understanding of cycling accessibility, making it possible to identify disparities in connectivity and accessibility across different neighborhoods.

One of the main advantages of this superposition is that it enables a comparative analysis of accessibility across various IRIS units. Instead of evaluating the network as a whole, this approach allows for the assessment of which areas have higher or lower cycling accessibility within a uniform spatial basis. Policymakers and urban planners can thus pinpoint regions where the cycling network is insufficient or where minor interventions, such as new cycling lanes or improved crossings, could significantly improve connectivity. The isochrone overlay also helps in understanding socio-spatial inequalities in cycling accessibility, ensuring that improvements are targeted toward areas with lower network efficiency.

Additionally, the superposition allows for the integration of cycling accessibility with demographic and socioeconomic data available at the IRIS level. By intersecting isochrone polygons with population density, employment centers, or public service locations, urban planners can assess whether the existing cycling infrastructure adequately serves key urban functions. This enables the identification of underserved areas where infrastructure improvements could improve mobility for a broader population, contributing to sustainable urban development. Furthermore, visualizing the isochrones on IRIS units facilitates effective communication with stakeholders. Decision-makers, residents, and transportation authorities can better interpret accessibility results when they are mapped onto familiar spatial divisions rather than abstract network representations. This improves transparency in urban mobility planning and supports evidence-based decision-making.

In this study, employment data is available at the IRIS level, providing valuable insights into the spatial distribution of employment opportunities across the MEL. Figure 6 illustrates the IRIS zones alongside the distribution of employment centers, highlighting areas with high employment density and those with relatively lower concentrations of workplaces. This spatial representation is essential for analyzing the relationship between cycling accessibility and employment opportunities, as it allows for the identification of well-connected areas as well as potential gaps in access.

By integrating employment data with isochrone-based accessibility analysis, this study assesses how effectively the existing cycling infrastructure supports commuting to key employment hubs, ultimately contributing to a more inclusive and efficient urban mobility strategy. Figure 7 presents the integration of employment data with bicycle isochrone-based accessibility analysis, illustrating the spatial distribution of employment centers at the IRIS level within the MEL alongside their accessibility. The figure highlights distinct connectivity patterns, demonstrating significant variations in accessibility across different parts of the metropolitan area. While some peripheral employment centers maintain strong connections to the central MEL employment hub, others exhibit weaker accessibility. Additionally, peripheral employment centers are often poorly connected to one another, reinforcing a predominantly radial commuting pattern that relies on the central urban core.

This uneven accessibility distribution indicates that while the current cycling infrastructure partially facilitates commuting between the periphery and the central city, it remains insufficient for decentralized commuting across the metropolitan region. These findings emphasize the need for urban planners to develop transversal cycling corridors that directly link peripheral employment centers, reducing reliance on central transit nodes for inter-suburban travel. Strengthening these connections through dedicated cycling infrastructure and multimodal integration would improve network efficiency, promote sustainable commuting, and expand overall employment accessibility, contributing to a more balanced and resilient urban mobility system within MEL.

4. Discussion

4.1. Principal Findings

This study presents a detailed computational approach to extracting and analyzing the MEL cycling network, with a specific focus on bicycle accessibility. Using geographical data, network and graph-based modeling, isochrone computation, and network analysis, the proposed data-driven framework identifies areas where the cycling network fails to provide efficient mobility options. The analysis concentrates on cycling accessibility within a 30 min travel threshold measured from each IRIS, offering a granular view of network accessibility across the study area.

By identifying the nearest network nodes from IRIS boundaries and simulating travel distances based on network constraints, the analysis delivers a detailed view of accessibility limitations, while filtering out non-cyclable roads ensures the results accurately reflect actual cycling conditions. The resulting maps visualize disparities and offer an objective foundation for evaluating potential infrastructure improvements.

A key finding is that the cycling network does not uniformly support continuous and efficient travel throughout the MEL region, as accessibility varies significantly depending on the underlying cycling network structure. Some IRIS zones exhibit high connectivity, with well-integrated cycling routes enabling cyclists to cover significant distances within 30 min, whereas others are disproportionately affected by network bottlenecks—areas constrained by missing links, lack of dedicated cycling paths, or inefficient layouts. These constraints manifest in fragmented accessibility, where discontinuities and indirect routes reduce travel efficiency and lengthen commuting times, particularly to employment centers.

Several zones remain underserved due to these structural issues, confirming a fragmented network design that disproportionately affects some areas and forces cyclists to take circuitous routes, limiting practical commuting options.

The polygon creation process reveals that accessibility zones often expand asymmetrically. These irregular shapes, arising from road segmentation and uneven infrastructure distribution, reflect disparities in network integration; the shape of the isochrone polygons consistently highlights areas with robust infrastructure compared to those with poor connectivity.

Furthermore, the superposition of isochrone polygons on IRIS units provides additional insights into accessibility at a localized scale, especially in relation to employment distribution. While some peripheral employment centers maintain strong accessibility to the central MEL employment hub, others remain weakly connected, limiting direct cycling commutes to job opportunities. Moreover, the peripheral centers themselves are poorly interconnected, reinforcing a radial commuting pattern and restricting cross-peripheral connectivity among outlying areas.

From a network ergodesign perspective, these irregular isochrone expansions underscore that the cycling network is not uniformly planned for accessibility, demonstrating how accessibility is highly sensitive to local infrastructure variations. Even minor additions—such as short cycling links or improved crossings—can substantially improve accessibility patterns. The transformation of isochrone edges into polygons further confirms that accessibility critically depends on localized infrastructure quality and route availability.

Together, these findings provide comprehensive insights into the structure and efficiency of the current cycling network, highlighting both its strengths and limitations and offering guidance for targeted improvements to support more equitable and effective cycling mobility across the MEL region.

4.2. Interpretation and Critical Analysis

The findings from this study highlight that conventional cycling infrastructure planning, which often focuses primarily on expanding network size, may fall short if it does not also address underlying structural inefficiencies within the network. Simply adding more routes without improving connectivity and removing bottlenecks is unlikely to optimize cycling accessibility and mobility in complex metropolitan environments such as the MEL region. This suggests a pressing need to shift the planning paradigm toward quality, integration, and strategic targeting of network weak points.

The fragmented accessibility and bottlenecks detected in the MEL cycling network align with broader findings in the literature that emphasize connectivity as a central challenge in urban cycling infrastructure [3,15,21,28]. However, our work advances this field by employing a computational, data-driven framework that moves beyond traditional qualitative or macro-analytical approaches [3,4] and delivers granular, spatially explicit diagnostics of accessibility gaps. Unlike earlier studies that primarily focus on network size or expansion [17], our modeling framework, which is a data-driven micro-analysis, identifies and quantifies the impact of small-scale discontinuities and infrastructure gaps on real-world accessibility, complementing and extending the literature on sustainable urban mobility [5,6].

By using isochrone polygons integrated with employment distribution data, the analysis reveals fine-scale infrastructure discontinuities and accessibility bottlenecks that conventional methods often overlook. This granular spatial insight enables more precise diagnosis of network inefficiencies and actionable guidance for targeted improvements. By focusing on isochrone-based accessibility polygons and integrating employment location data, our study responds to recent calls for methodologies that combine territorial equity analysis with ergonomic principles [7,8]. This offers a more realistic and actionable perspective on user experience and utilitarian cycling needs.

Moreover, the uneven distribution of cycling accessibility highlighted in the analysis indicates that the current network may inadvertently contribute to unequal opportunities for accessing key employment centers. Poor connectivity in certain areas can discourage cycling uptake among residents, especially in zones where infrastructure discontinuities force longer detours or fragmented routes. Improving these gaps is thus crucial not only for improving overall mobility but also for promoting equity in sustainable transportation access—ensuring all communities can benefit from convenient, safe, and efficient cycling options.

A significant advancement lies in integrating isochrone-based polygon mapping with employment center distribution—stepping beyond traditional distance-based or node-link network metrics [8,21]. This enables not only the identification of underserved areas but also the quantification of their practical ramifications on economic mobility, thus responding to calls in the recent literature for stronger links between infrastructure analysis and socioeconomic outcomes [7,12].

Isochrone polygons, which visually delineate areas reachable within a set cycling time, reveal striking asymmetries in route accessibility. These irregular shapes reflect how the infrastructure often fails to enable smooth, radial expansion of travel options from any given origin point. Instead, cyclists are frequently funneled onto specific corridors and constrained pathways, limiting route choice and flexibility. This phenomenon reinforces a radial commuting pattern focused on center-periphery travel to the MEL employment hub, while cross-peripheral or decentralized movement within the metropolitan area remains poorly facilitated. Such a pattern restricts connectivity between peripheral centers themselves and contributes to spatial inequalities in mobility. It substantiates established critiques of radial commuting bias in metropolitan transport [1], while also providing evidence that cross-peripheral cycling connections—an underexplored area—deserve more planning attention [8]. Furthermore, the shape and reach of isochrones are highly sensitive to even minor infrastructural upgrades, which challenges conventional wisdom that cycling network effects manifest only after large-scale changes. This finding is actionable for policymakers constrained by budgets and aligns with calls for more adaptable, iterative approaches to urban design [28].

Network bottlenecks, characterized by gaps in connectivity or inefficient road segments that constrain cycling flow, emerge as critical intervention points. These locations offer clear opportunities where targeted infrastructure improvements—such as closing missing links or upgrading crossing facilities—could have outsized positive effects on network efficiency and accessibility. Addressing bottlenecks is thus a strategic priority for creating a more cohesive and inclusive cycling network.

The identification of threshold sensitivity and micro-intervention potential introduces a novel, data-driven dimension to cycling network planning. This study demonstrates that small, context-specific upgrades—like short connecting links or improved crossings—can disproportionately improve accessibility and network functionality. While existing research underlines the cost-effectiveness of cycling infrastructure at large [21], our threshold sensitivity analysis demonstrates that targeted, micro-scale interventions—such as strategic links or intersection upgrades—can yield significant improvements. This provides a new, computational approach to prioritizing interventions for maximum impact rather than relying solely on large-scale investment [20].

From an ergonomic design perspective, this study underscores how road geometry and route layout deeply influence cycling viability. Routes that compel cyclists to undertake long detours or indirect paths reduce the overall efficiency of the network, potentially discouraging cycling as a practical mode of transportation. Moreover, the operationalization of ergodesign principles [29] through isochrone expansion asymmetry offers an innovative quantitative method to diagnose ergonomic inefficiencies within cycling networks. Introducing isochrone expansion asymmetry as a measurable indicator of ergonomic inefficiency represents a conceptual innovation. While previous work has highlighted ergonomic needs [18,19], it has rarely operationalized them spatially and quantitatively within a replicable framework.

This study not only fills critical knowledge gaps highlighted in recent research [2,10] but also provides a replicable methodological blueprint for other cities. It offers a framework that bridges sustainability and ergodesign, alongside new quantitative tools for diagnosing and remedying network gaps through fine-grained, location-specific interventions. By moving beyond aggregate statistics [28] to deliver actionable, spatially explicit strategies, this research advances the practice of sustainable urban mobility planning and equips cities with the analytical rigor needed to support more equitable and effective cycling networks.

A key ergonomic consideration in cycling network modeling is the extent to which design parameters reflect the diversity of user needs, capabilities, and conditions of use. In this study, a fixed benchmark cycling speed of 15 km/h was assumed for the MEL region. While this value aligns with reported averages for comparable French cities, it does not capture individual variability in speed, which can be influenced by age, riding experience, physical condition, and trip purpose. The literature indicates [31,32] that most observed speed data on live roadways have been collected in urban contexts, yet detailed integration of individual characteristics into accessibility modeling remains limited. Furthermore, seasonal and monthly variations—arising from factors such as weather, daylight, and traffic conditions—are known to affect average cycling speeds, which in turn influence isochrone shapes and accessibility outcomes. Future research should incorporate longitudinal speed data that reflect both individual-level attributes and temporal dynamics, enabling a more ergonomically robust and context-sensitive representation of cycling accessibility.

4.3. Specific Recommendations for Spatial Interventions in MEL (Process Domain)

Building on the identified weaknesses and critical interpretations, this subsection presents targeted spatial interventions designed to improve the overall practicality of the cycling network across the MEL metropolitan region. The interventions address structural inefficiencies, bridge network gaps, and work to correct disparities in access to employment centers.

Table 2. Spatial recommendations for network improvement.

Identified Issue	Recommended Intervention	Target Area/Basis
Fragmented network and missing links	Construct new cycling links to close connectivity gaps	IRIS zones with low connectivity and missing corridors
Lack of dedicated cycling paths	Upgrade existing roads to include protected bicycle lanes	High-traffic routes where cycling is constrained
Inefficient and indirect route layouts	Redesign problematic intersections and optimize routing	Locations with indirect routing that increases cycling time
Peripheral employment centers poorly connected to the MEL hub and to each other	Improve radial cycling access and develop cross-peripheral routes	Peripheral IRIS units and employment centers with weak accessibility
Unequal distribution of cycling infrastructure	Prioritize investment in underserved and disadvantaged zones	IRIS zones with systematically lower cycling accessibility
Safety concerns at major crossings	Install or upgrade cyclist-friendly crossings and signals	Known high-conflict intersections and bottlenecks
Asymmetric and irregular accessibility zones	Create short linking paths to improve spatial continuity	Areas where isochrone polygons show weak expansion
Over-reliance on radial corridors with limited alternatives	Develop secondary cycling corridors to diversify network use	Radial routes with congestion and insufficient alternatives

summarizes these recommendations, linking each identified issue to corresponding interventions and their specific spatial targets.

These proposed interventions emphasize precise, context-aware improvements rather than broad expansion alone. By resolving the identified network bottlenecks and addressing structural shortcomings with focused spatial strategies, the cycling network can substantially increase its inclusiveness, connectivity, and effectiveness—supporting a more balanced and sustainable urban mobility model for the MEL metropolitan area.

4.4. Implications for Sustainable Ergodesign of Cycling Network

The implications of this study for the advancement of sustainable cycling infrastructure are outlined in the following points:

• Data-Driven Identification of Accessibility Gaps—Unlike traditional expansion-focused planning, this approach systematically detects and quantifies accessibility disparities using computational modeling. It demonstrates that network gaps are not solely a function of distance but are strongly influenced by discontinuities, unsafe segments, and poor connectivity, which are often overlooked in conventional approaches.
• Cycling Accessibility Beyond Distance Metrics—This study demonstrates that accessibility is not just about adding more infrastructure, but rather optimizing network structure to reduce indirect routes, inefficient road layouts, and fragmented accessibility. This shifts the focus from network expansion to structural efficiency.
• Employment-Oriented Cycling Integration—Instead of evaluating cycling accessibility in isolation, the framework superimposes isochrone maps on employment centers, identifying gaps in economic mobility due to poor cycling access. This approach challenges the radial commuting bias by highlighting the need for cross-regional connectivity rather than just center-periphery links.
• Localized Micro-Interventions with High Impact—Traditional cycling network planning often involves large-scale interventions. This study’s findings emphasize how minor, strategically placed improvements (short links, crossings, or road reconfigurations) can drastically improve network efficiency, shifting from macro-scale expansion to micro-scale optimization.

The implications for ergodesign in cycling networks are:

5.

Isochrone-Based Ergonomic Cycling Design—The study introduces isochrone expansion asymmetry as an indicator of poor ergonomic design. It identifies how network bottlenecks disrupt the natural expansion of reachable zones, forcing cyclists onto inefficient routes, leading to physical strain and increased cycling time. This provides a quantifiable, computational method to assess and correct ergonomic inefficiencies.

6.

Threshold Sensitivity in Network Accessibility—The computational approach demonstrates that small structural modifications (e.g., adding a single missing link) can significantly alter cycling accessibility patterns. This introduces a new quantitative threshold-based framework for evaluating network improvements, allowing planners to prioritize interventions based on their projected impact.

7.

Cycling Flow Optimization for Even Network Distribution—Traditional approaches often assume that cycling networks should naturally expand in a uniform manner. This study’s findings, however, demonstrate that the current infrastructure forces asymmetrical, corridor-dependent travel, making some areas disproportionately well-connected while others remain inaccessible. This insight leads to a more balanced and flow-optimized network design.

8.

Polygon-Based Accessibility Mapping for Structural Redesign—Instead of relying on traditional node-link analysis, the study constructs accessibility polygons based on real-world cycling conditions. This provides a continuous, topologically accurate representation of cycling accessibility, allowing for a more precise ergonomic and structural assessment of the network.

5. Conclusions

This study systematically evaluates the accessibility and efficiency of cycling networks in MEL, addressing key gaps in sustainable and ergonomic design. The findings highlight that accessibility is influenced not only by distance but also by network discontinuities, unsafe road segments, and missing links, underscoring the need for structural optimization rather than mere expansion. The integration of employment accessibility analysis shows a radial commuting bias, emphasizing the necessity for cross-regional cycling connections.

Furthermore, the study demonstrates that small-scale, targeted interventions—such as short links and optimized crossings—can significantly improve network efficiency at a lower cost than large-scale expansions. The research introduces a holistic modeling framework, where computational modeling combines graph-based modeling, isochrone analysis, and ergonomic assessment, offering a data-driven method to quantify accessibility disparities and prioritize interventions.

By redefining cycling accessibility beyond distance metrics and introducing isochrone expansion asymmetry as an ergonomic design principle, this study advances urban mobility planning. The findings provide actionable insights for urban planners, advocating for a shift towards micro-scale improvements and employment-oriented cycling integration. Practical implications include data-driven decision-making for infrastructure planning, micro-interventions, and improved accessibility to employment hubs, contributing to a more inclusive and efficient cycling network.

The methodological framework developed is scalable and adaptable, allowing for its application to other metropolitan contexts. By bridging sustainability and ergonomic principles, this study contributes to the development of more connected, efficient, and cyclist-friendly metropolitan infrastructure, promoting a data-driven approach to urban mobility planning.

Building on this contribution, the present study applies the framework specifically to the User, Functional, and Physical domains to model and analyze urban cycling accessibility in the MEL region. While the Process Domain—focused on experimental validation and computational simulation of network improvements—is not fully implemented, the discussion section provides specific recommendations for spatial interventions, illustrating potential applications of this domain. Future work should incorporate real-world cycling behavior data, and cross-city comparative studies could help refine best practices and assess the adaptability of the proposed framework across diverse urban environments.