A Multidimensional Approach to Bike Usage in Barcelona: Influence of Infrastructure Design, Safety, and Climatic Conditions

Abstract

Promoting cycling as a sustainable mode of transport is a pressing priority in contemporary urban mobility planning. This study examines the infrastructure characteristics that most strongly influence bicycle use in dense metropolitan contexts. A mixed-methods approach was adopted, combining a systematic review of current design guidelines with a large-scale empirical analysis of Barcelona’s Bicing bike-sharing system. The dataset comprised more than 54 million recorded trips, enabling the identification of the most and least frequented routes and the subsequent assessment of their infrastructural attributes. The results indicate that network configuration, continuity, and adaptation to topographic conditions have the greatest influence on cycling uptake. By contrast, factors frequently emphasized in design recommendations, such as lane width, were not decisive, as several of the city’s most intensively used corridors did not conform to these standards. These findings suggest that the expansion of network coverage and the improvement of route connectivity are more effective strategies for increasing cycling adoption than isolated design optimizations. This study contributes evidence-based guidance for urban planners and policy-makers seeking to advance cycling as a principal component of sustainable urban mobility in Barcelona and other comparable urban environments.

1. Introduction

Sustainable urban mobility has become a central concern for cities worldwide as they confront the persistent challenges of traffic congestion, air pollution, and the need to improve public health and quality of life. In this context, cycling has emerged as one of the most promising alternatives to private motorized transport, offering a flexible, healthy, and environmentally sustainable mode of travel [1]. Beyond its individual benefits, cycling contributes to broader societal goals by reducing greenhouse gas emissions, mitigating noise pollution, and alleviating pressure on public transport systems [2,3]. Consequently, many municipal administrations are prioritizing the promotion of cycling as a key strategy to advance sustainable mobility and enhance urban livability.

Many studies [1,4] try to understand barriers and enablers of cycling adoption, both general and user-dependent. Some of these studies suggest that the successful adoption of cycling at scale depends to a great extent on the provision of infrastructure that ensures safety, continuity, and practicality for users [5,6]. However, most of these studies are based on user surveys that serve as a basis for the use of discrete choice models [7] or structural equation modeling (SEM) techniques [8,9]. Despite its important contribution, they should be complemented by others based on the analysis of de facto data, therefore completely free from any degree of subjectivity. Well-designed cycling networks are a powerful tool for enabling bicycles to serve as a daily mode of transport, but achieving this requires moving beyond anecdotal evidence and hypothesizing. There is an urgent need for data-driven approaches and empirically grounded best practices that can guide policy-makers and planners in designing infrastructures that genuinely encourage widespread and sustained adoption. By focusing on infrastructure characteristics that influence actual user behavior, administrations can maximize the effectiveness of their investments and deliver cycling networks that are both functional and appealing.

Barcelona provides a particularly valuable case study in this regard. Over the past two decades, the city has undertaken substantial efforts to promote cycling through investments in dedicated infrastructure and through the operation of Bicing, a large-scale public bike-sharing system. Bicing has generated an extensive dataset encompassing millions of trips, thereby offering a unique opportunity to analyze individual users’ travel patterns. This enables the systematic identification of the infrastructural features that most strongly determine whether routes are frequently used, while also allowing a comparison between existing conditions and the optimal characteristics identified in the academic literature.

The central objective of this research is therefore to identify and analyze the design features of urban cycling infrastructure that most effectively foster daily bicycle use while enhancing user safety. To achieve this, the study adopts a dual approach: first, a review of state-of-the-art knowledge on cycling infrastructure design, drawn from both international standards and academic research; and second, an empirical analysis of Barcelona’s cycling network based on usage data from the Bicing system. By comparing observed patterns of bicycle use with the infrastructural characteristics of the corresponding network segments, the study seeks to isolate the factors that most significantly influence cycling uptake. This provides both a diagnostic assessment of Barcelona’s current cycling network and actionable insights for future planning, offering evidence-based recommendations that can be generalized to other cities pursuing sustainable mobility transitions.

The remainder of the paper is structured as follows: Section 2 includes the background of the research; Section 3 describes the methodological approach; Section 4 applies the methodological framework to the city of Barcelona and presents the obtained results; Section 5 discusses the findings and provides some policy recommendations and, finally, Section 6 includes the main conclusions of the performed analysis.

2. Background

The evolution of urban cycling infrastructure has progressed from treating bicycles as marginal road users to recognizing them as central actors in sustainable transport systems. Also, contemporary urban planning increasingly employs a multidisciplinary approach, drawing on transport engineering, urban design, human factors research, and data science to address the interrelated objectives of enhancing user comfort and ensuring efficiency and safety. This holistic approach is especially important in the case of active mobility: like pedestrians, cyclists are vulnerable road users. A growing body of literature confirms that the decision to cycle is strongly shaped not only by objective risks but also by perceived security and convenience, particularly among vulnerable groups such as women, children, elderly riders, and novice cyclists [10,11].

2.1. Foundational Principles and Design Paradigms

International design manuals, such as the Dutch CROW guidelines [12]. and the North American NACTO standards [13], converge on the principle of designing from the perspective of the most vulnerable users. Safety, continuity, and intuitive usability are key parts of successful network design. Academic research further stresses the importance of perceived comfort, with fear of motor vehicle interaction identified as the greatest deterrent to adoption, particularly among the “interested but concerned” demographic [14,15].

2.2. Infrastructure Typologies and Spatial Requirements

Cycling infrastructure may take multiple forms, ranging from mixed-traffic streets in calmed zones (≤30 km/h) to painted lanes, buffered lanes, and fully protected cycle tracks. In Spain, transport legislation differentiates several typologies, including “carril bici” (painted lanes), “carril bici protegido” (physically segregated lanes), “pista bici” (independent cycle paths), and “senda cyclable” (shared-use paths in open spaces). Current best practice recommends minimizing sidewalk-based lanes, which generate pedestrian conflicts, and prioritizing protected lanes on the roadway or, where infeasible, traffic-calmed shared streets.

Effective design also requires consideration of cyclist dimensions and dynamic space requirements. National references such as those of the Spanish Ministry of Transport specify minimum clearances (e.g., 1.00 m dynamic width, 2.25 m vertical clearance) to accommodate lateral sway and ensure safety margins, particularly in bidirectional facilities [16].

2.3. Intersections and High-Conflict Zones

Intersections represent the most critical points of conflict. Both European and North American practice [17,18] emphasize “protected intersections,” incorporating corner refuge islands, setback crossings, and dedicated signal phases. Complementary tools include bicycle boxes, colored pavement to enhance visibility, and clear priority markings. Physiological studies, such as those using Galvanic Skin Response sensors, confirm that intersections and peak-hour conditions are among the greatest stressors for cyclists, underlining the importance of careful geometric and visual design [19,20].

2.4. Environmental and Network-Level Factors

Topography and climate have a significant influence on cycling uptake. Steep gradients are a strong barrier [21], while adverse weather conditions—rain, extreme heat, or cold—reduce trip frequency, with differential impacts across genders [22]. On the network scale, continuity and coherence are requested. Disconnected facilities force users into unsafe spaces, undermining confidence and reducing adoption [6,23]. Surveys in Barcelona, for instance, reveal that 91% of cyclists encounter discontinuities on their routes, with more than a quarter reporting resorting to prohibited areas such as sidewalks or arterial roads [24]. Accessibility is equally crucial: networks must follow “desire lines,” ensuring that entry points are located within short distances, and they should be designed for intermodality by integrating with public transport nodes to solve first- and last-mile barriers [25,26].

2.5. Safety, Perception, and Signaling

A clear distinction exists between objective and perceived safety. Even in environments with relatively low accident rates, the obligation to ride among motor traffic remains one of the strongest psychological barriers to cycling [1,27]. Physically segregated lanes, combined with consistent signaling, such as lane coloring, buffer markings, and advanced stop zones, have been shown to improve both actual safety and user confidence [28,29].

2.6. Behavioral Modeling and Big Data Approaches

Traditional transport planning has mainly employed discrete choice models to estimate the probability of cycling, with utility functions incorporating trip attributes (distance, time, slope) and infrastructure characteristics (safety, continuity, segregation). Recent advances aim to quantify the relative utility of comfort and safety, assigning coefficients to infrastructure attributes within these models. In research, in addition to discrete choice [30,31], structural equation modeling has also been applied, among others, to assess how latent factors such as attitudes towards cycling, perceived behavioral control, subjective norms, and environmental perceptions influence the intention and actual behavior of cycling among different groups [8,9,32,33].

However, the proliferation of automated bike-sharing systems (BSS) offers the opportunity of performing empirical analysis of cycling. Most of these systems generate high-resolution data on millions of trips, capturing origins, destinations, and temporal dynamics. This is, BSS data can provide robust evidence of how cyclists actually value infrastructure features or, in other words, to which point these features condition their cycling behavior. As with any other methodology, some findings are expected to be site-specific and user group-specific. For example, studies in cities such as Melbourne and Montreal have demonstrated that users frequently select longer routes in order to travel along protected lanes, thereby confirming the primacy of comfort and safety over strict travel efficiency [34,35]. However, some common trends could be found as well.

2.7. Research Gaps

Although the superiority of segregated infrastructure is well established, the literature often lacks fine-grained distinctions between specific designs, such as the comparative effectiveness of various separation types or the safety implications of bidirectional lanes. Moreover, most research originates from Northern European or North American contexts, leaving a relative paucity of studies focused on dense Mediterranean cities. Most importantly, there is a shortage of studies whose conclusions are based on big data analysis and that can be compared with others relying on statistical modeling approaches. The present study addresses these gaps by conducting a localized, data-driven analysis of Barcelona’s extensive BSS dataset, correlating microscopic infrastructural details with macroscopic usage patterns.

2.8. Conceptual Framework

Given the described literature review, the conceptual framework of this study draws on established theories of transport choice, risk perception, and policy-driven infrastructure effects. First, discrete-choice and travel-behavior theories affirm that individuals select transport modes by maximizing perceived utility subject to costs, comfort and constraints [36]. Second, perceived and observed safety influence mode choice: studies of bicycling risk perception [1,27] demonstrate that routes perceived as risky deter users, and that perceptions often correspond with empirical accident data. Third, policy-focused literature shows that investment in safe cycling infrastructure is positively associated with higher levels of cycling [37].

Based on this foundation, this paper proposes the following hypotheses, which guide the empirical analysis:

H1. 

Infrastructure effect: Cycling activity is positively associated with the availability and quality of dedicated cycling infrastructure (lanes, network density, protected segments).

H2. 

Accident/Safety effect: Higher rates of accidents or higher perceived risk in a spatial unit reduce cycling activity, controlling for infrastructure.

H3. 

Weather effect: Unfavorable weather (rain days, extreme temperatures) is negatively associated with cycling activity, reflecting its effect on positive utility of cycling.

3. Methodological Approach

This research adopted a mixed-methods approach designed to integrate theoretical best practices in cycling infrastructure design with empirical evidence derived from large-scale usage data. The methodological approach is then structured in two principal phases: a comprehensive systematic review of the state-of-the-art literature and international design standards, and a subsequent data-driven empirical analysis of cyclist behavior within a real-world environment. Figure 1 describes the methodological workflow.

3.1. Systematic Literature Review and Design Criteria

The initial phase involves an exhaustive review of related academic literature and well-known global design guides, such as those published by CROW [12] and NACTO [13], together with national design standards [16]. The purpose of this review is to establish a benchmark for the ideal characteristics of urban cycling infrastructure and to identify factors known to either promote or discourage cycling adoption. Key criteria examined include:

• Infrastructure typology: Analysis of different lane types (protected cycle tracks, shared lanes, cycle paths) and their efficacy in promoting safe, convenient travel, with a focus on prioritizing segregated facilities over shared space with pedestrians;
• Geometric and safety standards: Assessment of geometric design elements, including dimensions of the cyclist and required circulation space, appropriate turning radii, and best practices for the design and signage of network intersections, which are identified as critical points of conflict and cyclist stress [20]. There exist clear recommendations at the national level for all these features (

  Table 1. Summary of the Spanish National Guidelines (MITMA, 2023) [16] for cycling infrastructure.

  Key Infrastructural Features	MITMA Recommendations [16]
  Reference cyclist dimensions (vehicle + rider)	Length ≤ 1.90 m; width ≈ 0.70 m
  Basic operational space (single cyclist)	≈1.0 m + 0.2 m lateral clearance
  Two cyclists riding in parallel (same direction)	2.00 m
  Two cyclists in opposite directions (bidirectional)	2.20 m
  Minimum width of urban cycle lane (unidirectional)	2.00 m (min); 2.40 m (recommended)
  Minimum width of urban cycle lane (bidirectional)	2.20 m (min); 2.60 m (recommended)
  Cycle lane width in interurban areas	1.40 m (unidirectional); 2.60 m (bidirectional); separation from roadway: 1.50 m
  Minimum curve radii (free-flow sections)	20 km/h → R ≥ 10 m (+1.0 m widening); 30 km/h → R ≥ 20 m (+0.5 m); 40 km/h → R ≥ 30 m (+0.25 m)
  Minimum turning radius at intersections	≥3.0–3.2 m for standard bicycles; ≥ 5.0 m for cargo cycles
  Reference curb radius at intersections	≈4.5 m (standard); 6.0 m when frequent cargo/rear delivery is expected
  Signage and marking in cycling networks	Consistent use of regulatory and informative signs; clear pavement markings and pictograms; standardized colors
  Intersections (design practices)	Favor separated crossings where possible; clear signal phasing; ensure bicycle routes remain on carriageway, not sidewalks
  Lighting and visibility	Adequate illumination according to context; ensure nighttime/rain visibility of markings; avoid visual obstructions
  Intersection management best practices	Reduce turning radii for motor vehicles, apply traffic calming (neckdowns, raised elements), prioritize continuous bicycle routes

  );
• Network planning and contextual factors: Investigation into the importance of network coherence, continuity, and accessibility, including integration with other modes of transport (intermodality);
• Topographical and meteorological factors: The influence of topographical conditions and meteorological factors on route choice and cycling demand was also systematically reviewed to establish a comprehensive set of variables for empirical correlation.

For the sake of clarification, in this study continuity refers to the presence of uninterrupted cycling facilities, ideally maintaining consistent design and quality from origin to destination; directness expresses the ability to travel between two points along routes that are close to the shortest possible distance, minimizing unnecessary detours, and network integration relates to the degree to which the cycling network connects with other transport modes, thereby supporting multimodal and intermodal mobility.

3.2. Empirical Data Collection and Integrated Analysis

The second phase involves the collection and analysis of multiple, disparate big data sources related to bicycle use and urban mobility. The primary objective of this phase is to analyze revealed user preference by observing actual route choices and travel patterns, and to correlate these patterns with the physical characteristics of the corresponding infrastructure segments. This comparative analysis allows for a rigorous, data-driven assessment of whether established design guidelines truly influence user behavior in a particular context.

Data processing involves the application of specialized software, utilizing languages such as M and DAX, to perform data cleaning, integration, and the creation of supplementary metrics to facilitate the complex correlation tasks. Data harmonization techniques used were simple, in accordance with the detail of the available data. In the spatial domain, matching was accurately performed by geographic coordinates, while in the temporal domain, the different data were aggregated and processed by day. In this sense, climate was the limiting factor. Accident data differentiated between morning, afternoon, and night, and Bicing data provided the exact time when the user picked up the bike at one station or left it at another. However, weather information was only available per day.

4. Application to Barcelona

4.1. Barcelona Boundary Conditions

Barcelona, capital of Catalonia, is a dense metropolitan area with a population exceeding 1.69 million and an urban area of approximately 101.35 km². Population density is on the order of 16,637.5 inhabitants km² [38].

The city has established cycling as an explicit policy priority in its Pla de Mobilitat Urbana (PMU) 2024 and 2025–2030 [39,40]. Currently, the cycling network has around 268 Km and the target is to increase it in 55 km by 2030. Barcelona operates a large public bike-sharing system (Bicing), and the combination of municipal policy initiatives and comprehensive usage data makes the city an appropriate empirical case for studying how infrastructure features relate to bicycle uptake.

4.2. Barcelona Data

Several data types supplied by different sources were used: infrastructure data, cycling data, user data, accident data and meteorological data. Next subsections detail these databases. Data cleansing, integration, and preliminary visualization were mainly (but not only) performed using a specialized business intelligence software platform (Microsoft Power BI Pro v.2.117.286.0), which allowed for their fusion accounting for their varied initial formats.

4.2.1. Infrastructure Data

It consisted of a high-resolution geospatial layer of Barcelona’s cycling network, in which each segment was attributed with variables such as infrastructure type, directionality, separation method, segment length, and topological integration with the urban road system.

4.2.2. Mobility Data

The empirical analysis was grounded in Bicing trip records provided by the system operator. The available mobility dataset covered the period 2020–2023 and comprised more than 54 million individual trip records. These included start and end timestamps, origin and destination station identifiers, bike model (mechanical or electric), and anonymized user attributes (age, gender). Station metadata supplied geographic coordinates and station capacity.

Processing steps included an initial pre-filtering (Microsoft Access 2021, v.18.0) to reduce dataset volume, followed by further cleaning in Power Query (M language). These processes removed erroneous trips, defined as those with duration less than 60 s, and extreme outliers, this is, trips with duration longer than three times de standard deviation σ [41,42,43]. Operational trips (e.g., for reposition) were also removed.

Next, feature engineering and analytical variables were created in the Power BI data model using DAX; these included trip duration, a unique route identifier concatenating origin and destination station IDs, and an estimated travel distance between stations computed via the Haversine formula (Equation (1)). This equation estimates the great-circle distance between two points on a sphere based on their latitude and longitude. It accounts for the Earth’s curvature and provides a robust approximation for distances at both short and long ranges. It is widely used in transportation studies when trajectories are not available, as in our case.

$$d=2R·\mathrm{a}\mathrm{r}\mathrm{c}\mathrm{s}\mathrm{i}\mathrm{n}(\sqrt{{sin}^2\left({\displaystyle \frac{{\phi }_2-{\phi }_1}{2}}\right)+\mathit{cos}\left({\phi }_1\right)·\mathit{cos}\left({\phi }_2\right)·{sin}^2\left({\displaystyle \frac{{\lambda }_2-{\lambda }_1}{2}}\right)})$$

(1)

where:

1, 2: origin and destination points

R: Earth’s mean radius (commonly 6371 Km)

φ: latitude

λ: longitude

4.2.3. Accident Records

Accident data for 2020–2022 were sourced from the Barcelona Guardia Urbana OpenData repository. Records were cleaned and classified to isolate bicycle-involved incidents. Severity categories (mild, medium, severe) were derived from the level of medical assistance required, and victim ages were grouped into ranges for later analysis. Regarding accident categories, mild accidents were those in which no special medical care was necessary, medium accidents required immediate medical care or hospitalization within the day and those accidents implying longer hospitalizations or deaths were classified as severe.

4.2.4. Meteorological Data

Daily climatological variables were obtained from AEMET (Spanish State Meteorological Agency) OpenData. Extracted variables included precipitation, maximum and minimum temperature, wind speed, and total sun exposure. These data were used to quantify the influence of weather on daily trip counts.

4.3. Barcelona Results

Figure 2 shows the Spatial kernel density analyses (heat maps) of Bicing trip volumes from 2020 to 2022, while Figure 3 contains those of the first 8-month period (from January to August) of 2023. Next sections elaborate on the clear differences that can be seen in those figures and introduce a more detailed description of the Bicing cycling patterns during this period.

4.3.1. Temporal Evolution of User Patterns and the Influence of Electrification

Heat maps analysis and aggregation of trip counts showed a consistent year-over-year increase in Bicing usage from 2020 through 2023 (

Table 2. Evolution of trip volume, average duration and average distance per bike type and in total.

Year	Total Trips (M)			Average Duration (min)			Average Length (Km)
	Mechanical	Electric	Total	Mechanical	Electric	Total	Mechanical	Electric	Total
2020	9.26	3.23	12.49	14.32	16.62	14.91	1.17	1.63	1.29
2021 1	9.48	5.64	15.12	13.51	15.41	14.22	1.30	1.79	1.48
2022	7.60	8.69	16.29	12.56	14.31	13.49	1.34	1.84	1.61
2023 2	3.98	8.23	12.21	12.25	14.23	13.58	1.46	1.90	1.76

¹ Heavily influenced by COVID-19 pandemic. ² Only 8 months, from January to August.

).

It must be noted that mobility in Barcelona, as in the rest of Spain, was affected by the restrictions imposed in the context of the COVID-19 pandemic, from mid-March to mid-June 2020. The effects of these restrictions remained: reluctant public transport users shifted to private car use but also to cycling as main mode of transportation, which increased the number of trips in 2021. After a normalization period, public bike sharing promotion and awareness campaigns about climate change-related problems, among others, contributed to a new usage peak.

Over the analyzed period, the introduction and growing share of electric bicycles has been associated with longer average trip durations for the electric fleet in comparison to mechanical bikes. Trip mean durations were 13.24 min for mechanical bicycles and 15.14 min for electric bicycles. In addition, there was a broader spatial dispersion of trips associated with the electric fleet relative to mechanical bicycles, respectively, resulting in average distances of 1.3 and 1.8 km (Table 2).

Therefore, electric bicycles were found to be a significant factor in altering user behavior and expanding the serviceable area of the network. Travel patterns for electric bikes covered a significantly greater geographical surface area than those of mechanical bikes. This change in behavior included the willingness to undertake longer journeys and successfully navigate routes with greater altitudinal gain, specifically promoting travel to districts of higher elevation. In contrast, mechanical bike usage tended to cluster in the flatter zones closer to the metropolitan center and coastal areas, demonstrating the high sensitivity of conventional cycling to slope.

4.3.2. Accident Patterns and Safety Indicators

Analysis of bicycle-involved accidents revealed that the highest concentration of incidents occurred in the central, dense, L’Eixample district. The dominant age cohort involved was 21–35 years and males accounted for approximately 60.8% of the recorded bicycle incidents. The most frequent severity class was medium (incidents requiring on-site medical assistance without extended hospitalization); and Wednesday showed the greatest incidence by weekday. Notably, the majority of conflicts were reported at regulated intersections or signalized crossings, rather than at unregulated ones.

4.3.3. Influence of Climatological Variables

Quantification of weather effects indicates that rainfall had a modest negative effect on daily trip counts, accounting for a reduction of 2.88% on rainy days, whereas extreme temperatures have a much larger influence: days with temperatures below 5 °C or above 34 °C reduced daily trips by 26.42%.

4.3.4. Route-Level Usage: Most- and Least-Used Corridors

Aggregating OD flows and assigning trips to probable network paths produced a ranking of high- and low-demand routes (

Table 3. Infrastructural features of the most and less used cycling routes.

Features	Most Used Routes			Less Used Routes
	St. 94-79	St. 455-29	St. 390-41	St. 290-118	St. 276-142	St. 238-174
Lane type	Segreg.	Segreg.	Mixed	Mixed	Segreg.	Segreg.
Lane width (m)	1.2	2	Variable	Variable	1.9	2.1
Traffic direction	Unidirect.	Bidirect.	Variable	Variable	Unidirect.	Bidirect
Network continuity	Yes	Yes	No	No	Yes	Yes
Level difference (m)	−5	19	0	−60	−84	−10
Distance (Km)	1.7	1.5	1.9	6.3	3.3	3.2
Connection with public transport	Yes	Yes	Yes	Yes	Yes	Yes
Location of interest	Yes	Yes	Yes	No	Yes	Yes
Maintenance level	Low	Low	Low	Good	Good	Good
Signalized intersections	Yes	Yes	Mixed	Mixed	Yes	Yes

). The most-used routes tended to combine directness, flat terrain, and proximity to activity centers. Findings suggest that users often accept suboptimal segment-level infrastructure (for example inconsistent lane width or maintenance issues) when the route is direct and well-integrated into the network. Conversely, some under-used corridors feature physically segregated infrastructure, yet suffer from disadvantages such as long complex detours or downhill/uphill profiles that discourage usage (especially uphill return trips).

The top 3 most-used routes were:

• Station 94 (Pl. España) → Station 79 (Pl. Universidad);
• Station 455 (C/ Provença, 595) → Station 29 (C/ Provença, 388);
• Station 390 (C/ Comerç, 36) → Station 41 (Pl. del poeta Boscà).

Conversely, the top 3 least-used routes were:

• Station 290 (Pl. dels Jardins de Alfabia) → Station 118 (C/ Pujades, 1);
• Station 276 (Pl. Alfons X) → Station 142 (C/ Sancho Ávila, 104);
• Station 238 (C/ Espronceda, 298) → Station 174 (Pg. de Garcia Faria, 21).

5. Discussion

This section aims to interpret the empirical results obtained from the Barcelona case study and relate them to the theoretical principles and state-of-the-art considerations reviewed earlier. The goal is to derive implications for how cycling networks should be designed, prioritized, and managed in dense metropolitan contexts.

The evidence collected through large-scale Bicing usage data, accident records, and climatological influences converges on a set of consistent insights:

• Attractive network layout, defined by continuity, directness, and topographical feasibility, emerges as the decisive factor shaping adoption;
• Electric bicycles constitute a technological complement that significantly alters the accessibility profile of the city’s network by expanding the areas that can be reached with reasonable effort by bicycle and, therefore, contributing to an increase in users and/or journeys and;
• Accidents cluster at regulated intersections, highlighting persistent deficiencies in conflict management and signaling and potentially hindering some users’ safety perceptions.

Together, these findings support the suggested relationship between infrastructure design and user behavior, while they can also inspire particular measures to be included in mobility policy frameworks. Both the findings and these potential measures are elaborated in the next subsections.

5.1. Network Layout as the Dominant Determinant

Perhaps the most robust outcome of the analysis is that trip distribution is primarily explained by network-level properties rather than segment-level design quality. While international design manuals emphasize optimal characteristics for width, separation, and marking, the Barcelona case s that users prioritize continuity and directness over perfection. The most heavily used routes, such as the corridor linking Pl. España and Pl. Universidad, illustrate this principle: these connections are short, flat, and continuous, and they traverse zones of high urban activity. Even when infrastructure quality is inconsistent, whether due to narrow widths, surface irregularities, or intermittent markings, usage remains high as long as the route itself is coherent and efficient.

By contrast, the least-used routes are not rejected because of poor construction quality. In fact, some possess well-maintained segregated lanes. Instead, their lack of use is explained by unfavorable alignment: they are longer, involve significant elevation changes, and connect to fewer high-demand destinations. These findings support the conclusion that network layout has a stronger influence on behavior than the refinement of individual segments. This is consistent with the principle that a “good enough but continuous” network is functionally superior to a fragmented collection of high-standard but isolated paths.

5.2. The Modifying Role of Electric Bicycles

The second major insight concerns the integration of electric bicycles. Usage data between 2020 and 2023 shows a progressive rise in Bicing trips, with e-bikes playing a central role. Average trip duration is 14.3% longer for e-bikes than for mechanical bicycles, and their spatial distribution extends beyond the central flat districts to cover more varied terrain.

In this case, technology interacts with infrastructure by mitigating the barrier of slope, historically identified in the literature as the most significant topographical deterrent to cycling [21,30]. In practice, the introduction of e-bikes expands the effective service area of the network without requiring immediate physical modification of the infrastructure itself. This does not diminish the importance of network expansion but highlights that technological interventions can operate in parallel, accelerating adoption and promoting inclusive mobility.

5.3. Safety Outcomes and Intersection Design

Accident data between 2020 and 2022 reinforces long-standing concerns about intersections as focal points of conflict. The highest concentration of incidents occurred in L’Eixample, which is also the district of greatest cycling intensity, suggesting that exposure is a key factor. However, the prevalence of accidents at regulated intersections, where traffic lights and signs should theoretically manage conflicts, requires a detailed investigation. Possible causes could range from overconfidence and/or disobedience on the part of those involved, to insufficient signage in terms of quantity or clarity, or to inadequate traffic light timings [44,45]. Also, to higher overall mobility demands in the area that make user interaction more difficult.

The demographic distribution, with the 21–35 age cohort and male cyclists most frequently involved, is probably related with their higher use of the system: Bicing subscribers are predominantly men, and 6 out of 10 are between 25 and 44 years old. However, this trend also matches state-of-the art findings that link this group’s overconfidence and tendence to riskier behaviors [46,47]

5.4. Weather Effects and Behavioral Elasticity

The climatological analysis introduces an environmental dimension to cycling adoption. Results confirm that while rainfall has a small effect on daily trips, extreme temperatures sharply reduce demand. These values are consistent with international findings that comfort is a key determinant of cycling uptake [22]. The relatively minor impact of rain suggests that in Mediterranean climates like Barcelona’s, precipitation is not a primary barrier, but thermal extremes are. This underscores the importance of integrating climatological considerations into both infrastructure design (e.g., shading, resting points, which would also help pedestrians) and technological strategy (e-bikes enabling faster trips under uncomfortable conditions).

5.5. Comparison with Other Big Data Analyses on Factors Influencing Bike Use

The main findings of this research match trends observed in previous empirical studies using large-scale trip and GPS/bikeshare datasets, despite their different geographical scopes. For example, data-driven assessments in London, Boston, and Pittsburgh demonstrated that continuous, high-quality, and coherent cycling networks increase cycling participation, whereas fragmented or discontinuous infrastructure holds flows back [6,48]. Evidence from the UK showed that e-bikes facilitate increased cycling in hilly areas by reducing physical effort and extending trip distances [49]. Regarding safety patterns shaping cycling behavior, spatial analyses of accident data indicated that streets with higher crash incidence can experience reduced cycling volumes, whereas the presence of protected lanes or safer intersections encourages usage [50]. As for weather, multi-city analyses [51] showed a robust inverse-U relationship between temperature and bikeshare use. However, the station-level study of [52] demonstrated that the built environment and station design mediate weather sensitivity. City-specific investigations [53,54,55] confirmed that precipitation and extreme temperatures reduce trip counts and that the magnitude of these effects varies by climate, station type and the availability of transit alternatives.

Based on these similarities, it can be said that the trends in the relationships found are universal. However, there are local nuances. For example, in this study, the maximum temperature that led to a clear change in travel volume was 34 °C, while [56] found that temperatures above 28 °C significantly discouraged users from riding. These differences seem logical as Barcelona is significantly smaller but also denser than Brisbane, distances to essential destinations are much shorter and it allocates a higher percentage of its area to green spaces and vegetation cover compared to Brisbane. In addition, temperatures in Brisbane change significantly throughout the year, while citizens of Mediterranean cities are more used to long-lasting high temperatures. The influence of local conditions has also been addressed in the literature. For example, [57] analyzed a considerable amount of bike-sharing data and identified that factors like city population, proximity to the city center, and the presence of leisure-related establishments significantly influenced bike-sharing ridership, highlighting the importance of considering built environment characteristics when planning and implementing bike-sharing systems. Local nuances can also be cultural. For example, in cities like Copenhagen or Amsterdam and in nearby villages bikes have been the primary transportation mode for years. In these locations, users have preferences, but “barriers” to the use of bikes are almost nonexistent: no matter if there is a dedicated infrastructure or not, if it rains or snows, bike flows remain high. This is not the case for Barcelona, as shown by this piece of research. While the cycling culture is increasing, it can still be considered recent and fostered by common European measures to encourage sustainable mobility.

5.6. Implications for Policy and Planning

The synthesis of these results leads to several clear conclusions for mobility policy.

First, continuity and coverage should be prioritized above incremental refinements of existing segments. Expanding the network’s reach and ensuring that all districts have access to direct corridors will likely yield the greatest marginal gains in ridership. Plans for the systematic fill of missing links should be developed, and they should include prioritization criteria, e.g., based on space-based or demand-based connectivity gains. Enforcing consistent design and surface standards at transitions as well as intersection continuity treatments (continuous markings, protected crossings and signal modifications) would also be desirable. In order to support intermodality, bike routes should be coordinated with public transport stops and curb uses. More subjective metrics related to users’ stress level along routes could also be considered [58,59].

Second, electric bicycles should be actively supported and integrated into the public system, as they multiply the base accessibility provided by the existing infrastructure as well as expand user prototypes.

Third, targeted interventions at intersections should be treated as urgent priorities, since they represent both actual safety hazards and sources of perceived insecurity. Key, existing blind spots should be eliminated, and ongoing revisions should be made to ensure that new ones do not arise (due to vegetation, for example). In areas where most accidents occur, the timing of traffic lights should be checked. Measures such as cyclist-specific phasing, setback crossings, and targeted use of colored pavement can reduce ambiguity and make interactions more predictable. The inclusion of at least partially shaded cycle paths would also be desirable. And very important, sufficient lighting during dark hours should be ensured. Even if not infrastructure-related, more surveillance and higher fines for those users overlooking traffic rules would be worth considering.

Finally, data-driven prioritization, e.g., using trip volumes as a proxy for impact, could offer a rational method for guiding infrastructure maintenance and investment. As stated, perfect infrastructures are not needed to encourage adoption, but a minimum level of quality must be required for the sake of safety.

Overall, cycling uptake is co-produced by infrastructure, technology, and behavioral perception of effort and safety. Policies must therefore integrate these four dimensions rather than treating infrastructure as the sole variable.

6. Conclusions

This study examined the determinants of urban cycling adoption through a dual approach: a review of established design principles and a large-scale exploratory analysis of over 54 million trips from Barcelona’s Bicing system. The findings suggest that network-level properties—coverage, continuity, directness, and topographic feasibility—are more influential in determining use than segment-level design features such as lane width or pavement quality. The most-used routes were characterized by short, flat, and continuous connections between central activity nodes, whereas underused routes typically involved longer distances, elevation changes, and weaker functional integration. These results indicate that network layout is the primary driver of ridership, while local infrastructure quality, although relevant, plays a secondary role.

The analysis further highlights the significant contribution of electric bicycles. E-bikes mitigate topographical barriers, extend the feasible range of daily trips, and expand network accessibility to previously underused districts. Accident records confirm that intersections remain the principal points of conflict, suggesting that junction design, rather than corridor conditions, constitutes the greatest safety weakness. Weather sensitivity was also evident, with extreme temperatures substantially reducing demand, while moderate rainfall produced only a marginal effect.

The study has several limitations that should be acknowledged. The first concerns the reliance on origin–destination station data, which restricts the capacity to analyze detailed route choice behavior or micro-spatial variations in factors such as safety or comfort. Second, the use of Bicing data means that the analysis reflects the behavior of public bike-sharing users, who may differ from the broader population of cyclists in terms of socio-demographic characteristics, trip purposes, bicycle types, or route preferences; thus, the findings cannot be directly generalized to all cycling activity in Barcelona. Third, although the four-year period examined allows for the observation of temporal dynamics, external factors, such as the COVID-19 pandemic, infrastructure expansions, or changes in Bicing pricing and availability, may have influenced usage patterns, potentially introducing temporal biases. Furthermore, socioeconomic and localized environmental variables were not fully integrated into the analysis, which may have limited the explanatory depth of the results. Finally, the study adopts a largely descriptive and exploratory approach, without incorporating inferential statistical analyses to formally validate relationships between road link use and infrastructural features. This methodological choice, combined with the temporal mismatch between the infrastructure conditions of 2020–2022 and those of 2023 (used as reference), may have introduced additional sources of bias and limits the comparability of the findings across the entire study period.

Future research should aim to strengthen the methodological depth of the present analysis by incorporating inferential statistical techniques and expanding the range of data sources. Specifically, regression and correlation analyses could be applied to quantify the relationships between cycling intensity and infrastructural variables such as slope, continuity, and segregation. In addition, the spatial nature of the data should be explicitly addressed through the assessment of spatial autocorrelation effects, such as clustered usage patterns, using indicators like Moran’s I. Beyond these methodological enhancements, future studies should integrate dynamic network data and precise GPS-based route reconstructions to better capture actual route choices and micro-spatial variations in comfort or safety. Including datasets that encompass both public bike-sharing and private cyclists, combined with advanced discrete choice models, could allow for a more detailed representation of user heterogeneity. Another promising direction would be the integration of spatially detailed measures of the urban heat island (UHI) effect to complement the present city-level weather analysis. Combining cycling activity data with high-resolution micro-climatic information could enable more accurate, climate-adapted infrastructure planning for Barcelona. Finally, the integration of socioeconomic indicators and cost–benefit evaluations of infrastructure typologies, informed by empirical coefficients, should support a more comprehensive and equity-oriented assessment of cycling accessibility and guide evidence-based prioritization of future investments in urban mobility planning.