Urban Blue-Green Spaces and Everyday Well-Being in a High-Density Megacity: Evidence from Delhi

Abstract

Urban blue-green spaces (UBGS) are crucial nature-based solutions for enhancing urban resilience and improving public health. This study examined the experiential relationships linking BGS use to human well-being among users of five urban parks in Delhi, India. Using an integrated experience-centered framework, we collected in-situ survey data (n = 411) to profile usage patterns, assess environmental quality, and quantify restorative outcomes grounded in Attention Restoration Theory (ART) and Stress Reduction Theory (SRT). Advanced analytical techniques, including ordinal logistic regression and interpretable machine learning (SHAP), were used to identify the key factors associated with user satisfaction. The results revealed that for these respondents, BGS appeared to function as an essential neighbourhood, with over 40% visiting three or more times per week. Although visual attractiveness was rated positively, deficits in noise buffering and amenities indicated a gap between aesthetic and functional qualities. Restorative benefits, including emotional calmness, mood refreshment, and fatigue recovery, were consistently reported among respondents. Analyses showed that embodied experiences, particularly post-visit relaxation and physical comfort, were more strongly associated with user satisfaction. SHAP interpretation highlighted seating adequacy, routine use, and thermal comfort as prominent contributors, suggesting somatic relief may be particularly salient. This study provides exploratory evidence from a Global South megacity and context-sensitive insights into how restorative processes operate under high-density urban conditions. The findings show that routine accessibility, basic amenities, and thermal comfort are central to the everyday functioning of blue-green spaces as urban infrastructure, underscoring the need for experience-responsive and equity-oriented urban greening policies in high-density cities.

1. Introduction

Contemporary urban development, particularly within the rapidly expanding megacities of the Global South, faces interconnected crises of demographic pressure, environmental degradation, and public health burdens [1]. This growth exacerbates issues such as urban heat islands, noise pollution, and psychological stress while reducing routine human-nature contact [2], thereby undermining urban liveability and resilience. Urban blue-green spaces (UBGS), including parks, forests, and water bodies, are gaining recognition as essential multifunctional infrastructures that deliver critical ecosystem services [3]. In this study, UBGS refers specifically to publicly accessible parks and water-integrated green spaces within the urban fabric of Delhi, distinguishing them from the broader planning concept of blue-green infrastructure (BGI). As nature-based solutions, they provide synergistic benefits, including climate mitigation [4], biodiversity support, and direct health promotion through psychological restoration and physical activity [5], serving as a primary public conduit for regulating, supporting, and cultural ecosystem services in dense urban settings [6].

The delivery of ecosystem services and health benefits by blue-green spaces (BGS) is not guaranteed by their mere presence but is associated with human dimensions, such as accessibility, perceived environmental quality, and patterns of use [5,7]. Although research primarily from the Global North demonstrates strong links between UBGS exposure and well-being, particularly through cultural ecosystem services, this evidence is predominantly grounded in Attention Restoration Theory (ART) and Stress Reduction Theory (SRT), which emphasise cognitive recovery and physiological calming in low-density, resource-abundant settings [8,9,10]. Consequently, prevailing approaches prioritise visual aesthetics and cognitive outcomes while underrepresenting behavioural engagement, multisensory environmental quality, and the role of activity type in shaping restorative experiences [11].

Assessments of environmental quality often remain fragmented, focusing on visual attributes while overlooking acoustic comfort, thermal regulation, and infrastructure, thereby limiting the evaluation of UBGS as integrated systems delivering multiple ecosystem services [12]. User satisfaction, in this context, functions as a key evaluative outcome associated with experiential engagement into realised well-being benefits and sustained use [13]; however, empirical research rarely disentangles how activity type, social context, and patterns of use shape satisfaction and ecosystem service flows [14]. Addressing these gaps, this study adopts a human-centred approach to examine how UBGS ecosystem services are accessed and experienced in Delhi, focusing on use patterns, perceived environmental quality, and both cognitive and embodied restorative outcomes to inform experience-responsive and equitable BGI design.

The existing evidence linking UBGS to human well-being remains geographically skewed and conceptually limited, as predominant models derived from lower-density, resource-affluent contexts [15] assume a stable delivery of ecosystem services, an assumption of uncertain transferability to rapidly urbanising, environmentally stressed cities of the Global South. It is characterised by extreme density, infrastructure limitations, environmental stressors, and socio-spatial inequalities that fundamentally reshape how UBGS function and are experienced. In these contexts, visual greenness alone is an unreliable proxy for well-being benefits, as cumulative stressors may substantially limit experiential and restorative gains despite the presence of green cover [16]. This research integrates lived experiences with ecosystem services by focusing on embodied services, such as physical cooling and sensory relief, as well as equitable access, providing insights for designing effective and socially just urban greening strategies. However, this mismatch has rarely been empirically examined. Despite major urban greening initiatives in India, systematic research on urban parks examining environmental quality, restorative experience, usage, and user satisfaction is limited. However, no study has examined these features in Indian urban parks, especially in Delhi, using interpretable machine-learning (ML) techniques.

Delhi represents a critical and policy-relevant case for the exploratory investigation of how UBGS may operate as ecosystem service infrastructure under conditions of acute urban stress [17]. As one of the world’s largest megacities, Delhi faces extreme population density [18], rapid land-use change, rising heat exposure, chronic air and noise pollution, and limited access to private open spaces. Publicly accessible parks, urban forests, water bodies, and riverfronts provide essential regulating ecosystem services, such as thermal moderation, air purification, and noise buffering, as well as cultural services that support recreation, psychological restoration, and social interaction. These functions are particularly vital for populations with limited adaptive capacity in urban environments. However, the benefits of UBGS in Delhi are neither evenly distributed nor automatically realised; they depend on accessibility, mobility constraints, patterns of use, and users’ subjective perceptions of environmental quality [19].

This study addresses conceptual and geographical gaps by adopting an experience-centred framework that positions restorative experience as a central explanatory factor alongside environmental qualities, patterns of use, and user satisfaction in UBGS. The study integrates this framework with SHapley Additive exPlanations (SHAP), an interpretable ML approach enabling transparent assessment of socio-demographic factors, accessibility, behavioural engagement, perceived environmental quality, and restorative responses. Using SHAP, the analysis moves beyond linear models to reveal how multiple interacting factors shape restorative outcomes and satisfaction, an approach largely absent from UBGS research in the Indian and broader Global South context. The study is structured around three integrated objectives: (1) to profile user demographics and characterise behavioural patterns of access and use; (2) to assess perceived environmental quality and quantify restorative experiences grounded in ART and SRT constructs; and (3) to identify the relative contribution of experiential, behavioural, and environmental factors associated with overall user satisfaction using SHAP.

This study makes three central contributions to the literature on urban ecosystems and human well-being. First, it provides exploratory empirical evidence from five urban parks in Delhi, a quintessential Global South megacity, offering context-rich insights that begin to address the geographical bias in the existing research and generate theoretical propositions for testing in broader populations. By examining the context of acute environmental stress, infrastructural constraints, and high density, it extends theories on UBGS restorative benefits beyond the stable, resource-abundant settings from which they emerged. Second, the study advances an integrated analytical framework that links accessibility, perceived multisensory quality, behavioural use, restorative experience, and user satisfaction. This approach transcends single-factor analysis to model UBGS as a lived socio-ecological infrastructure. Our findings indicate that embodied relaxation emerged as the strongest contributing predictor of user satisfaction within the modelling framework, suggesting that sensory and physical relief from heat, noise, and congestion is closely associated with more positive evaluations in high-stress urban environments. This highlights an important but overlooked experiential mechanism. The insights suggest a shift in urban forestry and public health planning, emphasising that equitable outcomes rely on both the spatial distribution of resources and the experiential quality of BGI.

2. Analytical Framework for Assessing Everyday Use and Well-Being

2.1. Framework Rationale and Orientation

This study provides exploratory evidence suggesting that UBGS may serve as essential everyday infrastructure for active users in high-density Global South megacities. This suggests that well-being benefits are not derived solely from the presence of BGS but are associated with processes involving access, use, perception, and experience. The framework responds to critiques of supply-oriented and proximity-based metrics that tend to overestimate ecosystem service benefits by treating them as static landscape attributes while underplaying the human and contextual processes required to activate them [20,21]. By foregrounding lived experience, the framework emphasises the experiential realisation of cultural ecosystem services (CES), conceptualising restoration as an outcome that emerges when residents meaningfully access and engage with UBGS in their everyday urban lives. This orientation is particularly salient in Global South megacities, where high density, environmental stress, and socio-economic inequality strongly condition how green infrastructure is encountered and used [22].

The framework conceptualises user satisfaction with UBGS as shaped by interrelated social, behavioural, perceptual, and experiential factors, emphasising conditional associations rather than linear causality. It highlights conditional relationships rather than linear causality. The benefits of UBGS emerge through filtering shaped by urban conditions, where socio-demographic factors influence accessibility and mobility, affecting usage and satisfaction. At the foundational level, socio-demographic characteristics influence access by shaping mobility resources, safety perceptions, time availability, and recreational needs. Accessibility and mobility operate as gatekeepers, integrating travel time and transport modes with subjective assessments of ease and safety. Evidence shows that older adults and vulnerable groups face barriers, including uneven surfaces and poor infrastructure, while valuing the social, cultural, and restorative aspects of BGS use [23]. These constraints are intensified by spatial inequalities, with lower-income areas experiencing reduced access to green space, reinforcing environmental inequities in rapidly urbanising contexts [24,25].

When accessibility conditions are favourable, they facilitate diverse activities and frequent visitation, influencing user interactions with BGS. This interaction shapes the perception of environmental attributes, where perceived environmental quality, encompassing aesthetics, thermal comfort, acoustics, cleanliness, and infrastructure, is crucial for perceived restorativeness in public spaces [26,27]. This appraisal is associated with restorative experiences such as calmness, relaxation, mental refreshment, and recovery from fatigue, which are central experiential factors linked to well-being outcomes in high-stress urban settings [27]. Overall satisfaction reflects evaluative synthesis of realised restoration rather than spatial provision. The framework shows that BGS well-being value is sequentially realised: constraints at early stages, such as accessibility disparities or poor quality, can attenuate downstream restorative outcomes, highlighting the need for inclusive, high-quality BGS planning for urban populations [25].

2.2. Theoretical Foundations of Restorative Benefits

The restorative benefits of natural environments are commonly explained through Stress Reduction Theory (SRT) and Attention Restoration Theory (ART), which provide frameworks for understanding how green and blue environments support psychological well-being in urban contexts [28,29]. Both conceptualise restoration as a response to environmental stress and mental fatigue but differ in mechanisms: SRT emphasises unconscious affective and physiological responses to natural elements like vegetation and water, leading to emotional calming [9,30], whereas ART frames restoration as a gradual cognitive process where involuntary attention through “soft fascination” replenishes depleted directed attention and enhances cognitive functioning [8,31]. Although extensive empirical evidence links UBGS to reduced stress, improved mood, and enhanced attention [31,32], both frameworks were developed mainly in lower-density Global North contexts and underrepresent behavioural engagement, multisensory environmental quality, accessibility constraints, and socio-spatial inequities that shape restoration in dense Global South megacities. While SRT and ART provide foundational explanations of nature-based restoration, their focus on cognitive and affective processes offers limited insight into how access, behaviour, and experience shape well-being outcomes in densely populated megacities in the Global South (Figure 1).

2.3. Components of the Framework

2.3.1. Socio-Demographic Attributes

Socio-demographic characteristics, particularly age and gender, are positioned at the upstream end of the framework as foundational structuring factors that shape engagement with the UBGS. These attributes influence daily routines, physical capacity, safety perceptions, and leisure preferences, thereby conditioning who accesses the UBGS, how they travel, and which activities they undertake [14,33]. Within this framework, socio-demographic factors are not treated as direct predictors of well-being outcomes; rather, they operate indirectly by shaping the conditions under which engagement becomes possible, producing differentiated experiential pathways [34,35].

2.3.2. Accessibility and Mobility

Accessibility functions as a structural gateway through which potential interactions with UBGS are translated into actual use. It is conceptualised as a multidimensional construct that extends beyond proximity to encompass objective factors such as travel time and transport mode, as well as subjective perceptions of ease and safety [36]. In dense urban environments, short physical distances may still entail significant experiential barriers, including exposure to traffic, pedestrian infrastructure, perceived safety, and thermal discomfort [37,38]. Accessibility is essential for engagement and usage patterns, but does not directly lead to restoration or satisfaction with well-being outcomes.

2.3.3. Use Pattern and Activity Type

Use patterns constitute the behavioural core of the framework, capturing not only how frequently UBGS are used but, more critically, how engagement is enacted in everyday practice. While visit frequency reflects the quantity of exposure, activity type represents the qualitative mode of engagement that shapes how environments are encountered and the psychological functions they serve [39]. Empirical evidence indicates that the mode of engagement may exert a more substantial influence on restorative outcomes than frequency [40,41]. Accordingly, activity type is conceptualised as a moderating factor that shapes how environmental conditions, such as the acoustic environment, seating availability, and path quality, are perceived and translated into restorative experiences [42].

2.3.4. Perceived Environmental Quality

Perceived environmental quality in UBGS experiences involves sensory and contextual factors affecting urban comfort, including cleanliness, shade, thermal comfort, noise reduction, and amenities such as seating and paths [43,44]. Environmental quality is crucial for restorative experiences, with moderate quality in dense areas potentially offering more restoration than in urban settings. Visually appealing spaces require functional elements, such as thermal comfort and noise reduction, to deliver benefits. Multisensory engagement through visual, auditory, olfactory, and tactile stimuli enhances the restorative capacity of urban environments and reduces stress. Environments with diverse vegetation, natural sounds, water features, and biophilic design are more effective for restoration than those that focus on visual aspects [45,46]. Soundscape quality is vital; relief from traffic noise supports reflection and recovery, whereas continuous noise disrupts restoration. Users’ assessments show that maintenance, sensory diversity, and functional design influence perceived restorativeness, highlighting the importance of holistic, sensory-rich environments for psychological restoration in urban areas [47].

2.3.5. Restorative Experiences

Restorative experiences are positioned centrally within the framework, representing a key explanatory factor associated with how engagement with UBGS relates to well-being. Drawing on ART and SRT, restoration is conceptualised as a multidimensional experiential process encompassing emotional calmness, cognitive refreshment, mood improvement, physical relaxation, and recovery from fatigue [8,9]. In high-stress urban environments, the embodied and somatic dimensions of restoration, such as deep relaxation and relief from cumulative daily fatigue, are particularly salient to users [48,49]. Thus, the restorative experience serves as an integrative link through which accessibility, behavioural engagement, and perceived environmental quality are synthesised into perceived well-being.

2.3.6. Overall Satisfaction

Overall satisfaction is positioned as the final evaluative outcome of the framework, capturing users’ assessment of whether a BGS meets their needs for comfort and restoration. Satisfaction primarily emerges from the quality of restorative experiences during use [13]. Satisfaction reflects the successful realisation of cultural ecosystem services, integrating sensory perceptions, emotional responses, and perceived restoration into an evaluative judgement [26]. Satisfaction functions as a behavioural reinforcer promoting repeated visits to BGS, which are essential for maintaining well-being benefits [50]. Studies have shown that satisfaction, shaped by perceived restorativeness and environmental quality, influences use patterns and supports outcomes such as social cohesion, environmental awareness, and life satisfaction [51]. Overall satisfaction serves both as a measure of experiential performance and as a mechanism for sustaining cultural ecosystem services in urban contexts. Figure 2 presents the experience-centred conceptual framework, illustrating the relationships among accessibility, use, perceived environmental quality, restorative experiences, and overall satisfaction. The framework is presented as a theoretically informed interpretive model rather than a statistically tested mediation structure; given the cross-sectional design, the relationships should be understood as associative rather than causal.

2.4. Research Gaps and Positioning of the Present Study

Despite the increasing recognition of the experiential dimensions in UBGS research, key gaps persist. First, studies often conceptualise well-being outcomes as direct responses to environmental attributes, with comparatively limited attention to the role of use behaviour and restorative experience in shaping these associations. Second, evidence from Global South megacities remains sparse, particularly for everyday small- to medium-sized UBGS under extreme density, environmental stress, and infrastructural constraints. Third, research rarely integrates socio-demographic factors, accessibility, behavioural engagement, perceived environmental quality, restorative experience, and satisfaction within a single analytical framework, limiting understanding of how these elements interact to produce well-being outcomes. Fourth, this study adopts an experience-centred framework that conceptualises restorative experience as a central explanatory factor associated with environmental qualities, patterns of use, and user satisfaction. To operationalise this, the study combines in situ survey data with ML-based interpretation using SHAP to unpack complex, non-linear relationships. This study frames UBGS as a dynamic restorative infrastructure embedded in daily urban life rather than as a static spatial asset. This integrated methodological and conceptual approach provides insights into how UBGS functions, offering evidence for planning and managing equitable, experience-responsive, and health-supportive green infrastructure in high-density cities in the Global South.

3. Materials and Methods

3.1. Study Area

Delhi, the National Capital Territory of India, represents a quintessential case of a high-density Global South megacity and provides a critical context for examining how UBGS operate as everyday restorative infrastructure for active users (Figure 3). With a projected urban population exceeding 32.2 million by 2025, Delhi is among the world’s largest and most densely populated urban agglomerations [52], experiencing acute socio-environmental pressures typical of rapidly urbanising regions, including severe congestion, chronic thermal stress, air and noise pollution, and sustained strain on infrastructure and public amenities [7]. In this setting, accessible urban nature is not merely an amenity but a vital public-health resource. Although precise quantitative ratios of blue to green spaces are not consistently defined, UBGS in Delhi are increasingly understood as an integrated network of natural and semi-natural areas that deliver multiple ecosystem services and significant physical and mental health benefits [53,54,55]. Therefore, examining how residents’ access, use, and experience these spaces offers critical insights for evidence-based planning of equitable, health-supportive, and climate-resilient urban environments in Delhi and other high-density Global South cities.

3.2. Data Collection and Survey Design

3.2.1. Study Sites

The study was conducted across five urban parks in Delhi, which were selected based on ecological significance, scale, accessibility, and importance within the city’s UBGS network. The sites such as Aastha Kunj Park, India Gate Lawns, Lodhi Garden, Sunder Nursery, and Buddha Jayanti Park (Figure 3) represent diverse landscape types, vegetation structures, and user compositions (

Table 1. Characteristics of Study Sites.

Park	Approx. Area	Blue/Green Elements	Facilities	Management	Surrounding Land Use	Typical Crowding
Lodhi Garden	~90 acres	Mature trees (neem, banyan, amaltas); ornamental shrubs; water body (central pond); extensive lawns	Walking paths, seating benches, heritage tombs, toilet blocks, drinking water, and entry gates	NDMC	High-density mixed-use (residential, commercial, institutional); Lutyens’ Delhi	High (morning/evening peak)
Sunder Nursery	~90 acres	Mature trees; heritage trees; water bodies (lakes, ponds); flowering gardens; arboretum	Ticketed entry, cultural spaces, walking tracks, seating, café, toilet blocks, children’s play area, entry plaza	Aga Khan Trust and NDMC	High-density residential and institutional area with mixed land use	Moderate to High
Buddha Jayanti Park	~80 acres	Dense tree cover; flowering shrubs; water body; Japanese garden section	Walking paths, meditation area, seating, toilet blocks, parking	DDA	Medium-density institutional (Ridge area) and diplomatic enclave	Moderate
India Gate Lawns	NA	Lawns; avenue trees (neem, jamun); ornamental planting; memorial structures; water features	Open lawns, walking paths, informal seating (lawn), food vendors, night lighting, toilet blocks	CPWD (Central Public Works Department)	High-density commercial and institutional, Central Vista; major tourist destination	Very High
Aastha Kunj Park	~200 acres	Plantation woodland; flowering shrubs; informal green space	Walking paths (paved and unpaved), seating area, open lawns, entry gates, informal gathering spaces	DDA	High-density residential and institutional area	Moderate

). These parks serve as recreational and ecological assets, supporting walking, leisure, social interactions, and nature observation. Their intensive use makes them suitable for examining routine interactions with the UBGS and associated restorative experiences.

3.2.2. Survey Administration

Primary data were collected through structured questionnaire surveys conducted during October-November 2025, coinciding with early winter in Delhi. This period was intentionally selected for data collection as it represents the optimal use conditions in Delhi’s climatic cycle. It is characterised by pleasant daytime temperatures (25–30 °C) and reduced thermal stress following the monsoon; it typically records peak park visitation and longer outdoor stays. These favourable conditions provide an ideal setting to capture the most comprehensive expression of restorative experiences. Surveys were administered in the morning and evening on weekdays and weekends to capture temporal variations in visitation patterns, activities, and user profiles. Within each park, respondents were approached across multiple zones, including entrances, paths, seating areas and lawns, to capture a range of engagement contexts. Eligible participants were adult park users aged 18 or above who were present and willing to participate voluntarily.

A total of 411 completed questionnaires were considered valid for analysis. The site-level distribution was as follows: Lodhi Garden (n = 92, 22.4%), Sunder Nursery (n = 87, 21.2%), Buddha Jayanti Park (n = 84, 20.4%), India Gate Lawns (n = 75, 18.2%), and Aastha Kunj Park (n = 73, 17.8%). Although the study employed opportunistic in-situ sampling rather than quota-based stratification, spatial and temporal rotation strategies were implemented to enhance the coverage of diverse user experiences. While the sampling strategy does not permit statistical generalisation to all Delhi park users, it is appropriate for exploratory and explanatory analyses of perceptual and restorative outcomes central to this study’s objectives. This sample comprised park users present during the survey administration and was not statistically representative of all Delhi park users or the broader population. The findings reflect respondents’ experiences across the five study sites and should not be interpreted as statistically generalisable. The sampling approach was appropriate for exploratory and explanatory analyses aligned with the study objectives.

3.2.3. Sampling Strategy and Ethical Considerations

An opportunity sampling strategy was employed, consistent with established practices in in situ environmental perception and public space research [56,57]. This approach was selected to reduce self-selection bias associated with recruitment methods that rely on prior interest in environmental or well-being research, while enhancing inclusivity by engaging users during routine park use. The inclusion criteria were as follows: (i) age ≥ 18 years, (ii) apparent capacity to provide informed consent, and (iii) active use of the identified UBGS at the time of recruitment. Individuals aged < 18 years were excluded. The study objectives and procedures were explained after the initial contact, and written informed consent was obtained via a signed consent form. No identifying information was collected, with responses recorded anonymously. Data were collected for academic research, ensuring participant confidentiality and ethical compliance. Recruitment occurred within park boundaries, and the sample is conditioned on current UBGS users present during the survey period. Non-users, infrequent visitors, and individuals facing access barriers were not represented. The findings reflect perceptions of active users and do not capture unmet demand or structural access constraints beyond this group.

3.2.4. Questionnaire Structure and Variables

The survey instrument followed an experience-centred design with six thematic domains: (i) socio-demographic characteristics (age group, gender, education), (ii) accessibility and mobility (transport mode, travel time, perceived ease and safety of access), (iii) usage patterns (visit frequency, duration, and activities like walking, observing greenery, socialising, exercising), (iv) perceived environmental quality, (v) psychological well-being and restorative outcomes, and (vi) overall satisfaction (

Table 2. Measurement Framework and Construct Operationalisation.

Construct	Items	Conceptual Source	Scale Anchors	Scoring Approach	Reliability Evidence
Socio-demographic	Age, Gender, Education	Standard demographic variables	Categorical	Descriptive grouping	Not applicable
Accessibility & Mobility	Travel time, transport mode, perceived ease, perceived safety	Urban accessibility literature	Mixed (categorical and Likert 1–5)	Analysed individually	Not applicable
Usage Patterns	Visit frequency, duration, main activity	Recreation behaviour literature	Categorical	Analysed individually	Not applicable
Environmental Quality (Formative)	Visual attractiveness, cleanliness, shade, noise escape, seating adequacy, and path condition	Environmental perception research	Likert (1–5)	Analysed individually (not aggregated)	α = 0.039 (low; expected due to formative structure)
Restorative Experience (Reflective)	Calmness, mental refreshment, mood improvement, recovery from tiredness, relaxation	Conceptually informed by ART	Likert (1–5)	Analysed individually and intercorrelated	α = 0.432 (multidimensional)
Overall Satisfaction	Single-item satisfaction	Environmental satisfaction literature	Ordinal (Very dissatisfied-Very satisfied)	Ordinal outcome variable	Single item

).

Perceived environmental quality was measured using five-point Likert-scale items (1 = strongly disagree; 5 = strongly agree) that captured visual attractiveness, cleanliness and maintenance, shade and thermal comfort, escape from noise and traffic, seating adequacy and path condition. These items were conceptualised as perceptual evaluations of distinct environmental features within UBGS rather than indicators of a single latent construct.

Restorative experience indicators were conceptually informed by ART [8] and SRT [9]. This study employed concise outcome-oriented indicators to capture self-reported affective and recovery-related responses to UBGS exposure. These included perceived calmness, mental refreshment, mood improvement, recovery from fatigue, and post-visit relaxation. While ART emphasises cognitive restoration processes (e.g., directed attention recovery) and SRT focuses on affective stress reduction, the selected items reflect experiential outcomes associated with these theoretical traditions rather than directly measuring core ART dimensions such as being away, fascination, extent, or compatibility.

All restorative items were measured using a five-point Likert scale (1 = strongly disagree; 5 = strongly agree), with higher scores indicating stronger perceived restorative benefits. Overall satisfaction with the park experience was measured using a single ordinal item from very dissatisfied to very satisfied. A clear distinction was maintained between the two constructs based on their theoretical nature: environmental quality items were treated as formative indicators, each representing a distinct environmental attribute that does not derive from a shared cause, while restorative experience items were treated as related affective and recovery-oriented outcomes that are empirically distinct. This theoretical distinction governed analytical decisions. Restorative indicators were analysed separately and intercorrelated to examine their dimensionality, and environmental attributes were entered individually into the models, preserving the multidimensional structure consistent with their formative specification.

In contrast, environmental quality items were conceptualised as formative perceptual attributes representing heterogeneous environmental features (e.g., aesthetics, cleanliness, thermal comfort, infrastructure). As formative indicators do not assume inter-item covariance or unidimensionality, environmental quality variables were analysed individually rather than combined into a composite scale. The questionnaire used explicit, accessible language to ensure comprehension among diverse user groups with varying educational backgrounds.

3.3. Statistical Analysis

Statistical analyses were conducted to examine the descriptive patterns, interrelationships, group differences, and predictors of overall satisfaction among urban park users. All statistical analyses were performed using Python 3.10 (JupyterLab) software. Ordinal logistic regression modelling and associated assumption diagnostics (Test of Parallel Lines, VIF) were conducted using IBM SPSS Statistics Version 27.

3.3.1. Descriptive Statistics

Descriptive statistics were computed to summarise socio-demographic characteristics, usage patterns, environmental perceptions, restorative experiences, and satisfaction levels. For ordinal and Likert-scale variables, means, medians, standard deviations, skewness, and frequency distributions were reported to characterise central tendency and dispersion.

3.3.2. Construct Structure and Reliability Assessment

An exploratory factor analysis (EFA) was conducted to examine the empirical clustering of Likert-scale items assessing perceived environmental quality and restorative experience. This EFA aimed to examine inter-item relationships and identify empirical groupings prior to determining the treatment of each construct in the main analyses. The two constructs were treated differently according to theory: environmental quality items were conceptualised as formative indicators representing distinct, heterogeneous attributes, while restorative experience items were treated as reflective-like, multidimensional outcomes informed by ART and SRT. This distinction governed how the constructs were analysed. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (0.523) was marginally acceptable by conventional standards (minimum threshold = 0.50; Kaiser, 1974), indicating that the item set provides a limited but sufficient basis for factor analysis. This marginal KMO value reflects the heterogeneous nature of the items, spanning distinct environmental attributes and restorative outcomes, and should be treated as a limitation. Bartlett’s test of sphericity was significant (χ² = 274.763, df = 55, p < 0.001), confirming that the inter-item correlations were non-zero and factor analysis was appropriate.

A theory-guided two-factor solution was retained, accounting for 28.6% of the total variance. This is a limitation of the factor-analytic results: the low explained variance indicates the two-factor solution captures only a modest portion of total item variance, and the EFA should not be interpreted as providing strong structural evidence for the constructs. The limited explained variance stems from the heterogeneous nature of the items: environmental appraisal and psychological restoration constructs in urban research are not expected to show a strict unidimensional structure, and combining items from two distinct domains naturally constrains shared variance. The weak factor solution reinforces the decision not to treat either construct as a composite scale, and the analytical strategy of entering all items individually does not depend on the factor solution’s robustness. The EFA is reported for transparency rather than structural justification. The rotated solution showed clustering corresponding to environmental quality attributes and restorative experience outcomes, though several cross-loadings indicated partial overlap between environmental perceptions and psychological restoration. Given this structure, the constructs were not treated as strict unidimensional latent scales. The full rotated component matrices and internal consistency statistics are provided in Supplementary Tables S1 and S2.

Internal consistency reliability was assessed using Cronbach’s alpha for scale diagnostics, while recognising its relevance only for reflective (unidimensional) constructs. Restorative experience items showed limited internal consistency (α = 0.432), reflecting their conceptualisation as theoretically related but distinct dimensions of restoration (emotional, cognitive, and somatic) rather than a unidimensional reflective scale. As supported by weak-to-moderate inter-item correlations, these items capture overlapping but distinguishable outcomes. They were analysed individually and intercorrelated rather than aggregated into a composite score. The alpha value is reported for transparency, but should be interpreted in light of this multidimensional structure. Environmental quality items showed very low internal consistency (α = 0.039), which is expected and consistent with their conceptualisation as formative indicators representing distinct environmental attributes (aesthetics, infrastructure, acoustic conditions) that do not share a common cause. Cronbach’s alpha is not meaningful for formative constructs, as such items are not expected to co-vary; the low alpha value confirms the heterogeneous nature of the environmental quality items. The EFA was employed as an exploratory tool to inspect empirical item clustering; it was not used to impose a reflective model on environmental quality, and its use aligns with the formative treatment in the analyses. Environmental attributes were analysed individually rather than combined into a composite index.

3.3.3. Normality Testing

The distribution of perception and well-being indicators was assessed using the Shapiro-Wilk (W) and Kolmogorov-Smirnov (D) tests (Equation (1)):

$$W={\displaystyle \frac{{\left(\sum _{i=1}^n{a}_i{x}_{\left(i\right)}\right)}^2}{\sum _{i=1}^n(x_i-\overline{x}{)}^2}}$$

(1)

This indicates statistically significant departures from normality (p < 0.05) for most variables, justifying the use of non-parametric statistical methods.

3.3.4. Correlation Analysis

The associations among the restorative experience indicators were examined using Spearman’s rank-order correlation (ρ), which is suitable for ordinal and non-normally distributed data (Equation (2)):

$$\rho =1-{\displaystyle \frac{6\sum d_i^2}{n(n^2-1)}}$$

(2)

where $d_i$ is the difference between paired ranks, and $n$ is the number of observations. Correlation strength was interpreted as weak (ρ < 0.30), moderate (0.30–0.50), or strong (>0.50).

3.3.5. Group Differences in Satisfaction

Differences in overall satisfaction across activity types and user groups were tested using the Kruskal-Wallis H test (Equation (3)):

$$H={\displaystyle \frac{12}{N(N+1)}}{\sum }_{j=1}^k{\displaystyle \frac{R_j^2}{n_j}}-3(N+1)$$

(3)

where $R_j$ is the sum of ranks for group $j$, $n_j$ is the group size, and $N$ is the total sample size. This test was selected due to its robustness to non-normality and unequal group sizes.

3.3.6. Ordinal Logistic Regression

To identify predictors of overall satisfaction, an ordinal logistic regression (proportional odds model) was estimated (Equation (4)):

$$\mathrm{l}\mathrm{o}\mathrm{g}\left({\displaystyle \frac{P(Y\le j)}{P(Y>j)}}\right)={\alpha }_j-{\sum }_{i=1}^k{\beta }_i{X}_i$$

(4)

where $Y$ represents the satisfaction level, ${\alpha }_j$ are threshold parameters, and ${\beta }_i$ are coefficients for predictor variables $X_i$. Model significance was assessed using likelihood ratio tests; p < 0.05 was considered statistically significant.

Prior to the estimation, the key model assumptions were assessed. The proportional odds assumption was tested using the Test of Parallel Lines and was not violated (χ² = 3.216, df = 6, p = 0.781), indicating that the slope coefficients were consistent across cumulative response thresholds. Multicollinearity was examined using Variance Inflation Factors (VIFs), with values ranging from 1.03 to 1.33, well below the commonly accepted threshold of 5, indicating no problematic collinearity among predictors (see Supplementary Table S3). Likert-scale predictors were treated as continuous covariates, consistent with the standard practice in applied social science modelling. No missing values were present in the dataset, and all observations (N = 411) were retained for analysis. The model fit was further evaluated using pseudo R² statistics (Cox & Snell, Nagelkerke, and McFadden).

3.4. Machine Learning and SHAP Analysis

3.4.1. Random Forest Model

A Random Forest (RF) model was implemented using 20 indicators representing experiential, behavioural, environmental, and accessibility-related dimensions of urban park use to capture non-linear relationships and interaction effects. Unlike the ordinal regression model, which focused specifically on restorative experience variables as theoretically central predictors of satisfaction, the RF model incorporated environmental quality, accessibility, behavioural, and experiential indicators simultaneously in an integrative exploratory framework. This approach enables the assessment of relative feature contributions without presupposing direct linear causality.

The dependent variable, overall satisfaction (five-point Likert scale: 1 = very dissatisfied to 5 = very satisfied), was treated as a continuous numerical outcome and modelled using a RandomForestRegressor. The variable was not dichotomised or collapsed into binary or reduced-ordinal classes, thereby preserving ordinal-intensity information while enabling flexible non-linear regression modelling. The RF model was specified with the following hyperparameters: number of trees (n_estimators) = 100, maximum tree depth (max_depth) = 6, maximum features considered at each split (max_features) = default (all features), splitting criterion = squared error, and random_state = 0 to ensure reproducibility. Tree depth was constrained to six levels to reduce overfitting and enhance generalisation stability.

The dataset was split into 70% training and 30% testing data, with the training subset used to fit the model and the testing subset to evaluate predictive performance. A fixed random seed was applied to ensure the replicability of the train-test partition. A five-fold cross-validation was performed on the training dataset to assess model stability and reduce dependence on a single split. The RF model generates predictions by aggregating the outputs of multiple decision trees, expressed as

$$\widehat{y}={\displaystyle \frac{1}{B}}{\sum }_{b=1}^B{f}_b(x)$$

(5)

where $f_b(x)$ denotes the prediction of the $b^{\mathrm{th}}$ decision tree and $B$ represents the total number of trees in the ensemble. The model performance was evaluated using an independent test set with the coefficient of determination (R²), learning curve diagnostics, and residual analysis. Although the model was estimated as a regression, the predicted values were also rounded to the nearest Likert category to compute classification-equivalent metrics (confusion matrix, overall accuracy, and weighted F1-score), enabling a transparent evaluation of ordinal prediction consistency.

3.4.2. Model Validation and Performance Evaluation

Model stability was assessed using five-fold cross-validation on the training dataset (mean score = 0.959, SD = 0.024). The predictive performance was evaluated on an independent 30% test set using the coefficient of determination (R²), learning curve analysis, and residual diagnostics. The test set R² was 0.660, indicating that 66% of the variance in the overall satisfaction was explained. Although estimated as a regression model, the predictions were rounded to the nearest Likert category to assess ordinal consistency. A confusion matrix, overall accuracy, and weighted F1-score were computed. Misclassifications were primarily confined to adjacent categories, consistent with the ordinal structure of outcomes.

3.4.3. SHAP-Based Model Interpretation

To assess the relative importance and directional influence (positive or negative) of individual indicators on overall satisfaction among urban park users in Delhi, SHAP was applied to the trained RF model. SHAP provides a model-agnostic, game-theoretic framework that decomposes each prediction into additive feature contributions relative to the baseline expectation. For a given feature $i$, the SHAP value ${\varphi }_i$ is defined as:

$${\varphi }_i={\sum }_{S\subseteq N\setminus \left\{i\right\}}{\displaystyle \frac{\mid S\mid !(\mid N\mid -\mid S\mid -1)!}{\mid N\mid !}}\left[f(S\cup \{i\left\}\right)-f\left(S\right)\right]$$

(6)

where $S$ denotes a subset of features and $N$ represents the complete set of predictors. Positive SHAP values indicate that a feature increases the model’s satisfaction prediction, whereas negative values indicate a decrease from the baseline. Mean absolute SHAP values were computed to assess feature importance by comparing indicators based on model contribution. SHAP summary plots were used to examine the direction, magnitude, and variability of the feature effects.

4. Results

4.1. Socio-Demographic Characteristics of Respondents

A total of 411 valid questionnaire responses were analysed, with no missing socio-demographic data. The respondent profile captured a diverse cross-section of individuals using the five study parks during the survey period. In

Table 3. Socio-demographic characteristics of Delhi’s Blue Green Spaces.

Variable	Category	Frequency	Percentage (%)
Age group (years)	18–25	50	12.2
	25–35	80	19.5
	35–45	92	22.4
	45–55	85	20.7
	55–65	55	13.4
	65+	49	11.9
Gender	Male	238	57.9
	Female	173	42.1
Education level	Primary or less	2	0.5
	Secondary/High school	136	33.1
	Diploma/Vocational	1	0.2
	Graduation	141	34.3
	Above graduation	131	31.9

, participants were well distributed across adult age groups, indicating balanced representation across the life course. The largest share of respondents belonged to the 35–45 years age cohort (22.4%), followed by those aged 45–55 years (20.7%) and 25–35 years (19.5%). Younger adults aged 18–25 years comprised 12.2% of the sample, while older adults (≥55 years) accounted for approximately 25.3%, highlighting substantial participation among both early- and later-life urban residents. The gender distribution showed a moderate male predominance, with 57.9% male and 42.1% female respondents.

The respondents’ educational attainment was high. More than two-thirds (66.2%) reported having graduate-level education or higher, with 34.3% holding bachelor’s degrees and 31.9% having qualifications above the graduate level. Only 0.5% reported having received primary education or less. This distribution indicates a sample capable of engaging with environmental perceptions and urban-ecological issues. However, this reflects an overrepresentation of highly educated individuals compared to Delhi’s population, where 25–30% of adults hold graduate degrees. This pattern is common in on-site park user surveys conducted in Indian cities. This may be attributed to differential visitation patterns, where educated individuals show greater recreational awareness, flexible schedules, or residential proximity to UBGS, and to response bias, as those with higher education are often more willing to participate in academic surveys. Cross-tabulations (Supplementary Table S4) show that while visitation patterns and activity types vary across education levels, overall satisfaction does not show a clear monotonic gradient. These findings suggest that, although the sample is educationally skewed, core experiential outcomes are not driven by any single educational group.

4.2. Patterns of Use and Visitation of Urban Blue-Green Spaces

Visitation patterns among respondents indicate that, for these park users, UBGS serves as an everyday routine environment and an occasional recreational destination. A substantial proportion of respondents (43.1%) used UBGS three or more times per week, including 21.4% who visited three to four times weekly and 21.7% who reported five or more visits per week (

Table 4. Frequency of visits to urban blue-green spaces in Delhi.

Visit Frequency	Frequency	Percentage (%)
Less than once a month	112	27.3
1–3 times per month	71	17.3
1–2 times per week	51	12.4
3–4 times per week	88	21.4
5 or more times per week	89	21.7
Total	411	100.0

). This high frequency of use suggests a strong integration of such spaces into the daily urban life of a significant segment of the population. In contrast, 27.3% of respondents reported visiting less than once per month, while an additional 17.3% reported visiting once to three times per month. A further 12.4% engaged in moderate and regular use, visiting once or twice per week. These patterns reveal a polarised distribution of engagement within our sample, characterised by the coexistence of frequent users alongside a substantial group of infrequent visitors, underscoring pronounced heterogeneity in use intensity among the individuals surveyed.

Indicators of accessibility and mobility are summarised in

Table 5. Accessibility and mobility.

Variable	Category	Frequency	Percentage (%)
Mode of transport	Walking	184	44.8
	Bicycle	3	0.7
	Public transport	140	34.1
	Two-/four-wheeler	84	20.4
Travel time to blue-green spaces	<5 min	54	13.1
	5–10 min	90	21.9
	10–20 min	132	32.1
	20–30 min	68	16.5
	>30 min	67	16.3
Perceived ease of access	Very difficult	70	17.0
	Difficult	95	23.1
	Neither easy nor difficult	42	10.2
	Easy	106	25.8
	Very easy	98	23.8

. Walking emerged as the dominant mode of access (44.8%), followed by public transport (34.1%), while private motorised transport was used by 20.4% of respondents. Cycling accounted for a negligible share, indicating limited uptake of active transportation beyond walking. Travel times were generally short, with 67.2% of respondents reaching UBGS within 20 min, suggesting that many respondents reside within a short travel distance of the surveyed parks. However, a notable minority (32.8%) reported travel times exceeding 20 min, indicating that perceived access conditions are not uniform across respondents. The perceived ease of access exhibited marked variations. While 49.6% of respondents described access as easy or very easy, a substantial proportion (40.1%) reported access as challenging, indicating that physical proximity does not uniformly translate to perceived accessibility.

4.3. Perceived Environmental Quality

Perceived environmental quality was assessed using six indicators that captured the visual, functional, and comfort-related attributes of the UBGS (

Table 6. Perceived environmental quality.

Indicator	Mean	Standard Deviation
Visual attractiveness	3.50	1.27
Cleanliness and maintenance	3.04	1.43
Shade and thermal comfort	3.18	1.45
Escape from noise and traffic	2.86	1.35
Adequacy of benches	3.05	1.41
Condition of paths	3.03	1.37

). The mean scores ranged from 2.86 to 3.50 on a five-point Likert scale, indicating a moderate to moderately high overall perceived quality. Among the evaluated dimensions, visual attractiveness received the highest rating (M = 3.50, SD = 1.27), followed by shade and thermal comfort (M = 3.18). Indicators associated with maintenance and supporting infrastructure, including cleanliness, bench adequacy, and path condition, clustered around the midpoint of the scale (M ≈ 3.0–3.2), suggesting generally acceptable conditions with scope for improvement. The lowest mean score was recorded for escape from noise and traffic (M = 2.86, SD = 1.35), reflecting perceived limitations of acoustic buffering in the urban context. Across all indicators, relatively large standard deviations were observed, indicating substantial inter-individual variability in perceived environmental quality. These indicators characterise the contextual conditions of the restorative experiences. Although not direct predictors in the ordinal regression model, they were incorporated into the ML analysis to explore their relative contributions within a broader modelling framework.

4.4. Restorative Experiences

Restorative outcomes were assessed using five indicators that captured the emotional, cognitive, and recovery-related dimensions of the experience (

Table 7. Restorative experience indicators.

Indicator	Mean	Median	Standard Deviation
Feel calmer after the visit	3.66	4	1.25
Feel mentally refreshed	3.45	4	1.34
Recovery from tiredness	3.42	4	1.40
Mood improvement	3.41	4	1.44
Feel more relaxed after the visit	3.07	3	1.44

Note: 1 = strongly disagree, 5 = strongly agree.

). The mean values ranged from 3.07 to 3.66, indicating moderately high restorative benefits associated with visits to the UBGS. The highest mean score was recorded for feeling calmer after the visit (M = 3.66, Median = 4), followed by mental refreshment (M = 3.45). Indicators reflecting recovery from tiredness (M = 3.42) and mood improvement (M = 3.41) also received favourable evaluations. The median values were 4 for four of the five indicators, indicating a predominance of agreement-oriented responses across most restorative dimensions. All restorative indicators showed significant departures from normality, with distributions that were negatively skewed (i.e., skewed toward higher agreement values) (Shapiro-Wilk tests, p < 0.001), thereby justifying the application of non-parametric statistical techniques in subsequent analyses.

4.5. Interrelationships Among Restorative Experience Dimensions

To examine whether restorative outcomes operate as interrelated dimensions, Spearman’s rank-order correlation analysis was conducted among the five restorative indicators (Figure 4). The study revealed a pattern of statistically significant, yet weak to moderate, positive associations across the restorative dimensions. Mental refreshment was positively related to feeling calmer (ρ = 0.214, p < 0.01) and recovering from tiredness (ρ = 0.211, p < 0.01), indicating an overlap between cognitive restoration and affective and physical recovery processes. The strongest association was found between recovery from tiredness and post-visit relaxation (ρ = 0.243, p < 0.01), suggesting that somatic fatigue recovery is linked to embodied relaxation.

Recovery from tiredness was also significantly associated with mood improvement (ρ = 0.112, p < 0.05), although this relationship was weaker. Correlations between affective calming and cognitive refreshment highlight interconnected restorative pathways, while weak associations involving mood improvement (ρ = 0.009–0.130) suggest that mood enhancement functions as a distinct outcome within the restorative experience. The correlation patterns show restorative experiences from UBGS use are multidimensional yet interconnected, with fatigue recovery and relaxation linking emotional, cognitive, and physical responses. Although these restorative outcomes co-occur and form a coherent experience, they are not redundant.

4.6. Behavioural and Socio-Demographic Differentiation

Differences in satisfaction with UBGS visits were examined across primary activity types using the Kruskal-Wallis test (

Table 8. Kruskal–Wallis test for satisfaction across activity types.

Test	Test Statistic (χ2)	Degrees of Freedom	p-Value
Overall satisfaction × Primary activity type	43.593	6	<0.001

). The results revealed a statistically significant variation in satisfaction across activity categories (χ²(6) = 43.59, p < 0.001), indicating that satisfaction derived from the UBGS differs systematically depending on how individuals engage with these environments. In contrast, no significant differences were observed across social contexts, suggesting that activity type exerts a stronger influence on satisfaction outcomes than companionship.

Behavioural differentiation was further evident across socio-demographic groups. A significant association was identified between gender and primary activity type (χ² (6) = 18.30, p = 0.006;

Table 9. Association between gender and primary activity.

Test	χ2	df	p-Value	Effect Size (Cramér’s V)
Gender × Primary activity type	18.302	6	0.006	0.211

). The corresponding effect size (Cramér’s V = 0.211) indicates a small-to-moderate association, reflecting gender-based variation in engagement patterns within UBGS. Similarly, age group was significantly associated with visit frequency (χ² (20) = 36.64, p = 0.013;

Table 10. Association between age group and frequency of visits.

Test	χ2	df	p-Value	Effect Size (Cramér’s V)
Age group × Visit frequency	36.638	20	0.013	0.149

). Although the effect size was modest (Cramér’s V = 0.149), the results indicate systematic life-course differences in use intensity, with patterns of engagement varying across age cohorts. The observed effect sizes were small to moderate, indicating patterned behavioural differentiation without strong stratification in restorative outcomes or satisfaction.

4.7. Socio-Demographic Differences in Access and Restorative Outcomes

To further examine the distribution of experiential outcomes, subgroup analyses were conducted across gender, age group, and education level (used as a proxy for socioeconomic status). Gender was significantly associated with the primary activity type (χ²(6) = 18.30, p = 0.006; Cramér’s V = 0.211), indicating small-to-moderate variation in engagement patterns. Age group was significantly associated with visit frequency (χ²(20) = 36.64, p = 0.013; Cramér’s V = 0.149), reflecting modest life-course differences in use intensity. Descriptive cross-tabulations further indicated variations in visitation patterns and satisfaction distributions across education levels (Supplementary Table S5), although no consistent monotonic gradient in overall satisfaction was observed. Accessibility indicators (mode of transport, travel time, and perceived ease of access) also displayed descriptive differences across socio-demographic categories. Perceived environmental quality and restorative experience indicators did not display clear socio-demographic patterns in descriptive comparisons. The observed effect sizes for statistically tested associations were small to moderate (Cramér’s V ≤ 0.211), suggesting patterned behavioural differentiation without evidence of pronounced stratification in experiential outcomes within the surveyed user population.

4.8. Predictors of Overall Satisfaction with Urban Blue-Green Spaces

Prior to modelling, overall satisfaction responses indicated moderately positive evaluations, with approximately 45% of respondents reporting satisfaction or high satisfaction with their UBGS experience. An ordinal logistic regression was conducted to identify the key experiential predictors of overall satisfaction (

Table 11. Ordinal regression predicting overall satisfaction.

Predictor	Estimate (β)	Std. Error	Wald χ2	p-Value	95% CI
Mental refreshment (E2)	−0.063	0.066	0.918	0.338	−0.192, 0.066
More relaxed after visit (E5)	0.168	0.062	7.417	0.006	0.047, 0.289

Model fit statistics: Model χ² = 8.497, df = 2, p = 0.014; Nagelkerke R² = 0.021; Link function: Logit.

). The ordinal logistic regression model significantly improved the fit relative to the intercept-only model (−2LL intercept = 397.928; −2LL final = 389.432; χ²(2) = 8.497, p = 0.014). The pseudo-R² values indicated modest explanatory power (Cox & Snell = 0.020; Nagelkerke = 0.021; McFadden = 0.007), suggesting limited variance explained by the included predictors. This modest explanatory power reflects the inherent complexity of experiential satisfaction in urban environments, which is shaped by diverse contextual, situational, and subjective factors that cannot be fully captured by survey-based models alone. Among the examined variables, post-visit relaxation was the only significant positive predictor of overall satisfaction (β = 0.168, p = 0.006). In contrast, mental refreshment did not have a statistically significant independent effect (β = −0.063, p = 0.338). These findings indicate that affective restoration, particularly post-visit relaxation, is more strongly associated with overall satisfaction than cognitive refreshment.

The analysis shows that UBGS in Delhi are frequently used and form an integral part of residents’ urban environments. Many respondents reported visiting these spaces at least three times weekly, accessing them by walking and public transport, with travel times under 20 min. Despite their proximity, perceptions of access varied considerably, indicating uneven accessibility across the city. The perceived environmental quality of the UBGS was moderate, with higher ratings for visual attractiveness and shade comfort and lower assessments for noise reduction and infrastructure. Respondents consistently reported restorative experiences from UBGS use, particularly emotional calmness, mental refreshment, and fatigue recovery, with these dimensions significantly interrelated. Behavioural analyses showed significant differences in satisfaction across activity types, alongside socio-demographic variation in activity preferences and visitation frequency. Ordinal regression identified post-visit relaxation as the only significant predictor of overall satisfaction, highlighting the importance of experiential relaxation in user evaluations of the UBGS. From a planning perspective, using interpretable ML techniques such as SHAP enables the transparent identification of design and management priorities, moving beyond black-box prediction toward evidence-based urban decision-making.

4.9. SHAP-Based Interpretation of Satisfaction Factors and Restorative Mechanisms

Figure 5a presents the mean absolute SHAP values, which illustrate the relative importance of the indicators in explaining the overall satisfaction of urban park users in Delhi. The results revealed a hierarchy, with comfort-oriented and experiential factors having a greater influence than accessibility, social, or cognitive variables. Within the SHAP-based integrative model (Figure 5), bench adequacy displayed the highest mean absolute SHAP value, indicating a strong contribution within the model (mean |SHAP| = 0.0285). This is followed by feeling relaxed after visits (0.0211), suggesting that post-visit relaxation contributes substantially to satisfaction predictions within the modelling framework. Indicators of routine engagement showed notable contributions, suggesting that habitual interaction was associated with higher predicted satisfaction levels within the model. The second tier includes main activity, walking path condition, visual attractiveness, mood improvement, and physical activity, whereas accessibility indicators such as travel time and visit frequency are of moderate importance. Indicators of cognitive restoration, safety, and social context exhibited a lower influence, suggesting that predicted satisfaction is more closely associated with comfort, relaxation, and routine usability within this modelling framework.

Figure 5b (SHAP summary plot) further reveals the direction and variability of these effects, providing insight into the underlying restorative mechanisms. High values of seating adequacy and post-visit relaxation shift predictions toward higher satisfaction, whereas poor conditions exert negative effects, indicating asymmetric responses. Activity and environmental quality indicators showed greater dispersion, reflecting heterogeneity in user responses to activity type and infrastructure quality. Interpreted through restorative theory, the SHAP results aligned more strongly with SRT than with ART. The most influential indicators, such as seating adequacy, relaxation, routine use, and sitting duration, correspond to the mechanisms of physiological de-arousal and affective calming central to the SRT. ART-related cognitive outcomes, such as feeling mentally refreshed, exhibited weaker effects, suggesting that cognitive restoration operates as a secondary process. The SHAP analysis suggests that predicted satisfaction in dense contexts, such as Delhi, appears more closely associated with SRT-type stress relief mechanisms within the modelling framework.

4.10. Model Performance and Validation

To ensure transparent evaluation and reproducibility, the Random Forest model was assessed using cross-validation, an independent test set, and diagnostic analyses (Figure 6). Five-fold cross-validation of the training dataset yielded a mean score of 0.959 (SD = 0.024), indicating stable internal performance. Predictive accuracy was primarily evaluated using the independent 30% test dataset, which produced an R² of 0.660 (Figure 6b), indicating that 66% of the variance in the overall satisfaction was explained. The actual-versus-predicted plot showed a clear positive association with moderate dispersion at higher satisfaction levels.

The learning curve (Figure 6a) shows convergence between the training and validation mean squared errors (MSEs), suggesting limited overfitting in the observed dataset. Residual diagnostics showed errors centered around zero without a pronounced funnel-shaped pattern (Figure 6c). Minor vertical clustering reflects the discrete five-point ordinal structure of outcome variables. The residual density distribution is approximately symmetric, indicating the absence of systematic bias (Figure 6d). Although estimated as a regression model, the predictions were rounded to the nearest Likert category to assess ordinal consistency. The confusion matrix showed a strong diagonal dominance, with an accuracy of 0.887 and a weighted F1-score of 0.890. The misclassifications were largely confined to adjacent categories. The combined evidence from cross-validation, independent testing, and diagnostic evaluation indicated stable internal predictive performance within the sampled user population.

5. Discussion

This study advances the understanding of how UBGS may function as socio-environmental infrastructure in a high-density Global South megacity by examining the interplay between access, use patterns, perceived environmental quality, restorative experiences, and satisfaction among users of Delhi parks. By integrating access and use patterns with perceived environmental quality, restorative experiences, and satisfaction outcomes, the findings move beyond biophysical evaluations to demonstrate the social and experiential functioning of urban BGI. The results indicate that UBGS operates as urban infrastructure, delivering restorative benefits shaped by modes of use and embodied experiences, underscoring its relevance for urban planning and public health in rapidly urbanising contexts.

5.1. Key Findings and Their Significance

These patterns position UBGS in Delhi as neighbourhood-scale infrastructure rather than recreational destinations. The high frequency of use, primarily via walking and public transport, shows that UBGS are embedded in daily urban life as accessible spaces for stress relief. This integration highlights the importance of proximity and pedestrian accessibility in sustaining engagement and enabling well-being. The contrast between positive ratings for visual appeal and lower evaluations of noise and traffic attenuation reveals a gap between aesthetic provision and restorative performance. Although greenery enhances visual appeal, it cannot alone counteract urban stressors in congested areas. These findings show that restorative quality depends on design elements that mitigate sensory stress, particularly acoustic discomfort, rather than visual greening alone.

The respondents reported emotional calmness, mental refreshment, and physical recovery, reinforcing the UBGS as a health-supportive environment. However, variations in satisfaction across activities show that restorative benefits depend on how spaces are used. Passive, nature-oriented activities were associated with higher satisfaction, suggesting that opportunities for quiet observation and contemplation are critical alongside active uses. Multivariate analysis showed that post-visit relaxation was the main predictor of overall satisfaction, whereas mood refreshment had no independent effect when combined. This indicates that somatic restoration and relief from physical tension play a more decisive role in shaping user evaluations than cognitive restoration under urban stress. Gender and age differences in activity patterns and visitation reveal varied needs across population groups, reinforcing the need for an inclusive design that accommodates diverse life-course contexts.

5.2. Usage Patterns as Indicators of Everyday Urban Infrastructure

The high visitation frequency observed among our respondents, with over 40% visiting three or more times weekly, reinforces their role as an everyday urban infrastructure. This pattern aligns with research highlighting the importance of accessible UBGS in Global South cities, where private outdoor space is limited [58,59]. However, confirmatory studies with representative samples are needed to establish population-level patterns. Such frequent engagement indicates the regular delivery of urban ecosystem services, especially cultural services (stress reduction, social interaction, and psychological restoration) and regulating services (thermal comfort and air purification), which accrue through repeated exposure. In Delhi, this pattern is supported by walking and public transport as primary access modes, combined with short travel times for most users (67% within 20 min), indicating these spaces operate within neighbourhood-scale catchments [60]. This pattern suggests that proximity and pedestrian accessibility may be important factors in sustaining frequent use [61] and enabling the everyday realisation of ecosystem service benefits for urban health and well-being among active users [62].

This high-frequency use coexists with a significant paradox regarding the accessibility of ecosystem services [63]. Despite the short distances, many respondents reported difficulty accessing these spaces, highlighting a divergence between potential ecosystem service supply and experienced access. This gap points to qualitative barriers such as safety concerns, inadequate pedestrian connectivity, and poorly designed entrances that constrain people’s ability to benefit from ecosystem services, even when green spaces are nearby [36]. The presence of frequent and infrequent users reveals persistent heterogeneity in ecosystem service uptake, suggesting that physical availability alone cannot guarantee equitable benefits. Social, perceptual, and experiential factors play a decisive role in determining whether BGS integrates into daily life and functions as an adequate ecosystem service infrastructure [37]. These findings position usage patterns as a lens for understanding BGS not merely as recreational amenities [17], but as urban ecosystem service systems whose effectiveness depends on spatial provision, accessibility, quality, and lived experience.

5.3. Accessibility: The Proximity-Perception Gap in Urban Blue-Green Space Access

Our research indicates a significant disparity between the actual proximity of UBGS and users’ perceived experience of accessing them. Although the majority of respondents indicated short travel times, with 67% indicating times under 20 min, a significant percentage (40.1%) concurrently regarded access as challenging or very challenging. This discrepancy highlights a fundamental shortcoming in urban green infrastructure development: although spatial availability is crucial, it does not necessarily ensure equal access to ecosystem services [23]. The findings reinforce emerging scholarship that distinguishes between metric proximity and lived accessibility, where the latter determines whether the potential ecosystem services supplied by UBGS, particularly cultural and regulating services linked to health and well-being, are realised in everyday urban life [36,64].

Perceived accessibility is associated with journey quality and experiential conditions, which influence how ecosystem service supply relates to actual use and perceived benefits of the ecosystem. Deficiencies in pedestrian infrastructure, such as poor sidewalks, missing crossings, or a lack of shade and rest areas, can make short distances burdensome, discouraging use, and limiting ecosystem service benefits [65,66]. Safety perceptions, often more influential than objective safety, constrain access, particularly for women, older adults, and caregivers, and are shaped by environmental cues such as lighting, visibility, and vegetation maintenance [38,67]. Micro-scale design factors [68], such as uninviting entrances, poor visual connectivity, and weak integration with pedestrian networks, reduce urban space permeability, making nearby green spaces feel inaccessible [69,70]. These dynamics show that cities with extensive green spaces may still reproduce inequities in access to ecosystem services, highlighting the need for planning frameworks that move beyond distance-based metrics to prioritise functional and perceptual accessibility, ensuring that UBGS operates as inclusive urban infrastructure [71].

5.4. Environmental Quality: The Gap Between Aesthetic and Experiential Quality

Moderate scores across environmental quality indicators reflect a persistent challenge in UBGS design within densely populated cities such as Delhi: the mismatch between aesthetic ecosystem service provision and functional restorative performance [72]. The high ratings for visual attractiveness align with the literature, highlighting the role of greenery and landscape design in enhancing aesthetic appreciation and scenic quality, corresponding to cultural ecosystem services for visual enjoyment and place satisfaction [73]. These perceptions are reinforced by the regulating ecosystem services of vegetation, including microclimate regulation through shading and evapotranspiration, air purification, and humidity moderation, which contribute to improved thermal comfort and environmental pleasantness [44]. Evidence indicating reductions in surface and air temperatures of up to 5 °C underscores the importance of urban vegetation as a climate-regulating asset, supporting both physical comfort and perceived environmental quality in heat-stressed urban settings [44,74]. However, while these ecosystem services enhance sensory appeal, they do not automatically translate into more profound restorative benefits for humans.

This limitation is evident in the low scores for escape from noise and traffic, revealing a deficit in regulating and cultural ecosystem services for acoustic comfort and psychological restoration [75]. Acoustic regulation is central to the “being away” dimension of Attention Restoration Theory, enabling cognitive disengagement from urban stressors and supporting recovery [8,76]. Although vegetation can reduce perceived noise, evidence suggests it is often insufficient in high-traffic environments, with effectiveness varying by noise sensitivity and attitudes toward greenery [43]. Moderate ratings for seating availability, shading adequacy, and pathway conditions indicate infrastructural shortcomings that constrain the benefits of ecosystem services. Inadequate seating and shade can limit visit duration, disproportionately affecting older adults and caregivers who need resting opportunities, while poor pathways undermine accessibility across mobility levels. These findings suggest that while the UBGS in Delhi delivers visible and thermal ecosystem services, its restorative potential remains constrained by noise exposure and amenity deficits.

5.5. Restorative Experiences Are the Core Mechanism of Benefit

A central contribution of this study is to provide exploratory evidence for the multidimensional restorative experiences associated with UBGS’s restoration, based on the experiences of sampled park users. Respondents reported benefits across emotional, cognitive, and physical domains, including enhanced calmness, mental refreshment [77], mood improvement, and recovery from tiredness. Intercorrelations among these outcomes are consistent with the view that restoration may involve a synergistic psycho-physiological process rather than isolated effects, reinforcing the understanding of cultural ecosystem services as holistic benefits from human-nature interactions [11]. Within the ecosystem services cascade, UBGS provide biophysical structures (vegetation and water bodies) that enable regulating functions (microclimate moderation and sensory buffering), translating into experiential benefits such as stress relief, cognitive clarity, and bodily recuperation. These findings are consistent with the conceptualisation of restoration as a key cultural ecosystem service bridging environmental quality and public health outcomes, though representative studies would be needed to establish this more broadly.

Recovery from tiredness and post-visit relaxation emerged as pivotal dimensions of this restorative response. The strong association between these factors (ρ = 0.243) suggests that somatic relaxation may function as a bridging mechanism through which regulating ecosystem services support emotional and cognitive renewal. This interpretation aligns with SRT, emphasising physiological de-arousal as a pathway through which natural environments facilitate recovery [9], while ART highlights cognitive respite enabled by effortless attention and “being away” experiences [8,78]. The multisensory aspects of UBGS, including visual exposure to greenery, birdsong, water sounds, and tactile engagement, activate these pathways, enhance perceived restoration, and encourage health-promoting behaviours like walking and visitation [11,79]. Evidence shows that environments with green and blue elements provide stronger restorative ecosystem services, particularly for older adults and vulnerable groups. Studies have confirmed reduced attentional fatigue and negative affect after blue-green exposure [80]. These findings position UBGS as a critical mental health infrastructure, supporting psychological resilience and highlighting the need for urban planning that prioritises multisensory, accessible, and restorative environments for public health and sustainability [40].

5.6. Activity Type as a Behavioural Moderator of Satisfaction

Our analysis indicates that the type of activity undertaken in UBGS is a more influential determinant of user satisfaction than the social context of visitation. The significant variation in satisfaction across primary activities (Kruskal-Wallis p < 0.001), contrasted with the non-significant effect of visiting alone or with others, suggests that, among these respondents, what people do may have mattered more than who they visit with. Activities that facilitate bodily ease and sensory engagement, such as sitting, observing greenery, and quiet contemplation, are associated with higher satisfaction. From an ecosystem services perspective, these activities enable cultural ecosystem services, particularly psychological restoration and emotional regulation through sustained interaction with natural elements. This aligns with “soft fascination,” whereby effortless attention to nature supports recovery from cognitive fatigue and urban stress [81].

The preference for passive restorative activities over complex social interactions suggests that users primarily value BGS for their individual-level restorative ecosystem services [82]. These findings support design strategies like activity zoning, which separates quiet, contemplative areas from active or social zones to minimise functional conflict and enhance restorative quality [41]. Activity-based satisfaction varies across socio-demographic groups, reflecting different capacities and restorative needs over the life course. This underscores the need for urban forestry and park design to accommodate diverse behavioural rhythms and preferences, ensuring equitable access to restorative benefits. Designing with this activity-sensitive approach is central to developing multifunctional green infrastructure that maximises cultural ecosystem services and supports inclusive, health-promoting urban environments [13,41].

5.7. Relaxation as the Dominant Predictor of Satisfaction

Ordinal regression analysis showed that post-visit relaxation was a significant predictor of overall satisfaction, whereas mental refreshment lacked explanatory power when both dimensions were considered. This suggests that somatic and affective restoration may play a more immediate role in shaping users’ global evaluations of BGS experiences in dense urban contexts [48]. This suggests that the cultural ecosystem services most important to users are linked to physiological and emotional relief, such as stress reduction, fatigue recovery, and bodily calmness, rather than abstract cognitive benefits. These findings refine restoration frameworks, which often prioritise attentional recovery as the primary outcome of nature exposure [31], by showing that in routine, high-stress urban environments, functions supporting physical relaxation directly translate into perceived satisfaction.

A potential concern relates to the conceptual proximity between post-visit relaxation and overall satisfaction. While both reflect positive experiential responses, they were operationalised as analytically distinct constructs: relaxation captures a specific affective-restorative state, whereas satisfaction represents a broader evaluative judgment of the visit. Some degree of conceptual overlap cannot be entirely excluded, as affective states may inform global evaluations. However, the modest explanatory power of the model (Nagelkerke R² = 0.021) indicates that relaxation accounts for only a limited proportion of the variance in satisfaction, suggesting a partial rather than redundant association. The non-significant effect of mental refreshment further supports the interpretation that not all restorative dimensions are embedded in satisfaction judgments.

These findings suggest that urban forestry strategies should prioritise ecosystem service pathways that facilitate physical and emotional relaxation, strengthening user satisfaction [83]. Design elements that enhance thermal comfort, acoustic buffering, and ease of use, such as shade, noise attenuation, comfortable seating, and opportunities for passive use, are practical in dense megacity contexts [84]. While cognitive restoration remains essential, the findings suggest that BGS that supports embodied calmness and recovery [85] will deliver more perceptible benefits, reinforcing public support for urban green infrastructure and enhancing its role as mental health infrastructure.

5.8. Socio-Demographic Patterning and Equity Implications

The socio-demographic variations observed among our respondents suggest that access to UBGS is associated with physical proximity, social identity, life-course stage, and perceived safety, factors that may influence the distribution of ecosystem service benefits. Gendered activity patterns, in which women engage more in observational and caregiving activities while men dominate vigorous pursuits such as jogging, reflect social norms and differentiated perceptions of public spaces [86]. These dynamics align with research showing that women’s park engagement is constrained by safety concerns and caregiving duties, limiting their participation in restorative activities [87] and access to cultural ecosystem services, such as psychological restoration [34,88]. Enhancing perceived safety through improved lighting, clear sightlines, and visible park presence is therefore a critical equity intervention that enables a more inclusive realisation of ecosystem service benefits [33].

Age-related differences in structural patterns of use and ecosystem service uptake. Older adults are frequent users, relying on UBGS for gentle physical activity, social interaction, and daily routines, functions central to healthy ageing and well-being. Their sustained engagement depends on supportive infrastructure, including comfortable seating, accessible pathways, shade, and age-appropriate exercise facilities [89,90]. The absence of such features limits older adults’ ability to access regulating and cultural ecosystem services, transforming public spaces into sites of exclusion [91]. These findings underscore the need for an equity-centred, ecosystem-service-oriented approach to green space design that goes beyond one-size-fits-all solutions. Inclusive [92], multifunctional spaces are designed through participatory processes that engage women, older adults, and people with disabilities, which are essential for ensuring that the restorative, health-promoting, and social benefits of UBGS are equitably distributed, thereby strengthening social cohesion and urban resilience [34,35,91].

The educational composition of the sample is skewed toward highly educated respondents relative to the broader Delhi population. Educational attainment may influence environmental perceptions by increasing awareness of ecosystem services, biodiversity, and the health benefits of nature exposure, potentially heightening sensitivity to positive attributes (e.g., visual attractiveness) and perceived deficiencies (e.g., seating adequacy, noise intrusion). Higher education may shape response styles and recreational preferences, including an inclination toward contemplative and wellness activities, which may contribute to the prominence of passive restorative uses observed. Moreover, expectations regarding infrastructure quality may be elevated among educated users, influencing their evaluations of amenities such as seating, pathways, and maintenance. Consequently, certain findings, particularly the emphasis on seating adequacy and embodied relaxation, may reflect both the experiential realities and evaluative standards of this educationally skewed subsample.

A key caveat concerns sampling, as respondents were recruited during their visits to the five study parks. The findings reflect the experiences of current UBGS users and do not capture the perspectives of non-users or those facing access barriers. Residents who avoid or cannot visit parks, including those with mobility limitations, safety concerns, caregiving constraints, limited resources, or feelings of exclusion, were not included. As respondents have navigated existing barriers, their positive accessibility evaluations may reflect selection effects rather than broader conditions. Although many participants reported access challenges, this likely underrepresents individuals facing severe constraints who never appear in the park samples. This limitation affects equity analysis because populations most underserved by the UBGS may remain underrepresented. The study did not collect geocoded residential location data, neighbourhood-level deprivation indices, or objective distance-to-park measures. Consequently, spatial accessibility gradients and distributive socio-spatial inequities across Delhi could not be assessed. Accessibility characterisation is limited to self-reported travel time, mode of transport, and perceived ease of access among active users. Consequently, equity-related interpretations should be understood as exploratory and bounded by the study design, rather than indicative of city-wide distributional justice in UBGS provision.

The findings of this study are based on an opportunistic sample of 411 users across five UBGS in Delhi and do not permit statistical generalisation to all Delhi park users or megacities of the Global South. The observed patterns, including embodied relaxation, seating adequacy, and the proximity-perception gap in accessibility, should be interpreted as exploratory evidence in this specific sample and context. This study contributes to the literature by identifying experiential mechanisms and theoretical propositions regarding how density and environmental stress may shape restorative processes and satisfaction within UBGS. While the results remain sample-specific, the conceptual insights offer analytical transferability, suggesting hypotheses for further testing through representative, multi-site, and comparative research across diverse urban settings.

5.9. Implications for Urban Planning, Policy, and Blue-Green Infrastructure

The findings underscore the need for strategic reorientation of urban forestry and green infrastructure policy in rapidly urbanising cities of the Global South. Planning approaches that rely on quantitative indicators, such as green space area or canopy cover, are insufficient to address the realities of dense, stress-prone urban environments. Instead, policies could benefit from adopting an ecosystem service-based framework that recognises BGS as essential public health, climate adaptation, and social equity infrastructure [93,94]. In megacities such as Delhi, the potential of UBGS may lie partly in its capacity to deliver regulating and cultural ecosystem services, particularly thermal comfort, stress reduction, fatigue recovery, and psychological calmness. Aligning urban greening strategies with global agendas for healthy, inclusive, and resilient cities requires prioritising functional and experiential performance rather than spatial provision [95].

Operationalising this policy requires evidence-based design and governance standards. Urban greening interventions should enhance thermal comfort through mature tree canopies and microclimate-sensitive design [96]. In contrast, acoustic regulation requires layered vegetation, landform design, and water features to mitigate traffic noise and support restorative experiences [11,97]. Improving equitable access through safe, shaded pedestrian networks linking homes, transit, and BGS may be particularly important in areas where environmental and health burdens converge [98,99]. At the site scale, micro-design elements, including biodiverse planting, seating, and maintained water features, may support restorative ecosystem services and sustained use [100,101]. Governance improvements may also be valuable; participatory co-design and stewardship improve relevance and effectiveness [102], while monitoring frameworks should integrate perception-based indicators of safety, comfort, and restoration with biophysical metrics for adaptive management [103,104]. Framing BGS as core components of policies such as the 15-min city can enable coordination among environment, transport, and health agencies, ensuring that ecosystem service delivery contributes to public well-being and sustainable urban development [95].

5.10. Limitations and Future Research Directions

This study has some limitations that should be considered when interpreting the findings. First, the cross-sectional design limits causal inference regarding the relationships among UBGS use, restorative experience, and user satisfaction. The October-November data collection presents a seasonal limitation, as pleasant post-monsoon conditions may have elevated satisfaction levels and shaped activity patterns differently than during the extreme summer or monsoon seasons. Although core predictor relationships may remain relatively stable, the magnitude of the effects and the relative importance of environmental attributes (e.g., shade versus sun exposure) may vary seasonally. Therefore, multi-season or longitudinal studies are needed to confirm the temporal stability of these relationships and identify design adaptations that support year-round restorative benefits. Although SHAP-based analysis enhances the interpretability of key predictors, it does not establish a temporal causality.

Second, the study relied on self-reported measures of environmental quality, restorative experience, and satisfaction from a single survey instrument. This raises the possibility of common method variance (CMV), recall bias, and social desirability bias for psychological constructs in the SRT and ART frameworks. Although perception-based measures suit experience-centred research, integrating objective environmental monitoring, behavioural observations, and multi-source data would reduce shared-method bias and strengthen causal inference. The analysis does not include spatial or ecological attributes of parks, such as vegetation structure, biodiversity, or microclimatic variation, which may moderate restorative responses. The exploratory factor analysis yielded weak results: the KMO value (0.523) was marginally acceptable, and the two-factor solution explained only 28.6% of the total variance. These values reflect the heterogeneous item set spanning environmental attributes and restorative outcomes, and confirm that the EFA should be interpreted as an exploratory diagnostic rather than robust structural validation. This limitation supports the decision to treat all items individually rather than as composite scales.

Third, the sampling strategy limits generalisability. Participants were recruited during park visits, thereby excluding non-users, infrequent visitors, and individuals facing access barriers, consolidating the earlier repeated discussion of selection effects and unmet demand. This may introduce selection bias and relatively favourable evaluations of accessibility and satisfaction. The sample also overrepresents highly educated respondents, who may differ systematically in environmental awareness and park-use behaviour from underrepresented socio-economic groups. Consequently, the findings cannot be assumed to represent the broader population of Delhi, particularly the underserved communities central to equity-oriented BGI planning. Finally, the single-city focus provides contextual depth but limits applicability to cities with differing climatic, socio-cultural, and infrastructural conditions.

A limitation of this study is its reliance on perceived environmental quality indicators (noise buffering, thermal comfort, and seating adequacy) rather than objective measurements of temperature or noise levels. As predictors and satisfaction outcomes were self-reported in the same survey, common method variance cannot be excluded. While perception-based measures suit experiential satisfaction and restorative appraisal, integrating objective environmental monitoring would strengthen causal interpretation and differentiate subjective evaluation from physical exposure. Additionally, recruiting participants within parks during visits limits the ability to find current UBGS users. Non-users, infrequent visitors, and those facing access barriers were not represented, limiting insights into unmet demand and constraints. Consequently, positive evaluations of accessibility and satisfaction may reflect the selection effects. This limitation is crucial for equity assessments, as populations underserved by the UBGS may be under-represented.

Future research should adopt longitudinal or experimental designs to establish causal pathways between UBGS exposure and seasonal well-being outcomes. Integrating physiological indicators, environmental measurements, and behavioural observations with experiential survey data would strengthen the validation of restorative mechanisms. Comparative multi-city studies across the Global South that combine experiential assessments with GIS-based metrics are needed to identify the contextual moderators of restoration. More inclusive sampling strategies, such as household-based, stratified, or community-targeted recruitment, would help capture the perspectives of non-users and underserved populations. Mixed-methods approaches, including interviews and go-along methods, could enrich the understanding of diverse experiential contexts. Incorporating physical assessments of park quality and testing effect modification across socio-demographic groups would enhance robustness and support equitable, evidence-informed BGI planning.

5.11. Comparison with Existing Literature

The findings from this sample are broadly consistent with global evidence on the salutogenic role of UBGS, and align with the core propositions of ART and SRT [105]. Improvements in calmness, mental refreshment, and fatigue recovery reported by respondents are consistent with North American and European studies documenting stress reduction, mood enhancement, and cognitive renewal from exposure to urban nature [5]. The high-frequency, neighbourhood-scale use observed among respondents is consistent with research suggesting UBGS may serve as embedded daily infrastructure for well-being in dense cities with limited private outdoor space [58]. However, population-level confirmation would require representative sampling. Beyond contributing to the established evidence base, this study advances the literature by foregrounding the embodied dimensions of restoration and providing rare exploratory evidence from a high-density Global South megacity. In this sample, post-visit relaxation and activity type emerged as stronger satisfaction factors than visual preference or green cover, suggesting that somatic relief may outweigh cognitive restoration in stress-intensive environments, though this warrants testing in larger representative samples [83,106]. The observed dissonance between visual attractiveness and limited noise buffering suggests the potential need to adapt design principles from lower-density settings to contexts characterised by extreme density and environmental stress [107]. These findings contribute to a more geographically inclusive exploratory evidence base for planning urban BGI in rapidly urbanising cities [108].

6. Conclusions

This study provides exploratory evidence that UBGS serves as an essential everyday infrastructure for psychological restoration among users of five parks in Delhi, a high-density Global South megacity. The findings from this sample suggest that spaces of moderate environmental quality can deliver consistent restorative benefits, such as enhanced calmness, mood refreshment, and recovery from fatigue, through routine, neighbourhood-scale engagement. Among the respondents, satisfaction appeared more closely associated with experiential comfort and perceived accessibility than with visit frequency or visual aesthetics alone. Advanced analytical approaches, including ordinal regression and SHAP-based interpretation, revealed that embodied experiences, particularly post-visit relaxation, seating adequacy, thermal comfort, and ease of use, appeared to exert a stronger influence on satisfaction than cognitive restoration or scenic appeal in this sample, suggesting the salience of somatic relief in high-stress urban contexts. These findings point to the potential value of a strategic reorientation of urban greening policy away from quantitative provision targets toward experience-responsive and equity-oriented design that prioritises multisensory comfort, accessibility, and everyday usability, with the aim of better supporting public health, resilience, and well-being in rapidly urbanising cities.