Urban Green Space and Mental Health: Mechanisms, Methodological Advances, and Governance Pathways for Sustainable Cities

Abstract

Urban green space (UGS) is a critical component of sustainable cities and a modifiable determinant of mental health (MH). This review synthesizes 93 empirical studies and 929 bibliometric records to map theoretical advances, methodological evolution, and governance implications in the UGS–MH field. We integrate the following six validated pathways into a unified socio-ecological framework: attention restoration, stress recovery, behavioral activation, physiological regulation, social cohesion, and environmental buffering. Methodological trends indicate a shift from static greenness proxies to street-view and multimodal exposure measures, and from cross-sectional correlations to models that address spatial heterogeneity, causal identification, and AI-enabled prediction. Bibliometric mapping reveals increasing interdisciplinarity, geographic diversification, and growing attention to dynamic exposure science. Persistent challenges include spatial and temporal misalignment between exposure and outcome measures, reliance on single-modality indicators, limited causal inference, and constrained cross-cultural generalizability. Building on these findings, we propose a governance-oriented framework to support sustainable and healthy cities through equitable green access, behavior-informed planning, nature-based interventions, and data-driven decision support. Overall, this review strengthens the bridge from evidence to action at the interface of urban sustainability and population mental health.

1. Introduction

As global urbanization accelerates, mental health has emerged as a major public health concern with substantial global disease burden. The World Health Organization (WHO) estimates that roughly one billion people worldwide are affected by mental disorders, with depression and anxiety consistently ranking among the leading contributors to disease burden, which threatens individual well-being and sustainable development [1,2,3,4]. Against the backdrop of high-density built environments, fast-paced lifestyles and rising risks of social isolation, the use of urban nature to strengthen psychological resilience and mental well-being has emerged as a core interdisciplinary agenda [5,6]. Within urban ecosystems, urban green space (UGS), given its potential to alleviate stress, foster social interaction, and improve subjective well-being, has been increasingly regarded as a form of “infrastructure for mental health intervention” [7,8]. However, despite a growing evidence base, mechanistic explanations remain fragmented and research coverage uneven, underscoring the need for a more integrated synthesis [9].

Since Olmsted’s nineteenth-century notion of the “lungs of the city,” the role of UGS has expanded beyond esthetics and recreation to encompass health promotion, ecological regulation, and social equity [10,11,12]. Recent empirical studies consistently document positive associations between green-space exposure and mental health outcomes, particularly reductions in depression and anxiety and improvements in subjective well-being [13]. To explain these links, multidisciplinary models have been advanced, including Attention Restoration Theory (ART), Stress Reduction Theory (SRT), social capital perspectives, and behavioral-activation pathways [14]. Together, these frameworks suggest that natural environments may influence mental health through intertwined cognitive, affective, physiological, and social pathways, but these pathways have not yet been synthesized within a sufficiently coherent explanatory framework.

Concurrently, advances in remote sensing, spatial analytics, and artificial intelligence (AI) have shifted UGS exposure assessment from traditional static averages to high-resolution, multidimensional, individual-level dynamic modeling [15]. Researchers increasingly integrate multi-source indicators of greenness, accessibility, and perceived environmental quality, and in some cases combine mobile sensing with Ecological Momentary Assessment (EMA) to capture more temporally aligned links between environmental exposure and psychological states [16]. This methodological evolution helps mitigate exposure misclassification arising from the Uncertain Geographic Context Problem (UGCoP) and advances the field toward causal identification and theory-informed mechanism building. However, limited harmonization, incomplete causal identification, and weak cross-regional comparability continue to constrain external validity and policy translation [17].

Despite accumulating findings, three key gaps remain. First, mechanistic accounts remain insufficiently integrated, and a coherent synthesis of major pathways is still lacking [18]. Second, standards for spatial modeling remain heterogeneous, and dynamic exposure assessment has not yet been sufficiently integrated with subgroup heterogeneity and moderation analysis [19]. Third, the literature remains concentrated in high-income settings, with limited attention to pathways relevant to the Global South and vulnerable populations, thereby restricting generalizability and equity in policy applications [12,20].

To address these gaps, this review conducts a global synthesis of the UGS–MH literature from January 2013 to August 2025 using a dual-dataset design that combines bibliometric mapping (N = 929) with qualitative synthesis of included empirical studies (n = 93). Rather than treating mechanisms, methodological advances, and governance implications as separate topics, we organize the review around an integrative analytical chain structured by the following three linked questions: through what pathways UGS influences mental health, how these pathways have been measured and identified across evolving methodological traditions, and how the resulting evidence can inform more equitable and context-sensitive planning and governance.

Recent reviews have substantially advanced the field by examining specific dimensions of the UGS–MH relationship, including mechanisms, green-space quality, health equity, and methodological challenges [7,9,12,17]. However, these strands of work are less often integrated into a single framework that explicitly connects mechanism, method, and governance. The present review extends prior work in four ways. First, it links knowledge-structure analysis to evidence interpretation through a dual-dataset design. Second, it integrates six mechanistic pathways with methodological evolution and governance translation rather than discussing them in parallel. Third, it places particular emphasis on dynamic exposure science, mobility-aware assessment, causal identification, and AI-enabled analytical transitions. Fourth, it frames the UGS–MH field not only as an environmental-health topic, but also as an equity and planning challenge for sustainable cities.

2. Theoretical Framework and Mechanistic Pathways

2.1. Cognitive and Emotional Mechanisms

UGS may influence mental health through six recurrent pathways. These pathways should not be interpreted as fixed and independent routes; rather, they form an interrelated socio-ecological mechanism system. At the intrapersonal level, attention restoration [21], stress reduction [22], and physiological regulation [23] capture how exposure to green environments may facilitate cognitive recovery, affective stabilization, and autonomic–immune–endocrine adjustment. At the behavioral and interpersonal level, behavioral activation [24] and social interaction [25] help translate environmental opportunities into physical activity, routine use, social support, and neighborhood cohesion. At the environmental-contextual level, buffering processes—including reductions in air pollution, noise, and heat—shape the background conditions under which mental-health benefits may be amplified or constrained. Together, these pathways draw on multiple theoretical traditions, including ART, SRT, social-capital perspectives, the biophilia hypothesis, and the old friends hypothesis [26,27].

A large body of evidence links green-space exposure to improved attention, relief of depressive symptoms, and higher subjective well-being [28,29]. Observational and mechanism-oriented studies further suggest that the mental health benefits of UGS are often mediated by perceived environmental quality and restorative psychological processes, rather than by exposure quantity alone, with reduced rumination representing one plausible pathway [30,31]. Neuroimaging studies provide convergent and objective evidence: exposure to natural environments is associated with increased activation in the prefrontal cortex and anterior cingulate cortex [32,33], alongside reduced amygdala reactivity [34], thereby elucidating the neural underpinnings of ART and SRT, respectively. Longitudinal or repeated-exposure studies further suggest gains in the efficiency of emotion-regulation networks, implying experience-dependent plasticity [35]. Together, these findings strengthen the multi-level “environment-brain-mind” chain, extending ART/SRT explanations from psychometric outcomes to neural mechanisms [34,36]. That said, most neural phenotypes have been observed under short-term or repeated exposures; extrapolation to long-term structural change and clinical endpoints warrants caution and requires validation via longitudinal and quasi-experimental designs.

Taken together, these findings support the view that UGS functions as a social and environmental determinant of mental health through coordinated cognitive, affective, physiological, and social processes. The corresponding structural, perceptual, and functional attributes of UGS, along with their mapping to the six pathways and subgroup differences, are summarized in Figure 1.

2.2. Physiological Mechanisms

In addition to the cognitive-affective phenotypes discussed in Section 2.1, UGS may also influence MH through physiological pathways involving the autonomic–endocrine–immune axes [37]. A large body of evidence shows that nature exposure reduces stress-related biomarkers—such as salivary cortisol, heart rate, and blood pressure—and improves autonomic function indexed by heart rate variability (HRV) [38]. In addition, green-space exposure is associated with lower levels of inflammatory markers, including C-reactive protein (CRP) and interleukin-6 (IL-6), suggesting mitigation of chronic inflammation and psychological stress via immune–endocrine routes [39,40].

This inflammation-immune pathway has attracted increasing attention in recent years. Individuals with regular or more frequent contact with green-space show lower CRP and IL-6, higher HRV, and stronger indices of immunoregulation, suggesting that nature exposure can downregulate low-grade chronic inflammation via autonomic-immune interactions [41]. This pathway also provides a biological explanation for stress relief, lower depression risk, and fewer cardiovascular and metabolic comorbidities [42]. Evidence from acute single-exposure studies remains mixed, indicating that dose–response patterns and minimum effective exposure thresholds remain insufficiently resolved [43,44].

Physiological mechanisms therefore provide a biological basis for psychological restoration and emotion regulation [36]. Empirical work indicates that green-space not only relieves short-term stress but also reduces long-term risks of cardiovascular and metabolic disease, thereby indirectly promoting mental health [24]. Physiology therefore represents an important mediating link between external environmental exposure and internal psychological experience, forming one of the major evidentiary chains through which UGS may benefit mental health [45].

2.3. Social Pathways, Heterogeneity, and Contextual Moderation

Not all populations benefit equally from UGS [46]. Inter-individual heterogeneity in mental health gains is now a critical variable in the literature. Evidence indicates that sex, age, Socioeconomic Status (SES), physical health, and psychological vulnerability are each important moderators of the UGS–MH association. Emerging evidence suggests that heterogeneity concerns not only effect size, but also the dominant pathways involved, with restorative, stress-buffering, social-cohesive, and activity-related mechanisms varying across population groups and neighborhood contexts [30,47]. For example, women, who often exhibit greater sensitivity to affective cues and more frequent green-space use, tend to show stronger mood-alleviation effects; children and older adults, who depend more on outdoor activities, more readily obtain behavioral activation and social-interaction benefits from green-space [48]. Mechanism-focused analyses in older adults further support multi-pathway mediation, indicating that the mental-health benefits of UGS can operate through reduced relative deprivation, increased physical activity, and enhanced social trust [49]. Although low-SES groups may reside in green-poor neighborhoods, equitable access can yield larger marginal improvements in mental health. Likewise, psychologically vulnerable individuals (e.g., those with anxiety or elevated depression risk) appear more responsive to the restorative stimuli of nature, exhibiting faster stress relief and attention restoration. This “high-susceptibility-high-benefit” pathway highlights the targeted value of green-space for urban mental health interventions and underscores the need for policies that prioritize spatial equity and protection of socially vulnerable groups [50].

Heterogeneity is shaped not only by user characteristics but also by green-space type [51]. Forested environments are more often associated with deeper restorative experiences and cognitive recovery, whereas pocket parks or neighborhood green-spaces more commonly facilitate day-to-day social interaction and stress buffering [52,53]. In short, population-by-environment-type interactions jointly shape the mental-health effects of UGS, implying that planning and intervention should be differentiated by both population needs and green-space typology [54].

The integrated framework is illustrated in Figure 1, which summarizes the six mechanistic pathways, the structural-perceptual-functional attributes of UGS, and the sources of effect heterogeneity across users and contexts. Together, these components provide an analytical logic that moves from pathway identification to attribute configuration and differentiated effects across populations and settings.

These pathways should not be understood as isolated explanatory channels. In practice, restoration, stress buffering, physiological regulation, behavioral activation, and social interaction often operate in overlapping and mutually reinforcing ways, while their relative salience varies across urban settings [55,56,57]. For example, nearby neighborhood parks may primarily support everyday stress buffering, routine physical activity, and informal social contact, whereas larger and more naturalistic green environments may more strongly facilitate attention restoration, affective recovery, and physiological downregulation [58,59]. The proposed framework should therefore be understood as an integrative analytical scaffold rather than a universally invariant causal template. Its operation is likely to be shaped by climatic conditions, urban form, accessibility, safety, governance capacity, and cultural meanings attached to green-space use [60,61]. Accordingly, the UGS–MH relationship is better interpreted as layered and context-sensitive than as universally uniform [47,55].

3. Methods

3.1. Research Design and Overall Framework

We adopted a mixed-methods design integrating systematic review and bibliometric analysis to characterize the global landscape, methodological development, and knowledge structure of research on UGS-MH. The overall workflow followed PRISMA 2020 reporting standards [62,63] and proceeded through systematic identification, screening, and structured synthesis to support transparency, reproducibility, and auditability [64].

To capture evidence at both macro and micro levels, we constructed the following two complementary datasets: (i) a bibliometric corpus (N = 929) used to profile spatiotemporal publication trends, international collaboration networks, and topic clustering/evolution; and (ii) a qualitative synthesis sample (n = 93) composed of eligible studies with sufficient empirical or methodological detail for synthesis of mechanisms, methodological characteristics, and policy implications.

This dual-dataset design allowed us to connect field-level knowledge development with study-level evidence synthesis. Bibliometric analysis was used to map publication trends, collaboration structures, and thematic evolution across the broader UGS–MH literature, whereas the systematic review component focused on mechanisms, exposure assessment, modeling strategies, and governance implications within the included empirical studies.

Protocol and registration. This review was not registered in PROSPERO or OSF, and no public protocol was published prior to study selection. However, the review question, database scope, eligibility criteria, and core coding items were specified before full-text screening and applied consistently throughout the review. To support transparency despite the absence of formal registration, we report the search strategy, screening logic, PRISMA-based selection process, and coding framework in the Methods and Supplementary Materials.

3.2. Data Sources and Search Strategy

We queried the Web of Science Core Collection (WoSCC) and PubMed for records published from January 2013 to August 2025. The last search was conducted on 11 September 2025. To balance coverage and precision across environmental and public-health literature, we used Topic searches with Boolean combinations of synonyms and related terms. Green-space terms included “urban green space”, “greenspace”, “park”, and “vegetation”; mental health terms included “mental health”, “depression”, “well-being”, and “psychological health”. An illustrative query was: (“urban green space” OR “greenspace” OR “park” OR “vegetation”) AND (“mental health” OR “depression” OR “well-being” OR “psychological health”). To balance disciplinary breadth, metadata consistency, and relevance to public health, we selected WoSCC and PubMed as the two primary databases. WoSCC provides standardized bibliographic fields suitable for bibliometric mapping, whereas PubMed strengthens coverage of environmental health, public health, and mental-health-related research. Nevertheless, the exclusion of other databases, particularly Scopus and CNKI, may have led to underrepresentation of some regional, non-English, or social-science-oriented studies, and this limitation should be considered when interpreting the findings.

Inclusion criteria limited records to English, peer-reviewed journal articles; we excluded conference abstracts, editorials, commentaries, and the gray literature. Following pilot tests, we iteratively refined the search strings to balance recall and precision. After format- and content-level deduplication across the two databases, we obtained 929 unique records for downstream analysis.

Because PubMed frequently lacks complete affiliation metadata, country/institution collaboration analyses were conducted on WoSCC records (n = 499) to ensure accuracy, whereas other analyses drew on both databases. Full database-specific search strings, filters, and limits are provided in the Supplementary Materials.

3.3. Study Selection and Eligibility Criteria

Following the PRISMA 2020 four-stage process—Identification, Screening, Eligibility, and Inclusion [65]—we compiled 929 unique records (PubMed = 430; WoSCC = 499). Deduplication was conducted using exact DOI matching and, when DOI was unavailable, exact matching of title, first author, and publication year. No duplicate records were retained in the final merged dataset.

During title/abstract screening, 814 records were excluded as out of scope. Full texts were sought and assessed for 115 articles, with none unavailable (reports not retrieved = 0). At eligibility assessment, 22 reports were excluded, including ineligible article types or non-original empirical studies (n = 17) and studies with ineligible focus, outcome, or exposure for this review (n = 5). Consequently, 93 studies met the inclusion criteria and entered the qualitative synthesis, whereas all 929 records were retained for bibliometric analyses. The study selection process is summarized in Figure 2.

3.4. Bibliometric Analysis and Tool Application

We conducted bibliometric mapping and visualization using VOSviewer 1.6.20 (Centre for Science and Technology Studies, Leiden University, Leiden, The Netherlands) to identify keyword co-occurrence networks, country/institution collaboration networks, and topic-evolution trajectories [66,67]. To ensure input consistency and reproducibility, we first standardized fields and performed preliminary tabulation in Excel 2021 (Microsoft Corporation, Redmond, WA, USA), then used Python 3.9 (Python Software Foundation, Wilmington, DE, USA) to compute keyword frequencies and convert formats to meet VOSviewer import requirements.

Thresholds were set as follows: minimum keyword occurrences ≥ 6, country network publication count ≥ 5, and institution network publication count ≥ 10. For keyword co-occurrence analysis, the minimum-occurrence threshold was set to six in order to balance network interpretability with retention of substantively meaningful topics. Lower thresholds introduced excessive sparsity and noise, whereas higher thresholds disproportionately emphasized only the most dominant themes. Before network construction, keywords were manually standardized through spelling harmonization, singular/plural normalization, and synonym merging where conceptually appropriate. Despite these steps, bibliometric mapping remains subject to several limitations, including dependence on database coverage, English-language indexing bias, variation in author-supplied keyword practices, and the underrepresentation of emerging but low-frequency themes in threshold-based networks. We therefore interpret bibliometric outputs as structured representations of dominant knowledge patterns rather than exhaustive maps of all research trajectories. All figures were exported as SVG and post-processed in Adobe Illustrator 2026 to standardize font sizes, color palette, line weights, and legend scales for journal-ready publication.

3.5. Evidence Synthesis and Analytic Framework

We applied a structured coding scheme to the 93 included studies, recording study design, sample and region, mental health indicators, exposure measures, and modeling approaches. Additional coding items captured mechanistic interpretation, exposure-assessment type, treatment of spatial or subgroup heterogeneity, and any reported governance or intervention implications.

The evidence synthesis was then organized around the following four analytical dimensions: (i) exposure–outcome associations between UGS and mental health indicators; (ii) mechanistic interpretation, including attention restoration, stress reduction, behavioral activation, physiological regulation, social interaction, and environmental buffering; (iii) methodological characteristics, with particular attention to dynamic exposure metrics, spatially explicit modeling, and emerging machine learning applications; and (iv) governance relevance, including green social prescribing, spatial equity, accessibility optimization, and blue-green infrastructure coordination. Different methods were not treated as interchangeable, but were interpreted in relation to their inferential aims, temporal sensitivity, spatial granularity, and policy relevance. The main characteristics of the 93 included studies are summarized in

Table 1. Main characteristics of the 93 included studies.

Study	Country/Region	Design	Population/Sample	Mental Health Outcome(s)	UGS Exposure Measure (s)	Key Finding
Zuo (2024) [49]	China	Cross-sectional	Older adults aged ≥60 years in urban China (n = 2465; 119 cities)	Depressive symptoms/mental health	City-level urban green space coverage rate in built-up areas	Positive association; relative deprivation and physical activity mediated the association, whereas social trust was not a significant direct mediator
Zhang (2025) [68]	Nanjing, China	Comparative experiment/intervention study	Young adults aged 18–28 years, primarily students and a small number of early-career workers (n = 80)	Positive and negative affect; mood states; perceived restoration; restoration outcomes	Experimental exposure to pocket parks versus community parks with differing vegetation richness, hardscape activity space, and recreational facilities	Pocket parks can achieve restorative effects comparable to community parks; park quality mattered more than size, with activity space and vegetation richness as key restorative factors
Zhang (2024) [69]	Shanghai, China	Cross-sectional observational study	Urban residents in central Shanghai; stress-perception model trained on volunteer ratings (32 volunteers) and applied to large-scale street-view samples	Psychological stress perception	Street view–based green space exposure derived from Baidu street view images using semantic segmentation	Higher GS exposure was linked to lower stress; the association was nonlinear, with a diminishing stress-relief effect beyond an exposure threshold of about 0.35
Zhang (2019) [70]	Harbin, China	Field experiment/cross-sectional observational study	Young adults (graduate students aged 22–33 years; n = 36)	Emotional responses; cognitive responses; psychological restoration	In-park audio-visual exposure assessed via AV-walk, including acoustic comfort, visual comfort, visual green rate, sky visibility, paving visibility, and soundscape characteristics	Audio-visual comfort was strongly associated with better emotional and cognitive responses; acoustic comfort was more influential for emotion-related outcomes, whereas visual comfort was more influential for…
Zhang (2021) [71]	Guangzhou, China	Cross-sectional observational study	Urban adults in Guangzhou (n = 1003)	Mental well-being (WHO-5)	Green space coverage derived from home buffer (HB) and time-weighted activity and travel buffer (TATB); also included blue space coverage and built-environment exposures	Dynamic activity-space exposure (TATB) explained mental health better than static residential exposure (HB); green and blue space exposure and fitness/recreational facility density were linked to well-being…
Zewdie (2022) [72]	Addis Ababa, Ethiopia	Cross-sectional observational study	Young adults in Addis Ababa from the POFO cohort (n = 210)	Psychological health; emotional and behavioral difficulties (SDQ); depressive symptoms (PHQ-8)	Residential-area greenspace exposure measured by NDVI derived from Landsat 8 imagery and assigned at kebele/sub-city level	Higher greenspace exposure was linked to fewer emotional and behavioral difficulties, but not clearly linked to depressive symptoms; associations were stronger among males and participants with…
Yuen (2020) [73]	Alabama, United States	Pre–post field study/observational study	Park visitors from three urban parks (n = 94)	Subjective well-being; affect; life satisfaction	Short-term urban park visit exposure; visit duration; in-park physical activity tracked by accelerometer	Subjective well-being improved immediately after park visits; longer visit duration was linked to greater gains in life satisfaction, with ~20.5 min identified as the optimal threshold
Yue (2022) [74]	Dalian, China	Cross-sectional observational study	Older adults aged ≥60 years living in urban Dalian (n = 879)	Mental well-being	NDVI, vegetation coverage, park coverage, streetscape greenness, streetscape trees, and streetscape grasses measured within 300 m/500 m buffers and neighborhood boundaries	Multiple greenspace measures were linked to older adults’ well-being; overhead-view measures showed stronger associations than street-view measures, streetscape grasses were more strongly associated than streetscape trees…
Yu (2024) [75]	Hong Kong, China	Repeated-measures observational study	Adult residents from two Hong Kong communities (n = 210; 1408 EMA responses)	Momentary stress	Dynamic real-time greenspace exposure measured using NDVI, Green Space Area Ratio (GSAR), and eye-level Green View Index (GVI) based on GPS trajectories	Greenspace–stress associations varied by exposure metric and cumulative time frame; eye-level greenness (GVI) showed the most robust stress-reducing association, and cumulative exposure exerted stronger effects than…
Yu (2024) [16]	Hong Kong, China	Repeated-measures observational study	Adult residents from two Hong Kong communities (n = 221)	Overall mental well-being (WHO-5); momentary stress	Dynamic greenspace exposure measured using NDVI and eye-level Green View Index (GVI) across residential, workplace, and mobility contexts; total, distance-weighted, and time-weighted exposure metrics	Eye-level greenspace exposure was linked to lower momentary stress, whereas NDVI was not; distance-weighted dynamic greenspace exposure showed a stronger association with overall well-being than total…
Yoo (2022) [76]	New York City, United States	Ecological cross-sectional study	Census tracts across New York City; neighborhood-level counts of mental disorder-related ER visits (2010–2016)	Mental disorder-related emergency room visits; substance abuse; anxiety disorders; mood disorders; psychotic disorders; dementia	Proximity to nearest park, green space coverage (NDVI), and visibility of greenness (Green View Index) aggregated at census-tract level	Associations between greenspace exposure and mental disorder-related ER visits were evident in socially vulnerable neighborhoods but not in low-vulnerability neighborhoods; social vulnerability modified the mental health…
Yoo (2024) [77]	South Korea	Cross-sectional observational study	Older adults aged ≥60 years in South Korea (n = 617)	Depressive symptoms (CES-D-10)	Residential greenspace and blue-greenspace exposure measured using annual and monthly NDVI and land use/land cover–based blue-greenspace metrics across 250 m, 500 m, 1000 m, and 2000 m buffers around home addresses	Higher residential greenspace and blue-greenspace exposure were linked to lower odds of depressive symptoms, especially at smaller buffers near the home (250–500 m); associations were stronger…
Yen (2024) [78]	Taiwan	Cluster trial/intervention study	Healthy adults aged 20–50 years recruited from three college campuses (n = 92 randomized; n = 84 completed)	Quality of life; self-rated health	Actual park exposure and virtual park exposure delivered as 30-min sessions once weekly for 12 weeks	Actual park exposure improved social quality of life and light-intensity physical activity, while virtual park exposure improved mental quality of life; both intervention groups improved self-rated…
Yang (2019) [79]	China	Cross-sectional observational study	University students from 50 universities in 42 Chinese cities (n = 11,954)	Uncertainty stress; life stress	City-level green space measured by per capita public green land area and green land area as a proportion of built-up city area	Greater city-level green space was linked to lower uncertainty stress and, to a lesser extent, lower life stress among university students; associations remained after adjusting for…
Xu (2022) [80]	Fuzhou, China	Cross-sectional field study/observational study	Adult visitors to urban mountain parks in Fuzhou (593 valid questionnaires across 30 sampling sites)	Momentary positive affect; negative affect; anxiety	Urban mountain park exposure across low-, moderate-, and high-disturbance parks; site-level bird diversity metrics including richness, abundance, Shannon diversity, and Simpson diversity	Low-disturbance parks showed better momentary mental health and higher bird diversity, but bird diversity itself was not clearly linked to positive affect, negative affect, or anxiety…
Xu (2024) [81]	Nanjing, China	Cross-sectional observational study	Older urban green space users aged ≥60 years in Nanjing (n = 536)	Subjective well-being	Perceived satisfaction with spatial, green, and grey features of urban green spaces; overall satisfaction with urban green spaces; visit frequency	Spatial, green, and grey feature satisfaction were all associated with well-being; grey features showed the strongest association with older adults’ well-being, and overall satisfaction with urban…
Wood (2017) [82]	Perth, Western Australia, Australia	Cross-sectional observational study	Adults living in 73 new housing developments in Perth, Western Australia (n = 492)	Positive mental health/mental well-being (WEMWBS)	Number, total area, access, and types of public green spaces/parks within a 1.6 km road-network neighborhood around home	Greater number and total area of public green spaces were linked to better well-being, showing a dose–response relationship; both nature-focused parks and recreational/sporting green spaces were…
Wood (2018) [83]	Bradford, United Kingdom	Cross-sectional field study/observational study	Adult urban park users in Bradford (n = 128 survey participants across 11 analyzed parks; 12 parks assessed overall)	Psychological restoration/restorative benefit	Urban park biodiversity and site quality, including plant, bird, and bee/butterfly species richness, habitat number/diversity, tree cover, and site facilities	Biodiversity predicted psychological restorative benefit, whereas site facilities did not; restorative benefit was largely independent of age, gender, and ethnicity
Wang (2025) [84]	Tainan City, Taiwan, China	Randomized controlled trial (RCT)	Older adults aged 50–75 years (n = 90 completed)	State anxiety; perceived exertion; environmental preference	Cycling exposure under different Green View Index (GVI) levels: high GVI (60%), medium GVI (35%), and control (0%) in an urban park setting	Medium-to-high visible greenery (GVI ≥ 35%) enhanced psychophysiological benefits during cycling; medium GVI appeared optimal for balancing anxiety reduction and perceived exertion, whereas both very low…
Wang (2021) [31]	Guangzhou, China	Cross-sectional observational study	Adult residents from 26 inner-city neighborhoods in Guangzhou (n = 1003)	Mental well-being (WHO-5)	Greenspace quantity measured by NDVI and Street View Greenness-quantity (SVG-quantity); greenspace quality measured by Street View Greenness-quality (SVG-quality) and self-reported greenspace quality	Greenspace quantity and quality influenced well-being through different mechanisms: NDVI operated mainly through pollution mitigation (especially NO2), SVG-quantity through NO2, perceived pollution, and social cohesion, whereas…
Diana Vidal Yañez (2023) [85]	Barcelona, Spain	Quantitative health impact assessment/scenario modeling study	Adult residents of Barcelona aged ≥20 years (n = 1,235,375) across 1096 grid cells	Self-perceived poor mental health; visits to mental health specialists; antidepressant use; tranquilliser/sedative use	Grid-cell green space exposure measured by percentage green area (%GA) and NDVI under baseline (2015) and counterfactual Eixos Verds greening scenarios	Implementation of the Eixos Verds Plan was estimated to substantially increase urban greenness and prevent large numbers of adverse mental health annually, indicating that city-scale street-greening…
Torres Toda (2020) [86]	Spain (Valencia, Sabadell, and Gipuzkoa)	Cross-sectional observational study	Mothers from a Spanish birth cohort across three study areas (n = 1171)	Somatization; anxiety; broader psychopathological and psychosomatic symptoms assessed by SCL-90-R	Residential surrounding greenspace measured by satellite-based NDVI within 100 m, 300 m, and 500 m buffers around home addresses	Higher residential surrounding greenspace was linked to lower odds of somatization and anxiety symptoms; protective associations for other symptom dimensions were generally observed but were not…
Thompson (2025) [87]	New York City, United States	Quasi-experimental longitudinal study/difference-in-difference study	Low-income adults living within 0.3 miles of 31 renovated parks and 21 matched control parks in New York City (n = 313)	Perceived stress (PSS)	Exposure to renovated versus matched control neighborhood parks; post-renovation park use frequency	Overall perceived stress changes did not differ significantly between intervention and control groups, but park renovation was linked to lower perceived stress among divorced/separated/widowed adults and…
Taheri (2025) [88]	Sabzevar, Iran	Cross-sectional observational study	Women aged 15–45 years in Sabzevar, Iran (n = 741)	Depressive symptoms (CESD-20 total and subscales)	Residential greenspace exposure measured by distance to nearest green space, distance to nearest green space ≥ 5000 m2, NDVI, and MSAVI2 within 100 m, 300 m, and 500 m buffers	Higher residential greenness (especially NDVI/MSAVI2 at 300–500 m) was linked to lower depressive symptom scores, particularly somatic complaints and adverse social interaction components; greater distance to…
Stopforth (2024) [89]	Not specified	Randomized cross-over panel study	Healthy adults aged 25–70 years (n = 41)	Distress; state anxiety (STAI)	Short-term exposure to an urban park versus an adjacent built urban space during sitting and walking sessions	Visiting an urban park reduced self-reported distress and state anxiety compared with a built urban space; park exposure was also associated with higher HRV (SDNN) and…
Song (2013) [90]	Chiba, Japan	Cross-over field experiment	Young male university students (n = 13)	State anxiety; mood states; subjective psychological responses	Short-term walking exposure in an urban park versus a city area (15-min predetermined walking course)	Walking in the urban park reduced anxiety and negative mood and improved comfort, relaxation, and vigor relative to the city area; physiological responses also suggested greater…
Song (2015) [91]	Kashiwa, Chiba, Japan	Cross-over field experiment/within-subject study	Japanese male university students (n = 23; physiological analyses n = 20)	State anxiety; mood states; subjective psychological responses	Short-term walking exposure in an urban park versus a nearby city area during fall (15-min walk)	Walking in the urban park during fall reduced anxiety and negative mood and improved comfort, relaxation, naturalness, and vigor relative to the city area
Shen (2024) [92]	China	Cross-sectional observational study	University students/young adults from four Chinese universities (n = 27,755)	Anxiety (GAD-7)	Self-reported greenspace use frequency in the past 4 weeks	More frequent greenspace use was linked to lower odds of anxiety; interoception mediated a substantial part of the greenspace use–anxiety association
Ryan (2023) [93]	North Carolina, United States	Ecological spatial study	County- or area-level communities in North Carolina, including rural and urban areas (2009–2018)	Emergency department visits for six mental health conditions; suicide mortality	Public and private greenspace quantity	Greenspace was linked to better mental health in both rural and urban communities, but the strength and pattern of associations varied by rurality and by public…
Roe (2018) [94]	Kashiwa, Chiba, Japan	Cross-over field experiment/within-subject study	Japanese male university students (n = 23; physiological analyses n = 20)	State anxiety; mood states; subjective psychological responses	Short-term walking exposure in an urban park versus a nearby city area during fall (15-min walk)	Walking in the urban park during fall reduced anxiety and negative mood and improved comfort, relaxation, naturalness, and vigor relative to the city area
Ricciardi (2023) [95]	Bari, Italy	Cross-sectional observational study	Older adults aged 60–90 years in the metropolitan area of Bari, Southern Italy (n = 454)	Geriatric depression (GDS-15)	Self-reported greenspace use, measured by visit frequency and time spent in greenspace	Greenspace use was indirectly associated with lower geriatric depression through perceived social support; the direct path was not significant, but the indirect and total effects were…
Reimer (2024) [96]	United States and Canada	Cross-sectional observational study	Female-identified pregnancy planners aged 21–45 years in a North American preconception cohort (n = 9718)	Perceived stress (PSS-10); depressive symptoms (MDI)	Residential greenness measured by NDVI within 50 m, 100 m, 250 m, and 500 m buffers around home addresses	Higher residential greenness was linked to lower perceived stress and depressive symptoms, with stronger associations among women living in lower-SES urban neighborhoods
Razani (2018) [97]	United States	Randomized trial	Low-income parents of pediatric patients aged 4–18 years (n = 78)	Perceived stress; loneliness	Park prescription intervention with or without facilitated group park visits; park visits per week	Overall parental stress decreased over follow-up, and more park visits per week were linked to greater stress reduction; however, supported group park visits did not confer…
Petrunoff (2021) [98]	Singapore	Randomized controlled trial process evaluation/mixed-methods study	Community-dwelling middle-aged adults aged 40–65 years in Singapore (n = 160 randomized; 145 completed 6-month follow-up)	Psychological quality of life	Park prescription intervention promoting physical activity in parks, including counselling, park prescription materials, follow-up counselling, and weekly group exercise in parks	Activity in parks mediated the intervention’s beneficial effects on psychological quality of life, recreational physical activity, and park use; interactive components and social support facilitated participation…
Patwary (2024) [99]	Bangladesh and Egypt	Retrospective cross-sectional survey with repeated self-reported time points	Adults aged ≥18 years in Bangladesh (n = 556) and Egypt (n = 660); total n = 1216	Anxiety (GAD-2); depression (PHQ-2)	Changes in indoor plants, window views of nature, time spent outdoors, and residential greenness (NDVI within a 500 m buffer) from lockdown to post-lockdown	Mental health improved after lockdowns in both countries; increased time spent outdoors was linked to lower anxiety and depression in Bangladesh, whereas changes in indoor plants…
Pasanen (2023) [56]	18 countries/territories (international urban sample)	Cross-sectional observational study	Urban adults from 18 countries/territories; people living alone (n = 2062) and living with a partner (n = 6218)	Mental well-being (WHO-5); anxiety/depression medication use	Neighborhood greenspace coverage within a 1-km buffer around home; greenspace visits	Greenspace visits were indirectly associated with better well-being and marginally lower anxiety/depression medication use through relationship satisfaction and community satisfaction; these indirect associations were similarly strong…
Orstad (2020) [100]	New York City, United States	Cross-sectional observational study	Adult New York City residents from the Physical Activity and Transit Survey (n = 3652)	Mental distress/number of poor mental health days in the past month	Perceived walking time to the nearest park from home; frequency of using the nearest park for physical activity	Closer park proximity was indirectly associated with fewer poor mental health days through more frequent park-based physical activity, but only among residents not concerned about park…
Nutsford (2016) [101]	Wellington, New Zealand	Cross-sectional observational study	Adult residents aged ≥15 years in Wellington, New Zealand (n = 442)	Psychological distress (K10)	Residential visibility of blue space and green space measured using the Vertical Visibility Index (VVI)	Greater visibility of blue space, but not green space, was linked to lower psychological distress; the association appeared specific rather than a general socioeconomic artifact, as…
Nishigaki (2020) [102]	Japan	Multilevel cross-sectional study	Older adults aged ≥65 years in Japan (n = 126,878; 881 neighborhoods)	Depression (GDS ≥ 5)	Neighborhood proportions of trees, grasslands, and fields derived from high-resolution land-use/land-cover data	Greater neighborhood greenspace was linked to lower odds of depression overall; in urban areas, higher tree density was linked to lower odds of depression, whereas in…
Nawrath (2021) [103]	Kathmandu, Nepal	Sequential mixed-methods study (participatory video, focus groups, and Q-methodology)	Urban residents in Kathmandu, Nepal; participatory video involved 20 residents from two socioeconomically contrasting neighborhoods, and Q-methodology included 40 participants from diverse stakeholder groups across Kathmandu.	Perceived mental health pathways and outcomes, including stress reduction, attention restoration, relaxation, reduced anxiety, social cohesion, child development, and perceived harms such as fear, disease risk, and gender-related safety concerns.	Perceived exposure to and interactions with urban greenspaces, including parks, open green spaces, trees, biodiversity/wildlife, and culturally or spiritually significant green features and related ecosystem services/disservices.	Urban greenspaces were perceived to influence mental health through multiple pathways, including reducing harm, restoring capacities, building capacities, and in some cases causing harm
Naghibi (2024) [104]	Tehran, Iran	Experimental visual-stimulation study	Adult participants exposed to visual stimuli of transformed and untransformed urban spaces (sample size not provided in the abstract)	Psychological well-being; comfort; relaxation; mood	Visual exposure to small urban green spaces (SUGS) versus leftover/untransformed urban spaces	Viewing SUGS elicited more favorable psychological responses and stronger preference than untransformed spaces; EEG results showed significant gamma-wave changes consistent with emotional restoration
Murphy (2022) [105]	United States	Cross-sectional secondary analysis	Community-dwelling adults with chronic spinal cord injury in the United States (n = 313)	Positive affect; depressive symptoms	Residential natural greenspace and developed open space within 0.5-mile and 5-mile buffers around home addresses	Contrary to expectations, moderate levels of community natural greenspace and neighborhood developed open space were linked to lower positive affect and higher depressive symptoms, suggesting accessibility-related…
Mukherjee (2017) [106]	Delhi, India	Cross-sectional observational study	Adults with pre-existing chronic conditions in Delhi, India (n = 1208)	Major depression (MINI)	Park availability measured as distance to nearest park, area of nearest park, number of parks within 1 km, and total park area within 1 km	Greater area of the nearest park was linked to lower odds of major depression, whereas other park availability measures were not significantly associated; mediation through walking…
Mouly (2023) [107]	Australia	Longitudinal observational study	Women born 1973–1978 living in major Australian cities (n = 3938)	Anxiety symptoms	Residential NDVI and fractional non-photosynthesising vegetation (fNPV) within a 500 m buffer around home addresses	Higher residential NDVI was linked to lower odds of anxiety symptoms over time, while higher fNPV tended to be associated with higher anxiety; moving to areas…
Mollaesmaeili (2024) [108]	Isfahan, Iran	Cross-sectional observational study	Youth aged 15–24 years from 12 socioeconomically similar neighborhoods in Isfahan (n = 273)	Anxiety (GAD-7); depression (PHQ-9)	Perceived urban green space in residential neighborhoods; with perceived pollution, aesthetics, and fear of COVID-19 infection as mediating variables	Perceived UGS was inversely associated with youth anxiety and had an indirect inverse association with depression; perceived pollution, perceived aesthetics, and fear of COVID-19 infection mediated…
Min (2017) [109]	South Korea	Cross-sectional observational study	Adults from the 2009 Korean Community Health Survey (n = 169,029)	Self-reported depression; suicidal ideation; suicide attempt	Amount of parks and green areas per capita at the administrative-district level	Lower district-level park and green-area provision was linked to higher odds of depression and suicidal indicators; moderate physical activity was also associated with lower odds of…
Mayen Huerta (2021) [110]	Mexico City, Mexico	Cross-sectional observational study	Adult residents of Mexico City surveyed online during COVID-19 (n = 1954)	Subjective well-being (short Warwick–Edinburgh Mental Well-being Scale)	Frequency of urban green space use; perceived neighborhood UGS quality; walking time to nearest UGS	Using UGS once or more per week during the pandemic was linked to higher well-being than no use; perceived UGS quality and accessibility were also important…
Ma [111] (2025)	Beijing and Shanghai, China	Cross-sectional observational study	Adult residents of Beijing and Shanghai (n = 5895)	Depressive symptoms/mental well-being (CESD-10)	Perceived biodiversity of natural outdoor environment and nearest park; empirically measured biodiversity index derived from NDVI-based metrics and park characteristics	Perceived biodiversity was linked to better mental health through perceived psychological restoration, physiological restoration, and satisfaction with greenspace, whereas empirically measured biodiversity showed no significant association…
Liu (2024) [112]	Guangzhou, China	Cross-sectional observational study	Adults in Guangzhou (n = 719)	Mental well-being (WHO-5) as mediator	Network distance to nearest park; % green space; NDVI around residence/workplace	Greenspace was linked to lower blood pressure and hypertension risk, and mental health partly mediated the association
Li (2025) [113]	Nanjing, China	Cross-sectional observational study	Park visitors in 51 parks in central Nanjing (n = 1562)	Psychological health and well-being (PWB)	Park engagement, perceived restorativeness, sustained exposure, and park type (comprehensive, community, specialized, pocket parks)	Park engagement had a stronger total effect on PWB than perceived restorativeness; restoration experience and sustained exposure significantly mediated the associations
Li (2023) [114]	Shanghai, China	Cross-sectional observational study	Cancer survivors residing in Shanghai (n = 4195)	Anxiety symptoms (GAD-2); depressive symptoms (PHQ-2)	Residential greenness measured by NDVI and EVI within 250 m, 500 m, and 1000 m buffers around home	Higher residential greenspace, especially NDVI within 250 m, was linked to lower anxiety and depressive symptom scores; stronger associations were observed among adults aged 18–65 years…
Li (2023) [115]	China	Cross-sectional observational study	Urban residents recruited online (n = 668)	Mental well-being (WHO-5)	Perceived residential greenspace; residential NDVI within 500 m; green physical activity for leisure, transportation walking, and transportation cycling	Both perceived greenspace and NDVI-500 m were linked to well-being, but only perceived greenspace was linked to green physical activity; only leisure green physical activity mediated…
Lee (2019) [116]	South Korea	Cross-sectional observational study	Elderly urban residents aged ≥65 years from seven Korean metropolitan areas (n = 11,408)	Subjective stress; depressive symptoms	Proportion of urban green area per administrative area	Higher community green-space proportion was linked to lower prevalence of stress and depressive symptoms among urban older adults
Kothencz (2017) [117]	Szeged, Hungary	Cross-sectional observational study	Visitors of five urban green spaces in Szeged; city residents with valid questionnaires (n = 227)	Satisfaction with urban green space; self-reported quality of life	Perceived green space characteristics (nature, noise abatement, recreation capacity, microclimate regulation, habitat, air purification, visual appearance); crowd-sourced running trajectories and green-space aesthetic photographs	Perceived green space characteristics with direct well-being benefits were strong predictors of both satisfaction and quality of life, whereas regulating ecosystem service perceptions had weaker effects
Kondo (2022) [118]	Philadelphia metropolitan area, United States	Cross-sectional observational study	Non-institutionalized adults from the 2015 and 2018 Southeastern Pennsylvania Household Health Survey (n = 11,601)	Self-reported diagnosed mental health condition	Residential tree canopy cover within 250 m, 500 m, and 1000 m of home; perceived access to a comfortable local park or outdoor space	Greater greenspace exposure was linked to reduced socioeconomic inequalities in self-reported mental health conditions; both objective tree canopy and perceived park access showed equity-relevant associations
Kodali (2023) [119]	New York City, United States	Cross-sectional observational study	Adults living in low-income communities near 54 neighborhood parks in New York City (complete-case n = 650; sensitivity sample n = 1354)	Quality of life (QoL)	Park use frequency; park perception	Park perception had a stronger association with quality of life than park use; the park perception–QoL relationship was partly mediated by perceived stress and, to a…
Klompmaker (2019) [120]	The Netherlands	Cross-sectional observational study	Adults from the 2012 Dutch national health survey (n = 387,195; analytic sample after exclusions n = 354,827 for greenness analyses)	Severe psychological distress (K10); prescriptions of anxiolytics, hypnotics/sedatives, and antidepressants	Residential surrounding greenness/green space measured by NDVI and land-use data within 300 m and 1000 m buffers	Surrounding green was inversely associated with poor mental health, while air pollution was positively associated and traffic noise showed weaker outcome-specific associations; in multi-exposure models, greenspace…
Huang (2021) [121]	Fuzhou, China	Field experimental study	Young adults aged 22–28 years visiting 13 green-space sites in a low-density residential area (n = 33)	Perceived restorativeness/attention restoration; stress recovery	Exposure to different urban green-space sites varying by plant richness, water features, topography, road presence, and cultural landscape	Exposure to urban green spaces significantly improved perceived restorativeness and multiple physiological stress-related indicators; sites with high plant species richness, water features, uneven topography, cultural landscape…
Heo (2021) [122]	South Korea	Cross-sectional observational study	Adults living in South Korea recruited online during COVID-19 (analytic sample n = 322)	Probable major depression (PHQ-9); probable generalized anxiety disorder (GAD-2)	Self-reported change in greenspace visits after the COVID-19 outbreak compared with 2019; ZIP-code level greenness (EVI)	Decreased visits to greenspace during the pandemic were linked to higher odds of depression, but not significantly with generalized anxiety; barriers to greenspace use during outbreaks…
Helbich [123] (2025)	The Netherlands	Cross-sectional observational study	Adults in at least moderately urbanized Dutch municipalities from the 2022 Dutch Public Health Monitor (n = 180,949)	Psychological complaints (MHI-5); anxiety and depression symptoms (K10-based severe distress); psychological resilience	Adherence to the 3+30+300 urban green space rule and its components: ≥3 visible mature trees, ≥30% neighborhood tree canopy cover, and ≤300 m to a public green space	Meeting the 3+30+300 rule was linked to higher physical activity and lower odds of overweight, but showed no clear association with psychological complaints, anxiety/depression symptoms, or…
He (2023) [124]	China	Prospective cohort study	Older adults aged ≥65 years from the Chinese Longitudinal Healthy Longevity Survey (baseline n = 13,133)	Subjective well-being (8-item SWB scale)	Residential greenness measured by NDVI within 250 m around home	Higher residential greenness was linked to better well-being, and well-being partially mediated the protective association between greenness and all-cause mortality; stronger benefits were observed among cognitively…
Grigoletto (2023) [14]	Barcelona (Spain), Doetinchem (the Netherlands), Kaunas (Lithuania), and Stoke-on-Trent (UK)	Cross-sectional observational study	Adult residents from four European cities (n = 3134)	Restoration outcomes (ROS); baseline mental health measured by SF-36 mental health subscale	Self-reported use of the most frequently visited urban green space and frequency of activities performed there (e.g., walking/sport, tranquility, personal relaxation, meeting family/friends)	Visiting urban green spaces was linked to restorative outcomes, with stronger restoration among people with poorer baseline mental health; activity-related restoration patterns were broadly similar across…
Gascon (2018) [125]	Barcelona, Spain	Cross-sectional observational study	Adults aged 45–74 years from the ALFA cohort residing in Barcelona (n = 958)	Self-reported doctor-diagnosed anxiety and depression; history of benzodiazepine and antidepressant use	Residential greenness measured by NDVI, amount of green space, access to major green spaces, and access to blue spaces	Higher surrounding greenness and access to major green spaces were linked to lower odds of some mental health, particularly benzodiazepine use and depression
Gao (2019) [126]	Baoji, China	Cross-sectional observational study	Urban park visitors in People’s Park, Baoji (n = 827)	Self-reported stress level; stress-recovery preference/restorative experience	Perceived sensory dimensions of urban green space (e.g., serene, nature, culture, social, prospect) and habitat types within the park	Highly stressed individuals preferred multi-layered woodland areas adjacent to water with stronger serene and nature dimensions and weaker prospect, culture, and social dimensions; sports/leisure and quiet…
Fossa (2024) [127]	United States	Repeated cross-sectional analysis within a longitudinal cohort	Urban and suburban older adults from the Health and Retirement Study, 2008–2016 (n = 21,611)	Major depression (CIDI-SF, 12-month major depressive episode)	Residential greenness measured by annual maximum NDVI within a 1 km buffer around home	Higher residential greenspace was linked to lower prevalence of major depression among older adults; associations appeared stronger in tropical and cold climates than in arid and…
Feng (2022) [128]	Australia (Sydney, Newcastle, Wollongong)	Cross-sectional observational study	Adults aged >45 years from the 45 and Up Study living in houses (n = 66,453) or apartments (n = 13,196) in three NSW cities	Psychological distress (Kessler-10, high-risk distress)	Percentage total green space, tree canopy, and open grass within 1.6 km road-network buffers around residence	Tree canopy was linked to lower odds of psychological distress for both house- and apartment-dwellers, whereas open grass was linked to higher odds of distress
Felappi (2024) [129]	São Paulo, Brazil	Cross-sectional observational study	Urban park users in a Neotropical megacity	Perceived restorativeness/mental well-being	Urban park quality characteristics, including safety, naturalness, water bodies, tree canopy, habitat heterogeneity, maintenance, vegetation structure, and perceived biodiversity	Perceived restorativeness varied across parks and was mainly driven by safety and naturalness; water bodies benefited both well-being and wildlife, whereas high tree canopy favored restoration…
Dzhambov (2018) [130]	Plovdiv, Bulgaria	Cross-sectional observational study	Youth aged 15–25 years living in Plovdiv (analytic sample n = 399)	Mental health (GHQ-12)	Objective residential greenspace (NDVI, SAVI, tree cover density, distance to nearest greenspace) and self-reported greenspace (greenness, visible greenery, walking time, time in greenspace, perceived greenspace quality)	No direct association was found between objectively measured greenspace and mental health, but serial mediation models suggested that restorative quality contributed to better mental health through…
Dzhambov (2019) [131]	Plovdiv, Bulgaria	Cross-sectional observational study	University students aged 18–35 years residing in Plovdiv (n = 529)	Anxiety symptoms (GAD-7); depressive symptoms (PHQ-9)	Residential greenness measured by NDVI and tree cover density in multiple buffers (main analyses at 500 m), plus perceived greenspace	Higher residential greenspace was linked to lower anxiety and depression scores; This association was partially mediated through perceived greenspace, restorative quality, greater mindfulness, lower rumination, and…
Duan (2024) [132]	Xi’an, China	Experimental field study	College students (n = 400)	Perceived restorativeness/mental fatigue recovery; physiological stress response	Visual exposure to urban green spaces with different vegetation structures across winter and summer (single-layer woodland, tree-shrub-grass woodland, tree-grass woodland, single-layer grassland, and non-green concrete control)	Seasonal differences significantly affected perceived restorativeness but not skin conductance level; single-layer woodland showed the strongest restorative effect overall, with tree-shrub-grass woodland also performing well in…
Deng (2025) [133]	Christchurch, New Zealand	Prospective birth cohort/lifecourse study	Christchurch Health and Development Study birth cohort (original n = 1265; analytic samples varied by model)	Depressive symptoms, anxiety disorders, suicidal ideation	Greenspace availability measured as proportion of vegetated area within 100 m–3000 m buffers around geocoded residential addresses across birth to age 40	No clear associations were found between childhood greenspace and adolescent mental health, but higher greenspace availability in adulthood—especially within 1500 m and 2000 m buffers—was linked…
Dang (2025) [134]	China	National cross-sectional study	Adolescents aged 10–18 years from the 2019 Chinese National Survey on Students’ Constitution and Health (n = 149,697)	Mental well-being, psychological distress, and Dual-Factor Model mental health categories (vulnerable, symptomatic but content, troubled)	School-based greenness measured by 3-year mean NDVI within a 3000 m buffer around school	Higher NO2, O3, and especially oxidative potential (OX) were linked to worse mental health profiles; these adverse associations were significantly stronger among adolescents in low-greenness school…
Cronshaw (2025) [135]	England, UK	Longitudinal cohort study	Millennium Cohort Study children continuously residing in England, ages 3–11 years (analytic sample n = 6946)	Conduct problems, hyperactivity/inattention, peer problems, emotional symptoms (SDQ), and cognitive ability	Neighbourhood greenspace quantity at ward level (MEDIx greenspace deciles based on GLUD and CORINE data)	Greenspace quantity was not associated with child outcomes at the intercept (~age 7), but was linked to the slope of conduct problems and cognitive ability trajectories…
Cleary (2019) [136]	Brisbane, Australia	Longitudinal observational study	Mid-aged urban residents from the HABITAT study who remained at the same address across two waves (n = 5014)	Psychological well-being (short Warwick–Edinburgh Mental Well-being Scale)	Perceived quantity of urban green space in the suburb (greenery and tree cover along footpaths)	Higher perceived quantity of urban green space was linked to better psychological well-being cross-sectionally, and within-person increases in perceived green space over two years were linked…
Chu (2021) [137]	Taipei, Taiwan	Cross-sectional multilevel study	Park users aged ≥55 years in 19 neighborhood parks (n = 380)	Well-being and depression	Urban park quality assessed with the Neighborhood Green Space Tool; environmental perception; leisure activity frequency	Associations of environmental perception and leisure activity with well-being and depression were moderated by park quality dimensions, especially accessibility, amenities, and incivilities, indicating that park quality…
Cheng (2024) [60]	Wuhan, China	Cross-sectional multilevel study	Rural-to-urban migrants in Wuhan from 60 neighborhoods (n = 716)	General mental health (GHQ-12)	UGS quality (satisfaction, park social-media ratings, greenspace disorder, plaza-type parks) and quantity (park density, park area ratio, green view index) within neighborhood buffers	Both UGS quality and quantity were linked to migrants’ mental health, with quantity showing a stronger overall association; Park density and green view index were linked…
Chen (2024) [138]	China	Prospective cohort study	Adults aged ≥50 years from the Chinese WHO SAGE cohort (n = 8481)	Incident depression	Residential greenness measured by annual maximum NDVI averaged over 2 years within 100 m, 250 m, 500 m, and 1000 m buffers around home addresses	Higher residential greenness was linked to lower depression incidence; Each IQR increase in NDVI within 500 m was linked to a 40% lower hazard of incident…
Cao (2023) [139]	Chengdu, China	Experimental repeated-exposure study	Healthy adult participants (n = 50)	Positive affect, negative affect, perceived restorativeness, pulse/blood pressure	Different UGS environmental types varying by naturalness, water presence, and sky-view factor under sunny vs. cloudy weather	Restorative benefits of UGS were stronger on sunny than cloudy days; Spaces with greater sky exposure showed stronger physiological restoration on sunny days, while spaces with…
Camargo (2017) [140]	Bucaramanga, Colombia	Cross-sectional study	Urban park users aged >12 years from 10 parks (n = 1392)	Quality of life (EUROHIS-QOL-8)	Park accessibility, active vs passive park use, companionship, perceived safety, perceived quality of green areas, trees, and walking paths	Better QoL was linked to years of schooling, visiting parks with a companion, active park use, higher perceived quality of trees and walking paths, and perceived…
Bustamante (2022) [141]	United States	Longitudinal mixed-methods study (baseline cross-sectional quantitative analysis + qualitative thematic analysis)	Adults aged ≥55 years from the COVID-19 Coping Study; quantitative sample n = 6913 with valid ZIP code	Depressive symptoms (CES-D-8), anxiety symptoms (BAI-5), loneliness (UCLA 3-item)	Number of neighborhood parks within ZIP Code Tabulation Area; sensitivity analysis using park area	Among urban residents, higher neighborhood park availability was linked to lower prevalence of depression and anxiety; Qualitative findings showed that engagement with parks and outdoor spaces…
Bu (2022) [142]	England, United Kingdom	Longitudinal panel study during COVID-19	19,848 urban adult participants from the UCL COVID-19 Social Study, followed for 20 months (March 2020–October 2021)	Anxiety symptoms measured repeatedly using GAD-7	Area-level greenspace coverage (%) within residential LSOAs based on the 2019 UKCEH Land Cover Map; sensitivity analyses used proximity to nearest public parks/playing fields and subjective satisfaction with local greenspace	Higher residential greenspace coverage was linked to fewer anxiety symptoms during the pandemic, independent of major confounders; There was limited evidence that greenspace influenced the trajectory…
Bray (2022) [143]	Multiple countries/international systematic review (studies published 2000–2020)	Systematic review with narrative synthesis and conceptual framework development	48 included sources on young people aged 14–24 years living in urban settings; studies covered multiple countries, exposure types, and study designs	Anxiety, depression, and related psychological outcomes/mechanisms including mood, wellbeing, life satisfaction, quality of life, mindfulness, physical activity, social cohesion, self-esteem, and stress	Broad green/blue space exposure measures, including access to and contact with urban green space, forests and blue spaces accessed by urban youth, street trees, sensory components of nature, and green-space-based activities/interventions	Experimental studies suggested that being in or walking through green space can improve mood and reduce state anxiety immediately; Observational and non-randomized studies suggested that physical…
Boudier (2022) [144]	Europe (19 centres in 8 countries: Belgium, England, France, Italy, Norway, Spain, Sweden, Switzerland)	Longitudinal cohort analysis using repeated measures from ECRHS-II (2000–2002) and ECRHS-III (2011–2013)	6542 participants at ECRHS-II and 3686 at ECRHS-III; repeated data available for a multicentre adult population; mean follow-up 11.3 years	Health-related quality of life (HRQOL), specifically SF-36 Mental Component Summary (MCS) and Physical Component Summary (PCS)	Residential greenness by mean NDVI within 300 m buffer; presence of green spaces within 300 m buffer, including urban green spaces, forests, and agricultural land (binary indicators for subset)	Higher residential greenspace was linked to better mental HRQOL, while no consistent association was found for physical HRQOL; Higher NDVI and presence of forests were linked…
Bojorquez (2018) [145]	Tijuana, Mexico	Cross-sectional study combining adult women survey data (2014) with public-space data (2013)	Adult women living in Tijuana, Mexico (female community sample; exact sample size not provided in the excerpt)	Depressive symptoms measured by the Center for Epidemiologic Studies-Depression scale (CES-D)	Urban public park coverage within 400 m and 800 m buffers around participants’ homes; park use considered as mediator; park and personal characteristics considered as moderators	Higher urban public park coverage was linked to lower CES-D scores; The 400 m effect was moderated by age and was significant mainly for younger women…
Boakye (2025) [146]	Sunyani Municipality, Ghana	Hospital-based cross-sectional study	420 mothers attending post-natal clinic at Sunyani Teaching Hospital; eligible if child aged 0–24 months	Perceived stress, anxiety, and depression (measured using adapted PSS-9, HAM-A, and CES-D-based scales)	Self-reported green space exposure only: perceived amount of neighborhood greenery, visibility of vegetation from residence, ease of access to closest park/green space, frequency of spending time in natural vegetation, perceived quality of nearby green spaces, use of green spaces for leisure/physical exercise, feelings after spending time in green spaces, and belief that green spaces improve well-being	Higher stress, anxiety, and depression were common in the sample; Use of green spaces for leisure and physical exercise was significantly associated with lower perceived stress
Barry (2023) [147]	Metro-Philadelphia, United States	Secondary analysis of an ongoing prospective pregnancy cohort	283 Black pregnant patients in the MOMENTUM cohort; greenspace survey at <16 weeks’ gestation	Perceived stress measured by Perceived Stress Scores at 16–20 weeks’ gestation; analyzed as continuous score and dichotomized high stress (≥30) vs. low stress (<30)	Self-reported greenspace exposure/use: hours per week in greenspace, analyzed continuously and categorically as none, moderate, and high (median split among exposed at 1.125 h/week); also included neighborhood cohesion as a contextual social variable	In adjusted models, each additional hour of self-reported greenspace exposure per week was linked to 0.48 points lower perceived stress (95% CI: −0.92, −0.04)
Bao (2022) [148]	Changchun, China	Cross-sectional field survey conducted in community parks during summer weekends	254 valid questionnaires from children aged 4–15 years in 15 community parks; 300 distributed, 83% valid response rate; younger children (4–7 years) completed with guardians	Children’s social anxiety measured using the Social Anxiety Scale for Children-Revised (SASC-R), including fear of negative evaluation (SAD-FNE), social avoidance/distress to novelty (SAD-NEW), and general social avoidance/distress (SAD-GEN)	UGS metrics included NDVI, 10-min walk accessibility/service area, and park-scale indicators; children’s play-space qualities assessed using the Woolley and Lowe play-space assessment tool, including service facility diversity, surfacing materials, attractiveness, challenge, and environmental/play-space characteristics	Higher NDVI and better 10-min accessibility were significantly associated with lower children’s social anxiety; Greater diversity of service facilities, better surfacing materials, and more attractive and…
Bai (2025) [149]	Hunan Province, China (Tianjiling National Forest Park/Hunan Provincial Botanical Garden)	Cross-sectional questionnaire study in an urban forest park	504 valid respondents from 530 distributed questionnaires; surveyed visitors in July 2024 in an urban forest park/botanical garden setting	Subjective well-being (primary), plus related psychological constructs including mental recovery, restorative environment perception, and positive emotions	Primarily subjective exposure/perception measures rather than GIS greenness metrics: natural environment perception (including natural attribute/form perception), restorative environment perception (fascination/charm, compatibility, being away/distance, extent/expansiveness), and visit frequency to the urban forest park	Natural environment perception had significant positive effects on both mental recovery and restorative environment perception; Mental recovery significantly improved well-being as a mediating variable
Astell-Burt (2019) [150]	Australia (Sydney, Wollongong, Newcastle)	Prospective cohort study using baseline (2006–2009) and follow-up (2012–2015) data from the Sax Institute’s 45 and Up Study	46,786 residentially stable, city-dwelling adults; mean age 61.0 ± 10.2 years; 53.8% female	(1) Psychological distress measured by Kessler-10 (K10); (2) self-reported physician-diagnosed depression or anxiety; (3) fair to poor self-rated general health	Percentage of total green space, tree canopy, grass, and other low-lying vegetation within a 1.6-km road network buffer around residential address/mesh block centroid at baseline	Greater total green space (≥30%) and especially tree canopy (≥30%) were linked to lower incidence of psychological distress; higher tree canopy was also associated with lower…
Arifwidodo (2023) [151]	Bangkok, Thailand	Cross-sectional study conducted during the COVID-19 lockdown (March–May 2020) using a telephone survey	579 adults in Bangkok; respondents were drawn from a 2019 survey on park visits and physical activity	Mental health well-being measured by the WHO-5 Well-Being Index; dichotomized with <50 indicating poor well-being	Self-reported urban green space visitation within the last 2 weeks during lockdown; exposure defined as visiting accessible/open urban green spaces such as neighborhood parks, gardens, community plazas, closed streets, and other non-government-managed green areas	Visiting urban green spaces during lockdown was significantly associated with higher mental health well-being; Socioeconomic characteristics and healthy behaviors were also related to better WHO-5 scores
Allaouat (2021) [152]	Helsinki region, Finland	Cross-sectional study based on the Helsinki Capital Region Environmental Health Survey 2015–2016	5895 adults living in the Helsinki region; 377 reported doctor-diagnosed or treated depression	Prevalence of depression, defined as self-reported doctor-diagnosed or treated depression	Not a UGS-focused exposure study. UGS was treated as an adjustment covariate in the models; the main environmental exposures were long-term ambient PM2.5 from road traffic and PM2.5 from residential wood combustion, linked to home addresses via high-resolution urban-scale emission and dispersion modeling	No convincing evidence that long-term exposure to traffic-related PM2.5 or wood-combustion-related PM2.5 was linked to depression prevalence; Estimates for traffic PM2.5 were elevated but non-significant, especially…
Ahmed (2022) [153]	Australia	Cross-sectional analysis within the 2016/17 MatCH sub-study of the Australian Longitudinal Study on Women’s Health; exposure history reconstructed prospectively across pregnancy and childhood	3048 mothers reporting on 5799 children aged 0–12 years (mean age 7.0 ± 3.2 years)	Health-related quality of life (HRQoL) assessed by PedsQL: total score, psychological health summary score, and physical health summary score	Residential greenspace around maternal/child home addresses using NDVI (green vegetation) and fractional non-photosynthetic vegetation, fNPV; calculated within 100 m and 500 m buffers. Multiple exposure windows: during pregnancy, first year of life, lifetime average, maternal long-term exposure before birth, and year preceding survey; sensitivity analysis also used parkland percentage	In crude models, higher NDVI at 500 m during early life/childhood showed some positive association with total and psychological HRQoL, but associations disappeared after adjustment
Afrad (2020) [154]	Tangier, Morocco	Cross-sectional study	388 participants from three densely populated neighborhoods in the Beni-Makada district, Tangier; face-to-face survey, systematic random sampling	Depression level measured by PHQ-9 (score 0–27)	Informal urban green space exposure operationalized as ownership of a potted street garden (PSG); additional PSG attributes among owners included PSG publicness (whether seen as a public amenity), daily care duration, weekly care frequency, PSG size (number of pots), PSG age, and recreational activity near PSG	PSG ownership was linked to higher depression scores in this disadvantaged, dense neighborhood context; However, among owners, perceiving the PSG as a public amenity was linked…

, while extended extraction details are provided in Supplementary File S1.

Detailed analytical approaches, extended findings, and identifiers are provided in Supplementary File S1.

3.6. Methodological Reliability and Quality Control

We implemented dual independent screening at both the title/abstract and full-text stages. Inter-rater agreement was documented and discrepancies were resolved through discussion to consensus, with adjudication by a senior reviewer when necessary. Data extraction was conducted using a standardized form, and key fields were cross-verified item by item against full texts to ensure internal consistency of variables and coding outputs. In addition, we performed a random audit (n = 100 sampled from the bibliometric corpus, N = 929) as a quality-control check of screening consistency; this audit was used solely for QC purposes and did not change the PRISMA flow counts.

For bibliometric mapping and visualization, we disclosed primary parameters (e.g., thresholds and inclusion rules) and conducted threshold perturbation/sensitivity checks to assess robustness [64].

Risk of bias/quality appraisal. Given substantial heterogeneity in study designs, exposures, and outcomes and our objective of mapping mechanisms and methodological trends (rather than estimating pooled effects), we did not apply a formal risk-of-bias tool to rate each included study. Instead, we strengthened methodological transparency via standardized coding, cross-verification, and explicit reporting of study characteristics.

In line with reproducibility principles, the manuscript provides the main characteristics of the 93 included studies in Table 1, while Supplementary File S1 provides the full search strategies, extended extraction details, the full-text exclusion list with reasons, and the random QC audit of screening/extraction consistency.

4. Research Results and Thematic Evolution

Building on the two datasets introduced in Section 3—the bibliometric corpus (N = 929) and the evidence-synthesis sample (n = 93)—this chapter presents temporal trends, spatial patterns, collaboration networks, keyword structures, and the methodological progression of research on UGS and mental health. All trend interpretations map one to one to Figure 3, Figure 4 and Figure 5 and

Table 2. Method–theme matrix of UGS–MH research (2013–August 2025).

Theme	Analytical Approach	Representative Techniques/Models	Typical Use Context/Application Focus
Spatial Distribution and Clustering [88,89,90,91,92,93]	Spatial statistics and spatial autocorrelation	Global/Local Moran’s I; LISA; Getis-Ord Gi *	Best suited for city-/regional-scale disparity detection and equity-oriented spatial planning
Exposure–Outcome Relationships [91,92,93,94,95]	Regression and spatial regression	OLS; GWR; MGWR; SEM	Best suited for estimating average and spatially varying exposure–outcome associations
Temporal Dynamics and Trends [83,91,92,93,96]	Time-series and trend analysis	Mann–Kendall; Sen’s slope; AutoRegressive Integrated Moving Average	Best suited for temporal pattern detection and trend estimation in exposure–mental health relationships
Mechanisms and Mediation Effects [76,77,91,92,93]	Structural-equation and path analysis	SEM; Partial Least Squares–SEM; multilevel models	Best suited for testing mediators, pathway structure, and subgroup differences
AI and Methodological Innovation [91,92,93]	Machine and deep learning	RF; XGBoost; CNN; GCN	Best suited for nonlinear exposure characterization, individualized profiling, and predictive modeling
Policy Simulation and Scenario Modeling [78,79,80,91,92,93]	Multi-criteria and system-dynamics modeling	Multi-Criteria Decision Analysis; System Dynamics; agent-based modeling	Best suited for intervention comparison, planning trade-offs, and scenario-based governance evaluation

* this table summarizes the major methodological categories across four dimensions: theme, analytical approach, representative techniques/models, and typical use context.

and

Table 3. Methodological evolution and modeling characteristics in UGS–MH research (2013–August 2025) *.

Stage	Data and Exposure Metrics	Methods and Theoretical Orientation	Key Outcomes/Findings	Illustrative Studies
Stage I (2013–2016): Macro-level Exposure Identification	Vegetation indices (NDVI); cross-sectional population and health surveys	Correlation and regression; macro-spatial description	Established macro-level associations between “green-space and mental health”; limited causal mechanisms	[155,156]
Stage II (2017–2020): Mechanism Integration	Street-view greenness (GVI); accessibility; social/psychological covariates	SEM and multilevel models; mediation/moderation analysis; pathway identification	Identified pathways such as stress buffering, attention restoration, social cohesion, and physical activity	[155,156]
Stage III (2021–2025): Intelligent Modeling	Dynamic exposure (GPS-EMA, TWCE); multimodal data (remote sensing, street view, wearable, social)	Machine learning/deep learning (RF, XGBoost, CNN, GCN); toward stronger causal inference; policy scenario modeling	Individualized exposure characterization and mental-state prediction; progress toward causal inference and planning translation	[19,157,158]

* stages I–III progress from macro-level exposure identification to mechanism integration to intelligent modeling. Methods emphasize dynamic exposure (GPS-EMA, TWCE/WCE), multimodal inputs, and machine/deep learning approaches (RF, XGBoost, CNN, and GCN).

, providing the empirical basis for the subsequent discussions on mechanisms and policy.

4.1. Global Distribution and Temporal Trends

From 2013 to 2025 (through August 2025), annual publications increased overall (Figure 3). The 2013–2018 period marks an embryonic stage centered on broad UGS accessibility–MH associations; 2019–2021 shows an acceleration phase; and 2022 reaches a stage peak (104 papers) amid heightened public mental health concerns. Growth moderates thereafter, but the thematic scope becomes visibly more diversified, indicating a shift from “whether an association exists” toward mechanism-oriented modeling and evaluation. By database, WoSCC remains dominant for interdisciplinary and environmental-science outputs, whereas PubMed rises rapidly after 2017, reflecting convergence between environmental health and public health.

4.2. Geographical Patterns and International Collaboration

As shown in Figure 4A,B, the global distribution of studies is spatially uneven. Europe and North America—notably the United Kingdom (UK), the Netherlands, Germany, and the Nordic countries—built an early comparative advantage that integrates urban ecological planning, restorative environments, and public health. In recent years, the UK has advanced green social prescribing (GSP), embedding nature contact within the national mental health intervention system; its nationwide “Test and Learn” pilots (seven areas; >8500 participants) report positive effects on mental health improvement and mitigation of health inequalities [159], with joint evaluations issued by NHS England, the public health service of the UK and the European Center for Environment and Human Health [160,161].

At the evidence-synthesis level, recent reviews frame urban nature as a nature-based solution, emphasizing multi-pathway mechanisms such as affective restoration, stress reduction, social cohesion, and equitable access, and translating findings into actionable public health and planning measures [162]. The WHO Regional Office for Europe further recommends integrating blue-green-spaces into urban public health and mental health action frameworks, promoting a closed loop from evidence to planning to intervention [163].

Regionally, East Asia started later but has accelerated in the past five years, becoming active in geographic information system (GIS)-based exposure modeling, forest-therapy experiments, and dynamic trajectory analysis; room remains for advances in intervention translation, policy practice, and equity evaluation. Low- and middle-income countries show lower overall output; during the pandemic, disparities in accessibility and mobility restrictions widened socioeconomic gaps in green-space use and associated health gains [99,164].

Collaboration networks indicate that the United States (US), UK, and Italy are major producers; China (including Hong Kong) is expanding both publication volume and international co-authorship, with Korea and Japan exhibiting steady growth (Figure 4A). At the institutional level, the University of Oxford and Universiti Teknologi MARA (Malaysia) make notable contributions (Figure 4B). Overall, the structure reflects a Euro-Atlantic core, East Asian catch-up, and underrepresentation of low- and middle-income countries. Collaboration is shifting from single-center to multi-polar coordination, a transition that promotes thematic and methodological convergence globally and sets the stage for the subsequent analysis of knowledge-structure evolution.

4.3. Research Hotspots and Network Structure

Under a minimum-occurrence threshold of six, VOSviewer identified 66 high-frequency keywords from the N = 929 corpus to construct a knowledge network (Figure 5). The map exhibits a triangular core centered on UGS (174 occurrences), Normalized Difference Vegetation Index (NDVI) (146), and mental health (108), corresponding to the three principal dimensions of exposure quantification, ecological indicators, and health outcomes. Peripheral nodes such as Stress, Quality of Life, Physical Activity, and Parks radiate outward and articulate a multilayer linkage from environmental exposure to psychophysiological and socio-behavioral processes (Figure 5A,D). Although Parks and Stress are not the most frequent, their high total link strength (TLS) and cross-cluster ties indicate a bridging function. The appearance of Urban Planning, Ecosystem Services, Green Infrastructure, and GIS signals a clear expansion toward governance and planning domains.

A temporal overlay (Figure 5B) shows a progression from macro-level exposure assessment to intelligent modeling and dynamic prediction. In the early stage (2013–2016), NDVI/vegetation-based exposure measures were paired with outcomes such as Depression, emphasizing general associations. The middle stage (2017–2020) introduced Stress, Well-being, Physical Activity, and Parks, marking a shift toward mechanism identification and multivariate interaction. The recent stage (20217–present) features Urban Planning, Ecosystem Services, AI modeling, and Dynamic Exposure, reflecting a turn to spatial equity, governance evaluation, and AI-enabled modeling.

The density view (Figure 5C) confirms UGS, NDVI, and mental health as stable anchors. The following two high-density bands emanate from these cores: a behavioral-experiential band (Parks, Quality of Life, Physical Activity, Nature) and a psychophysiological-environmental band (Stress, Vegetation, Temperature), coupled through the triangular core. GIS sits adjacent to the core as a methodological hub.

Synthesizing the top 20 keyword profile and nine computational clusters (Figure 5D), the following four functional channels emerge: (i) environmental measurement and exposure assessment (NDVI, vegetation, and GIS); (ii) psychological and physiological mechanisms (stress, restoration, and well-being); (iii) behavioral interventions and health impacts (parks, physical activity, and quality of life); and (iv) spatial equity and planning translation (urban planning, ecosystem services, green infrastructure, and climate). Across these channels, mental health outcomes (mental health, depression, and anxiety) form an integrative axis that links exposure, mechanisms, interventions, and policy.

In summary, Figure 5A–D jointly indicate an orderly evolution from quantitative assessment to mechanistic explanation, behavioral intervention, and policy and governance translation. Together, they establish a health outcome feedback loop and signal a shift from static correlation to dynamic modeling and causal identification.

4.4. Methodological Evolution and AI Modeling Trends

Methodological development in the UGS–MH field has increasingly linked exposure quantification, mechanism-oriented analysis, and policy-relevant interpretation. Table 2 summarizes six methodological families and their representative techniques, whereas Table 3 outlines their staged evolution. Building on these two summaries, this section highlights three major methodological turning points.

First, modeling spatial heterogeneity and spatial dependence has been pivotal for moving beyond simple average associations toward more spatially sensitive and context-aware inference. Studies have progressed from ordinary least squares (OLSs) to spatial regressions and multilevel models—including geographically weighted regression (GWR) and multiscale geographically weighted regression (MGWR) and spatial error/lag specifications—allowing effects to vary by place and scale and thereby capturing nonstationarity and spatial dependence in the exposure–outcome relationship [165]. Rather than treating the UGS–MH relationship as spatially uniform, these approaches allow researchers to identify where associations appear stronger, weaker, or more unstable across urban contexts. Spatially explicit modeling can therefore reduce aggregation bias and provide a stronger basis for mechanism testing and context-sensitive interpretation.

Second, dynamic exposure approaches represent more than a technical refinement of static greenness metrics; they mark a substantive shift toward mobility-aware and time-sensitive exposure science. GPS-based tracking, Ecological Momentary Assessment, activity-space methods, and time-weighted cumulative exposure metrics can reduce exposure misclassification relative to residential proxies and improve temporal alignment between environmental context and psychological-state measurement [76,157]. In parallel, machine and deep learning approaches—including street-view-based convolutional models and tree-based algorithms such as random forest (RF) and XGBoost—have improved the capacity to characterize nonlinear exposure patterns and support individualized exposure profiling [16,17,19]. However, greater technical sophistication does not in itself resolve causal ambiguity, sample-selectivity concerns, or generalizability limits. These methods are therefore best understood as tools that can strengthen inference when embedded in longitudinal, quasi-experimental, and theory-informed research designs rather than as substitutes for causal identification itself.

Third, the field has increasingly moved from analytical explanation toward planning support and scenario-based governance evaluation. Multi-criteria decision analysis helps structure trade-offs and distributional impacts when comparing intervention portfolios [166]; system dynamics represents feedbacks and delays in urban health–environment systems to test policy robustness [167]; and agent-based modeling explores heterogeneous behaviors and neighborhood effects under alternative scenarios [168]. From a temporal perspective, this broader methodological development can be grouped into three stages (Table 2). Stage I relied mainly on NDVI-based vegetation indicators and cross-sectional designs to establish macro-level associations between UGS–MH [155,156]. Stage II incorporated street-view greenness (GVI), accessibility metrics, and socio-psychological covariates, adopting SEM and multilevel models to test mediation or moderation and to identify pathways such as stress buffering, attention restoration, social cohesion, and physical activity [169]. Stage III integrates multimodal data—including remote sensing, street-view imagery, wearables, and social-media inputs—with dynamic exposure metrics such as Global Positioning System–Ecological Momentary Assessment (GPS–EMA) and time-weighted cumulative exposure/weighted cumulative exposure (TWCE/WCE), and increasingly employs machine and deep learning approaches, including random forests (RFs), XGBoost, convolutional neural networks (CNNs), and graph convolutional networks (GCNs), to support individualized exposure characterization, prediction, and more policy-relevant modeling [19,157,158].

Taken together, these developments indicate a shift from static, average-based association studies toward more spatially explicit, temporally aligned, and policy-relevant analytical frameworks. However, methodological advancement should not be equated with full inferential resolution; remaining challenges concerning causal validity, representativeness, and cross-context comparability are addressed further in Section 5.

5. Research Challenges and Methodological Reflections

This section synthesizes the major methodological bottlenecks in UGS–MH research, identifies their underlying causes, and discusses practical directions for improving causal interpretation, transparency, and policy relevance. Figure 6 provides a schematic summary of this logic, moving from key methodological challenges to reorientation strategies and future development pathways.

5.1. Root Causes of the Methodological Pitfalls

Over the past decade, UGS–MH research has expanded rapidly, but important forms of design–theory–data misalignment remain. Cross-sectional designs and single-shot exposure measures (such as a one-time NDVI) are ill-suited to long-term outcomes, yielding mostly associational findings; heterogeneity in spatial resolution, scales, and buffer choices undermines comparability; and misaligned timing between psychological assessments and environmental exposures further weakens inference. More fundamentally, many analytic pipelines under-integrate psychological and social theory, so that the cognition–behavior–environment mechanism is only partially specified.

In summary, the current methodological impasse is shaped by four interlocking challenges: exposure misclassification arising from UGCoP and spatiotemporal mismatch [170]; limited model interpretability and fairness, with social heterogeneity often underexamined [171,172]; weak standardization and reproducibility, including inconsistent protocols and scarce open materials [173,174]; and insufficient cross-disciplinary integration and policy translation, which constrains the formation of actionable guidance [175]. In addition, digitally mediated data sources—such as smartphones, wearables, street-view platforms, and social-media-derived signals—may systematically overrepresent younger, healthier, more connected, and higher-SES populations, thereby introducing sample-selectivity and representativeness concerns into ostensibly high-resolution analyses. Methodological progress in this field should therefore be judged not only by finer resolution or algorithmic complexity, but by gains in causal interpretability, representativeness, reproducibility, and policy relevance.

5.2. Spatial and Temporal Uncertainty

Urban residents move continuously through space, and their psychological states fluctuate over time, making the exposure–response relationship inherently uncertain. Even with GPS or wearable devices, differences in sampling frequency, buffer radii, and dwell-time rules can introduce systematic bias, while high-rise obstruction, signal loss, and varied daily activities further amplify measurement error. In parallel, psychological assessments often lag behind environmental exposure, so affective changes are not synchronized with the surrounding context, weakening the stability of dose–response estimation.

This combination of UGCoP and spatiotemporal mismatch is among the most underestimated error sources in the field [170,176]. For example, evidence from older adults in South Korea reports clear spatiotemporal variability in the greenspace–depression association, implying that effect estimates can shift across time windows and contexts and thus require time-aligned exposure definitions [77]. At the design stage, studies should specify spatiotemporal alignment rules in advance and use stratified or multilevel models, together with sensitivity analyses, to identify likely sources of error. Parameter settings and uncertainty bounds should be explicitly reported to enhance transparency and reproducibility. Along the exposure–response link, TWCE/WCE can improve temporal characterization and maintain consistency with mobility contexts [157]. These issues illustrate why the field is moving from static proxy-based exposure measures toward more temporally aligned and mobility-sensitive exposure modeling.

5.3. The Paradox of Intelligent Analytics

The introduction of AI has widened the research frontier: deep learning approaches can automatically extract green-space features from street-view and remote-sensing imagery, generating high-dimensional exposure data with finer spatial precision and stronger predictive performance. However, when predictive accuracy becomes the overriding objective, explanatory clarity may be weakened; models may become more accurate without clarifying why associations arise, thereby limiting their value for causal interpretation.

A further concern is algorithmic bias stemming from imbalanced samples: data-rich central districts and higher-income groups tend to be overrepresented, whereas the exposures and psychological responses of vulnerable populations are undercounted [171,172]. To mitigate these risks, studies should balance interpretability and fairness: incorporate explainable AI techniques such as Shapley Additive explanations and Local Interpretable Model-agnostic Explanations to enhance transparency [172,177], and routinely report fairness audits and ethical-governance assessments [171]. Going forward, closer integration of theory-driven and data-driven approaches will be essential if AI-assisted methods are to contribute not only to prediction, but also to causal interpretation and policy-relevant inference.

5.4. Pathways Toward Integration and Standardization

Further maturation of the field depends on stronger data standardization, open-science practices, cross-disciplinary integration, and clearer policy translation [173,174]. Accordingly, future development should advance along three complementary directions. First, multimodal data should be integrated within causal-inference and explainable-AI frameworks so that psychological, social, and spatial processes can be modeled jointly, thereby supporting more realistic multilevel socio-ecological representations and stronger causal interpretation [178]. Second, standardize and share data and algorithms by adopting principles that ensure information is findable, accessible, interoperable, and reusable, and by establishing a minimum-information protocol for reproducible reporting in UGS–MH research. Such a protocol should specify data sources, parameter settings, and algorithm versions, enabling open verification and continual updating [173,174]. Third, strengthen cross-disciplinary collaboration and policy alignment to channel evidence from health, planning, and ecology into decision-making processes, forming a durable feedback loop between evidence and decisions [175].

Together, these directions indicate that future methodological progress should prioritize integration, transparency, and inferential discipline over technical expansion alone. More fundamentally, the field is moving from asking whether UGS and mental health are associated with asking under what conditions, through which pathways, and with what planning relevance such associations can be interpreted and acted upon.

5.5. Areas of Inconsistency and Unresolved Debate

Although most studies report beneficial associations in the UGS–MH literature, a smaller number report null, weak, or even adverse associations. These inconsistencies may reflect poor green-space quality, safety concerns, residual confounding by neighborhood deprivation, reverse causality, or mismatches between greenness quantity and actual usability. In addition, several substantive questions remain unresolved, including whether benefits follow nonlinear or threshold-based dose–response patterns, and whether highly designed or artificial green-spaces produce effects comparable to those of more naturalistic environments. These unresolved issues underscore the need for longer-term, context-sensitive, and cross-regional research capable of distinguishing between apparent inconsistency and genuine heterogeneity.

6. Strategic Pathways for Green Interventions and Policy Implications

The governance implications discussed in this review are not intended as generic policy recommendations, but as strategic directions derived from the reviewed mechanism and evidence base. More specifically, restorative pathways support design strategies that enhance psychological recovery; stress-buffering evidence supports neighborhood greening and everyday exposure opportunities; social-cohesion findings support community-oriented green interventions; and subgroup/equity evidence supports priority targeting toward populations with lower effective access and higher vulnerability. Governance translation should therefore be understood as evidence-informed, mechanism-aligned, and context-sensitive rather than universally prescriptive.

Table 4. From mechanism and evidence to governance translation in UGS–MH research.

Mechanistic/Empirical Basis	Methodological Support	Governance Implication
Daily accessibility and stress buffering	Dynamic exposure and local-access evidence	Pocket parks, walkable greening, accessibility-oriented provision
Restorative quality and therapeutic design	Therapeutic-landscape and intervention evidence	Psychologically informed green-space design
Vulnerability, inequity, and uneven benefit	Subgroup, equity, and GSP evidence	Priority targeting for high-need populations
Cross-sector dependence on intervention effects	Socio-ecological and multi-actor governance evidence	Integrated governance across planning, health, and community systems
Precision targeting with ethical constraints	AI, IoT, and spatiotemporal modeling evidence	Adaptive intervention with transparency and fairness safeguards

makes this evidence-to-governance logic explicit by summarizing how major mechanistic and empirical insights map onto methodological support and governance implications.

6.1. Spatially Equitable Provision

From a public-health perspective, a primary goal of green interventions is to improve spatial equity in mental-health-relevant exposure. Simply increasing total green area does not guarantee population-wide mental health benefits; what matters is accessibility and balanced distribution [163]. To avoid overestimating benefits based solely on nominal residence-based accessibility, equity metrics should be aligned with observed or behaviorally relevant exposure.

Higher-resolution exposure assessment can improve equity-oriented planning by identifying where nominal provision diverges from actual use and mental-health-relevant exposure [157]. In this sense, the policy focus shifts from simple quantity expansion toward accessibility, usability, and quality optimization.

6.2. Therapeutic Design Functions

Conventional UGS design has prioritized landscape esthetics and ecological services while giving less systematic attention to psychological therapeutic effects. The Therapeutic Landscape framework foregrounds multisensory experience, spatial privacy, and cultural symbolism, enabling psychological restoration through cognitive recovery and affect regulation [179]. This orientation is consistent with core psychological mechanisms—particularly attention restoration and stress recovery—that link environmental qualities to mental health outcomes [21].

Across East Asia and Europe, structured nature-based programs—such as forest therapy and sensory gardens—have demonstrably reduced anxiety, improved sleep, and enhanced well-being [180]. Together, these findings indicate that ecological remediation and psychological therapy are not parallel tracks but mutually reinforcing design logics. These findings suggest that green-space planning should more explicitly incorporate psychological principles into design, implementation, and maintenance, and should move toward evidence-based therapeutic design standards.

6.3. Prioritizing Vulnerable Populations

Equity in UGS–MH research should not be treated merely as a general normative concern. Unequal accessibility, variable usability, and differential vulnerability may generate uneven mental-health benefits across children, older adults, low-income communities, and other socially marginalized groups. Priority setting in green interventions should therefore be guided not only by where green-space is lacking, but also by where mental-health need and structural disadvantage are concentrated. In this sense, equity-oriented governance is not only a matter of fairness, but also a strategy for improving the efficiency and public-health relevance of UGS–MH interventions under conditions of uneven exposure, uneven usability, and uneven benefit realization.

One operational pathway for translating this logic into practice is green social prescribing (GSP), which links health and community-service systems with structured nature-based activities for populations at elevated risk of mental-health burden or social isolation [58]. In this way, GSP helps translate equity goals into practical intervention delivery, increasing the likelihood that vulnerable groups can access, use, and benefit from green interventions.

Several countries have begun to institutionalize this approach. In the UK NHS, a dual model combining physician referral with community-based nature activities has reached more than 8500 participants, while national “test-and-learn” pilots have been formally evaluated and linked to improved mental health outcomes and reduced health inequalities [161,181]. Comparable models are also being scaled in Australia and explored in East Asian cities within broader social-prescribing and community-health systems [182,183] (

Table 5. Selected international cases of green interventions *.

No.	Country/Region	Core Features	Target Outcomes
(P1) Spatial equity	US; China; EU	Pocket parks; walkable green corridors; rooftop greening; high-resolution exposure assessment	Improved accessibility; reduced spatial disparities; more equitable mental-health benefits
(P2) Therapeutic design	Japan; Korea; EU	Diverse tree species; immersive sensory experience; zoned healing spaces	Stress reduction; better sleep; enhanced well-being
(P3) Social prescribing	UK (NHS)	Healthcare referral pathways; link-worker model; community nature activities	Improved mental health; reduced inequalities
(P4) Collaborative governance	Australia; China	NGO–health department collaboration; community–health service integration	Stronger social support; enhanced community cohesion
(P5) Smart green infrastructure	Korea	AI- and IoT-enabled green monitoring; spatiotemporal hotspot detection; intervention prioritization	Targeted interventions; optimized policy design

* Organized around five pathways—(P1) spatial equity, (P2) therapeutic design, (P3) social prescribing, (P4) collaborative governance, and (P5) smart green infrastructure—this table presents illustrative rather than exhaustive international examples consistent with the main text.

).

6.4. Integrating Governance Systems

The effectiveness of green interventions depends on cross-sector coordination and institutional integration. The EU’s Nature-based Solutions strategy, for example, emphasizes embedding health and mental-health indicators within a coordinated architecture spanning urban planning, education, and ecological management [184]. At the conceptual level, the Social-Ecological Systems framework provides a basis for linking multi-level actors, infrastructures, and feedbacks [185]. Pilot programs in Australia and selected Chinese cities further suggest that collaboration among non-governmental organizations, health agencies, and community organizations can generate measurable social and psychological benefits (Figure 7; Table 5).

6.5. Intelligent Intervention Tools

The convergence of AI and the Internet of Things (IoT) is reshaping the technical landscape of urban green interventions. Smart Green Infrastructure integrates environmental sensing, behavioral data, and algorithmic modeling to identify high-stress micro-areas in real time and support targeted recommendations, forming a more adaptive exposure–response–feedback governance loop [186]. For example, several cities have piloted AI + IoT green-space monitoring by linking air/noise/heat sensors with mobility and wearable data and using spatiotemporal modeling for hotspot detection and intervention prioritization [19].

At the same time, the governance use of intelligent tools requires safeguards for transparency, fairness, and ethical oversight, so that technological innovation strengthens rather than undermines equitable mental-health promotion [171,177]. Taken together, these strategic pathways suggest that the value of UGS–MH evidence lies not only in documenting beneficial associations, but in translating pathway-specific evidence into differentiated, equity-sensitive, and operationally feasible urban interventions.

7. Conclusions and Scholarly Outlook

This review synthesizes more than a decade of research on UGS and mental health and shows that the field has moved from broad eco-environmental associations toward more explicit attention to mechanisms, heterogeneity, and health equity. Using a dual-dataset design, we linked bibliometric mapping with evidence synthesis and organized the literature within an integrative framework connecting mechanistic interpretation, methodological development, and governance translation.

Across the reviewed literature, methodological development has shifted from static, average-based exposure assessment toward more dynamic, mobility-aware, and spatially explicit approaches, and from single-pathway interpretation toward multi-level cognitive, affective, physiological, and social explanations. At the same time, more policy-relevant and individualized analytical strategies are emerging. However, greater technical sophistication should not be equated with full inferential resolution: stronger causal interpretation still depends on theory-informed design, temporal alignment, longitudinal or quasi-experimental evidence, and routine attention to interpretability, fairness, and reproducibility [87,187].

The central contribution of this review is therefore not simply to confirm that UGS matters for mental health, but to clarify how this relationship can be interpreted across mechanisms, methods, equity conditions, and governance contexts. At the same time, the resulting framework should be understood as an integrative analytical scaffold rather than a universally invariant template: its explanatory strength is likely to vary across urban forms, cultural settings, green-space types, and levels of methodological rigor.

Important limitations remain. The current evidence base is geographically uneven, overweights English-language outputs, and is still dominated in many areas by cross-sectional or otherwise observational designs [188]. Alignment between space–time exposure measures and mental-health assessment also remains inconsistent, and cross-context comparability is still limited. Future work should therefore strengthen multimodal and mobility-aware exposure assessment, causal-identification strategies, reproducible and fairness-aware analytic practice, and cross-cultural validation. More broadly, the value of UGS–MH research lies not only in documenting whether green-space matters, but in clarifying for whom, through which pathways, under what conditions, and with what governance implications it matters most.