The relationship between greenspace and the mental wellbeing of adults: A systematic review

Introduction The view that interacting with nature enhances mental wellbeing is commonplace, despite a dearth of evidence or even agreed definitions of ‘nature’. The aim of this review was to systematically appraise the evidence for associations between greenspace and mental wellbeing, stratified by the different ways in which greenspace has been conceptualised in quantitative research. Methods We undertook a comprehensive database search and thorough screening of articles which included a measure of greenspace and validated mental wellbeing tool, to capture aspects of hedonic and/or eudaimonic wellbeing. Quality and risk of bias in research were assessed to create grades of evidence. We undertook detailed narrative synthesis of the 50 studies which met the review inclusion criteria, as methodological heterogeneity precluded meta-analysis. Results Results of a quality assessment and narrative synthesis suggest associations between different greenspace characteristics and mental wellbeing. We identified six ways in which greenspace was conceptualised and measured: (i) amount of local-area greenspace; (ii) greenspace type; (iii) visits to greenspace; (iv) views of greenspace; (v) greenspace accessibility; and (vi) self-reported connection to nature. There was adequate evidence for associations between the amount of local-area greenspace and life satisfaction (hedonic wellbeing), but not personal flourishing (eudaimonic wellbeing). Evidence for associations between mental wellbeing and visits to greenspace, accessibility, and types of greenspace was limited. There was inadequate evidence for associations with views of greenspace and connectedness to nature. Several studies reported variation in associations between greenspace and wellbeing by life course stage, gender, levels of physically activity or attitudes to nature. Conclusions Greenspace has positive associations with mental wellbeing (particularly hedonic wellbeing), but the evidence is not currently sufficient or specific enough to guide planning decisions. Further studies are needed, based on dynamic measures of greenspace, reflecting access and uses of greenspace, and measures of both eudaimonic and hedonic mental wellbeing.


Introduction
The view that interacting with nature enhances mental wellbeing is commonplace, despite a dearth of evidence or even agreed definitions of 'nature'. The aim of this review was to systematically appraise the evidence for associations between greenspace and mental wellbeing, stratified by the different ways in which greenspace has been conceptualised in quantitative research.

Methods
We undertook a comprehensive database search and thorough screening of articles which included a measure of greenspace and validated mental wellbeing tool, to capture aspects of hedonic and/or eudaimonic wellbeing. Quality and risk of bias in research were assessed to create grades of evidence. We undertook detailed narrative synthesis of the 50 studies which met the review inclusion criteria, as methodological heterogeneity precluded metaanalysis.

Results
Results of a quality assessment and narrative synthesis suggest associations between different greenspace characteristics and mental wellbeing. We identified six ways in which greenspace was conceptualised and measured: (i) amount of local-area greenspace; (ii) greenspace type; (iii) visits to greenspace; (iv) views of greenspace; (v) greenspace accessibility; and (vi) self-reported connection to nature. There was adequate evidence for associations between the amount of local-area greenspace and life satisfaction (hedonic wellbeing), but not personal flourishing (eudaimonic wellbeing). Evidence for associations between mental wellbeing and visits to greenspace, accessibility, and types of greenspace was limited. There was inadequate evidence for associations with views of greenspace and PLOS

Background
Urbanisation is increasing at an unprecedented rate, and with over half the world's population now residing in cities [1], many people may not have access to the green landscapes in which the human species evolved [2,3]. Greenspace may provide human benefits, such as facilitating exercise, social activities and connecting with nature [4], and it is suggested that urban greenspaces are critical to healthy living, both physically [5,6] and mentally [7,8]. There may also be salutogenic effects on mental health and wellbeing, such as increased attention, feelings of happiness and reduced stress [9,10]. The United Nations Sustainable Development Goals emphasise the importance of greenspace provision "to foster prosperity and quality of life for all" [11]. The World Health Organisation stated that urban greenspaces (including parks, woodlands, and sports facilities) are a "necessary component for delivering healthy, sustainable, liveable conditions" [12], while highlighting the dearth of evidence to support planning advice [12]. In the UK, local authorities are responsible for providing access to the natural environment [13], and guidelines recommend that all residents should live within 300m of at least 2 hectares of greenspace [14,15], despite limited evidence for the wellbeing benefits of these recommendations.

Measuring greenspace
One of the reasons for this dearth of evidence is the lack of consensus regarding the definition of the terms 'nature' and 'natural' [10,16], and features that may appear 'natural' are often artificially constructed [8]. Hartig et al. provide the most detailed definition of nature, as the "physical features and processes of nonhuman origin. . ., the 'living nature' of flora and fauna" [8].
Furthermore, 'nature' and 'greenspace' are often used interchangeably [17][18][19][20][21]; 'greenspace' is more inclusive, referring to areas of grass, trees or other vegetation [22], and can be used to describe both surrounding greenness in the countryside, and spaces managed or reserved in urban environments [14]. Greenspace was therefore chosen as the focus of this review. We chose not to include studies of water (blue space), as this is generally considered separately to greenspace [5,[23][24][25].
of shortlisted studies retrieved for formal inclusion/exclusion. It was agreed that any disputed studies would be cautiously retained for full text evaluation.

Study eligibility criteria
Criteria for inclusion were: (a) Population: adults aged over 16 (or all ages, but not wholly or mainly children); (b) Exposure: any measure of greenspace, defined as areas of grass, trees or other vegetation. Studies measuring personal connectedness to nature were included. As we were interested in all greenspace characteristics, we included both urban and rural studies; (c) Control: Comparators must include a control group which differed in the type/degree of exposure to greenspace, or direct comparison before and after an intervention; (d) Outcome: mental wellbeing, ascertained using a validated measure of hedonic and/or eudaimonic mental wellbeing, or one or more aspects of these (e.g. life satisfaction, happiness, quality of life. The General Health Questionnaire (GHQ) is designed to measure psychological distress, but includes several positive items, and is prevalent in the literature; studies using this outcome were therefore included. Instruments designed to capture only symptoms of mental distress were not included; (e) No study designs were explicitly excluded.

Evaluation of evidence
After identifying eligible papers, one reviewer (VH) evaluated study contents by extracting: authors, publication date, country, study design, age of participants, sample size, greenspace measures, methods, outcomes, confounders, and a results summary, including effect sizes (regression coefficients/risk ratio and confidence interval/standard error).
For quality appraisal, risk of bias was assessed using Cochrane-recommended criteria [32]: the Newcastle-Ottawa Scale (NOS), adapted for longitudinal and cross-sectional studies, alongside the Cochrane Risk of Bias (RoB) tool for controlled studies [34,35]. The criteria cover potential risk of bias arising from: representativeness of the sample, participant awareness of the intervention, control factors, and selection of reported results.
We used established Quality Assessment thresholds to categorise each article [36]. For those assessed using the Cochrane RoB tool, a Good quality study met all criteria (low RoB), while those of Fair quality had moderate RoB not meeting one criterion; Poor quality studies

Stratification by characterisation of greenspace
We identified six types of study, according to the characterisation of greenspace: (a) amount of local-area greenspace, most commonly the proportion of local areas covered by greenspace; (b) greenspace type; (c) views of greenspace; (d) visits to greenspace; (e) accessibility, in terms of proximity to greenspaces and self-reported 'access'; and (f) subjective connection to nature. We conducted a narrative review of evidence, as methodological heterogeneity precluded meta-analysis. Evidence for associations between each type of greenspace characteristic and mental wellbeing was classified according to the consistency, strength and methodological quality of the findings, and study design. Evidence of association was categorised using established guidelines used by other studies in the field [37]: Adequate (most studies, at least one Good quality, reported an association between greenspace and mental wellbeing); Limited (more than one study, at least one Good, reported an association, but with inconsistent findings); Inadequate (associations reported in one or more studies, but none Good quality); and No association (several Good quality studies reported an absence of a statistically significant association between greenspace and mental wellbeing).

Results
Titles and abstracts of 485 records were screened, and 75 chosen for full-text evaluation; 42 were found to be eligible. During this process, 10 additional papers were found via Auto-Searching the databases and recommendations. Therefore, 52 papers were finally included in this review (Fig 1).
Among these, 4 were controlled case studies and a further 6 were longitudinal cohort studies; there was one ecological analysis, 4 uncontrolled case studies, the remaining 37 were crosssectional surveys. Two studies were international, 31 were restricted to Europe, 15 just in the UK; 5 were based in the USA with another 6 in Canada, 10 in Australia. Analyses were confined to urban areas in 22 cases, 9 included only rural greenspace. Sample size ranged from 25 to 30,900 participants, but was not specified in 3 cases. Age ranges were fairly consistent, covering young adults to past retirement age, although 1 focused on 'youths' (aged [16][17][18][19][20][21][22][23][24][25], 3 studies recruited university students and two included mainly people aged over 55; however, 11 studies did not specify participants' age. After quality assessment, the majority of studies (n = 27) were determined to be Good, 13 were Fair, and 12 Poor. For Poor studies, Table 2 provides further justification. For full details of the risk of bias for each study, heat maps are presented in S1 and S2 Tables. Table 3 provides further detail on the typologies of greenspace measures implemented for each study.

Mental wellbeing measures
Only 14 studies were found to measure both hedonic and eudaimonic mental wellbeing, of which the most commonly used measure was the Warwick-Edinburgh Mental Well-Being Scale (WEMWBS) [19,46,52,53,61,66]. WEMWBS includes 14 positively worded questions, regarding individual feelings over the past 2 weeks, including "feeling relaxed", "interested in new things", and "close to others" [81]; there is also a reduced 7-item version, known as SWEMWBS (Shortened-WEMWBS) [82]. The recent Personal Wellbeing ONS4, applied in to one study [72], measures individuals' life satisfaction, happiness and anxiety (hedonic wellbeing) and sense of worth (eudaimonic wellbeing) [83].
The remaining 32 studies assessed outcomes considered to be aspects of mental wellbeing, such as quality of life, life satisfaction, and affect, but did not report both hedonic and eudaimonic wellbeing. The WHO-5 Well-Being Index, used in 2 studies [48,70], asks how frequently individuals have felt "cheerful and in good spirits" and "calm and relaxed", over the previous 2 weeks, but focusses on hedonic rather than eudaimonic wellbeing [84].
Eight studies reported affect scores [62, 63, 66-69, 78, 79], which include positive feelings (happiness, interest), and negative emotions (anger, sadness), using the 20-item Positive and Negative Affect Scale (PANAS) [90]. Similarly, The Profile of Mood States (POMS) asks about experiences of 65 different emotions, including some positive items, such as "lively" and "relaxed" [91], and was used in one study [20]. The General Health Questionnaire (GHQ) was used in 14 studies [5, 17, 19, 23-25, 38, 45, 47, 49, 54, 58, 60, 73]. It contains some positively worded items ("In the last 2 weeks I have. . . been able to concentrate", "felt I have been playing a useful part" and "feeling reasonably happy") but was designed and validated as a screening tool for psychiatric disorders, with higher scores indicative of greater distress [92]. Other studies which measured on poor mental health were excluded from this review.
Full details of the included studies are presented in Table 2, which is ordered by greenspace characteristic. Where articles cover multiple characteristics, the study appears under multiple headings. The relationship between greenspace and mental wellbeing

Greenspace characteristics
Amount of local-area greenspace. 21 studies examined associations between quantities of local-area greenspace and mental wellbeing, 2 of which were longitudinal. Most calculated the proportion of greenspace for each Lower-Layer Super Output Area (LSOA, a geographic area generated for being as consistent in population size as possible, with a minimum population of 1000 and the mean of 1500), Census District (CD), or within a defined radius of residents. Two articles measured greenspace area per capita. Of 15 studies, one was restricted to public greenspace [39], and 14 included only urban areas.
Only four (cross-sectional) studies measured hedonic and eudaimonic mental wellbeing (SWEMWBS and ONS4). No statistically significant association was reported between greenspace and mental wellbeing in three studies [46,52,72], although urban residents who reported "sufficient local greenspace" showed significantly higher SWEMWBS scores [52]. However, Wood et al.'s study found that a 1ha increase in park area within a 1.6km walk of an individual's home showed a 0.070-point increases in SWEMWBS score [53]; this suggests that examining greenspace around individuals, rather than aggregating to local area, may better detect associations.
Five studies, 4 of which were Good quality and based in urban areas, found that life satisfaction was significantly higher in areas with more greenspace [24,39,41,51], albeit with small linear effect sizes of 0.002-0.003 [24,39]. The study by White et al. included a large sample, over 10,000 individuals, demonstrating a slight but significant association between LSOA greenspace proportions and life satisfaction. Another large study by the same authors found no significant association between mental wellbeing and the amount of rural local-area greenspace [25], suggesting that associations may differ between urban and rural environments.
An ecological analysis of over 172 countries measured the amount of green land cover per km 2 , adjusted for the nation's size, finding a significant association with better life satisfaction. Despite the large sample size and strong odds ratios (2.450), the study was of poor methodological quality, due to its ecological design and hence inability to adjust for individual-level confounding [51]. Four studies also found the quantity of urban greenspace was associated with quality of life or mental health, characterised by the SF-36 scale and its sub-components [41][42][43]74]; however, three others, which included only public urban greenspace, found no association [39,44,50]. Taylor et al. observed mixed results: the amount of urban greenspace was positively and significantly associated with hedonic wellbeing for two cities in Australia, but not two others in New Zealand [48].
Based on these Good quality studies, we conclude that there is adequate evidence for an association between local-area urban greenspace and life satisfaction, but not rural greenspace. Mixed results provide inadequate evidence for associations with quality of life, mental health, and multi-dimensional mental wellbeing.
GHQ was the outcome in 8 studies, of which 6 were Good quality and 3 were confined to urban areas. All but one [45] found an inverse association between the amount of greenspace and GHQ score [5, 17, 23-25, 47, 49, 50], implying reduced mental distress; again, linear regression coefficients varied considerably, from 0.003 to 0.431. The Fair quality study by Dzhambov et al., however, found no statistically significant association for objective greenspace quantities, but observed significantly lower GHQ scores for those with higher perceived greenness in their neighbourhood [45]. In a longitudinal study, Alcock et al. found that people moving to areas with higher greenspace proportions had significantly lower GHQ score after relocating, averaging 0.430 points lower 3 years post-move [23]. Therefore, there was adequate evidence for the inverse association between the amount of local-area greenspace and (lower) GHQ score.
Greenspace types. A total of 8 Good and 2 Fair quality studies classified greenspace according to greenspace types, using bespoke classification systems; no consensus was observed regarding greenspace typology. Four of these were longitudinal studies.
Only one Fair study measured hedonic and eudaimonic wellbeing, with WEMWBS, comparing linear associations between the amount of sport, recreational and 'natural' spaces within a 1.6km buffer of the individual [53]. The strongest associations were observed for sports (0.430 increase in WEMWBS for each additional space), followed by recreational and natural spaces (0.110 each).
One research group conducted four studies (3 longitudinal) using the longitudinal Swedish Health Survey (SHS), based in suburban and rural areas. They classified public greenspace within 300m of each residents' home into 5 aspects: Serene (quiet, audible 'nature'), Wild (undeveloped, no visible human impact), Lush (biodiversity), Spacious (large cohesive area) and Cultural (cultural heritage, old trees) [18,54,58]. Two studies measured GHQ: the first found associations between Serene or Spacious greenspace and slightly, but significantly, lower GHQ scores for physically active individuals; however, associations with Spacious greenspace held only for women [54]. In the second, only women moving to areas with Serene greenspace had significantly lowered GHQ scores, but with much higher odds than in Annerstedt et al.'s work [58]. In a cross-sectional analysis, these authors found that the total number of green aspects (Serene, Wild, Lush, Spacious, Cultural) was associated with slightly better SF-36 Vitality scores for women [18]. The third longitudinal study found marginally but significantly lower GHQ scores for greater numbers of different green aspects, including those moving between areas [60].
In a cross-sectional study, based on 12,697 observations from 2,020 residents of rural England, no association was found between LSOA land cover classes and GHQ scores. However, individuals who relocated to areas with more arable, grass, 'natural', mountainous and heath land had significantly lower GHQ scores post-move [38]. Among 3 cross-sectional studies, urban residents with higher amounts of local vegetative or 'natural' greenspaces reported better mental wellbeing: vegetation density and cover, from field surveys and satellite imagery in Australia, were strongly and significantly associated with life satisfaction [55]. The number of trees, or an indicator of how 'green' the neighbourhood is, were significantly associated with better mental health (SF-36 Mental Component) and life satisfaction [57,59]. Residents' ratings of the 'quality of their local natural environment', on a scale of 0-10 (very dissatisfied to very satisfied), was associated with higher SF-36 Mental Component Summary scores [59].
A large cross-sectional study in the UK used app data on users' self-reported feelings, while their phones' GPS linked their location to a land-cover database; this novel study therefore benefits from measuring happiness in situ. Being in mountainous, woodland or 'semi-natural grassland' areas, as opposed to urban, was associated with approximately 2-points higher happiness, on a scale of 0-10, although no additional factors were controlled for [56].
While most of these studies were Good quality, interpretation is difficult due to lack of consensus in greenspace classification; in addition, four reports were based on data from the same survey. All but one were restricted to either urban or rural areas, so comparisons between these environments is not possible; however, larger effect sizes were observed in rural studies. Two of the Swedish studies concluded that green aspects were associated with lower GHQ scores for women, while 6 others highlighted that Serene (quiet, 'natural') and 'natural' rural greenspaces were associated with improved life satisfaction, SF-36 and lower GHQ scores, although none defined the term 'natural'. Additionally, two studies reported an association between subjective perceptions of local greenspace and mental wellbeing. Evidence is therefore limited.
Visits to greenspace. Seventeen papers reported studies of visits, either comparing mental wellbeing scores before and after an intervention (n = 7), or testing cross-sectional associations with greenspace visiting patterns (n = 10).
Fair quality studies compared happiness and positive affect for those walking in 'natural' versus indoor environments [69], and walks in urban versus green areas [66]. The former reported a statistically significant difference in favour of greenspace walking, the latter did not. In a further Fair quality cross-sectional study, Marselle et al. reported a positive association between perceived restorativeness of the walking environment and positive affect and happiness [67].
Duvall and Kaplan observed 73 individuals on a wilderness expedition; attention and affect were improved post-expedition, persisting for 3-4 weeks [62]. Although effects were quite large (score changes of 0.270 to 0.340), participants were not blind to the intervention. A Fair quality uncontrolled study encouraged individuals to engage with 'nature' for 30 days by noticing/protecting wildlife, sharing experiences, or connecting with 'nature'. Participants reported greater happiness following the programme [71]. Similarly, Molsher and Townsend noted mental wellbeing improvements following engagement with environmental volunteering projects [68], although their study displayed high risk of bias. Jakubec, however, reported no association between visiting greenspaces and Quality of Life Inventory score, in a Poor quality study [64].
A further 10 cross-sectional studies of varying quality examined self-reported greenspace visit frequency. Three studies measured both hedonic and eudaimonic wellbeing, with mixed findings [61,63,72]. In the first Fair study, university students who claimed greater typical contact with nature reported better mental wellbeing using Ryff's Scale of Psychological Wellbeing [63,93]. These findings were not replicated in a Good study by Glichrist et al., who examined associations between mental wellbeing (SWEMWBS) and greenspaces surrounding workplaces in Scotland [61]. White et al.'s Good study, measuring ONS4, found that those spending time outdoors and in nature every day, compared to never, had strong odds (OR 1.960) of a high sense of worth, the effect size decreasing with visit frequency. No associations were detected for visit frequency and hedonic wellbeing, although those reporting visiting greenspace the previous day had higher happiness scores, with no associations for life satisfaction, anxiety or worth [72].
A further 5 studies, one of which was Poor, showed that quality of life and mental health were improved, and GHQ scores reduced, with the number of greenspace visits [19,21,50,52,65]; Triguero-Mas et al. also noted that associations with mental health were stronger for males than females [50] In a Good study, Mitchell found that those who more regularly visited a local park had lower GHQ scores [19]. However, although Panno et al. observed that greater frequency of greenspace visits was associated with higher hedonic wellbeing, these results were not statistically significant [70], and Dzhambov et al. found no association between time spent in greenspace and GHQ [45].
Due to the mixed quality and inconsistent results, evidence for an association between greenspace visit frequency and mental wellbeing is considered limited.
Views of greenspace. Association between views of greenspace and mental wellbeing was reported in 3 papers. Gilchrist et al.'s Good quality study found that workers' satisfaction with their office views, particularly of trees, lawns and flowering plants, was associated with improved mental wellbeing (SWEMWBS) scores [61]. Similarly, urban residents reporting greater visibility of trees from their home had slightly better life satisfaction [59]. Pretty et al. observed increases in self-esteem for those viewing rural pleasant scenes, while both unpleasant urban and rural scenes could be detrimental; however, they did not control for potentially confounding factors [20]. The mixed quality and small study sample leads us to classify the evidence here as inadequate.
Greenspace accessibility. We identified 8 cross-sectional studies, mostly Good quality, which tested associations between greenspace accessibility and mental wellbeing. Two studies measured mental health using the SF-12 Mental Component, with significant positive findings [57,74]. In the first, a weak association was found with Euclidean (direct) distance from homes to the nearest public greenspace [74]. In the second, Sugiyama et al. used the Neighbourhood Environment Walkability Scale, which measures selfreported greenspace access. Access to the highest of levels of greenspace (perceived neighbourhood greenness, terciles) was associated with strong odds (OR 1.270) of better mental health [57].
Only one, Fair study compared public greenspace within different Euclidean buffers around individuals' postcodes [49]. Triguero-Mas et al. found greater amounts of greenspace within 300m were significantly associated with reduced risk of high GHQ scores (dichotomised around 3), with consistent results for control buffers of 100m, 500m, and 1km [49]. Bos et al. found that greenspace within 3km, but not 1km, of homes was significantly associated with greater quality of life [43], although this study was rated as Poor study because of limited statistical reporting. In a larger study, scores on the SF-36 Vitality scale were associated with rural greenspace, but this was only significant for women and within 300m (but not 100m), of their home [18]. Maas et al.'s large cross-sectional study showed that those with more greenspace within 1km, but not 3km, had slightly lower GHQ scores, contrary to findings by Bos et al. [43,47]. Dadvand et al. also measured GHQ (dichotomised around 3), finding strong odds of low GHQ scores for the amount of greenspace within 100m of homes (OR 1.320), effect sizes reducing with distance (OR 1.250 for 250m, 1.170 for 500m); stronger associations were also noted for subjective, than objective, proximity to greenspace, measured as self-report and calculated presence of a greenspace within a 10-minute walk [73]. Dzhambov et al. also found a significant association between subjective accessibility (time to walk to nearest greenspace) and lower GHQ, although associations for objectively measured Euclidean distance were not statistically significant [45].
Although several of these studies reported an association between greenspace accessibility and aspects of mental wellbeing, different measures of both were used and findings were inconsistent, providing limited evidence of an association.
Subjective connectedness to nature. We identified 7 cross-sectional studies examining associations between subjective connection to nature and mental wellbeing. The Connectedness to Nature Scale measures the extent to which individuals 'feel nature is part of their identity', with particular emphasis on sense of care for nature; this has been linked to the theory of biophilia: that humans possess an innate desire to affiliate with other forms of life [3,69]. Of these studies, 5 were of Poor quality, with no controls for potential confounding. Four studies demonstrated that self-reported 'connection to nature' was positively associated with mental wellbeing [76][77][78][79]. Effect sizes were moderate and consistent across the studies, although lower methodological quality means their results have limited generalisability; only one was of Good quality, and adjusted for potentially confounding factors. Similarly, meaning in life, quality of life, happiness and affect were higher for those who reported greater connection to nature [65,75,76,79]. Life satisfaction was also positively related to nature connectedness in two studies [79,80], with moderate effect sizes, although Zhang et al. revealed that the association only held for those who actively engaged with nature [80]. While consistent in their findings, poor study quality means that the evidence is inadequate.

Summary of findings
While both the World Health Organisation and United Nations agree that greenspace is vital for healthy, liveable environments [11,12], it remains unclear which amounts, types and uses of greenspace are most beneficial to mental wellbeing. Previous reviews have focussed on associations between greenspace (or nature) and general health or mental distress [7,8,12,31], but we are not aware of any previous systematic reviews of published evidence specifically for associations between greenspace and validated, positive measures of mental wellbeing in adults. Even after stratifying our review according to the six main ways in which greenspace was conceptualised and measured, methodological heterogeneity precluded meta-analysis. We therefore undertook a narrative synthesis.
The largest number of studies were concerned with the amount of local-area greenspace, although few used detailed hedonic and eudaimonic wellbeing measures. Consistent results revealed adequate evidence for an association between urban local-area greenspace and life satisfaction. This result did not hold for rural greenspace, however. There was also adequate evidence for an association between local-area greenspace and lower GHQ scores.
Inconsistencies in the categorisation of greenspace types, and dearth of definitions, made it difficult to synthesise results; limited evidence was therefore found for associations between mental wellbeing and variety and 'nature' in land cover. Evidence was similarly limited for greenspace accessibility, with results generally concluding that nearer greenspace has the strongest associations, but with results differing according to the mental wellbeing measure; limited evidence was also found for associations between greenspace visits and mental wellbeing.
However, while there was some evidence for an association between mental wellbeing and views of greenery and connectedness to nature, this was considered inadequate, due to the mixed quality and small sample sizes of studies. Table 4 provides full details of the evidence summary and implications for research and policy.

Mental wellbeing measures
Only 14 of the 52 studies used a measure of mental wellbeing that captured both hedonic and eudaimonic dimensions, while others measured aspects such as life satisfaction, happiness and quality of life. GHQ, which was designed as a psychiatric screening tool, was included as a prevalent surrogate in the literature, which includes some positive items. Papers using other psychiatric screening tools were excluded if they covered only symptoms, ie mental distress.

Greenspace definitions and indicators
We identified 6 types of assessment in greenspace studies: amount of local-area greenspace, greenspace types, visits to greenspace, views of greenspace, greenspace accessibility and selfreported connection to nature.
The amount of local-area greenspace was most commonly measured as the proportion of greenspace in a resident's local area, or more specifically within a set radius of participants' homes. Most of these studies were restricted to urban areas. Most researchers quantified greenspace objectively, while a small number of studies reported associations with perceptions of the adequacy of the amount of local greenspace provision. All studies used either linear or logistic regression, which may overestimate associations in spatial data. Although a number of studies examined different types of greenspace, no consensus was observed for a typology, and as such conflicting results were observed.  One of the UN's Sustainable Development Goals is to "provide universal access to. . .green and public spaces" [11]; most studies assessed accessibility by distance to greenspace. While the EU and UK recommend that individuals should have access to a greenspace within 300m of their home [14,94], only one study conducted sensitivity analysis to test this guideline [49]; no difference in associations was observed for buffers of 100m, 300m, 500m and 1km. One study used buffer radii of 100m and 300m, reporting a significant association between the latter and mental wellbeing, while a second found that associations with GHQ decreased with distance, at 100m, 250m and 500m buffers. Others found contradictory results using radii of 1 and 3km. Another drawback was the use of Euclidean distance, which doesn't account for access routes. Application of network distance and consideration of pedestrian routes may give a greater indication of accessibility on foot.
Greenspace visiting patterns were measured inconsistently, in small or cross-sectional studies. Individuals who visited greenspace more often reported greater mental wellbeing, though a second study found this held only for eudaimonic wellbeing; no associations were found in an analysis of greenspace adjacent to workplaces. This study did however report a positive association with views of greenspace from the workplace. This is in keeping with previous research showing that green views reduce the effects of stress [8][9][10]95]. While two studies highlighted that the perceptions of greenspace quality were more strongly associated with mental wellbeing than quantity [52,59], the size of this difference was not estimated.
Individual connection to nature, assessed in seven studies, relied on self-report for both the greenspace and wellbeing measures, thereby carrying a high risk of reporting bias, especially since few controlled for potentially confounding factors.

Strengths and limitations
We conducted a comprehensive database search, thorough screening of articles, risk of bias assessment, and detailed narrative synthesis of the 50 studies which met our inclusion criteria. We identified six different ways in which greenspace was conceptualised and measured, and by which we stratified our review. We believe this is the first review to systematically appraise the evidence for associations between greenspace and adult mental wellbeing, using only validated measures of positive mental health.
Selection criteria were designed to ensure results of sufficient quality and relevance, and we consulted an information specialist to maximise search efficiency. Screening was undertaken by two independent reviewers, to minimise potential bias. While our criteria were designed to be inclusive, an element of subjectivity means there was a possible risk of excluding potentially interesting studies; we attempted to minimise this by appraising each study with assessments recommended by the Cochrane Handbook, which provides guidance for internationally recognised highest-standard research [32,[34][35][36][37].
We considered all greenspaces, not restricting our criteria to studies specifically in urban areas, although some studies were confined to urban or rural locations. Nationwide studies were likely to have included both, without stratifying for setting. It was difficult, therefore, to draw clear conclusions about interactions between urban and rural location and associations with mental wellbeing. Although there is interest in understanding how urban greenspaces should best be designed and constructed, it was not possible to draw conclusions specifically for those living in cities.
Only one-quarter of included studies measured both hedonic and eudaimonic mental wellbeing; the majority focused only on aspects such as life satisfaction, affect and vitality, while others used measures (such as the GHQ) which combined positive and negative (distress) items.
While several studies implied that 'nature' was associated with aspects of mental wellbeing, none provided a definition of this term. To further complicate matters 'nature' and 'greenspace' were sometimes used synonymously [17-19, 21, 72]. Vegetative or 'natural' greenspaces, such as those described as 'serene' (quiet, 'natural'), or with more trees, were most strongly associated with aspects of mental wellbeing, although one study found a stronger association for sports facilities. However, there were few direct comparisons between greenspace types. While Government Guidance provides a standardised greenspace typology for urban planning in the UK [15], no studies used this classification.
Studies that considered greenspace accessibility were limited to estimates of Euclidean distances from home rather than access routes [96]. These studies did not take account of participants' routines, or where they spent their time. None of the included studies assessed greenspace quality (such as captured by the Green Flag Award [97], or the social contexts in which greenspaces are situated [7,98]. Only 6 out of 50 papers reported longitudinal studies. Cross-sectional analyses cannot distinguish between reverse causality and associations which may be causal in nature, and, like all observational studies, are prone to confounding (especially by indication) and bias. Although 26 studies were deemed to be of Good quality, 12 were Fair, and the remaining 12 were Poor; this was mostly due to lack of control for potentially confounding, minimal statistical reporting, and, in 3 cases, lack of participant blinding to an intervention.

Conclusions
We sought to synthesis and appraise the evidence for associations between greenspace and mental wellbeing, but found few studies measuring both hedonic and eudaimonic wellbeing. Results suggest associations between greenspace and mental wellbeing, particularly hedonic wellbeing. We discovered adequate evidence for associations between urban greenspace and life satisfaction; however, the evidence for the remainder of the greenspace characteristics, including greenspace (land use) type, accessibility, viewing and visiting patterns, was limited or inadequate. Although not a true measure of mental wellbeing, studies using the GHQ were prevalent in the literature. This measure includes some positive items, and we further concluded that there was adequate evidence for associations between greenspace and lower GHQ scores. While our review was limited by the lack of available data to conduct a meta-analysis, we were able to highlight key areas for future research through our narrative synthesis.
Government guidelines for greenspace provision require robust evidence, but evidence is currently not sufficient for informed, specific planning recommendations. Further methodological work in this field is needed, including the development of operational definitions of 'nature' and 'natural', and agreement on a land use typology. Measures of greenspace quality are also needed. More studies are required to measure both hedonic and eudaimonic mental wellbeing. Greenspace accessibility should also be measured more specifically, using individual travel distances, using spatial methods of analysis, to better understand how greenspaces should be designed and incorporated into environments. Further research is needed that considers differences in associations between greenspace and mental wellbeing in urban versus rural settings.
Supporting information S1