Exploring the role of exposure to green and blue spaces in preventing anxiety and depression among young people aged 14 – 24 years living in urban settings: A systematic review and conceptual framework

Despite the growing problem of anxiety and depression amongst young people aged 14 – 24 years living in urban settings, reviews about the role of exposure to green and blue spaces or nature in preventing anxiety and depression tend to focus on children, adults or sometimes adolescents. This review aims to explore whether exposure to green and blue spaces reduces the risk of anxiety and depression among young people aged 14 – 24 years living in urban settings and provide a conceptual framework. The academic databases CINAHL plus, Global Health, MEDLINE, ProQuest: Dissertations and Theses, PsycINFO, Scopus and OpenGrey were searched for research published in English between January 2000 and June 2020. All study designs were eligible. All included studies were assessed for quality. Searches identified 9208 sources with 48 meeting the inclusion criteria for the review. Experimental studies provided evidence that walking or being in a green space improves mood and state anxiety immediately following the intervention. Non-randomised evaluations and observational studies suggest that social interaction, physical activity, and mindfulness mediate the relationship between exposure to green space and mental health. We propose that the absence of noise and restorative qualities of green spaces promotes mindfulness and interrupt rumination, which in turn reduce the risk of anxiety disorders and depression. This review and the resulting conceptual framework provide evidence to healthcare professionals about the value of contact with nature and green social prescribing. For policymakers, it provides evidence about the value of bringing the benefits of forests, vegetation and nature into cities, and ensuring that these spaces are

that demonstrate the extent to which prevention of clinical disorder is achieved.Rather, they focus on reduction of risk and pre-clinical distress.
The scoping review informed this systematic review.Here we take a novel and empirical approach to examining the evidence about the role of exposure to green and blue spaces in preventing depression and anxiety in young people aged 14-24 years living in urban settings, and the pathways involved.We build on the conceptual framework provided by Hartig et al. (2014) and refined by Nieuwenhuijsen et al. (2017) but focus specifically on anxiety and depression in young people.This framework includes measures of green space (e.g.neighbourhood vegetation, access to green space), exposure to green space (e.g.frequency and duration of use), possible mechanisms to achieving health outcomes (e.g.environmental exposure, physical activity, social contacts and cohesion, biogenics, and restoration and stress reduction), and health and wellbeing outcomes (Fig. 1).Nieuwenhuijsen et al. (2017) summarise the relative strength of evidence across the mechanisms, and emphasise that studies often measure either mechanisms (e.g.physical activity) or the health and wellbeing outcomes.
For the purposes of this review, our definition of exposure to green space encompasses (i) access to urban green spaces such as parks, or vegetation (for example, as measured by the normalized difference vegetation index (NDVI; Ekkel and de Vries, 2017) and (ii) exposure through activities or interventions that take place in green or natural environments (e.g.walking or adventure training) and (iii) simulations of green or natural environments (e.g.recorded sounds of nature).We also include 'blue space' (such as lakes), although this field of research is more recent (Reece et al., 2021).
Following Wolpert et al. (2019), we use a broad definition of anxiety and depression, including self-report of stress or low mood.We include studies that focus on primary prevention, when the aim is to prevent anxiety or depression from developing in the first place, as well as studies of secondary prevention, which aims to reduce the impact of anxiety and depression early once it has developed.Prevention of depression and anxiety can be thought of as reduction in symptoms of these "disorders" that disrupt functioning (Kutcher et al., 2015;Lawrence et al., 2017), or as a reduction in processes associated with anxiety/depression (such as stress, rumination or lack of positive activity) or improvements in mental wellbeing (O'Connell et al., 2009).For example, in young people, experiences of sub-clinical levels of low mood are strongly linked to greater risk of developing major depressive disorders (Davey and McGorry, 2019).We therefore included these broad terms to identify studies that may improve mental health or reduce sub-clinical symptoms of mental "disorder", as indicating reduction in risk of anxiety and depression (Kutcher et al., 2015).As such, interventions that reduce low-mood may reduce risk and increase prevention of depression.Similarly, stress reduction focused interventions form part of the prevention initiatives for both depression and anxiety (Brown et al., 2019).
Fig. 1.Socio-ecological framework for the relationship between greenspace and health and well-being (adapted from Nieuwenhuijsen et al., 2017).
I. Bray et al.The inclusion of secondary as well as primary prevention is appropriate, given the high prevalence of anxiety and depression amongst young people and the need to explore the mechanisms through which green space can impact on health and wellbeing (Nieuwenhuijsen et al., 2017).We did not include tertiary prevention, that is studies of clinical samples that are not representative of the general population.
Rather than conducting a traditional systematic review with narrow inclusion criteria (e.g.focusing on specific mechanisms or mental health outcomes), we sought to include a wide range of evidence from different disciplines and using any study design, in order to understand causal mechanisms and develop a conceptual framework.

Specifying the research question and search strategy
Our research question was "Does exposure to green and blue space reduce the risk of anxiety and depression among young people aged 14-24 years living in urban settings and, if so, what are the causal mechanisms?"We aimed to search for evidence about a wide range of exposures, outcomes and study designs, and to then select studies which would address each aspect of the research question, including mechanisms.
Key terms for three main concepts were created and used in the literature search: population, exposure, and outcome.Medical Subject Headings (MeSH) were also used where possible.The search terms and strategy for each database can be found in Supplementary Material.The protocol was not registered due to time constraints.

Database searches
The following databases were searched; CINAHL plus, Global Health, MEDLINE, ProQuest: Dissertations and Theses, PsycINFO, Scopus and OpenGrey.All databases were searched in July 2020.The searches in databases; CINAHL plus, MEDLINE, ProQuest: Dissertations and Theses, PsycINFO, and Scopus, were last updated 6 th July 2020.Global Health and OpenGrey were searched 7 th July 2020.Further relevant studies were added from reference lists of previously published systematic reviews (McCormick, 2017;Tillmann et al., 2018;Vanaken and Danckaerts, 2018;Zhang et al., 2020).We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009).

Selecting the evidence
We included sources from any country, published between 2000 and July 2020, written in English, and of any study design (including systematic reviews, experimental, observational and qualitative).For studies to be eligible for inclusion, (i) at least half of the participants had to be within the target age range (14-24 years) old, or (ii) at least half of the participant age-range had to overlap with this age range, or (iii) the reported mean or median age had to be within this age range.Participants had to be living in urban settings.We included studies that explored both access to green and blue spaces, including those in rural settings accessed by urban populations (e.g.forests), and contact with green space through activities or interventions that take place in these environments (e.g.running, walking).We also included components of being in these environments (e.g.sounds, sights, smells), and built environments that could include green components (e.g.street trees).The outcome measures we included were anxiety and depression, as well as those psychological variables which may provide insights into the mechanisms associated with anxiety and depression.These included mood, life satisfaction, happiness, wellbeing, and quality of life.In order to answer our research question and refine the conceptual framework developed by Nieuwenhuijsen et al. (2017), papers reporting on potential additional mechanisms such as connectedness to nature, mindfulness, physical activity, social cohesion, and self-esteem, were also included.We excluded sources published before 2000 and not published in English.Participants in exclusively rural settings or diagnosed with other mental health or developmental conditions (e.g.addiction, behavioural problems, autism) and other clinical samples (e.g.children or young people diagnosed with diabetes) were excluded.We excluded sources focused on athletes or competition-level sport, treatment or management of mental health outcomes, and exclusively physiological outcome measures.
The lead author carried out the title screening supported by Mendeley software.Full-text screening was carried out by four authors in Mendeley and Excel.Following screening, a list of included sources was transferred into an Excel spreadsheet for data charting.Data charting was carried out by four authors.We charted the authors of the source, title, journal, year of publication, source type, country of publication, whether the country was developed or developing, study design, population age, general or clinical population, intervention/exposure type, the outcome measure and whether that included anxiety and depression, or potential mechanisms.These potential mechanisms included stress, wellbeing, quality of life, mood, and mental health/illness/condition/ disorder.A broad range of evidence was required to refine the existing conceptual framework.We therefore assessed eligible studies against four key inclusion criteria.(If we had simply applied stricter inclusion criteria with respect to age, study design, exposure or outcome, this would not have allowed us to explore the mechanisms and modifiers in the conceptual framework).These criteria (C) were chosen to select sources most relevant to our original aims, while retaining breadth of evidence.They related to: the population of interest (C1: all participants within the 14-24 years age range), study design (C2: an experimental study design, including before-and-after studies, to assess effectiveness), exposure (C3: access to, or contact with, urban green space) and primary outcome (C4: assessing anxiety and/or depression).Studies were selected for inclusion if they met at least two of these criteria.For example, the relationship between access or contact with urban green space (C3) and mood in a sample of only 14-24-year-olds (C1), or an experimental study (C2) which involved taking participants, of whom 50% were aged 14-24 years, to a forest and measuring their anxiety (C4).

Data extraction and quality assessment
Variables extracted for each included study were authors, year of publication, location of study (country), sample (age, sex, other characteristics), study design, exposure (or intervention) and how this was measured, results relevant to our outcomes of interest and any information regarding mediating factors or mechanisms.
Quantitative sources were assessed for quality, including risk of bias, using the Effective Public Healthcare Panacea Project's Quality Assessment Tool for Quantitative Studies (Effective public healthcare, 2020).EPHPP evaluates selection bias, study design, confounders, blinding, data collection method, withdrawals and dropouts as being 'weak', 'strong' or 'moderate'.The EPHPP provides a global quality rating for each study of 'strong' (no weak ratings), 'moderate' (one weak rating), or 'weak' (more than one weak rating).We also derived a summary score for the number of criteria scored as 'Strong' in each study.The EPHPP is considered suitable to assess the quality of a wide range of quantitative study designs, including interventions (Jackson and Waters, 2005;Deeks et al., 2003).The EPHPP has excellent inter-rater reliability for overall scores (when compared to the Cochrane Collaboration Risk of Bias Tool) (Armijo-Olivo et al., 2012).It also has good construct and content validity (Jackson and Waters, 2005).For other study designs, CASP tools were used (CASP, 2018) to create a similar summary score for quality assessment.For example, the CASP qualitative tool produces a quality rating for each study from the assessment of nine closed questions relating to rigour, credibility and relevance, and is widely used for quality assessment in systematic reviews (Dixon-Woods et al., 2007).Two researchers completed quality assessments and data extractions for I. Bray et al. each paper, with benchmarking between each pair of assessors.If the global score differed, this was discussed until agreement was reached.Given the broad range of disciplines that we intended to include in our review, we did not use the quality scores to exclude studies, but did consider these in the synthesis of results and development of the conceptual framework.

Data synthesis and development of conceptual framework
Given the wide range of study designs and outcome measures (e.g.difference in means, incidence rate ratio) including qualitative outcomes, a narrative approach to data synthesis was adopted rather than meta-analysis.This included an analysis of variability in terms of setting, population, exposure/intervention, and outcomes (Popay et al., 2006).
The included studies are organised into six groups based on type of exposure.
We tabulated the characteristics of each paper, including the exposure, outcome measure, results, and information from each paper about possible mediators or causal pathways, along with a summary quality score.Initial synthesis for each group of studies was undertaken by the primary reviewer in regular discussion with other members of the team.The synthesis gave greater weight to higher quality evidence.
This information was used to build on the conceptual framework from Nieuwenhuijsen et al. (2017), describing potential causal mechanisms linking exposure to green spaces to anxiety and depression, including possible intermediary psychological processes.The preliminary conceptual framework (Nieuwenhuijsen et al., 2017) was based on categories for exposure to green spaces on the one hand (e.g.I. Bray et al. exposure to parks), and the psychological effect (anxiety and depression) on the other.While there were few studies that used anxiety disorders or clinical depression as outcomes, many studies included symptoms or pre-cursors such as mood and state anxiety.The second stage was therefore to add in these mechanisms to the framework.We had also extracted data on other mechanisms reported in the Nieuwenhuijsen et al. (2017) framework.These included behaviours (e.g.physical activity) and environmental factors (e.g.noise) that had been related to green spaces, and which contributed to the causal pathway.A small number of studies included information on perceptions of these spaces (e.g.danger), which modified these associations.The final step was to ensure the framework linked green spaces with these behavioural, environmental and psychological mechanisms and then to the psychological outcomes of anxiety and depression.Each component of the framework was reviewed by the inter-disciplinary team, and by a PPI panel of seven young people with lived experience of anxiety or depression.Finally, we summarise the available evidence for different subgroups of the population of interest.

Results
After removal of duplicates, we screened 9208 titles and abstracts.702 sources were eligible for full text screening (Fig. 2).After further exclusions (76% of which were due to the age range not being relevant), 89 sources remained.The 89 sources were considered against four key criteria (C1-C4), relating to the population of interest, study design, exposure and outcome, as described in the Methods.Forty-eight sources met at least two of these criteria and were included in the review.

Study location
Of the 48 studies, 46% were located in Asia, including Japan (21% of all included studies), China, Taiwan, Indonesia, Turkey and Iran.Just over one quarter of studies were conducted in Europe (27%), including the United Kingdom, Germany, Austria, Switzerland, France, the Alps, Bulgaria, Czech Republic, Sweden, and Norway.Some studies were located in North America (19%), with 7 studies from the United States and two from Canada.Three studies were based in New Zealand and one study in South Africa.No studies were located in South America.

Sample population
Over half (58%) of the studies included in this review had a sample population of university students.Of the 48 studies, 8 (17%) had a sample population of adults which overlapped with the target age range (14-24 years).Some of the studies focused on students in high school education (15%), and two studies included youth participants (4%).Lastly, some studies had a mixed sample of university students and staff (4%) and university students and graduates (2%).

Environment exposures
A wide range of environments were represented in the included studies.The largest proportion reported exposure to greenspaces through being in forests, including bamboo forests (21%).Other types of green space exposure included parks, sports fields, botanical gardens, horse trekking through fields and forests, outdoor employment in greenspaces, and generally being outdoors in a green space.Only three studies reported NDVI as a measure of exposure to green space.Some of the experimental conditions involved participants being near an indoor plant or touching natural materials (such as wood).Several of the studies used simulated exposure to nature.These studies involved participants listening to recorded soundscapes of natural sounds, or images of nature such as scenes of forests, mountains, water, wilderness, gardens, and agricultural landscapes.Two studies included used videos of outdoor environments and green spaces as simulated exposure to nature.Other studies included evaluations of outdoor adventure programs situated near lakes and mountains, as well as outdoor rock climbing.Exposure to blue spaces was also represented with some studies evaluating interventions such as canal walkways and sailing on the ocean.

Outcome measures
There were a wide range of outcomes measured and several studies used more than one.Outcomes ranged from measures of mood/affect and stress/anxiety experiences, closely linked to the concepts of clinical anxiety and depression.Further measures assessed constructs that may support the prevention of anxiety and depression, for example mechanisms of mindfulness, motivation, self-determination and self-efficacy.Measures of more general mental wellbeing outcomes were common, such as life satisfaction and wellbeing.

Study design
The majority of studies (n = 47, 98%) were quantitative.These included randomised trials (n = 18), and non-randomised intervention studies (n = 18), cohort studies (n = 5) and cross-sectional surveys (n = 6).While some of these studies included qualitative elements, only one purely qualitative study was included.

Quality assessment
Assessment of publication bias via a funnel plot was not possible due to the heterogeneity of studies.Quality scores are reported for each paper in Tables 1-6.There was strong initial agreement between the reviewers for global quality scores for the 47 quantitative studies (90%).Overall, areas of strength identified included approaches to dealing with confounders, use of validated outcome measures and low drop-out.This low attrition reflects that very few studies did follow-up post-intervention.Other areas in which quality was not generally highly rated were in representativeness of the sample (e.g.response rates were not reported), and blinding.This is important given the self-reported nature of outcomes.Quality scores were higher on average for those studies which compared urban streets with urban parks and forests, which were all experimental (mean score 3.1/6), and studies that compare physical activity in a green/blue environment and indoors, or evaluate physical activity programmes, which were also intervention studies (mean score 2.8/6).The studies which evaluated outdoor adventure programs, and education/training/employment in green environments, had a mean score of 2.3/6 for the quantitative studies and included one high quality qualitative study (scoring 7/9).Of lower quality were the studies which assessed particular aspects of being in green/blue environments (mean score 1.4/6).These were particularly weak on selection bias and reporting blinding.Also, the observational studies of residential exposure to vegetation, which tended to be cross-sectional (mean score 2.2/ 6), and the studies about young people's perceptions of green spaces (mean score 2.0/6) had lower quality scores.

Studies that compare urban streets, urban parks and forests
Thirteen of the studies described experimental designs that compared (walking or being in) urban settings with neighbourhood green space or forests (Table 1; scales used to measure outcomes are listed in Supplementary Material).Over the last 10 years a body of research has developed comparing exposure to forests and urban environments (Lee et al., 2011(Lee et al., , 2014;;Mao et al., 2012;Tsunetsugu et al., 2013;Song et al., 2018Song et al., , 2019;;Lyu et al., 2019;Zeng et al., 2020;Hassan et al., 2018).This has been complimented in more recent years by studies on urban parks (Franek, 2013;Wallner et al., 2018;Song et al., 2014Song et al., , 2015)).The majority of interventions were for 15 min (Tsunetsugu et al., 2013;Song et al., 2014Song et al., , 2015Song et al., , 2018Song et al., , 2019) ) but ranged up to three days (Lee et al., 2011;Lyu et al., 2019.These experimental studies measured outcomes immediately after the intervention.They provide evidence that a brief exposure to a green environment elicits a    For the POMS scores, viewing the scenery in the urban areas increased tensionanxiety (p = 0.00).
The landscapes of the forests were evaluated as being significantly more "comfortable" (p = 0.00, r = 0.51), "soothing" (p = 0.00, r = 0.53), and "natural" (p = 0.00, r = 0.59).Viewing the scenery in the forests induced significantly higher refreshment (p = 0.00, r = 0.55).short-term increase in mood and reduction in state anxiety.One study reported longer-lasting effects following exposure to a forest than a park (Wallner et al., 2018) and estimated reductions in state anxiety are greater following visits to forests (Lee et al., 2011;Song et al., 2019) than parks (Song et al., 2014(Song et al., , 2015)).

Studies that assess particular aspects of being in green/blue environments
A further eight intervention studies (two of which were randomised) assessed particular aspects of being in green or blue environments and reported some positive changes in various outcome measures (see Table 2).The aspects considered were the colour green (Akers et al., 2012), natural soundscapes (Benfield et al., 2014;Tedja and Tsaih, 2015), natural scenery (Chang et al., 2008;Chan KLE, 2015), touching natural materials (Ikei et al., 2018), indoor exposure to plants (Buttelmann and Römpke, 2014 and to animals (Hassan et al., 2017), Levels of anger were lower when viewing green (compared with red) scenery (Akers et al., 2012), listening to natural sounds was shown to be more restorative than anthropogenic sounds (Benfield et al., 2014), and touching wood induced feelings of comfort (compared with marble; Ikei et al., 2018).The interventions were typically brief (e.g. 5 min) and outcomes were measured immediately after the intervention.

Studies that compare physical activity in a green/blue environment and indoors, or evaluate physical activity programmes
Three experimental studies compared exercise in a green or blue environment with doing the same exercise indoors (Table 3).The first intervention was a walking route (approximately 17 min) along a canal (Nistbet and Zelensji, 2011).Compared to an indoor walking route, this led to increased connectedness with nature, mediated by positive affect.It also showed that participants tended to underestimate the positive hedonic effects of the outdoor walk.The second involved cycling on a static bike on the edge of a green field for 15 min and found greater increases in vigour, but not self-esteem, when exercising in the green   environment (Flowers et al., 2018).The third paper compared horse-trekking for 30 min in a green environment with riding on an indoor simulator.This involved contact with the animal as part of the intervention.and the authors reported improvements in feelings of anxiety and depression (Matsuura et al., 2011).Two further randomised trials assessed reported improvements in mood and sleep (Kalak et al., 2012;Walter et al., 2013) following outdoor running programmes.One involved running twice a week for ten weeks for 30-60 min, the other five times a week for three weeks for 30-37 min.A before-and-after study reported reductions in social anxiety following a two-month climbing programme (three times a week; Ozen, 2015).Finally, one study compared the effects of doing 15 min outdoor exercise on mood in an induced goal-oriented state compared with baseline state.The results showed that improvements in mood occurred only in the baseline group (Legrand and Thatcher, 2011).

Studies that evaluate outdoor adventure programs, and education/ training/employment in green/blue environments
A further set of studies (Table 4) employ a range of study designs to evaluate the mental health benefits of outdoor adventure opportunities.They indicate that wilderness excursions of 9-10 days lead to increased mindfulness and reduced stress (Mutz and Müller, 2016) and engender greater self-esteem through group belonging (Scarf et al., 2018).Studies evaluating longer (3-15 weeks) outdoor education programs also found measurable improvements in stress (Opper et al., 2014), self-esteem (McAnally et al., 2018) and social anxiety (Kardjono, 2017).Much shorter interventions also resulted in positive changes in the autonomy, competence, relatedness and enjoyment subscales of the Intrinsic Motivation Inventory (Mackenzie et al., 2018) and a reduction in anxiety (Kanters et al., 2002).Some of these evaluations included follow-up.Although benefits are not consistently sustained (Mackenzie et al., 2018), there were some longer-term effects on general mood (Opper et al., 2014) and depression symptoms (Kanters et al., 2002).A park-based prospective cohort study compared a violence prevention and mental health promotion after-school program with other after-school clubs, using youth crime rates as a proxy for violent behaviour (D'Agostino et al., 2020).They reported lower adjusted youth arrest rates three years post-intervention for the park-based programme.A comparison of mindfulness training in a botanical garden greenhouse with conventional (indoor) mindfulness training concluded that the green environment particularly helped beginner meditators to overcome problems with concentration and stress (Lymeus, 2019).Two studies also evaluated employment interventions in green or natural environments (Falxa-Raymond et al., 2013;Wolf and Housley, 2017).The first is a qualitative study of a group of young people who completed a training programme and then undertook 2-3 months' work experience in an entry-level "green job" (Falxa-Raymond et al., 2013).More than half of them described a positive change in their attitude towards themselves during this time.94% of them cited gaining new knowledge and skills as a positive aspect.The second was a before-and-after study of a one-year conservation program.Quantitative findings suggested improvements in anxiety and social functioning.Qualitative themes included new skills and training, developing friendships, co-operation and the benefits of being outdoors in terms of health and happiness (Wolf and Housley, 2017).
In contrast to the studies of tightly controlled interventions delivered to homogeneous samples (Tables 1-3), these studies include more complex interventions, with multiple components potentially contributing towards outcomes.There is a common thread, of exposure to a green space or natural environment, but each intervention includes additional elements, comprising either education or training (50%) or outdoor adventure (50%).Some also involve reflection on these        experiences (Opper et al., 2014;Mutz and Müller, 2016;Scarf et al., 2018), which could be another important element.Evaluations of such interventions have studied group dynamics but have less often included mental health and subjective wellbeing as outcomes (Mutz and Müller, 2016).

Studies of exposure to residential vegetation and blue spaces
Five observational studies (Table 5) consider the relationship between residential exposure to vegetation or blue spaces and a variety of outcomes.Exposure to vegetation is commonly measured using NDVI, 'an indicator of greenness based on land surface reflection of visible (red) and near-infrared parts of the spectrum' (Ekkel and de Vries, 2017).Dzhambov (2018) provides a longitudinal analysis of NDVI and mental health, including symptoms of depression and anxiety, at one-year follow-up.The authors found a relationship between exposure to neighbourhood green/blue space and mental health, as measured by the 12-item General Health Questionnaire (Goldberg and Williams, a References for measures are given in Supplementary Material.b Quality assessment score out of 6. c Quality assessment score out of 9 (the number of yes's for CASP qualitative quality assessment).Bar-On EQ-I: The Bar-On Emotional Quotient Inventory; Bar-On EQ-I-YV: The Bar-On Emotional Quotient Inventory Youth Version; GSES: General Self-efficacy Scale; LSAS: The Liebowitz Social Anxiety Scale Mindful Attention and Awareness Scale; NEP: The New Ecological Paradigm; PRS: Perceived Restorativeness Scale; PSQ: Perceived Stress Questionnaire; PSS: Perceived Stress Scale; POMS: Profile of Mood States; ROPELOC: The Review of Personal Effectiveness with Locus of Control assessment; SDQ: Strengths and Difficulties Questionnaire; SDQ: III Self-Description Questionnaire III; SCAS: Swedish Core Affect Scale; SFSS: Short Flow State Scale; TMS: Toronto Mindfulness Scale.
I. Bray et al. Environmental Research 214 (2022) 2018) also provide longitudinal evidence, which may offer an insight into mechanisms.They report that greater levels of residential greenness when participants were aged 16-18 years are associated with lower levels of work stress when they reach age 20-23 years.This suggests a protective effect when transitioning to university or working life.In cross-sectional analyses the association between NDVI and mental health appears to be mediated by physical activity and restoration, though this does not hold in longitudinal analyses (Dzhambov, 2018).A later cross-sectional study found that NDVI was negatively associated with both anxiety and depression symptoms.The following variables were reported as mediatorsperceived greenness, the restorative quality of the neighbourhood, dispositional mindfulness, rumination and resilience to stress (Dzhambov et al., 2019).Another cross-sectional study reported a negative association between NDVI and serious psychological distress.The outcome measure was the Kessler 6 scale (Kessler et al., 2003), which measures symptoms of anxiety and depression.The authors found that the relationship was mediated by pollution and social cohesion (Wang et al., 2019).A final cross-sectional study provides further evidence that NDVI is associated with reduced symptoms of depression but an association was not found for presence of blue space (Bezold et al., 2018).

Studies of young people's perceptions of green spaces
Another group of studies provided insight into how green spaces are perceived by young people, and how this could mediate the relationship between green space and mental health (Table 6).An experimental study compared perceptions of bamboo forests with those of an urban environment (Zeng et al., 2020).The forest was rated as providing a better environmental experience in terms of sensory perception, atmosphere, climate, space and place.An observational study compared different views in Tokyo (Asgarzadeh et al., 2014).More trees in the view increased spaciousness and monotony.Oppressiveness was reduced in views with more trees and more sky and increased in views with more buildings.Finally, increasing distance to trees reduced perceived oppressiveness and danger.This would suggest that the optimal environment in one in which trees and sky dominate rather than buildings, but with some distance to the trees.An experimental study by Wilson et al. (2016) found that walking in an urban park is perceived as being more restorative than walking on a busy street near traffic.It also reported that perceived noise mediates perceived restoration.Yang et al. (2011) showed that plants cause 'psychological noise reduction'.In other words, a green environment reduces perceived noise levels.Alizadeh et al. (2018) examined personal preferences for different green (forests, agricultural) and natural (mountains) landscapes and predictors of these.They found that personality type and the subject studied are associated with preference, suggesting that it is important to consider personal preferences if the impacts of green spaces are reliant on their use.Taken together, this set of studies suggest that greener environments are preferred, and are associated with restoration, partly by reducing perceived noise levels.

The conceptual framework
Based on the literature reviewed, we built on the existing literature to provide a conceptual framework that links particular elements of exposure to green spaces to the psychological outcomes of anxiety disorders and depression (Fig. 3).The model includes evidence about the modifying effects of young people's perceptions of exposure to vegetation and trees, mechanisms including the behaviours facilitated by green spaces (e.g.physical activity) and environmental exposures (e.g.noise and air pollution).We suggest tentative psychological mechanisms to explain how the restorative qualities of green spaces might lead to increased mindfulness and interrupted rumination (Bratman et al., 2015).Our model also suggest that a reduction in negative stimulation could enable restoration and possibly a reduction in the risk of anxiety disorder and depression.The framework demonstrates that there is experimental evidence linking elements of exposure to green spaces to mechanisms related to reduced environmental exposure to noise and air pollution, and transitory psychological states such as mood and relaxation.It also highlights a lack of evidence linking these to the outcomes of anxiety disorder or depression.Observational studies and evaluations of interventions provide some evidence of links between other potential mechanisms (e.g.improved social cohesion and resilience) and these outcomes.

Analysis by subgroup
Few studies compared results for males and females, despite clear differences in the epidemiology of mental health disorders in adulthood which emerge during adolescence.One exception, Kardjono (2017), suggests that a 4-week hiking programme reduced social anxiety for males, but induced pre-intervention anxiety for females.In an evaluation of a climbing programme, Ozen (2015) found that a climbing intervention reduced social anxiety overall, with no differences between men and women.Of the thirteen experimental studies described in Table 1, eight included only males.The one study that included only females reported similar results, suggesting that exposure to forests or parks have similar effects for males and females (Song et al., 2019).In terms of age groups, most of the experimental studies (Tables 1-3) were carried out with student samples of very similar ages.This means that  results cannot be compared by age.Educational or adventure programmes were positively evaluated for both school-age teenagers and students (Table 4).There is evidence that neighbourhood green space is less strongly linked to depression for high school students than middle school (Bezold et al., 2018).There was no evidence of effect modification by sex.Few studies in this review considered ethnicity as a moderator.While 61% of the experimental studies (Tables 1-3) were carried out in Asia, the predominantly observational studies (Tables 4-6) have a global spread.Included studies did not consider effectiveness in different socioeconomic groups.Many of the experimental studies used students as participants, who are unlikely to represent all socioeconomic groups.Two US studies reported ethnic and sociodemographic diversity within their sample, but neither of these studies report results for different groups (Falxa-Raymond et al., 2013;D'Agostino et al., 2020).
It is notable that much of the experimental research considered in this review focuses on forests as the exposure or intervention.Some studies reported elevated state-anxiety immediately prior to a walk in a forest, which suggests that this can be an unfamiliar or even threatening environment.
One study found some evidence that young people's preferences for mountains, forest and agricultural landscapes is linked to personality type (Alizadeh et al., 2018).We also found some evidence that the impacts of exposure to green spaces and outdoor exercise might be moderated by personality type (Song et al., 2018) or by psychological state (Legrand and Thatcher, 2011).For example, those with high-trait anxiety levels experienced a greater reduction in feeling of "depression-dejection" after walking through forest areas than those with normal and low-trait anxiety levels (Song et al., 2018).Differences in goal-orientation explain why individuals do not all respond in the same  way to exercise (Legrand and Thatcher, 2011).

Summary
There is good evidence that a 15-min walk in a forest or park can improve mood and state anxiety compared to an urban environment (Song et al., 2018(Song et al., , 2019)).It is not clear, however, whether this corresponds to a reduced risk of developing anxiety or depression.The impacts of different components of nature can be broken down, and such studies provide evidence of the restorative qualities of viewing or listening to nature, with associated improvements in mood and relaxation (Benfield et al., 2014;Tedja and Tsaih, 2015;Chang et al., 2008;Chan KLE, 2015).There is less good evidence about the social and psychological processes, such as activation or rumination, resulting from the opportunities for physical activity and restoration afforded by green space, or how these link to psychological disorders.There is limited evidence about how young people perceive green spaces, and how this affects their use of green space.There is also limited evidence of poor quality about the effects of exposure to blue spaces for young people's mental health.

Discussion
We synthesised a wide range of evidence about the role of access to green space in preventing anxiety and depression amongst 14-24 year olds, and developed a tentative conceptual framework linking exposures to outcomes, via a number of mechanisms including psychological processes.We found evidence that exposure to forest environments leads to greater momentary mental wellbeing compared to being on an urban street, and that urban parks can deliver similar benefits to forests.These studies did not provide evidence about longer-term outcomes such as symptoms or diagnoses of anxiety or depression.However, the evidence from observational studies that residential exposure to vegetation is associated with reduced risk of depressive symptoms for young people is crucial in terms of implementing change that will have broad reach and lasting benefits (Vanaken and Danckaerts, 2018).Observational studies also provide some insights into causal mechanisms, such as links between green space and restorative properties leading to reduced rumination (Dzhambov, 2018;Dzhambov et al., 2019).While adolescents spend more of their time further from home than children, young people are less likely to own cars or have access to private gardens.Neighbourhood green space and vegetation is therefore crucial to their well-being (as recognised in the Sustainable Development Goals; UN Environment Programme, 2020).
The importance of green space is partly because it enables many other activities that benefit mental health.Hartig et al. (2014) and then Nieuwenhuijsen et al. (2017) suggest four mechanisms through which exposure to natural environments may affect mental health.These are environmental exposure/air quality, physical activity, social contacts and cohesion, and restoration and stress reduction.Less researched mechanisms include microbial exposure and biogenic volatile organic compounds.It is hard to disentangle the impacts of green space and such mediators, and the psychological processes by which green space may reduce the risk of anxiety or depression are unclear.Many of the included studies cite Attention Restoration Theory (Kaplan and Kaplan, 1989;Ohly et al., 2016).This theory suggests green space engages indirect attention and thus provides rest for directed attentional capacity, and therefore presumably interrupts unhelpful rumination or worry.The papers reviewed suggest that green spaces promote mindfulness, mediated by physical activity and restorative qualities.We hypothesize that this in turn reduces rumination (Short et al., 2020) and improves psychological outcomes.Others have suggested that greenspace reduces the tendency to ruminate and thereby increases adaptive coping through enhanced psychological resilience (Marselle et al., 2019).Similar benefits might be afforded by the practice of mindfulness (Chi et al., 2018), but this requires training and effort.It seems that green environments encourage 'effortless mindfulness'.This is supported by one study that found a green setting when learning mindfulness was particularly helpful for beginners (Lymeus, 2019).An evaluation of a hiking programme encouraged participants to reflect on the natural surroundings and engage in 'active mindfulness'.It has been found that regular mindfulness can promote trait mindfulness (Quaglia et al., 2016).This provides a possible mechanism to a sustained reduction in the risk of psychological disorders.

Recommendations for research
In order to understand if exposure to green space prevents anxiety and depression amongst people aged 14-24 years, it is essential that more studies examine longer-term follow-up.Further research must explore the type and frequency of exposure associated with longerlasting impacts indicative of prevention of anxiety and/or depression.Another recommendation for future research is to measure changes in psychological processes such as rumination and activation that are key to anxiety and depression.Such research should utilise psychological theory to underpin explorations of the impact of green space on mental health.
The studies used a wide variation of interventions/exposures, considering different types of green space, different durations and varying activities.To allow an improved dialogue with mental health science, these interventions could adopt a more structured way of reporting the contents of the exposure, similar to those laid out in the TIDieR guidelines (Hoffmann et al., 2014).Systematic reporting would then allow investigations into impacts of green space on mental health to be more readily evaluated, replicated and potentially implemented.
Although it was included in our search terms, we found few studies about the benefits of blue spaces for young people's mental health, and these studies tended to be of poorer quality.This reflects the finding of our earlier scoping review (Reece et al., 2021), in which only 2% of sources related to blue spaces.However, this is a rapidly growing area of research.In terms of potential mechanisms, more work is needed on the importance of microbial exposure.Future studies should also consider the role of demographic variables, previous experiences (e.g. with outdoor activities) and individual preferences (for different environments).

Recommendations for practice
Recommendations for practice should be interpreted with some caution, as there is limited evidence of mixed quality regarding prevention.Recommendations might include consideration of social prescribing of adventure interventions for young people at risk of anxiety and depression and, at a population level, the integration of outdoor adventure opportunities into educational curricula.Evaluations of such interventions provide some limited evidence of sustained effects on general mood and feelings of depression (Opper et al., 2014;Kanters et al., 2002).However, there is as yet no clear evidence about the type, length or frequency required to have a lasting impact, which may be a barrier to prescribing. .High-quality longitudinal studies are still needed to estimate the long-term effects of regular exposure to green space on clinical measures of anxiety and depression.However, based on the studies we reviewed and the insights of our panel, we conclude that access to green space is likely to enhance other interventions to improve mental health, such as physical activity (Nistbet and Zelensji, 2011;Matsuura et al., 2011), mindfulness practice (Lymeus, 2019) and problem-solving.It also has multiple additional benefits (Hunter et al., 2019) in terms of health and wellbeing for the rest of the urban population, improving thermal comfort and biodiversity in cities, and reducing pollution and risk of flooding.These all have implications for the future health of today's young people.
I. Bray et al.

Limitations
We have taken a novel approach to reviewing the evidence about exposure to green spaces and mental health for young people, drawing on a wider range of literature than previous reviews.While this approach has helped bridge a gap (from immediate psychological responses to changes in clinically measurable depression or anxiety disorders), we acknowledge several limitations.First, the search terms and databases we used may have missed relevant papers.Second, screening was carried out by one researcher.Third, though we set out to include all study designs, the review is heavily weighted towards quantitative research.This is partly due to the limited qualitative research on this topic with young people, but also partly due to the additional exclusion criteria at the full text screening stage, which sought to select the most relevant sources from a large number of eligible studies.Nevertheless, it is worth noting that many of the included studies applied quantitative approaches to qualitative descriptions of attitudes or feelings, and some used mixed methods.Fourthly, thirteen of the studies (27%) include a proportion (<50%) of participants outside of the target age range (14-24 years).Due to the heterogeneous nature of the included studies, we did not formally assess publication bias.Although publication bias is a possibility, we did include a search of grey literature which did not reveal unpublished non-significant findings.We must also acknowledge that the studies were of varying quality and that selective reporting within studies may have biased our findings.

Fig. 3 .
Fig. 3. Conceptual framework based on the literature in the review (Thicker lines represent evidence from experimental studies and thinner lines represent evidence from nonrandomised evaluations and observational studies.Colour coding differentiates pathways from each feature of green space).(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Table 1
Main characteristics and results of studies that compare urban streets, urban parks and forests.

Table 1
(continued ) 3(continued on next page) I.Bray et al.

Table 1
(continued ) a References for measures are given in Supplementary Material b Quality assessment score out of 6; POMS: Profile of Mood States; POMS: Profile of Mood States -Shortened Version; SDM: semantic differential method; STAI: State-Trait Anxiety Inventory.I.Bray et al.

Table 2
Main characteristics and results of studies that assess particular aspects of being in green/blue environments.

Table 2
(continued ) 2(continued on next page) I.Bray et al.

Table 2
(continued ) a References for measures are given in Supplementary Material.bQualityassessment score out of 6. AAS: Audience Anxiousness Scale; BMIS: Brief Mood Introspection Scale; PAS: Pet Attitude Scale; POMS: Profile of Mood States; PRS: Perceived Restorativeness Scale; SDM: semantic differential method; STAI: State-Trait Anxiety Inventory -German Adaptation.I.Bray et al.

Table 3
Main characteristics of studies evaluating physical activity programmes activity in a green/blue environment.

Table 3
(continued ) Shortened Version; PSS: Perceived Stress Scale; SAS-A: Social Anxiety Scale for Adolescents; STAI: State-Trait Anxiety Inventory; TESI: Tension-Effort Stress Inventory; TSM: Telic State Measure.Main characteristics and results of studies evaluating outdoor adventure programmes, and education/training/employment in green/blue environments.
a References for measures are given in Supplementary Material.b Quality assessment score out of 6. BAGE: Belief About Green Exercise questionnaire; INS, Inclusion of Nature in Self scale; PANAS: Positive and Negative Affect Scale; POMS, Profile of Mood States -

Table 5
Main characteristics and results of studies of exposure to residential vegetation and blue spaces.
bI.Bray et al.

Table 6
Main characteristics and results of studies of young people's perceptions of green spaces.

Table 6
(continued ) a References for measures are given in Supplementary Material.b Quality assessment score out of 6; 16 PF: Cattell Sixteen Personality Factor Questionnaire; SRSS: Short-version Revised Restoration Scale.