& Place 39 (2016) 70–78

Contents lists available at ScienceDirect

Health & Place

journal homepage: www.elsevier.com/locate/healthplace

Residential exposure to visible blue space (but not green space) associated with lower psychological distress in a capital city

Daniel Nutsford a, Amber L. Pearson b,c,n, Simon Kingham a, Femke Reitsma a a University of Canterbury, Department of Geography, Christchurch, New Zealand b Michigan State University, Department of Geography, 673 Auditorium Road, East Lansing, MI 48824, USA c University of Otago, Department of Public Health, 23A Mein Street, Wellington 6242, New Zealand article info abstract

Article history: As urbanisation escalates globally, urban neighbourhood features which may improve physical and Received 29 July 2015 mental health are of growing importance. Using a cross-sectional survey of adults and the application of Received in revised form novel geospatial techniques, this study investigated whether increased visibility of nature (green and 8 February 2016 blue space) was associated with lower psychological distress (K10 scores), in the capital city of Well- Accepted 3 March 2016 ington, New Zealand. To validate, we also tested whether visibility of blue space was associated missing Available online 11 March 2016 teeth in the same sample. Cluster robust, linear regression models were fitted to test the association Keywords: between visibility of nature and K10 scores, adjusted for age, sex, personal income, neighbourhood Mental health population density, housing quality, crime and deprivation. Higher levels of blue space visibility were Visibility analysis associated with lower psychological distress (β¼0.28, po0.001). Importantly, blue space visibility was Blue space not significantly associated with tooth loss. Further research is needed to confirm whether increased Green space visibility of blue space could promote mental well-being and reduce distress in other cities. & 2016 Elsevier Ltd. All rights reserved.

1. Introduction heart disease (Pretty et al., 2005). As such, exploration of the ways in which urban neighbourhood features, such as exposure to green The last decade has seen mounting evidence to suggest that the and blue spaces, may promote mental health, is of great im- presence of natural environments in urban neighbourhoods is portance to reducing this major source of global disease burden. associated with positive physical and mental health. These en- In our previous work (Nutsford et al., 2015) we outlined three vironments, often called “green space” and “blue space”, are places primary causal pathways through which green and blue spaces for recreational opportunities (Barton and Pretty, 2010), social may directly or indirectly influence mental health. First, there is connection and enhanced social ties (Nutsford et al., 2013), often strong evidence of an indirect pathway whereby increased phy- providing respite in urban settings to assist mental and physical sical activity has been associated with improved mental health fi recuperation (White et al., 2010; De Ridder et al., 2004). While the (Barton and Pretty, 2010; Pretty et al., 2005). The bene ts of definition of green space may vary slightly between users, this physical activity include reduced blood pressure, increased self- esteem and reduced anxiety (Pretty et al., 2005), all of which may term tends to include open areas of vegetation (e.g. , sports improve mental health. Nearby green spaces, particularly useable fields) and conservation areas (e.g. forests), but can also include green spaces, provide recreational opportunities and are asso- backyard gardens, farms or any other space predominantly cov- ciated with higher levels of physical activity (Sugiyama et al., ered in vegetation. Blue space includes water bodies (e.g. lakes, 2008). Particularly relevant to the research here, the proximity to oceans, rivers, etc.) but rarely includes human-made features (e.g. both green (Pretty et al., 2005), and blue space (Barton and Pretty, fi water fountains or sculptures). The mental health bene ts of green 2010) has shown a significant impact on self-esteem, especially and blue spaces have become increasingly of interest, with mental amongst the mentally ill. A weakness of this research is that fi health contributing signi cantly to the global disease burden (Lim proximity to green or blue space is often used as a proxy for use or et al., 2012). Anxiety and depression are also precursors for chronic engagement in physical activity. In fact, a recent review associa- physical conditions such as , arthritis, diabetes, stroke and tions between proximity, in particular to green spaces, and phy- sical activity concluded that findings are inconsistent across stu-

n dies (Bell et al., 2014). Corresponding author at: Michigan State University, Department of Geography, 673 Auditorium Road, East Lansing, MI 48824, USA. Second, green and blue spaces offer opportunities for social E-mail address: [email protected] (A.L. Pearson). interaction, whether planned or not, which is also linked to http://dx.doi.org/10.1016/j.healthplace.2016.03.002 1353-8292/& 2016 Elsevier Ltd. All rights reserved. D. Nutsford et al. / Health & Place 39 (2016) 70–78 71 improved moods (Sugiyama et al., 2008; Miles et al., 2012). In- willingness-to-pay for houses with ocean views. In 2005, Putra creased social interaction is particularly beneficial for the elderly, and Yang (2005) developed a GIS-based 3D visibility analysis in where increased social interactions improve community cohe- the hope that it would map environmental perceptions of the built siveness and has been associated with lower suicide rates, lower environment. Similarly Miller et al. (2009) employed spatial ana- fear of crime and better physical health (Kweon et al., 1998). lysis tools to quantify the visual perception of green spaces for a However, Bell et al., (2014) identify a number of limitations un- case study in Edinburgh, UK. Their work illustrates the potential derpinning extant research and conclude that there is limited for including visual measures in future research and concludes evidence to support the social interaction pathway. The vast ma- that three aspects could influence access to green spaces: loca- jority of research makes the assumption that green space use is tional access, cultural access and visual exposure. These studies correlated with locational access measures, although actual use os highlight that the visual structure of residential environments is often not directly measured. Much of this work argues that such important and quantifiable, however, to date these advances in use of green spaces are also driven by personal characteristics and visibility quantification have not been applied to understanding the amentities and quality of the green spaces. the relationship between the visual exposure of green and blue Last and most salient to the present study, green and blue spaces and mental health. spaces are recognised as therapeutic or salutogenic places and This study serves as the first of its kind, to our knowledge, and may lower psychological distress by serving as calming backdrops investigated whether self-reported psychological distress was as- in residential neighbourhoods, in contrast to built-up features of sociated with visibility of green and blue spaces in residential the urban setting (White et al., 2010; Pretty et al., 2005; Ulrich neighbourhoods for a sample of adult New Zealanders in Well- et al., 1991). This backdrop is theorised to reduce sensory stimuli ington, using the New Zealand Health Survey (NZHS). To validate, and thus promote mental relaxation (Ulrich et al., 1991). In fact, it we also tested whether visibility of blue space was associated with is theorised that humans have an innate imperative to favour missing teeth in the same sample (a theoretically unrelated out- natural environments in contrast to urban environments due to come). The study made use of a novel geospatial method for their evolutionary importance as key resources (Newell, 1997). As measuring the visibility of natural environments (Nutsford et al., a result of this evolutionary connection, the human brain pro- 2015). cesses natural environments more efficiently than built-up en- vironments, thereby further increasing opportunity for relaxation (Ulrich et al., 1991; Heerwagen and Orians, 1986) and combating 2. Methods the onset of stress (Igarashi et al., 2015). Kearney (Kearney, 2006) conducted a qualitative study which found that prisoners in Eng- 2.1. Ethical approval land with non-natural views had a 24% higher frequency of sick calls than prisoners with a view of farm land. Similarly, it was Approval to access the NZHS data was granted by the New observed that increased views of nature from residential windows Zealand Ministry of Health. Spatial data were linked to survey increased neighbourhood satisfaction, which has been linked to responses, then participant identifying information was removed, improved psychological stress (Kearney, 2006). A literature review both conducted by the Ministry of Health. As such, all data used in conducted by Völker and Kistemann (2013), identified numerous analyses were anonymized prior to our use and did not require benefits of visible waterscapes including emotional, recreational University ethical approval for use. All de-identified data were and direct health benefits. Specifically, they identify work by Kar- password protected and kept in a secure computer facility. manov and Hamel (2008) conducted in the Netherlands that found blue spaces to stress-reduce and enhance mood in urban en- 2.2. Spatial data and the creation of the visibility measures vironments. Other research by Laumann et al. (2003) observed increased levels of attentiveness in study participants when ex- Spatial data on green spaces and oceanic blue spaces were posed to a simulated coastal environment as opposed to an urban compiled from three national datasets (the Land Class DataBase II, setting. Department of Conservation land register, and the Land Informa- The majority of existing studies evaluating relationships be- tion New Zealand parcel database), as previously used by Ri- tween exposure to green and blue spaces and indicators of mental chardson et al. (2010). Freshwater blue space features such as lakes health have employed basic measures of proximity (e.g. distance and rivers were made available by Land Information New Zealand from home) (Wheeler et al., 2012; White et al., 2013) or the pro- (LINZ) and obtained from an online spatial data archive (www. portion of parks within a given distance from home (Nutsford koordinates.com, downloaded April 2013). See Fig. 1 for the spatial et al., 2013; Richardson et al., 2013, 2010). These measures of lo- distribution of green and blue spaces across the study area. cational access are useful when exploring the potential pathways Visibility measures were generated for population-weighted of physical activity and social interactions in reducing psycholo- centroids of meshblocks, called viewpoints hereafter, in which gical distress, as easy access to a from one's home could fa- study participants resided (New Zealand's finest administration cilitate such processes. However, proximity measures are unable to boundary, study area n¼46, national n¼46,263, mean area accurately characterise one's visual exposure to natural environ- ¼0.1 km2 in 2006) across each cell in a 5 m resolution gridded ments and thus to explore the potential salutogenic aspects of digital elevation surface and extended 15 km from each viewpoint. green or blue backdrops. Advances in geospatial technologies A novel method of visibility was applied, termed the Vertical permit more precise quantification of the visibility of green and Visibility Index (VVI), which accounts for the slope, aspect, dis- blue spaces (Nutsford et al., 2015) and thus offer a new way to test tance and elevation of visible areas relative to the observer loca- the effect of visibility of green and/or blue space on psychological tion (see Fig. 2). This method is described in detail elsewhere distress, to complement existing qualitative evidence (Herzog, (Nutsford et al., 2015). 1985; Rose, 2012). In brief, the VVI is a visual summation of green and blue spaces Quantitative measures of green and blue space visibility have expressed as degrees of visibility. For observers, green and blue been employed in various disciplines. However, they have not yet spaces are visible across a broad spectrum of degrees, both verti- been employed in health research. For example, Hamilton and cally and rotationally. Observers surrounded by buildings, for ex- Morgan (2010) incorporated views of blue space into house va- ample, are likely to have limited visibility of green and blue spaces luation price models and found a clear favouritism and and this visibility may be limited to specific angles. These 72 D. Nutsford et al. / Health & Place 39 (2016) 70–78

Fig. 1. Distribution of natural environments throughout Wellington City and the greater region.

Fig. 2. Cross-sectional diagram of the VVI measure, highlighting how visibility can be represented as degrees of visibility. The observer on the left (indicated by the dot) has higher visibility of the grassy hill (indicated by a higher VVI value). differences are captured with the VVI while accounting for the impact of that cell is. For example, a hill sloping down towards a impacts of slope, aspect, distance and elevation, unlike standard viewer is of much greater visual significance than a surface sloping viewshed analysis. The VVI is generated through an iterative pro- away perpendicular to a viewer or indeed, to a level surface with cess where the elevation, aspect and slope of each visible cell in no slope. The summation of these visibility angles across each the Digital Elevation Model is computed relative to the viewpoint. visible cell for a single viewpoint provides a more accurate mea- This measure provides information about whether a cell is visible sure of visibility from a human perspective. The influence of (standard viewshed analysis), but also how significant the visual buildings in visibility analysis is incorporated in the initial stages D. Nutsford et al. / Health & Place 39 (2016) 70–78 73 of analysis where building heights are added in the standard quality ratings from Quotable Value New Zealand (QV) data. QV is viewshed analysis. Using line of sight algorithms, buildings effec- a crown-owned independent entity which holds and maintains tively obstruct cells which are candidates, thus preventing them data on all New Zealand properties and rates properties as “su- from being included in the further visibility analysis. These perior” (1), “average” (0) to “poor” (1). Most ratings are assigned methods were applied to generate quantities of visible green space on the basis of exterior inspection from the street, as closer in- and blue space for each meshblock in the study area, expressed as spections only occur following work requiring a building consent. degrees of visibility. While there is no formal validation of these ratings, QV ratings for Each measure was also divided into four distance bands, visible overall dwelling condition have been found to be broadly similar areas: (i) r300 m; (ii) 300 m to 3 km; (iii) 3–6 km; and (iv) 6– to standardised assessment-based ratings assigned by the Building 15 km. These bands were selected to categorise visible areas as Research Association of New Zealand (BRANZ, 2005). A Housing foreground objects, background objects, or objects somewhere in Quality Index (HQI) was then created at CAU level, by averaging between. Through this notion questions such as “does distant the quality value over the CAU, then ranking these averages and green space have a stronger influence on health compared to dividing them into deciles, where higher values indicate better nearby green space?” can be answered. After preliminary analyses, quality. Each participant was assigned the HQI for the CAU in blue space distance bands were not included in further analyses which they resided. due to the lack of blue space in nearby distance bands (i.e., almost Last, to validate our findings, we also tested whether visibility all visible blue space was at distances43 km). Furthermore, there of blue space was associated missing teeth in the same sample. We was a lack of variation between the distant bands simply because selected tooth loss due to its high correlation with socioeconomic locations with plenty of visible blue space between 3 and 6 km status while being theoretically unrelated to blue space exposure. were also likely to have visible blue space beyond 6 km due to the In studies from a variety of countries, tooth loss has been asso- expansive nature of oceanic blue space. Each visibility score (de- ciated with measures of individual socioeconomic status, including grees of visibility) was finally transformed to an ordinal scale be- income and education (Ito et al., 2012; Buchwald et al., 2013; tween 1 and 10, where 1 represents the lowest 10% of visibility Wennström et al., 2013; Wu et al., 2014). Many such studies were scores and 10 represents the highest 10%. conducted in older adult and elderly populations (Wu et al., 2014; Gaio et al., 2012). A recent longitudinal study among adult (38–50 2.3. Health and demographic data years old), but not elderly, women in Sweden also found that fewer remaining teeth was significantly higher for women of Kessler Psychological Distress Scale (K10) scores were obtained lower social group, or living alone, in all analyses over a 36 year- for Wellington adults (n¼442, 15 years and older) who partici- period (Wennström et al., 2013). In this Swedish study, various pated in the 2011/2012 New Zealand Health Survey (NZHS) which logistic regression models were fitted to compare those with 1 or covers population health, long-term conditions, health service more teeth to those with none, those with 11 or more teeth to utilisation and patient experience, health risk and protective fac- those with 0–10 teeth, those with 21 or more teeth to those with tors, health status and socio-demographics (Ministry of Health, 0–20 teeth, etc. In their large sample, they were able to find sig- 2013). The national survey uses a multi-stage, stratified, prob- nificant results for each of these comparisons (Wennström et al., ability-proportional-to-size (PPS) sampling design. K10 scores are 2013). In a Japanese study, those with the lowest household in- a simple measure of psychological distress, designed for large come had increased risk of tooth loss (OR¼1.25) compared to sample population studies where scores range from 0 to 40 those with the highest income (Ito et al., 2012). (higher values indicate increased distress). This measure involves In our sample (n¼442), most respondents reported having all 10 questions about personal feelings over the previous month and their teeth (n¼239) and very few reported having no teeth has proven to be an accurate predictor of anxiety and mood dis- (n¼14). In fact, only 5% of the sample reported having less than orders (Oakley Browne et al., 2010) and diagnosed mental dis- half of their teeth (n¼23). For these reasons, we compared those orders (Andrews and Slade, 2001). In 2001, the Victorian Popula- with missing teeth to those without missing teeth, as done in si- tion Survey determined thresholds to represent the likelihood of a milar research (Ito et al., 2012). mental health disorders based on K10 responses, which are now widely accepted (Oakley Browne et al., 2010; Kessler et al., 2003). 2.4. Statistical analyses Scores of 0–5 were considered'none or low’;6–11 were ‘moder- ate’;12–19 were ‘high’ and 20–40 were ‘very high’ in regards to Approximately 20% of the Wellington NZHS respondents had the likelihood of having a mental health disorder. K10 scores were missing personal income data (n¼95). Thus, multiple imputations treated as an ordinal dependent variable in regression models. by chained equations were used to impute these missing values Other demographic variables from the NZHS included sex (cate- using Stata v12 software (College Station, TX, USA). Follow- gorical), age groups (ordinal), ethnicity (categorical), and personal ing White et al. (2011) recommendations: (1) 20 replicates of the income groups (ordinal). dataset were generated to roughly reflect the percentage of To account for other features of the urban environment which missing data; (2) data were assumed missing at random; and might be important in understanding the relationship between (3) all independent and dependent variables included in the final visibility of natural environment and mental health, we also analytical regression models were used as variables for the mul- compiled data for neighbourhood socioeconomic status, popula- tinomial logistic chained equation model. tion density, and crime rates. We obtained deciles of neighbour- After imputation, analytical regression models were fitted to hood deprivation (ordinal) and population density (continuous) examine associations between visibility of green and blue spaces for 2006 from Statistics New Zealand at the meshblock level. We and K10 scores while controlling for individual-level and area-le- calculated average annual neighbourhood crime rates 2007–2010 vel covariates. In total, separate linear regression models for six (continuous) based on New Zealand Police data at the census area visibility measures were fitted, namely: (a) green space; (b) blue unit level (meshblocks are nested within census area units or space; (c) green space (o300 m); (d) green space (300 m to CAUs). To further adjust for socioeconomic position, we included a 3 km); (e) green space (3–6 km); and (f) green space (6–15 km). variable to represent the housing quality of the neighbourhood. Initially, a variable representing combined blue and green space Many poor neighbourhoods are characterised by poor quality visibility was examined, but results were not included here, as housing (Braubach and Fairburn, 2010). We compiled dwelling green space tended to dominate nearby views and blue space the 74 D. Nutsford et al. / Health & Place 39 (2016) 70–78 distant views, resulting in a lack of variation between the observer Table 1 locations. Each of these models included K10 scores as the de- Descriptive statistics of the survey participants and their neighbourhoods in pendent variable and was adjusted for sex, age, income, neigh- Wellington City, by sex. bourhood deprivation, neighbourhood housing quality, population Characteristic Females Males Total density and crime rate. Last, we fitted a logistic model to validate our findings which included tooth loss as the dependent variable N 260 182 442 of interest and visibility of blue space as the independent variable of interest, adjusted for the same confounders as all other models. Demographic characteristics In all analyses, adjustments were made for the complex multi- of participants, % level sampling design of the survey. The clustered sampling design Sex 58 41 100 Age was specified and Taylor series variance estimations were used. All 15–44 yr 56 54 55 primary analytical models were cluster robust, linear regression 45–64 yr 32 35 33 with full adjustment for the above covariates and were fitted using 65þ yr 12 12 12 Stata v13. Ethnicity, Māori (indigenous 10 9 10 In comparing final analytical model results, it was found that population) fi o Personal income beta coef cients changed 15% between models using non-im- $0–$40,000 38 30 34 puted versus imputed data. Thus, while descriptive statistics were $41,000–$70,000 26 15 21 reported using the non-imputed dataset, all final analytical re- $71,000þ 16 36 24 gression results were derived using the imputed data. Missing 20 19 20

VVI measures for participants, mean (sd) ** 3. Results Visible green space 30 (28) 33 (29) 31 (27) Visible green space r300 m** 9 (18) 9 (20) 8 (19) Visible green space 300 m to 3 km** 29 (28) 29 (27) 29 (28) 3.1. Descriptive statistics Visible green space 3–6km** 8 (15) 9 (14) 9 (14) Visible green space between 6 km and 5 (17) 9 (24) 7 (20) Throughout the study area of Wellington City (including a 15 km** ** 15 km buffer to encapsulate visible green and blue spaces beyond Visible blue space 4 (17) 7 (24) 5 (20) the city limit) there was a total of 2076 km2 of green and blue spaces (green space¼795 km2 and blue space ¼1280 km2). Of the Neighbourhood characteristics of participants, mean (sd) *** visible blue space, the vast majority is oceanic with fresh water Neighbourhood deprivation 5 (2) 4 (2) 5 (2) Population density (km2) 5535 (3827) 5020 5323 contributing o1% of the total. The age composition of participants (3571) (3729) was similar between males and females, with most participants Crime rate per 100,000 10 (78) 10 (8) 10 (8) aged less than 45 years (Table 1). Neighbourhood characteristics and visibility exposure variables did not vary substantially by sex. ** Possible value ranges 0–100, with higher values indicating higher levels of The visibility measures were also similar across sex, and showed visibility. high variation across the study area, as indicated by the standard *** Deciles from 1 to 10, 1 is least deprived, 10 is most deprived. deviations. The indigenous population (Māori) comprised about 10% of the sample, which is slightly less than the national popu- lation proportion (15%) (Statistics New Zealand, 2013). Table 2 in both models (pr0.05), yet effects were weak (βo0.001). shows that, on average, psychological distress was slightly higher When assessing green space visibility by distance bands amongst females compared to males (K10¼6.1 and 5.5, respec- (Table 4), none of these measures were independently associated tively). The 15–44 year old age group exhibited the highest aver- with K10 scores. However, personal income, neighbourhood age psychological distress (K10¼6.4), followed by respondents 65 housing quality, population density and crime rate were all sig- years and older (K10¼6.0) and those aged 45–64 years (K10¼4.8). nificantly associated with k10 scores. Psychological distress was higher on average amongst Māori In the validation model (Table 5, Model 7), we found that compared to non-Māori (K10¼8.9 and 5.5, respectively). Psycho- personal income and age were significantly associated with tooth logical distress exhibited a gradient by personal income level loss, in the expected directions. Importantly, blue space visibility (K10¼6.8 in the lowest income group and 4.2 in highest income was not significantly associated with tooth loss. group), with a high average K10 score in among respondents with missing personal income data (K10¼6.7). 4. Discussion 3.2. Regression results Increased visibility of blue space was significantly associated In Table 3, no significant association was detected between with lower psychological distress after controlling for covariates, green space visibility (independent variable of interest) and K10 in our sample of adults in the capital city of New Zealand. Yet, scores (dependent variable) (Model 1, β¼0.09, p¼0.455). tooth loss was not – as theorised. Our findings suggest that re- However, higher visibility of blue space was significantly asso- sidents living in neighbourhoods with increased views of blue ciated with lower psychological distress (β¼0.28, po0.001) spaces (predominantly oceanic) had lower distress levels. This (Model 2). This suggests that for a 10% increase in visibility of blue finding is in accordance with qualitative studies which demon- space, a 0.28 lower K10 score could be expected. strated visible waterscapes evoke positive emotions (White et al., In both models, personal income exhibited a significant, in- 2010; Herzog, 1985; Ulrich, 1981; Völker and Kistemann, 2011). dependent association with K10 scores, with higher income Blue space may be of more importance in regards to psychological earners exhibiting lower levels of distress. Neighbourhood housing distress reduction, a notion that is supported by (Ulrich, 1981; quality was negatively, significantly associated with K10 scores, Ashbullby et al., 2013) who found scenes of blue space may have a whereby better quality housing was associated with lower levels stronger influence on psychological distress than views of green of distress. Population density and crime rate were also significant space. Richardson et al., (2010) offer theoretical support that blue D. Nutsford et al. / Health & Place 39 (2016) 70–78 75

Table 2 Wellington City also has a significant amount of greenery in Survey participant characteristics, by average level of psychological distress (K10 backyard gardens which were not included in measures of green scores). space. It is possible that residents across the city are exposed to Variable N K10, mean (sd) varying levels of greenery within their immediate neighbourhood. It is also possible that too much localised green space may have Total study population 442 5.8 (4.9) less beneficial connotations, as suggested in other research such as Sex being intrusive, creating a feeling of crowdedness or reduce light Female 260 6.1 (5.2) and airflow (Kuo et al., 1998). Still, our findings are inconsistent Male 182 5.5 (4.5) with a number of other qualitative studies which found improved Age moods to be associated with green space immediately visible and 15–44 yr 243 6.4 (5.1) in close proximity (Kearney, 2006; Moore, 1981; Kaplan, 2001). 45–65 yr 146 4.8 (4.3) This study may differ for a host of reasons. As this study cap- 65þ yr 53 6.0 (5.3) tured visibility at the neighbourhood or small area level and ap- Ethnicity plied quantities to individuals, measures of greenery are likely to Māori 42 8.9 (7.5) Non-Māori 400 5.5 (4.5) be less representative for each individual within a neighbourhood. Additionally, the majority of the qualitative studies relied on Personal income $0–$40,000 152 6.8 (5.1) comparisons in mental health for those viewing green spaces and $41,000–$70,000 95 5.4 (4.1) those viewing starkly contrasted urbanised spaces (i.e. open pas- $71,000þ 108 4.2 (3.3) ture views versus building facades and prison courtyards). Our Missing 87 6.7 (6.5) study takes advantage of a real life, city scenario. Furthermore, our study did not measure individual orientations that may influence K10 scores are values between 0 and 40 indicative of psychological distress, where use inclinations (i.e., self-identification with nature) or the possi- higher values represent increased distress. Scores above 6 represent a moderate likelihood of a mental health disorder. ble dynamic nature of the relationship between exposure to visible green space and mental health over time (with changing life cir- space may be of more significance in New Zealand than green cumstances), which were limitations identified in a recent review space due to the country's island geography. Other work in New of the green space and health research (Bell et al., 2014). Studies in Zealand has highlighted how island spaces offer unique settings other cities could help confirm whether the findings of this study for the therapeutic benefits of blue space (Kearns et al., 2014). are generalizable. Also, it is possible that visibility of blue space is simply a better There are a number of limitations to this study. Due to the representation of visibility of ‘natural’ environments than green cross-sectional nature of this study, it is unknown whether people space, especially in urban settings where sports fields and open who have lower distress levels choose to live in areas with higher parks fall under the category of green space. Since almost all of the levels of blue space visibility or the reverse. Additionally, this blue space in Wellington is oceanic, it is unclear whether these study did not account for the length of residence in the neigh- results would be replicated if the blue space was fresh water. If the bourhood and its possible influence on the relationship between type of water is irrelevant, similar findings could potentially be visibility of natural environments and psychological distress. It can evaluated on large fresh water bodies, such as the North American be posited that the effects of visible natural environments on Great Lakes. If the type of water is salient, this may relate not only psychological distress and well-being are not instantaneous and to the visibility of the ocean, but the other sensory stimuli related that prolonged visual exposure might lead to the theorised ben- to the ocean, including the sound of waves and the smell of air efits. As a result of issues relating to data confidentiality, this study passing over the ocean. While this study was not able to assess was not able to use the exact home address locations for study these questions, these could be fruitful areas of future investiga- participants for measuring visibility of natural environments, ra- tion. Studies similar to this one could also confirm whether in- ther the population-weighted centroid of small areas were used creased visibility of blue space could promote mental well-being (meshblocks). Likewise, some continuous demographic variables and reduce stress in other cities. for participants such as income and age were supplied as ordinal Levels of visible green space were not associated with psy- groupings to conform to Ministry of Health confidentiality re- chological distress. The quality or composition of the vegetation quirements, leading to a loss of precision. In addition, the study within green spaces was not assessed in this work. It is possible did not include private green spaces such as backyard gardens, that native or high quality vegetation, for example, evokes differ- which have been found to be important for distress reduction ent responses from a viewer, compared to a sports fields. (Stigsdotter and Grahn (2004)). Nor did this study separately

Table 3 Results from models showing the association between 10 percentiles of visible green space and visible blue space and psychological distress.

Outcome¼K10 score Model 1: Visibility of green space Model 2: Visibility of blue space

Variables β p 95% CI β p 95% CI

All visible green space 0.09 0.455 0.32 0.14 All visible blue space 0.28 o0.001 0.41 0.15 Sex 0.51 0.324 1.51 0.50 0.40 0.412 1.37 0.56 Age 0.06 0.461 0.23 0.10 0.04 0.592 0.20 0.12 Income 0.99 0.003 1.70 0.34 0.89 0.006 1.53 0.25 Neighbourhood deprivation 0.12 0.428 0.18 0.43 0.21 0.079 0.02 0.45 Neighbourhood housing quality 0.45 0.002 0.73 0.17 0.33 0.021 0.61 0.05 Population density o 0.001 0.003 0.005 o 0.001 o 0.001 0.002 o 0.001 o 0.001 Crime rate o0.001 0.047 o0.001 o0.001 o0.001 0.023 o0.001 o0.001

Note: Items in bolded font indicate po0.05. 76 D. Nutsford et al. / Health & Place 39 (2016) 70–78

Table 4 Results from models showing the association between visible green space (i) within 300 m, (ii) at distances 300 m to 3 km, (iii) at distances 3–6 km, and (iv) at distances 6– 15 km, and psychological distress.

Outcome¼K10 score Model 3: Visibility of green space (o300 m) Model 4: Visibility of green space (300 m to 3 km)

Variables β p 95% CI β p 95% CI

Green space o300 m 0.09 0.299 0.08 0.26 Green space 300 m to 3 km 0.11 0.239 0.08 0.31 Sex 0.47 0.363 1.48 0.54 0.49 0.340 1.48 0.51 Age 0.05 0.540 0.22 0.12 0.07 0.429 0.23 0.10 Income 1.05 0.002 1.70 0.40 1.01 0.003 1.66 0.25 Neighbourhood deprivation 0.18 0.163 0.07 0.43 0.25 0.099 0.05 0.55 Neighbourhood housing quality 0.49 0.003 0.80 0.17 0.37 0.010 0.65 0.09 Population density o 0.001 0.006 o 0.001 o 0.001 o 0.001 0.004 o 0.001 o 0.001 Crime rate o0.001 0.008 o0.001 o0.001 o0.001 0.006 o0.001 o0.001

Outcome¼K10 Score Model 5: Visibility of green space (3–6 km) Model 6: Visibility of green space (6–15 km)

Variables β p 95% CI β p 95% CI

Green space 3–6km 0.13 0.096 0.29 0.02 Green space 6–15 km 0.57 0.05 0.591 0.22 0.13 Sex 0.42 0.406 1.41 0.57 0.48 0.340 1.48 0.51 Age 0.06 0.492 0.22 0.11 0.06 0.461 0.23 0.10 Income 0.99 0.003 1.65 0.34 0.99 0.004 1.66 0.32 Neighbourhood deprivation 0.12 0.373 0.14 0.38 0.17 0.157 0.07 0.41 Neighbourhood housing quality 0.47 0.001 0.75 0.18 0.39 0.019 0.72 0.07 Population density o 0.001 0.004 o 0.001 o 0.001 o 0.001 0.006 o 0.001 o 0.001 Crime rate o0.001 0.045 o-0.001 o0.001 o0.001 0.014 o0.001 o0.001

Note: All items in bolded font indicate po0.05.

between personal income and distress), we included an indicator Table 5 of home value as a covariate which attenuated all effects of Results from regression model showing the association between visible blue space greenspace on psychological distress. Still, even if the effect was and missing teeth. halved when accounting for residual confounding, our findings Outcome¼missing teeth Model 7: Visibility of blue space would suggest that for a 20% increase in visibility of blue space, a 0.28 lower K10 score could be expected. Furthermore, our finding β Variables p 95% CI that exposure to higher levels of blue space did not exhibit an All visible blue space 0.03 0.456 0.05 0.11 effect on tooth loss (while the effect of personal income remained) Sex 0.37 0.217 0.20 0.94 provides some evidence that the observed relationship between Age 0.32 o0.001 0.22 0.42 blue space and distress may not be purely a result of residual Income 0.41 0.037 0.79 0.03 confounding. Further limitations include the lack of adjustment for Neighbourhood deprivation 0.08 0.244 0.06 0.22 fl Neighbourhood housing quality 0.08 0.289 0.23 0.07 other individual characteristics which likely in uence mental Population density o 0.001 0.184 o 0.001 o0.001 health, including household composition (Simon, 2002), physical Crime rate o0.001 0.649 o 0.001 o0.001 health status, and immigrant status (Beiser et al., 2002) and other neighbourhood characteristics such as social fragmentation (Ivory Note: All items in bolded font indicate po0.05. et al., 2011) and isolation (Pearson et al., 2013). While other studies have used qualitative measures, such as photographical response analysis (White et al., 2010) and surveys evaluate types or quality of greenery (e.g., native vegetation). (Velarde et al., 2007) to assess positive effects of natural environ- There may, in fact, be differing effects of type or quality of green ments, this is the first identifiable study to quantitatively examine ‘ ’ space on mental health, whereby more natural green spaces and the relationship between visibility of blue and green spaces and those with native, dense or lush vegetation may be more bene- distress. Testing both green and blue spaces overcomes many lim- fi fl cial. The study only evaluated the in uence of visibility of natural itations of studies which either ignore blue space or treat it as a environments in residential settings and was unable to assess component of green space. This study separately assessed the ef- visibility in other settings, such as work, school or while travelling fects of visibility of green space and blue space on psychological (along routes). Future studies could quantify green and blue space distress. Future research improvements include a longitudinal study visibility in these other settings. If paired with data on time spent design which would allow more accurate temporal characterisation in each setting, a more precise measure of visual exposure to green of exposure to visible natural environments and subsequent out- and blue spaces could be usefully generated. comes. Longitudinal data could potentially assist in investigating Importantly, because personal income and neighbourhood the effect of changes in exposure on distress levels, via residential housing quality were significantly associated with distress in all mobility (Bell et al., 2014). Also, automated satellite image classifi- models, there is likely residual confounding related to income over cation techniques could be used as a method to incorporate private and above that measured in this study. Since housing prices tend and backyard green spaces as well as measures of streetscape and to be higher when accompanied by views of the ocean (Hamilton the techniques are already in place (De Ridder et al., 2004; Miller and Morgan, 2010) (and since we observed significant associations et al., 2009; Taylor et al., 2011). Satellite images could also assist in D. Nutsford et al. / Health & Place 39 (2016) 70–78 77 differentiating between ‘natural’ native vegetation and other out- scitotenv.2004.04.054. door green spaces. Inclusion of home address, work address and Gaio, E.J., Haas, A.N., Carrard, V.C., Oppermann, R.V., Albandar, J., Susin, C., 2012. Oral health status in elders from south Brazil: a population-based study. Ger- commuting patterns as well as amount of time spent in each setting odontology 29 (3), 214–223. could also improve the individual characterisation of exposure to Hamilton, S.E., Morgan, A., 2010. Integrating lidar, GIS and hedonic price modeling visible blue and green spaces throughout the day. Finally, future to measure amenity values in urban beach residential property markets. – studies using similar methodologies could help confirm whether Comput. Environ. Urban Syst. 34, 133 141. http://dx.doi.org/10.1016/j. compenvurbsys.2009.10.007. increased visibility of blue space could promote mental well-being Heerwagen, J.H., Orians, G.H., 1986. Adaptations to windowlessness: a study of the and reduce stress in other cities. use of visual decor in windowed and windowless offices. Environ. Behav. 18, 623–639. http://dx.doi.org/10.1177/0013916586185003. Herzog, T.R., 1985. A cognitive analysis of preference for waterscapes. J. Environ. Psychol. 5, 225–241. http://dx.doi.org/10.1016/S0272-4944(85)80024-4. 5. Conclusion Igarashi, M., Aga, M., Ikei, H., Namekawa, T., Miyazaki, Y., 2015. Physiological and psychological effects on high school students of viewing real and artificial – fi pansies. Int. J. Environ. Res. Public Health 12, 2521 2531. http://dx.doi.org/ This study, in the capital city of New Zealand, identi ed an 10.3390/ijerph120302521. association between increased views of blue space and decreased Ito, K., Aida, J., Wakaguri, S., Takeuchi, K., Noguchi, Y., Osaka, K., 2012. Socio- psychological distress while adjusting for covariates. In addition to economic inequalities in tooth loss among Japanese. In: Sasaki, K., Suzuki, O., Takahashi, N. (Eds.), Interface Oral Health Science 2011. Springer, Japan, theoretical evidence that improved health can be gained in re- pp. 291–293. sponse to natural background environments within cities, this Ivory, V.C., Collings, S.C., Blakely, T., Dew, K., 2011. When does neighbourhood research offers encouragement to further our understanding of the matter? Multilevel relationships between neighbourhood social fragmentation and mental health. Soc. Sci. Med. 72, 1993–2002. http://dx.doi.org/10.1016/j. relationship between visible blue spaces and mental well-being. socscimed.2011.04.015. Confirmation of these findings in other locations is important and Kaplan, R., 2001. The nature of the view from home: psychological benefits. En- would be of value to the urban health field. If confirmed, there viron. Behav. 33, 507–542. http://dx.doi.org/10.1177/00139160121973115. may be opportunities to promote mental health through the Karmanov, D., Hamel, R., 2008. Assessing the restorative potential of contemporary urban environment(s): beyond the nature versus urban dichotomy. Landsc. designation of state housing and designate a proportion of af- Urban Plan. 86, 115–125. http://dx.doi.org/10.1016/j.landurbplan.2008.01.004. fordable homes in locations with oceanic views or to designate Kearney, A.R., 2006. Residential development patterns and neighborhood sa- – some higher levels of multi-family high rise buildings to affordable tisfaction: impacts of density and nearby nature. Environ. Behav. 38, 112 139. http://dx.doi.org/10.1177/0013916505277607. ownership or creative to maximise exposure to blue Kearns, R.A., Collins, D., Conradson, D., 2014. A healthy island blue space: from spaces. space of detention to site of sanctuary. Health Place 30, 107–115. http://dx.doi. org/10.1016/j.healthplace.2014.08.005. Kessler, R.C., Barker, P.R., Colpe, L.J., Epstein, J.F., Gfroerer, J.C., Hiripi, H., et al., 2003. Screening for serious mental illness in the general population. Arch. Gen. Competing interests Psychiatry 60, 184. http://dx.doi.org/10.1001/archpsyc.60.2.184. Kuo, F.E., Bacaicoa, M., Sullivan, W.C., 1998. Transforming inner-city landscapes: trees, sense of safety, and preference. Environ. Behav. 30, 28–59. http://dx.doi. None declared. org/10.1177/0013916598301002. Kweon, B.-S., Sullivan, W.C., Wiley, A.R., 1998. Green common spaces and the social integration of inner-city older adults. Environ. Behav. 30, 832–858. http://dx. Acknowledgements doi.org/10.1177/001391659803000605. Laumann, K., Gärling, T., Stormark, K.M., 2003. Selective attention and heart rate responses to natural and urban environments. J. Environ. Psychol. 23, 125–134. We thank the Ministry for the Environment, Land Information http://dx.doi.org/10.1016/S0272-4944(02)00110-X. New Zealand and the Department of Conservation for access to Lim, S.S., Vos, T., Flaxman, A.D., Danaei, G., Shibuya, K., Adair-Rohani, H., et al., 2012. A comparative risk assessment of burden of disease and injury attributable to their land use datasets. This research was carried out as part of the 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic Geohealth Laboratory work programme at the University of Can- analysis for the global burden of disease study 2010. Lancet 380, 2224–2260. terbury, funded by the New Zealand Ministry of Health. http://dx.doi.org/10.1016/S0140-6736(12)61766-8. Miles, R., Coutts, C., Mohamadi, A., 2012. Neighborhood urban form, social en- vironment, and depression. J. Urban Health 89, 1–18. http://dx.doi.org/10.1007/ s11524-011-9621-2. References Miller, D., Horne, P., Donnelly, D., Morrice, J., 2009. Mapping physical and visual accessibility of Greenspace. Environ. Hum. Health 44, 1–8 http://www.hutton. ac.uk/sites/default/files/files/projects/GreenHealth-InformationNote6-Map Andrews, G., Slade, T., 2001. Interpreting scores on the Kessler psychological dis- ping-physical-and-visual-access-to-greenspaces.pdf. tress scale (K10). Aust. N. Z. J. Public Health 25, 494–497. http://dx.doi.org/ Ministry of Health, 2013. New Zealand Health Survey Methodology Report 2012/13. 10.1111/j.1467-842X.2001.tb00310.x. Ministry of Health, Wellington http://www.health.govt.nz/publication/new- Ashbullby, K., Pahl, S., Webley, P., White, M., 2013. The beach as a setting for fa- zealand-health-survey-methodology-report-2012-13. milies' health promotion: a qualitative study with parents and children living in Moore, E., 1981. A Prison environment's effect on health care service demands. J. coastal regions in Southwest England. Health Place. http://dxdoi.org/10.1016/j. Environ. Syst. 11, 17–34. healthplace.2013.06.005. Newell, P.B., 1997. A cross-cultural examination of favorite Places. Environ. Behav. Barton, J., Pretty, J., 2010. What is the best dose of nature and green exercise for 29, 495–514. http://dx.doi.org/10.1177/001391659702900403. improving mental health? A multi-study analysis. Environ. Sci. Technol. 44, Nutsford, D., Pearson, A.L., Kingham, S., 2013. An ecological study investigating the 3947–3955. http://dx.doi.org/10.1021/es903183r. association between access to and mental health. Public Beiser, M., Hou, F., Hyman, I., Tousignant, M., 2002. Poverty, family process, and the Health. http://dxdoi.org/10.1016/j.puhe.2013.08.016. mental health of immigrant children in Canada. Am. J. Public Health 92, Nutsford, D., Reitsma, F., Pearson, A.L., Kingham, S., 2015. Personalising the view- 220–227. http://dx.doi.org/10.2105/AJPH.92.2.220. shed: visibility analysis from the human perspective. Appl. Geogr. 62, 1–7. http: Bell, S.L., Phoenix, C., Lovell, R., Wheeler, B.W., 2014. Green space, health and //dx.doi.org/10.1016/j.apgeog.2015.04.004. wellbeing: making space for individual agency. Health Place 30, 287–292. http: Oakley Browne, M.A., Wells, J.E., Scott, K.M., McGee, M.A., 2010. The Kessler psy- //dx.doi.org/10.1016/j.healthplace.2014.10.005. chological distress scale in Te Rau Hinengaro: the New Zealand mental health BRANZ, 2005. New Zealand 2005 House Condition Survey. BRANZ, Wellington, p. survey. Aust. N. Z. J. Psychiatry 44, 314–322. http://dx.doi.org/10.3109/ 146. 00048670903279820. Braubach, M., Fairburn, J., 2010. Social inequities in environmental risks associated Pearson, A.L., Griffin, E., Davies, A., Kingham, S., 2013. An ecological study of the with housing and residential location—a review of evidence. Eur. J. Public relationship between socioeconomic isolation and mental health in the most Health 20 (1), 36–42. deprived areas in Auckland, New Zealand. Health Place 19, 159–166. http://dx. Buchwald, S., Biffar, K.T., Harb, R., Holtfreter, A., Meisel P, B., 2013. Tooth loss and doi.org/10.1016/j.healthplace.2012.10.012. periodontitis by socio-economic status and inflammation in a longitudinal Pretty, J., Peacock, J., Sellens, M., Griffin, M., 2005. The mental and physical health population-based study. J. Clin. Periodontol. 40, 203–211. outcomes of green exercise. Int. J. Environ. Health Res. 15, 319–337. http://dx. De Ridder, K., Adamec, V., Bañuelos, A., Bruse, M., Bürger, M., Damsgaard, O., et al., doi.org/10.1080/09603120500155963. 2004. An integrated methodology to assess the benefits of urban green space. Putra S.Y., Yang P.P.-J, 2005. Analysing mental geography of residential environment Sci. Total Environ. 334–335, 489–497. http://dx.doi.org/10.1016/j. in Singapore using GIS-based 3D visibility analysis. In: Proceedings of a 78 D. Nutsford et al. / Health & Place 39 (2016) 70–78

Conference ‘Doing, Thinking, Feeling Home: the Mental Geography of Re- Psychol. 11, 201–230. http://dx.doi.org/10.1016/S0272-4944(05)80184-7. sidential Environments’. Delft, The Netherlands. Völker, S., Kistemann, T., 2011. The impact of blue space on human health and well- Richardson, E., Pearce, J., Mitchell, R., Kingham, S., 2013. Role of physical activity in being – Salutogenetic health effects of inland surface waters: a review. Int. J. the relationship between urban green space and health. Public Health 127, Hyg. Environ. Health 214, 449–460. http://dx.doi.org/10.1016/j. 318–324. http://dx.doi.org/10.1016/j.puhe.2013.01.004. ijheh.2011.05.001. Richardson, E., Pearce, J., Mitchell, R., Day, P., Kingham, S., 2010. The association Völker, S., Kistemann, T., 2013. “I'm always entirely happy when I'm here!” Urban between green space and cause-specific mortality in urban New Zealand: an blue enhancing human health and well-being in Cologne and Düsseldorf, ecological analysis of green space utility. BMC Public Health 10, 240. http://dx. Germany. Soc. Sci. Med. 78, 113–124. http://dx.doi.org/10.1016/j. doi.org/10.1186/1471-2458-10-240. socscimed.2012.09.047. Rose, E., 2012. A psychoanalytic approach for understanding how therapeutic Velarde, M.D., Fry, G., Tveit, M., 2007. Health effects of viewing landscapes – fi – landscapes bene t health and wellbeing. Health Place 18, 1381 1387. http://dx. landscape types in environmental psychology. Urban Urban Green 6, 199–212. doi.org/10.1016/j.healthplace.2012.07.002. http://dx.doi.org/10.1016/j.ufug.2007.07.001. Simon, R.W., 2002. Revisiting the relationships among gender, marital status, and Wennström, A., Ahlquist, M., Stenman, U., Björkelund, C., Hakeberg, M., 2013. mental health. Am. J. Sociol. 107, 1065–1096. http://dx.doi.org/10.1086/339225. Trends in tooth loss in relation to socio-economic status among Swedish wo- Statistics New Zealand, 2013. Narrowing gap between Māori and non-Māori life men, aged 38 and 50 years: repeated cross-sectional surveys 1968-2004. BMC expectancy [press release]. Wellington. http://www.stats.govt.nz/browse_for_ Oral Health 13 (63), 1–8. stats/health/life_expectancy/NZLifeTables_MR10-12.aspx. Wheeler, B.W., White, M., Stahl-Timmins, W., Depledge, M., 2012. Does living by the Stigsdotter, U.K., Grahn, P.A., Garden at your doorstep may reduce stress – private coast improve health and wellbeing? Health Place 18, 1198–1201. http://dx.doi. gardens as restorative environments in the city. In: Proceedings of Open Space, org/10.1016/j.healthplace.2012.06.015. People Space: International Conference on Inclusive Environments OPENspace White, I., Royston, P., Wood, A.M., 2011. Multiple imputation using chained equa- Research Centre. Edinburgh, Scotland, UK. 2004. Available: 〈http://friskina – turen.org/media/terapihaven_a_garden_at_your_doorstep.pdf〉. tions: issues and guidance for practice. Stat. Med. 30, 377 399. http://dx.doi. Sugiyama, T., Leslie, E., Giles-Corti, B., Owen, N., 2008. Associations of neighbour- org/10.1002/sim.4067. hood greenness with physical and mental health: do walking, social coherence White, M., Alcock, I., Wheeler, B.W., Depledge, M., 2013. Coastal proximity, health and local social interaction explain the relationships? J. Epidemiol. Community and well-being: results from a longitudinal panel survey. Health Place 23, Health 62. http://dxdoi.org/10.1136/jech.2007.064287. 97–103. http://dx.doi.org/10.1016/j.healthplace.2013.05.006. Taylor, B.T., Fernando, P., Bauman, A., Williamson, A., Craig, J.C., Redman, S., 2011. White, M., Smith, A., Humphryes, K., Pahl, S., Snelling, D., Depledge, M., 2010. Blue Measuring the quality of public open space using Google Earth. Am. J. Prev. space: the importance of water for preference, affect, and restorativeness rat- Med. 40, 105–112. http://dx.doi.org/10.1016/j.amepre.2010.10.024. ings of natural and built scenes. J. Environ. Psychol. 30, 482–493. http://dx.doi. Ulrich, R., 1981. Natural versus urban scenes: some psychophysiological effects. org/10.1016/j.jenvp.2010.04.004. Environ. Behav. 13, 523–556. http://dx.doi.org/10.1177/0013916581135001. Wu, B., Hybels, C., Liang, J., Landerman, L., Plassman, B., 2014. Social stratification Ulrich, R., Simons, R.F., Losito, B.D., Fiorito, E., Miles, M.A., Zelson, M., 1991. Stress and tooth loss among middle-aged and older Americans from 1988 to 2004. recovery during exposure to natural and urban environments. J. Environ. Community Dent. Oral Epidemiol. 42 (6), 495–502.