Few people would deny an intuitive sense of increased wellbeing when spending time in beautiful locations. Here, we ask: can we quantify the relationship between environmental aesthetics and human health? We draw on data from Scenic-Or-Not, a website that crowdsources ratings of “scenicness” for geotagged photographs across Great Britain, in combination with data on citizen-reported health from the Census for England and Wales. We find that inhabitants of more scenic environments report better health, across urban, suburban and rural areas, even when taking core socioeconomic indicators of deprivation into account, such as income, employment and access to services. Our results provide evidence in line with the striking hypothesis that the aesthetics of the environment may have quantifiable consequences for our wellbeing.
Few people would deny that spending time in areas of beautiful scenery results in a sense of increased wellbeing. Yet, what if scenic environments had an impact on our health? While the question of how the aesthetics of the environment might relate to our wellbeing has been of interest to researchers for many years, studies to date have been limited to investigations of differences in reactions to urban and natural scenes1,2,3,4,5,6,7,8,9, research into the role of greenspace and vegetation in urban environments7,10,11,12,13,14,15,16,17,18,19,20, or analyses of small-scale survey data21,22,23 due to the impracticality of gathering large-scale data on humans’ perception of the environment.
The ubiquitous presence of the Internet in today’s society, however, has led to the creation of a new source of information on human behavior: large datasets of online activity. Analysis of data obtained from platforms such as Google, Flickr, Wikipedia and Twitter has already led to a range of new insights into human behavior in the real world24,25,26,27,28,29,30,31,32,33,34,35,36,37.
Here, we use data from Scenic-Or-Not38, a website that crowdsources ratings of “scenicness,” in order to develop a better understanding of how the aesthetics of the environment may impact our health. Scenic-Or-Not users rate random geotagged photographs of Great Britain on an integer scale of 1–10, where 10 indicates “very scenic” and 1 indicates “not scenic”. The Scenic-Or-Not website comprises 217,000 images covering nearly 95% of the 1 km grid squares of Great Britain. As of August 2014, the Scenic-Or-Not dataset contained 1.5 million votes.
We first explore the nature of photographs rated as scenic, including their color composition. Next, we evaluate to what degree scenicness relates to an objective measurement, green land cover, as several studies analyzing green land cover data indicate that an abundance of greenspace results in increased human wellbeing11,12,13,14,15,16,17,18,19,20.
Finally, we investigate whether scenicness can explain more of the variance in geographic self-reports of health than data on green land cover are able to alone. In previous studies, self-reported health has been shown to be a strong predictor of subsequent mortality rates39,40.
We analyze this relationship across urban, suburban and rural areas, to take into account that the definition of scenicness may vary depending on environmental context. As less scenic areas may also be areas of higher air pollution, we run an additional analysis including modeled estimates of concentrations of the following pollutants: sulphur dioxide (SO2), oxides of nitrogen (NOx), particles and fine particles (PM10 and PM2.5), benzene (C6H6), carbon monoxide (CO) and ozone (O3)41.
Examining a sample of Scenic-Or-Not photographs helps us understand what constitutes scenicness (Fig. 1). Visual inspection of photographs with the highest rating for scenicness reveals that they tend to contain landscapes with broad open areas and essentially no manmade structures (Fig. 1A). Unscenic photographs appear to contain mostly manmade structures including roads, buildings and cars (Fig. 1B). Crucially, we observe that even unscenic images can have large areas of greenspace, though often with manmade structures that obstruct the view of the greenspace. The existence of such photographs provides initial evidence that the presence of green in an image is not sufficient for it to be considered scenic.
We therefore assess how perceived colors, in general, correspond with objective reports of the environment. We examine each image from Scenic-Or-Not on a per-pixel level, with each pixel being allocated to one of eleven colors that constitute principal colors in the English vocabulary (black, blue, brown, grey, green, orange, pink, purple, red, white, yellow). As color naming varies from one individual to another42, we draw on crowdsourced data generated through an online survey of 1.5 million participants to determine to which color a pixel should be allocated. More details of this procedure can be found in the Supplementary Information.
Figure 1C depicts the relationship between scenicness ratings and the proportion of each color found in the images. Although all colors are significantly correlated with scenicness (all |τ |s > 0.01, all ps < 0.001, N = 206,873, Kendall’s rank correlation), Fig. 1C suggests that the association between color and scenicness is complex. While one may expect scenicness ratings to steadily increase as the proportion of green in images increases, visual inspection of the data instead reveals that highly scenic images tend to have a high proportion of blue, grey and brown. This may be due to open skies and mountains in highly-rated images. Less scenic images tend to be mainly grey with higher proportions of black and white, but also more green pixels than the highly-rated scenic images.
Scenicness therefore does not appear to constitute a simple predominance of green areas. In order to further explore how scenicness compares to greenspace, we compare green land cover, as reported in the Generalised Land Use Database Statistics for England 200543, to the scenicness ratings extracted from Scenic-Or-Not. The Generalised Land Use Database covers all land in England. We analyze land use data describing England at the level of Lower Layer Super Output Areas (LSOA), which are geographic areas with an average population size of 1,600, defined by the Office of National Statistics for statistical analyses, with areas ranging between 0.018 square km to 684 square km. To combine the land use data with the Scenic-Or-Not ratings, which are provided as ratings for one geotagged photograph in each 1 km grid square, we calculate the average scenic rating of all Scenic-or-Not photographs taken within each LSOA. Using this method, scenicness ratings are available for 16,907 out of all 32,844 English LSOAs.
We find that scenicness and green land cover are significantly correlated (τ = 0.2, p < 0.001, N = 128,213, Kendall’s rank correlation). The relationship between scenicness (Fig. 2A) and green land cover (Fig. 2B) is apparent upon inspection of the two maps. However, the correlation is not very strong in terms of effect size, suggesting that scenicness and green land cover are not necessarily the same. For example, in the East of England, green land cover and scenicness diverge considerably. We therefore investigate to what extent these two different variables can help us understand geographic differences in health.
We draw on geographic data from the 2011 Census for England and Wales capturing respondents’ classification of their health as “Very good or good”, “Fair” or “Bad or very bad”44. Following Mitchell and Popham14, we calculate health rates using the Standardized Morbidity Ratio (SMR), which is the ratio of the observed to the expected number of cases of bad health for a particular population, taking the age and gender of inhabitants into account. Following standard practice in other studies13,14,15,16,17, we control for socioeconomic characteristics that may be linked with reports of health. We use deprivation data from the relevant domains of the 2010 English Indices of Deprivation45: Income Deprivation, Employment Deprivation, Education Skills and Training Deprivation, Barriers to Housing and Services, Living Environment Deprivation and Crime. The value of these indices increases in line with the proportion of people who experience deprivation in each domain.
Finally, in order to explore whether there is any variation in the association between health and scenicness across urban, suburban and rural areas, we use the 2011 Rural-Urban Classification46. For the purposes of this study, “urban” is defined using the category “Urban Major Conurbation” from the 2011 Rural-Urban Classification. The remaining urban categories are deemed suburban. Scenicness data is available for 3,945 urban LSOAs, 7,781 suburban LSOAs and 5,182 rural LSOAs.
We investigate to what extent geographic differences in self-reported health, as measured using the Standard Morbidity Ratio (SMR), can be explained by scenicness and greenspace. To carry out this analysis, we build a Conditional Auto Regressive (CAR) model, which takes spatial autocorrelation into account. As with time series data, where observations that are closer in time may be correlated and hence violate the linear regression assumption that observations are independent, spatial data may also exhibit autocorrelation where neighboring areas may be more or less alike47.
We confirm the need for this approach by initially building two different linear regression models to predict poor reports of health at the level of LSOAs. Both models include the socioeconomic variables describing estimates of deprivation across income, employment, education, housing, crime and living conditions. The first model additionally includes scenicness only, while the second model additionally includes greenspace only. We then run a Moran’s I test on the residuals of both models to test for spatial auto-correlation. Both models exhibit significant spatial autocorrelation in the residuals of the linear regression models (Scenic model: Moran’s I = 0.143, p < 0.001, N = 16,907; Greenspace model: Moran’s I = 0.136, p < 0.001, N = 16,907).
We therefore investigate to what extent geographic differences in reports of health can be explained by scenicness and greenspace, by running three different CAR models. Again, all models include the socioeconomic variables mentioned above. The first CAR model additionally includes both scenicness and greenspace. However, as our previous correlation analysis indicates that scenicness is significantly correlated with greenspace (τ = 0.2, p < 0.001, N = 128,213), we run two additional CAR models: one that includes greenspace only and a second that includes scenicness only.
Table 1 provides results of the CAR model that includes both scenicness and greenspace. Across the entire English dataset, we find that lower values of scenicness are significantly associated with reports of worse health (β = −0.008, p < 0.001, N = 16,907), even when taking a wide range of deprivation variables into account. We also find that this relationship holds across urban, suburban and rural areas (Urban: β = −0.007, p = 0.012, N = 3,944, Suburban: β = −0.005, p = 0.007, N = 7,781, Rural: β = −0.012, p < 0.001, N = 5,182). However, in this model, while greenspace is not associated with reports of poor health in general (all ps > 0.22 for urban, suburban and rural areas as well as England as a whole), more greenspace is significantly correlated with reports of poor health in suburban areas (β = −0.020, p = 0.024, N = 7,781).
In the second CAR model (Table 2) with scenicness removed, less greenspace is significantly correlated with reports of worse health when considering England as a whole (β = −0.019, p = 0.003, N = 16,907). However, this effect is not significant for urban areas, suburban areas or rural areas when considered separately.
In the third CAR model (Table 3), with greenspace removed, the results are similar to the first model that includes both scenicness and greenspace. Lower ratings of scenicness are significantly associated with reports of worse health across the entire English dataset (β = −0.008, p < 0.001, N = 16,907), as well as urban, suburban and rural areas when considered individually (Urban: β = −0.009, p = 0.010, N = 3,944, Suburban: β = −0.004, p = 0.028, N = 7,781, Rural: β = −0.011, p < 0.001, N = 5,182).
Finally, in order to determine which of the three models provides the best fit for predicting reports of poor health, we rank all three models in terms of their Akaike Information Criterion (AIC) values. This provides a measure of the likelihood of a given model and its free parameters. In order to compare the fit of the models to each other, AIC values are then transformed to Akaike weights (AICw) following the method proposed by Wagenmakers and Farrel48. These weights can be interpreted as the probability of each model, given the data. This model comparison indicates that models including scenicness have more explanatory power than the model with only greenspace (Fig. 2D).
One final concern could be that less scenic areas may also be areas of higher pollution, which could impact the health of local residents. Our examination of water pollution data held by the World Bank suggests that 100% of the United Kingdom population has access to an “improved water source” (defined as water that has been modified or protected from outside contamination)49. Furthermore, in accordance with the guidelines specified by the EU Drinking Water Directive for England and Wales, the Drinking Water Inspectorate (the independent regulator of public drinking water supplies in England and Wales) reports that only 0.04% of the 1.9 million tests conducted in 2011 failed to meet one of the chemical or microbiological standards50. Due to the overwhelmingly high level of water quality in the UK, we therefore conclude that further analysis of water pollution is unwarranted.
In our original model, a measure of air pollution is included in the Living Environment Deprivation variable. However, as air pollution is an ongoing health concern, particularly in urban areas, we run a further analysis including modeled estimates of concentrations of the following pollutants: sulphur dioxide (SO2), oxides of nitrogen (NOx), particles and fine particles (PM10 and PM2.5), benzene (C6H6), carbon monoxide (CO) and ozone (O3). Air quality data at a 1 km2 resolution was obtained from the UK Air Information Resource for the year 2011.
We build a CAR model that includes both scenicness and greenspace, as well as the air pollutant variables (Table S2). Even after explicitly including the air pollutant measurements, we continue to find that lower scenicness is significantly associated with reports of worse health across the entire English dataset (β = −0.006, p < 0.001, N = 16,907) as well as urban, suburban and rural areas when considered individually (Urban: β = −0.008, p = 0.010, N = 3,944, Suburban: β = −0.004, p = 0.0259, N = 7,781, Rural: β = −0.008, p = 0.007, N = 5,182).
We also rank three models in terms of their Akaike Information Criterion (AIC) values: model 1 which includes scenicness only; model 2 which includes greenspace only and model 3 which includes scenicness and greenspace. AIC values are again transformed to Akaike weights (AICw) in order to facilitate interpretation. This model comparison confirms our earlier findings that the models including scenicness have more explanatory power than the model with only greenspace (Fig. S4).
In conclusion, we use crowdsourced data from the website Scenic-Or-Not, where people rate the “scenicness” of geotagged photographs of Great Britain, in order to explore whether a relationship exists between the scenicness of an environment and the reported health of its inhabitants. We also analyze the color composition of each image to help us assess how scenicness differs from greenspace, traditionally measured by green land cover. Crucially, we investigate whether scenicness may offer new insights into geographic differences in reports of health, beyond the variation explained by green land cover alone.
We find that inhabitants of more scenic environments report better health, across urban, suburban and rural areas. This result holds even when taking core socioeconomic indicators of deprivation, such as income and data on air pollution into account. Importantly, we find that differences in reports of health can be better explained by the scenicness of the local environment than by measurements of greenspace.
Our color analysis also reveals that scenicness does not simply constitute large areas of green. Indeed, we find that the most scenic areas do not contain the most green, but rather contain high proportions of blue, grey and brown. It seems that scenic environments can include large areas of water, open blue skies or mountainous landscapes. Green areas congested with manmade objects such as buildings and roads may deter the enjoyment of greenspace and may cause a decrease in scenicness ratings.
Our analysis of online crowdsourced datasets indicates that the interaction between the built and natural environment and human experience may be more complex than can be explained by studies limited to using objective datasets such as greenspace. We present evidence that aesthetics, as measured by “scenicness,” may play a central role in the environment’s ability to affect our health. These findings provide evidence that the aesthetics of the environment may have a greater practical impact than previously believed. In order to ensure the wellbeing of local inhabitants, it may therefore be valuable to consider the aesthetics of the environment when embarking upon large projects to build new parks, housing developments or highways.
Scenic-Or-Not scenic ratings
Scenic-Or-Not presents users with random geotagged photographs of Great Britain, most of which have been taken at eye level. Users can rate photographs on an integer scale of 1–10, where 10 indicates “very scenic” and 1 indicates “not scenic”. The Scenic-Or-Not dataset comprises images sourced from Geograph (http://www.geograph.org.uk/) covering nearly 95% of the 1 km grid squares of Great Britain. Images included in this dataset may have been submitted via Geograph at any point since March 2005. Scenicness ratings can be retrieved from Scenic-Or-Not for images that have been rated three times or more. As of August 2, 2014, 1,529,927 ratings were available for 212,057 images.
We retrieved data on scenicness ratings by accessing the Scenic-Or-Not website (http://scenic.mysociety.org) on August 2, 2014. Color naming data was retrieved from XKCD (http://blog.xkcd.com/2010/05/03/color-survey-results/) on May 23, 2014. Data on self-reported health from the 2011 Census for England and Wales was retrieved on July 16, 2014 from Nomis (https://www.nomisweb.co.uk/census/2011). Green land cover data was retrieved on June 30, 2014 from Neighbourhood Statistics (http://www.neighbourhood.statistics.gov.uk/). Data on the English indices of deprivation was retrieved on July 1, 2014 from Neighbourhood Statistics (http://www.neighbourhood.statistics.gov.uk/). Data on the Rural Urban Classification was retrieved on June 24, 2014, also from Neighbourhood Statistics (http://www.neighbourhood.statistics.gov.uk/). Data on “improved water sources” was retrieved on May 19, 2015 from the World Bank (http://data.worldbank.org/). Data on the quality of drinking water for England and Wales was retrieved on May 19, 2015 from the Drinking Water Inspectorate (http://dwi.defra.gov.uk/). Modeled pollution data was retrieved on May 19, 2015 from the Air Information Resource (http://uk-air.defra.gov.uk/). The boundary data for LSOAs in England was retrieved on December 17, 2013 from the UK Data Service Census Support (http://census.edina.ac.uk/).
Conditional Auto Regressive (CAR) model
When working with spatial data, it is reasonable to assume that observations in neighboring areas may be more or less alike simply due to their proximity, and hence exhibit autocorrelation47. We confirm this by first running a Moran’s I test, which measures whether spatial autocorrelation is present in the data. Due to this autocorrelation, we cannot run a simple linear regression analysis, as spatial dependencies would exist in the error term. Hence, we run our analysis using a conditional autoregressive prior (CAR), as initially proposed by Besag and colleagues51,52, which captures spatial dependence between neighbors through an adjacency matrix of the areal units.
The CAR model quantifies the spatial relationship in the data by including a conditional distribution in the error term for area i, ei. The conditional distribution of ei is thus represented as:
where ej~i is the e–i vector including only neighboring areas of i; e–i is the vector of all the errors terms except for e–i itself; and cij are dependence parameters used to represent the spatial dependence between the areas.
Akaike weights (AICw)
In order to determine which model best captures variance in the data on poor health, we calculate the Akaike weights (AICws), following the method proposed by Wagenmakers and Farrel48, as the AIC values themselves are difficult to interpret on their own. We derive AICws by first identifying the model with the lowest AIC. For each model, we then calculate an AIC difference, by determining the difference between the lowest AIC and the model’s AIC. We next determine the relative likelihood of each model, following the method described in Wagenmakers and Farrel48. To determine the AICws we normalize these likelihoods, such that across all models they sum to one. The resulting AICws can be interpreted as the probability of each model given the data.
Color composition of images
We examine each image from Scenic-Or-Not on a per-pixel level, with each pixel being allocated to one of eleven colors that constitute the principal colors in the English vocabulary (black, blue, brown, grey, green, orange, pink, purple, red, white, yellow). As color naming varies from one individual to another42, we draw on crowdsourced data generated through an online survey of 1.5 million participants to determine to which color a pixel should be allocated. In this survey, participants were shown an area filled with a random fully-saturated color on both black and white backgrounds and asked to name the color. These responses were then used to create a list of the dominant color names corresponding to fully saturated RGB (Red, Green, Blue) values. We use this data in order to determine where color boundaries should be drawn: for example, where “brown” ends and “green” begins. The RGB colors are converted to the HSV (Hue, Saturation, Value) color space and each pixel is matched to the closest corresponding color, based on its hue parameter. The nature of the relationship between HSV and RGB space is such that all possible hues are covered by all fully saturated RGB colors. As black, grey and white do not have a defined hue, these color boundaries were determined based on a combination of the levels of “Saturation” and “Value” (Fig. S3).
How to cite this article: Seresinhe, C. I. et al. Quantifying the Impact of Scenic Environments on Health. Sci. Rep.5, 16899; doi: 10.1038/srep16899 (2015).
Ulrich, R. S. Visual landscapes and psychological well‐being. Landscape Res. 4, 17–23 (1979).
Ulrich, R. View through a window may influence recovery. Science 224, 224–225 (1984).
Parsons, R. The potential influences of environmental perception on human health. J. Environ. Psychol. 11, 1–23 (1991).
Ulrich, R. S. et al. Stress recovery during exposure to natural and urban environments. Journal of Environmental Psychology 11, 201–230 (1991).
Kaplan, S. The restorative benefits of nature: Toward an integrative framework. J. Environ. Psychol. 15, 169–182 (1995).
Kaplan, R. The Nature of the View from Home: Psychological Benefits. Environment and Behavior 33, 507–542 (2001).
Abraham, A., Sommerhalder, K. & Abel, T. Landscape and well-being: a scoping study on the health-promoting impact of outdoor environments. Int. J. Public Health 55, 59–69 (2010).
Bratman, G. N., Daily, G. C., Levy, B. J. & Gross, J. J. The benefits of nature experience: Improved affect and cognition. Landscape Urban Plan. 138, 41–50 (2015).
Bratman, G. N., Hamilton, J. P., Hahn, K. S., Daily, G. C. & Gross, J. J. Nature experience reduces rumination and subgenual prefrontal cortex activation. P. Natl. Acad. Sci. USA 112, 8567–8572 (2015).
Smardon, R. C. Perception and aesthetics of the urban environment: Review of the role of vegetation. Landscape Urban Plan. 15, 85–106 (1988).
Mitchell, R. & Popham, F. Effect of exposure to natural environment on health inequalities: an observational population study. Lancet 372, 1655–1660 (2008).
Ellaway, A., Macintyre, S. & Bonnefoy, X. Graffiti, greenery and obesity in adults: secondary analysis of European cross sectional survey. BMJ 331, 611–612 (2005).
White, M. P., Alcock, I., Wheeler, B. W. & Depledge, M. H. Would you be happier living in a greener urban area? A fixed-effects analysis of panel data. Psychol. Sci. 24, 920–928 (2013).
Mitchell, R. & Popham, F. Greenspace, urbanity and health: relationships in England. J. Epidemiol. Commun. Health 61, 681–683 (2007).
Maas, J., Verheij, R. A., Groenewegen, P. P., de Vries, S. & Spreeuwenberg, P. Green space, urbanity and health: how strong is the relation? J. Epidemiol. Commun. Health 60, 587–592 (2006).
van den Berg, A. E., Maas, J., Verheij, R. A. & Groenewegen, P. P. Green space as a buffer between stressful life events and health. Soc. Sci. Med. 70, 1203–1210 (2010).
de Vries, S., Verheij, R. A., Groenewegen, P. P. & Spreeuwenberg, P. Natural environments—healthy environments? An exploratory analysis of the relationship between greenspace and health. Environ. Plann. A 35, 1717–1731 (2003).
Sugiyama, T., Leslie, E., Giles-Corti, B. & Owen, N. Associations of neighbourhood greenness with physical and mental health: do walking, social coherence and local social interaction explain the relationships? J. Epidemiol. Commun. Health 62, E9 (2008).
Takano, T., Nakamura, K. & Watanabe, M. Urban residential environments and senior citizens’ longevity in megacity areas: the importance of walkable green spaces. J. Epidemiol. Commun. Health 56, 913–918 (2002).
Kardan, O. et al. Neighborhood greenspace and health in a large urban center. Sci. Rep. 5, 11610–11610 (2015).
Giles-Corti, B. et al. Increasing walking: How important is distance to, attractiveness and size of public open space? Am. J. Prev. Med. 28, 169–176 (2005).
Humpel, N., Owen, N., Iverson, D., Leslie, E. & Bauman, A. Perceived environment attributes, residential location and walking for particular purposes. Am. J. Prev. Med. 26, 119–125 (2004).
Ball, K., Bauman, A., Leslie, E. & Owen, N. Perceived Environmental Aesthetics and Convenience and Company Are Associated with Walking for Exercise among Australian Adults. Prev. Med. 33, 434–440 (2001).
King, G. Ensuring the Data-Rich Future of the Social Sciences. Science 331, 719–721 (2011).
Vespignani, A. Predicting the Behavior of Techno-Social Systems. Science 325, 425–428 (2009).
Watts, D. J. A twenty-first century science. Nature 445, 489–489 (2007).
Lazer, D. et al. Computational Social Science. Science 323, 721–723 (2009).
Moat, H. S., Preis, T., Olivola, C. Y., Liu, C. & Chater, N. Using big data to predict collective behavior in the real world. Behav. Brain. Sci. 37, 92–93 (2014).
Goodchild, M. F. Citizens as sensors: the world of volunteered geography. GeoJournal 69, 211–221 (2007).
Bollen, J., Mao, H. & Zeng, X. Twitter mood predicts the stock market. J. Comput. Sci. 2, 1–8 (2011).
Choi, H. & Varian, H. Predicting the Present with Google Trends. Econ. Rec. 88, 2–9 (2012).
Moat, H. S. et al. Quantifying Wikipedia Usage Patterns Before Stock Market Moves. Sci. Rep. 3, 1801 (2013).
Preis, T., Moat, H. S. & Stanley, H. E. Quantifying Trading Behavior in Financial Markets Using Google Trends. Sci. Rep. 3, 1684 (2013).
Preis, T., Moat, H. S., Bishop, S. R., Treleaven, P. & Stanley, H. E. Quantifying the Digital Traces of Hurricane Sandy on Flickr. Sci. Rep. 3, 3141 (2013).
Botta, F., Moat, H. S. & Preis, T. Quantifying crowd size with mobile phone and Twitter data. R. Soc. Open Sci. 2, 150162 (2015).
Manrique, P. et al. Context Matters: Improving the Uses of Big Data for Forecasting Civil Unrest, In Intelligence and Security Informatics (ISI) IEEE International Conference 169–172 (IEEE, Washington, 2013).
Barchiesi, D., Moat, H. S., Alis, C., Bishop, S. & Preis, T. Quantifying international travel flows using Flickr. PLOS ONE 10, e0128470 (2015).
Scenic-Or-Not. Scenic ratings. Available at: http://scenic.mysociety.org (Accessed August 2, 2014).
Idler, E. L. & Benyamini, Y. Self-Rated Health and Mortality: A Review of Twenty-Seven Community Studies. J. Health. Soc. Behav. 38, 21–37 (1997).
DeSalvo, K. B., Bloser, N., Reynolds, K., He, J. & Muntner, P. Mortality Prediction with a Single General Self-Rated Health Question: A Meta-Analysis. J. Gen. Intern. Med. 21, 267–275 (2006).
Air Information Resource. Modelled background pollution data. Available at: http://uk-air.defra.gov.uk/(Accessed May 19, 2015).
Ratliff, F. On the Psychophysiological Bases of Universal Color Terms. Proc. Am. Philos. Soc. 120, 311–330 (1976).
DCLG. Generalised Land Use Database Statistics for England 2005 (DCLG, London, 2007).
ONS. 2011 Census Data for England and Wales. Available at: https://www.nomisweb.co.uk/census/2011 (Accessed July 16, 2014).
DCLG. English indices of deprivation 2010 (DCLG, London, 2011).
ONS. The 2011 Rural-Urban Classification For Small Area Geographies (ONS Publications, London, 2013).
Bivand, R. S., Pebesma, E. & Gómez-Rubio, V. Applied Spatial Data Analysis with R (Springer, New York, 2013).
Wagenmakers, E.-J. & Farrell, S. AIC model selection using Akaike weights. Psychon. Bull. Rev. 11, 192–196 (2004).
The World Bank. Improved water source data. Available at: http://data.worldbank.org/(Accessed May 19, 2015).
Drinking Water Inspectorate. Drinking Water 2011. Available at: http://dwi.defra.gov.uk/(Accessed May 19, 2015).
Besag, J., York, J. & Mollié, A. Bayesian image restoration, with two applications in spatial statistics. Ann. I. Stat. Math. 43, 1–20 (1991).
Besag, J. Spatial Interaction and the Statistical Analysis of Lattice Systems. J. Roy. Stat. Soc. B. Met. 36, 192–236 (1974).
T.P. and H.S.M. acknowledge the support of the Research Councils UK Grant EP/K039830/1. C.I.S. is grateful for support provided by a Warwick Business School Doctoral Scholarship. This research utilized Queen Mary’s MidPlus computational facilities, supported by QMUL Research-IT and funded by EPSRC grant EP/K000128/1.
The authors declare no competing financial interests.
Electronic supplementary material
About this article
Cite this article
Seresinhe, C., Preis, T. & Moat, H. Quantifying the Impact of Scenic Environments on Health. Sci Rep 5, 16899 (2015). https://doi.org/10.1038/srep16899
Using greenspace and nature exposure as an adjunctive treatment for opioid and substance use disorders: Preliminary evidence and potential mechanisms
Behavioural Processes (2021)
The Scenic Beauty of Geosites and Its Relation to Their Scientific Value and Geoscience Knowledge of Tourists: A Case Study from Southeastern Spain
Assessing experienced tranquillity through natural language processing and landscape ecology measures
Landscape Ecology (2021)
Reviewing the reliability of Land Use and Land Cover data in studies relating human health to the environment
Environmental Research (2021)
Sustainable Tourism and Natural Protected Areas: Exploring Local Population Perceptions in a Post-Conflict Scenario