Urban green space and happiness in developed countries
Oh-Hyun Kwon, Inho Hong, Jeasurk Yang, Donghee Yvette Wohn, Woo-Sung Jung, Meeyoung Cha
UUrban green space and happiness in developed countries
Oh-Hyun Kwon, ∗ Inho Hong, ∗ Jeasurk Yang, DongheeYvette Wohn, Woo-Sung Jung,
1, 5, 6, † and Meeyoung Cha
7, 8, ‡ Department of Physics, Pohang University of Science and Technology, Pohang 37673, Republic of Korea. Center for Humans and Machines, Max Planck Institute for Human Development, Berlin 14195, Germany. Department of Geography, National University of Singapore, Singapore 119260, Singapore. Department of Informatics, New Jersey Institute of Technology, Newark, NJ 07103, USA. Department of Industrial and Management Engineering,Pohang University of Science and Technology, Pohang 37673, Republic of Korea. Asia Pacific Center for Theoretical Physics, Pohang 37673, Republic of Korea. Data Science Group, Institute for Basic Science, Daejeon 34126, Republic of Korea. School of Computing, Korea Advanced Institute of Science and Technology, Daejeon 34141, Republic of Korea.
Urban green space has been regarded as contributing to citizen happiness by promoting physicaland mental health. However, how urban green space and happiness are related across many countriesof different socioeconomic conditions has not been explained well. By measuring urban green spacescore (UGS) from high-resolution Sentinel-2 satellite imagery of 90 global cities that in total cover179,168 km and include 230 million people in 60 developed countries, we reveal that the amountof urban green space and the GDP can explain the happiness level of the country. More precisely,urban green space and GDP are each individually associated with happiness; happiness in the 30wealthiest countries is explained only by urban green space, whereas GDP alone explains happinessin the 30 other countries in this study. Lastly, we further show that the relationship between urbangreen space and happiness is mediated by social support and that GDP moderates the relationshipbetween social support and happiness, which underlines the importance of maintaining urban greenspace as a place for social cohesion in promoting people’s happiness. I. INTRODUCTION
The advantages of urban green space for public healthand urban planning have been of great interest in recentyears. Green spaces such as parks, gardens, street trees,riversides, and even private backyards facilitate physi-cal activity, social events, mental relaxation, and relieffrom stress and heat, thereby leading to direct and indi-rect benefits for mental and physical health [1, 2]. Thus,worldwide policy changes and efforts have been made tobuild more urban green space to create sustainable andcomfortable living environments [3].Urban green space and happiness are known to havean implicit positive correlation. Although this associa-tion is still unclear, five pathways through which greenerymight have beneficial effects have been reported: reliev-ing stress, stimulating physical activity, facilitating socialinteractions, generating aesthetic enjoyment, and facili-tating a sense of shelter from and adjustment to environ-mental stressors [2, 4, 5]. Studies have suggested that thesame pathways exist in numerous countries [6]. Amongthem, social interaction facilitation has been confirmedwith strong evidence. Studies [7, 8] have shown that opengreen space promotes social cohesion by providing placesfor social contact; people can naturally encounter neigh-bors in local green spaces while walking dogs, gardening, ∗ These two authors contributed equally. † Electronic address: [email protected] ‡ Electronic address: [email protected] and having outdoor parties, which enhances communityengagement. Moreover, larger green areas such as parkscan hold larger events and activities, enabling social mix-ing between communities.The amount of urban green space can be capturedmainly by three kinds of measurements: qualitative rat-ings of observers [4, 9], national land-use and land-coverdatabase [10–12], and geographic information system(GIS) techniques. Among these measurements, GIS tech-niques are the most recently developed method. One ex-ample is utilizing the normalized difference vegetation in-dex (NDVI), a vegetation index computed from Landsatseries satellite images (30 m resolution) [5, 6, 13]. Studiessuch as by Tsai et al . [14] introduced multiple landscapemetrics based on GIS and showed a strong association be-tween green space and mental health in U.S. metropolitanareas. These studies assume the distance from an individ-ual’s residence to the nearest green space has associationswith health data [2, 15]. The green space level was thenmeasured as the fraction of areas with NDVI values abovea certain threshold (e.g., 0.2 to 0.4 for sparse vegetationand 0.6 for highly dense vegetation) [16]. However, thismethod raises the question of how to set an appropriateNDVI threshold for global cities.Despite the rich literature on green space’s mental ben-efits, they still have limitations as global-scale compar-ative research. First, the analytical settings are basedon a limited number of Western countries [5]; most ofthese studies have been conducted in the United States[13, 14] and Europe [2, 6]. Moreover, only a few arebased on multi-country settings that enable comparativeanalysis [17]. As a result, it is unclear whether the asso- a r X i v : . [ phy s i c s . s o c - ph ] J a n ciation between green space and mental health is robustin developing countries or only in developed countries.The main limitation arises because there is no globalmedical dataset providing reliable and standardized men-tal health surveys from different countries. Moreover,no studies have established which green space measure-ment is appropriate for analysis across countries. Variousmethods of measuring green space – questionnaires, qual-itative interviews, satellite images, Google Street Viewimages, and even smartphone technology [18] – still relyon individual-level measurements (e.g., calculating thegreenery level around residential buildings) and hence arenot scalable to the global level.This paper presents a new way to analyze the effectsof green space on happiness at the planetary scale, in-corporate the different countries’ different contexts, andachieve robust results. First, we measure the amount ofurban green space from high-resolution satellite imagesfor different countries by developing a globally compa-rable green space metric. Our metric based on the totalNDVI of built-up areas enables this comparison as it doesnot require an arbitrary threshold that varies for differ-ent regions. It also overcomes the limitations of officialstatistics based on national land-use land cover data thattend to have different criteria by countries and often in-clude only official parks and open space. Our analysison high-resolution (10 m) Sentinel-2 satellite images pro-vides more accurate information of urban green spacethan the previous studies on the Landsat series images(i.e., the resolution of 30 m) [2, 5, 6, 13].Next, this study uses selected happiness scores fromthe World Happiness Report [19], which provides reli-able and standardized data on multiple countries’ men-tal health and allows comparisons among nations. Ashappiness is a criterion of emotional well-being, it isinterconnected with mental health. From the perspec-tive that economic studies distinguish between emo-tional well-being (happiness) and life satisfaction (lifeevaluation)[20], we focus on the impact of green spaceon emotional happiness. Specifically, we study this rela-tionship in the developed countries of the highest HumanDevelopment Index (HDI), where green environments incities are considered more important for well-being.Using these datasets from satellite imagery, we explorethe relationship between urban green space and happi-ness globally. Additionally, we identify conditional in-direct effects by national wealth and social support byemploying a moderated mediation regression model onsocioeconomic indicators. II. URBAN GREEN SPACE AND HAPPINESSIN COUNTRIES
We examine the global relationship between urbangreen space and happiness in 60 developed countriesranked by the Human Development Index. Using theSentinel-2 satellite imagery dataset, we define each coun- try’s urban green space score (UGS) as a logarithmic to-tal vegetation index per capita in the most populatedcities (i.e., those that include at least 10% of the nationalpopulation). Among the various vegetation indices avail-able, NDVI [21] is used based on the robustness of the re-sults for different tested indices. The happiness score andthe gross domestic product based on purchasing powerparity (GDP (PPP)) per capita of each country are fromthe World Happiness Report [19] and the InternationalMonetary Fund (IMF) estimation [22], respectively (seethe Methods section and the Supplementary Informationfor details).Figure 1(a) shows an overall view of urban green spaceand the happiness of countries around the world. Thismap highlights regional differences in the green space dis-tribution due to climate; countries near the equator intropical climates have relatively high UGS values, whilecountries located in the 20-30 ◦ latitude range have ex-ceptionally low UGS values due to the dry climate. UGSincreases with latitude in higher-latitude regions. Onthe other hand, Northern and Western European andNorth American countries display relatively large happi-ness. Western Asian countries also show relatively highhappiness with low UGS value, indicating that the rela-tionship between happiness and green space is not trivial.Figure 1(b-d) shows the distribution of happiness,UGS, and log-GDP, and they all show unimodal distribu-tions with low skewness, which is appropriate for linearregression analyses. Note that the probability distribu-tions of NDVI per capita and GDP per capita converge toa normal distribution after logarithmic scaling. Our com-parison of several green space measures shows that thelogarithmic NDVI per capita is most suitable for the fol-lowing analysis in terms of its distribution and explana-tory power. We hence choose the logarithmic NDVI percapita as the primary green space indicator in this re-search. (see Supplementary Information). We also usethe logarithmic GDP per capita (PPP) (hereinafter re-ferred to as the log-GDP) as a measure of the wealth ofthe country, as noted in the Happiness Report [19].As per-country wealth is an important indicator of itscitizens’ quality of life, wealth (i.e., log-GDP) should beconsidered in analyzing urban green space and happi-ness. Our regression analysis finds that UGS, togetherwith log-GDP, explains happiness. We make new obser-vations from Table I. Although UGS is not substantiallycorrelated with happiness in the simple linear regression(i.e., model (2)), the multilinear model with log-GDP(i.e., model (3)) has a substantial increase in predictionability compared to the simple regression analysis on log-GDP (i.e., model (1)). Therefore, urban green space addsexplanatory power to the correlation between wealth andhappiness across countries. The regression analyses withother green space-variant measures further confirm thisresult’s robustness, confirming a substantial increase inthe adjusted R-squared value when including UGS in theregression. Specifically, UGS based on the logarithmicNDVI per capita shows the best regression performance FIG. 1:
The distributions of urban green space and happiness over the world. (a)
The map of urban green spaceand happiness in 60 developed countries. The size and color of circles represent the level of happiness and urban green spacein a country, respectively. The markers are placed on the most populated cities of each country. (b-d)
The histograms of (b)happiness, (c) urban green space (UGS) and (d) logarithmic GDP per capita (log-GDP). We use the logarithm of the totalNDVI per capita as an indicator of urban green space, and the logarithm of GDP per capita as a measure of wealth. (see the Supplementary Information for the results forthe different measures).
III. URBAN GREEN SPACE IS EFFECTIVE INRICH COUNTRIES
Our results show that happiness is correlated with ur-ban green space and the GDP of a country. But, is thisgreen space-happiness effect uniform across all countries?Previous studies on the marginal effect of income on hap-piness suggest that happiness may have a nonlinear re-lationship with GDP, presumably showing saturation af-ter a specific GDP — a concept known as the Easterlinparadox [23]. This paradox tells us that increases in hap-piness through GDP reach a saturation point, yet whatfactors promote happiness beyond the saturation point isunknown.To test the Easterlin paradox, we repeated the analysisover clusters of countries grouped by GDP. Figure 2(a) shows a high correlation between GDP and happiness inthe 30 lower-GDP countries (i.e., ρ = 0 . ρ = − . ρ = 0 . , p < . ρ = 0 . , p = 0 . Countries All Lower 30 Top 30Model (1) (2) (3) (4) (5) (6) (7) (8) (9)log-GDP 1.0120 *** - 1.1319 *** ** - 0.8517 * -0.0809 - 0.2581(0.6603) (0.6234) (1.6305) (1.7493) (1.3559) (1.0314)UGS - 0.1165 0.2249 *** - 0.1497 0.0567 - 0.2785 *** *** (0.3545) (0.2643) (0.6042) (0.6051) (0.2313) (0.2403)Const -4.2945 ** *** -6.4709 *** -3.3428 5.1767 *** -3.0629 7.7712 ** *** R The regression analysis for happiness, UGS, and log-GDP.
The values denote the regression coefficientsand the confidence intervals of each independent variable with its significance (i.e., *** p¡0.01; ** p¡0.05; * p¡0.1). The regressionmodel (1-3), model (4-6) and model (7-9) are examined for the data of all countries, the lower 30 countries and the top 30countries ranked by GDP, respectively.FIG. 2: The effect of GDP on the green-happiness relation. (a, b)
The relations of (a) log-GDP and happiness, and(b) urban green space (i.e., UGS) and happiness across 60 developed countries. The top 30 and the lower 30 countries rankedby GDP are sized by the population size and colored by red and black. The dotted lines are the linear fit for each GDP group. (c)
Changes of coefficients between urban green space and happiness for different sets of GDP rank with increasing windowsize from top 10 to 60. (d)
The rank correlations between UGS and happiness for the groups of increasing countries in theGDP rank order. a factor that further increases the happiness of a countryafter its GDP reaches a certain level.The regressions for each of the 60 countries ranked byGDP in Table I confirm the individual effects of urbangreen space and GDP on happiness. GDP is the only sub-stantial factor explaining happiness in the 30 lower-GDPcountries (models 4-6). In contrast, for the 30 higher- GDP countries, happiness is explained only by the UGS(7-9). These findings suggest that GDP is critical forhappiness until it reaches a certain GDP threshold (i.e.,the Easterlin paradox), after which urban green spaceexplains happiness better.The correlation between UGS and happiness also cor-roborates the effect of UGS in rich countries. The corre-lation in Figure 2(d) decreases as more countries in thedecreasing order of GDP are added. The correlation issubstantial (i.e., ρ is approximately 0.8) among the coun-tries excluding the top 30. Figure 2(c) summarizes theeffects of urban green space and GDP that cross over eachother around the 30th wealthiest country. For the top 30countries, urban green space has positive coefficients, butthe GDP effect is not significant. These relationships arereversed for less affluent countries.In summary, economic support seems to promote hap-piness until the essential requirements and living stan-dards are met. However, economic success alone fails toadd persistent promotion of happiness. After some level,urban green space appears to be related to other socialfactors that can further promote happiness. IV. URBAN GREEN SPACE FOR SOCIALCOHESION
Our findings highlight urban green space as an indica-tor that might be correlated with social factors promotinghappiness beyond the achievement of economic success.The question then arises, which social factors connect ur-ban green space with happiness? To identify this connec-tion, we first examine the correlation between UGS andsocioeconomic variables reported in the World HappinessReport: GDP per capita, social support, life expectancy,freedom, generosity, and perceptions of corruption. Ofthese six variables, only “social support” has a signifi-cantly positive correlation ( ρ = 0 . , p < .
01) with UGSas we can see from Fig. 3(a), implying that social supportcould mediate between urban green space and happiness.This relationship is consistent with several existing stud-ies that suggested urban green space as a place of socialcohesion [7, 8]. On the other hand, as indicated by lifeexpectancy, physical health does not display a signifi-cant relationship with green space ( ρ = 0 . , p = 0 . H = β + β M + β S + β SM, (1) S = β + β ln G. (2)where H , M , S , and G represent happiness, log-GDP, so-cial support, and NDVI per capita, respectively, and the β values denote the coefficients of the regression models(see the Supplementary Information for details).Our moderated mediation model can be used to esti-mate the amount of urban green space required to in-crease happiness by a certain amount according to∆ H = ( β + β M ) ln G f G i . (3)In the equation 3, the required ratio of urban green spacesin a country decreases as its log-GDP increases. The re-quired increase in urban green space per capita can beestimated for each country based on its current GDPvalue. For example, the United States needs an addi-tional 36.1908 NDVI of urban green space per capita toincrease its happiness score by 0.0546. In contrast, 3,416USD per capita is required to achieve the same incrementin happiness. Here, we used a 0.0546 happiness score asa reference value of ∆ H , which is the average value be-tween happiness ranks. Note that the NDVI per capita isinterpreted as a weighted area of green space, with a unitof m . Similarly, Qatar needs 0.4981 NDVI per capita or7,556 dollars per capita, and South Korea needs 4.1332NDVI per capita or 2,315 dollars per capita to achievethe reference happiness score increase. V. DISCUSSION
This paper revealed a global relationship between ur-ban green space and happiness in over 60 countries usinghigh-resolution satellite imagery. Urban green space hasa higher impact in developed countries (i.e., countrieswith higher GDPs), which suggests urban green spaceas a key to promoting happiness beyond economic suc-cess. Our moderated mediation model further elucidatesthis relationship as social support mediates the green-happiness relation, and GDP moderates social supportand happiness. This sophisticated model could estimateadditional green space needed to promote happiness foreach country.The current study newly defined the concept of UGS(urban green space score), which can be used to calculatethe amount of green space at any spatial scale accountingfor population density. We compared several green space
FIG. 3:
The moderated mediation model for UGS, happiness and socioeconomic indicators. (a)
Scatter plot ofsocial support and UGS across countries. (b)
Diagram for the moderated mediation model. The boxes denote the modelvariables. Solid black arrows denote a statistically significant relationship between a pair of variables with the regressioncoefficient and the p-value (i.e., ***p¡0.01). The gray dashed arrow represents a non-significant relationship. Note that thecoefficients are calculated with z-scores of the variables to compare the effect size directly. measures and proposed to use the logarithmic NDVI percapita as a preferred measure of UGS. This index wasvalidated through experiments and it makes it possibleto investigate green space at a global level, allowing us toperform cross-sectional research on green space. Further-more, the method obtaining UGS can be utilized to inves-tigate any spatial areas such as blue space (i.e., aquaticenvironments such as lake and shore) [25, 26].Our findings have multiple policy-level implications.First, public green space should be made accessible tourban dwellers to enhance social support. In doing so,one critical aspect is public safety. If public safety in ur-ban parks is not guaranteed [27, 28], its positive role insocial support and happiness may diminish. The mean-ing of public safety may change; for example, ensuringbiological safety will be a priority in keeping the urbanparks accessible during the COVID-19 pandemic [29].Infact, the high indoor transmission rate of the virus [30]will increase awareness and importance of open spaceslike urban parks. While some urban parks may be closedduring lockdowns, some reports suggest that viewingthem from home could also help relax stress during thepandemic [31]. Second, urban planning of public greenspace is needed for both developed and developing coun-tries. While our findings confirmed a strong impact ofurban green space on happiness in developed countries,the same positive effect holds for developing countries,albeit to a smaller degree. Furthermore, it is challeng-ing or nearly impossible to secure land for green spaceafter built-up areas are developed in cities. Therefore,urban planning for parks and green recovery (new green-ing in built-up areas) should be considered in developingeconomies where new cities and suburban areas rapidlyexpand [32, 33].In addition to the above, recent climate changes cancreate substantial volatility in sustaining urban greenspace. Extreme events such as wildfires, floods, droughts,and cold waves could endanger urban forests around the world [34]. On the other hand, global warming couldalso accelerate tree growth in cities more than in ruralareas due to the urban heat island effect [35]. In the end,the environmental influence is bidirectional; urban greenspaces affect local climates by reducing carbon dioxidelevels [36] and providing a cooling effect inside the citythat indirectly affects people’s well-being. Thus, we needmore attention to predicting climate changes and dis-covering their impact on public places since the extremechanges could hamper the benefits of urban green space.As an exciting future direction, satellite images ofhigher spatiotemporal resolutions can be used to com-pute urban green space scores. This paper focused onthe correlation across countries fixed in time, given theshort span of the Sentinel-2 dataset launched in 2015.A causal analysis could be done with satellite imagerydata for a longer span. Also, our dataset does not coverall the countries in the world. Fortunately, our obser-vations from the 30 lower-income countries anticipatethe substantial effect of GDP in other developing coun-tries excluded in our analysis. We have analyzed thehighest-resolution public dataset of satellite imagery inthis study. However, our method still has room for ap-plication to higher-resolution non-public datasets such asthe household level (less than 10m resolution) availablein the national-scale health dataset [37]. Since satelliteimagery cannot account for green space inside buildings(such as green walls), future research could quantify theeffect of these mini-scale green spaces using computer vi-sion [38].
VI. METHODSA. Collecting happiness and remote sensing data
To identify the relationship between happiness andgreen space, we use happiness scores from the World Hap-
FIG. 4:
Measuring urban green spaces. (a)
Measurement methods to compute the size of urban green space in eachcountry. First, we find cities occupying more than 10 percent of the total population in a country. Then, we extract thebuilt-up area of the cities with Copernicus global landcover data. Finally, we calculate vegetation indices (e.g., NDVI) withinthe area using Sentinel-2 satellite images. (b)
Urban green space measured by the UGS (upper row) and the vegetation ratio(lower row) in four world cities.The red areas in the upper row indicate vegetation for the NDVI threshold of 0.4. The lowerrow images show the adjusted NDVI per capita (i.e., UGS) for every 10m by 10m pixel. piness Report [19] and the NDVI scores from Sentinel-2 satellite imagery as remote sensing data. The WorldHappiness Report from 2018 covered 156 countries. Thereport provides an annual survey of how happy citizensperceive themselves to be and ranks the countries by happiness score . The score is the average of the par-ticipants’ responses asked to rate how happy they are ona scale from 0 and 10. While many socioeconomic in-dicators (e.g., unemployment and inequality) may affecthappiness, not all of these factors are measured annuallyacross 156 countries. The report instead describes hap- piness with six primary socioeconomic indicators: GDPper capita, social support, life expectancy, freedom tomake life choices, generosity, and perceptions of corrup-tion. For example, the social support variable is based onbinary responses (yes/no) on a Gallup World Poll ques-tion: ”If you were in trouble, do you have relatives orfriends you can count on to help you whenever you needthem, or not?”To quantify urban green space in global cities, we usethe Sentinel-2 dataset that provides the highest spatialresolution (10 m) among the publicly available satelliteimagery datasets (e.g., 30 m resolution in Landsat se-ries) [2, 5, 6, 13]. With this high resolution, we can iden-tify granular green space, including street vegetation andhome gardens that could not be detected in other publicdatasets. When using satellite imagery to detect smallvegetation, it is critical to consider the season in whichthe images were obtained [2, 5, 6, 13]. We use the imagesfrom summer: June to September 2018 for the NorthernHemisphere and December 2017 to February 2018 for theSouthern Hemisphere. Satellite images with below 10%cloud cover were used; when such images could not beobtained for the study period, data from 2019 were usedinstead.Normalized difference vegetation index (NDVI) is awell-known remote sensing indicator of green vegetationareas in satellite images [21]. It detects vegetation asthe difference between near-infrared and red light, in thevalue range from -1 to +1. In general, high NDVI scoresinclude urban green spaces such as official parks, back-yards, street trees, mountains, riverbanks, golf courses,and urban farmlands. There are a few well-known vari-ants of NDVI [18], such as the soil-adjusted vegetationindex (SAVI) [39], which is corrected for soil brightness,and the enhanced vegetation index (EVI) [40], whichis corrected for atmospheric effects. All NDVI, SAVI,and EVI2 scores can be calculated from the two spectralbands of Sentinel-2, red (band 4) and near-infrared (NIR,band 8), as follows:
N DV I = N IR − REDN IR + RED , (4)
SAV I = (1 + L )( N IR − RED ) N IR + RED + L , (5)
EV I . N IR − REDN IR + 2 . RED + 1 . (6)The robustness of the results for the three green spacemeasures was verified using NDVI as the primary metric. B. Measuring the amount of green space
The vegetation indices are measured in three steps,as illustrated in Fig. 4(a). The first step is to identifytarget cities containing at least 10% of each country’s total population and represent the country’s overall hap-piness. The second step is to extract only the built-upareas within the identified cities’ administrative bound-aries. As cities’ boundaries are historically and culturallyconstructed and often arbitrary, the cities’ size needs tobe standardized; some cities include vast suburban areas(e.g., Istanbul) or natural areas (e.g., deserts in Dubai).Thus, referring to the global land cover data from theEU’s Copernicus Programme [41], we focus on urbanbuilt-up areas to quantify the urban green space. Fi-nally, the vegetation indices (NDVI, EVI2, and SAVI)are calculated for the extracted urban areas.The final step is to compute the amount of green spacein each country, determined from the measured vegeta-tion indices. Here, we define the amount of green spaceas the logarithm of the total NDVI of built-up areas inthe target cities divided by the cities’ total population,called UGS, as a metric for urban green space. UGS iscalculated as follows:
U GS = log P c P b ( c ) N DV I ( b ) P c N c ! , (7)where N DV I ( b ) is the value of NDVI of pixel b withinbuilt-up areas b ( c ) in city c and N c is the population ofcity c . In this calculation, we adjusted negative NDVIvalues to zero [18] to prevent errors caused by the ac-cumulation of negative values in areas next to bodies ofwater (see the Supplementary Information for the entiredataset). Acknowledgements
The authors thank to Farnoosh Hashemi, Ali Behrouz,and Taekho You for useful comments. M.Cha work wassupported by the Institute for Basic Science (IBS-R029-C2).
Author contributions statement
M.C. and D.Y.W. conceived the research, I.H., W.-S.J. and M.C. designed the research, O.-H.K. and J.Y.collected the data, O.-H.K. performed the research, O.-H.K., I.H. and M.C. analysed the data, O.-H.K., I.H.and J.Y. wrote the manuscript. All authors reviewed themanuscript. [1] S. de Vries, R. A. Verheij, P. P. Groenewegen, andP. Spreeuwenberg, Environment and Planning A: Econ-omy and Space , 1717 (2003).[2] P. Dadvand, X. Bartoll, X. Basaga˜na, A. Dalmau-Bueno,D. Martinez, A. Ambros, M. Cirach, M. Triguero-Mas,M. Gascon, C. Borrell, et al., Environmental Interna- tional , 161 (2016).[3] UN, Sustainable Development Goals (2015), available at https://sdgs.un.org/goals . Date accessed 4 November2020.[4] S. de Vries, S. M. E. van Dillen, P. P. Groenewegen,and P. Spreeuwenberg, Social Science & Medicine ,
26 (2013).[5] Y. Liu, R. Wang, G. Grekousis, Y. Liu, Y. Yuan, andZ. Li, Landsacpe and Urban Planning (2019).[6] A. Dzhambov, T. Hartig, I. Markevych, B. Tilov, andD. Dimitrova, Environmental Research , 47 (2018).[7] J. Maas, S. M. E. van Dillen, R. A. Verheij, and P. P.Groenewegen, Health & Place , 586 (2009).[8] V. Jennings and O. Bamkole, International journal ofenvironmental research and public health , 452 (2019).[9] B. S. Kweon, W. C. Sullivan, and A. R. Wiley, Environ-ment and Behavior , 832 (1998).[10] J. Maas, R. A. Verheij, P. P. Groenewegen, S. de Vries,and P. Spreeuwenberg, Journal of Epidemiology & Com-munity Health , 587 (2006).[11] G. MacKerron and S. Mourato, Global Environmentalchange , 992 (2013).[12] I. Alcock, M. P. White, B. W. Wheeler, L. E. Fleming,and M. H. Depledge, Environmental Science & Technol-ogy , 1247 (2014).[13] K. M. M. Beyer, A. Kaltenbach, A. Szabo, S. Bogar,F. J. Nieto, and K. M. Malecki, International Journalof Environmental Research and Public Health , 3453(2014).[14] W.-L. Tsai, M. R. Mchale, V. Jennings, O. Marquet,J. A. Hipp, Y.-F. Leung, and M. F. Floyd, InternationalJournal of Environmental Research and Public Health (2018).[15] U. K. Stigsdotter, O. Ekholm, J. Schipperijn, M. Tof-tager, F. Kamper-Jørgensen, and T. B. Randrup, Scan-dinavian Journal of Public Health , 411 (2010).[16] EOS, NDVI FAQ: ALL YOU NEED TO KNOWABOUT NDVI (2019), available at https://eos.com/blog/ndvi-faq-all-you-need-to-know-about-ndvi/ .Date accessed 22 June 2020.[17] M. van den Berg, M. van Poppel, I. van Kamp, S. An-drusaityte, B. Balseviciene, M. Cirach, A. Danileviciute,N. Ellis, G. Hurst, D. Masterson, et al., Health & Place , 8 (2016).[18] I. Markevych, J. Schoierer, T. Hartig, A. Chudnovsky,P. Hystad, A. M. Dzhambov, S. de Vries, M. Triguero-Mas, M. Brauer, M. J. Nieuwenhuijsen, et al., Environ-mental Research , 301 (2017).[19] J. F. Helliwell, R. Layard, and J. D. Sachs, New York:UN Sustainable Development Solutions Network (2018).[20] D. Kahneman and A. Deaton, Proceedings of the Na-tional Academy of Sciences , 16489 (2010).[21] T. Miura, S. Nagai, M. Takeuchi, K. Ichii, and H. Yosh-ioka, Scientific Reports (2019).[22] IMF, World Economic Outlook Database (2018),available at . Dateaccessed 16 November 2020. [23] R. A. Easterlin, L. A. McVey, M. Switek, O. Sawangfa,and J. S. Zweig, Proceedings of the National Academy ofSciences , 22463 (2010).[24] K. J. Preacher, D. D. Rucker, and A. F. Hayes, Multi-variate Behavioral Research , 185–227 (2007).[25] R. Foley and T. Kistemann, Health & Place , 157–165(2015).[26] C. M. Raymond, S. Gottwald, and M. Kytta, Landscapeand Urban Planning , 198–208 (2016).[27] E. Groff and E. S. McCord, Security journal , 1 (2012).[28] B. Han, D. A. Cohen, K. P. Derose, J. Li, andS. Williamson, American journal of preventive medicine , 352 (2018).[29] F. Ugolini, L. Massetti, P. Calaza-Mart´ınez,P. Cari˜nanos, C. Dobbs, S. K. Ostoi´c, A. M. Marin,D. Pearlmutter, H. Saaroni, I. ˇSaulien˙e, et al., UrbanForestry & Urban Greening , 126888 (2020).[30] S. Lolli, Y.-C. Chen, S.-H. Wang, and G. Vivone, Scien-tific Reports , 16213 (2020).[31] M. Hedblom, B. Gunnarsson, B. Iravani, I. Knez,M. Schaefer, P. Thorsson, and J. N. Lundstr¨om, ScientificReports , 10113 (2019).[32] R. H. Ewing, Characteristics, Causes, and Effects ofSprawl: A Literature Review (Springer US, Boston, MA,2008).[33] X. Liu, Y. Huang, X. Xu, X. Li, X. Li, P. Ciais, P. Lin,K. Gong, A. D. Ziegler, A. Chen, et al., Nature Sustain-ability , 564 (2020).[34] C. D. Allen, A. K. Macalady, H. Chenchouni,D. Bachelet, N. McDowell, M. Vennetier, T. Kitzberger,A. Rigling, D. D. Breshears, E. H. Hogg, et al., ForestEcology and Management , 660 (2010).[35] H. Pretzsch, P. Biber, E. Uhl, J. Dahlhausen, G. Sch¨utze,D. Perkins, T. R¨otzer, J. Caldentey, T. Koike, T. vanCon, et al., Scientific Reports (2017).[36] D. J. Nowak, E. J. Greenfield, R. E. Hoehn, and E. La-point, Environmental Pollution , 229 (2013).[37] V. Houlden, S. Weich, and S. Jarvis, BMC public health , 460 (2017).[38] I. Seiferling, N. Naik, C. Ratti, and R. Proulx, Landscapeand Urban Planning , 93 (2017).[39] A. R. Huete, Remote Sensing of Environment , 295(1988).[40] Z. Jiang, A. R. Huete, K. Didan, and T. Miura, RemoteSensing of Environment , 3833 (2008).[41] M. Buchhorn, B. Smets, L. Bertels, B. De Roo, M. Lesiv,N.-E. Tsendbazar, M. Herold, and S. Fritz, Coperni-cus Global Land Service: Land Cover 100m: collec-tion 3 epoch 2015, Globe (2020), available at https://lcviewer.vito.be/download . upplementary Information Oh-Hyun Kwon
1, * , Inho Hong
2, * , Jeasurk Yang , Donghee Yvette Wohn , Woo-SungJung
1, 5, 6, † , and Meeyoung Cha
7, 8, ‡ Department of Physics, Pohang University of Science and Technology, Pohang 37673, Republic of Korea. Center for Humans and Machines, Max Planck Institute for Human Development, Berlin 14195, Germany. Department of Geography, National University of Singapore, Singapore 119260, Singapore. Department of Informatics, New Jersey Institute of Technology, Newark, NJ 07103, USA. Department of Industrial and Management Engineering, Pohang University of Science and Technology, Pohang37673, Republic of Korea. Asia Pacific Center for Theoretical Physics, Pohang 37673, Republic of Korea. Data Science Group, Institute for Basic Science, Daejeon 34126, Republic of Korea. School of Computing, Korea Advanced Institute of Science and Technology, Daejeon 34141, Republic of Korea. * These authors contributed equally to this work: Oh-Hyun Kwon and Inho Hong † [email protected] ‡ [email protected] a r X i v : . [ phy s i c s . s o c - ph ] J a n ection S1. Data Section S1.1. Data description
Table S1 and S2 describe the dataset used in this study. The happiness scores were obtained from theWorld Happiness Report, which was averaged over three years to adjust for short-term fluctuations. Theaverage happiness score is 6.373, with a maximum of 7.769 for Finland and a minimum of 4.549 for Iran.UGS is calculated from Sentinel-2 satellite imagery data, and GDP per capita (PPP) data is obtainedfrom the IMF estimation.In this research, we used the data of 60 developed countries selected by comparing the HDI of the coun-tries. Andorra, Bahamas, Barbados, Brunei, Cyprus, Lichtenstein, Palau, and Seychelles are excludedfrom the analysis due to a lack of data for happiness.Country City counts Population [%] Happiness UGS log-GDPFinland 1 28.23 7.77 5.73 10.70Iceland 1 38.05 7.49 5.47 10.87Lithuania 1 19.25 6.15 5.46 10.44New Zealand 1 34.57 7.31 5.33 10.60Slovenia 1 13.99 6.12 5.32 10.45Croatia 1 19.82 5.43 5.23 10.11Montenegro 1 31.07 5.52 5.21 9.85Italy 1 7.21 6.22 5.17 10.56Slovakia 2 12.25 6.20 5.16 10.46Estonia 1 33.12 5.89 5.15 10.39United States 3 12.76 6.89 5.13 11.03Latvia 1 32.95 5.94 5.05 10.28Sweden 2 15.01 7.34 5.00 10.88Switzerland 4 10.87 7.48 4.98 11.04Norway 1 12.80 7.54 4.97 11.19Canada 1 18.26 7.28 4.96 10.80Serbia 1 23.99 5.60 4.93 9.67Poland 4 10.09 6.16 4.88 10.34Germany 5 10.54 7.02 4.79 10.85Hungary 1 17.89 5.82 4.78 10.31Czech Republic 1 12.13 6.85 4.75 10.50Portugal 3 11.67 5.69 4.72 10.32Bulgaria 1 18.69 5.01 4.70 10.02Australia 1 19.55 7.23 4.69 10.86Netherlands 3 10.71 7.49 4.52 10.91Luxembourg 1 30.40 7.09 4.49 11.59Ireland 1 11.62 7.02 4.36 11.24United Kingdom 1 13.42 7.05 4.28 10.72Trinidad and Tobago 1 12.76 6.19 4.25 10.46Uruguay 1 39.42 6.29 4.16 10.06
Table S1.
Data used in the study. Countries are ordered by UGS. We aggregate city-level data to coverat least 10% of total population. 2ountry City counts Population [%] Happiness UGS log-GDPSpain 1 13.97 6.35 4.15 10.59Russia 2 12.50 5.65 4.12 10.24Belarus 1 20.82 5.32 4.12 9.82Austria 1 21.42 7.25 4.11 10.83Panama 1 27.74 6.32 4.06 10.15Kazakhstan 1 11.13 5.81 4.06 10.19Albania 1 23.32 4.72 4.06 9.51Mauritius 1 29.03 5.89 3.94 10.05Costa Rica 1 32.68 7.17 3.93 9.79Belgium 1 10.58 6.92 3.91 10.76Denmark 1 10.78 7.60 3.89 10.82Romania 2 10.93 6.07 3.87 10.13France 1 10.62 6.59 3.72 10.72Malaysia 1 12.19 5.34 3.64 10.31Argentina 2 10.13 6.09 3.33 9.98Turkey 1 18.34 5.37 3.28 10.05Greece 1 24.11 5.29 3.28 10.30Malta 3 13.05 6.73 3.17 10.64Chile 1 30.54 6.45 3.05 10.15Japan 1 10.63 5.89 3.03 10.63Iran 1 10.86 4.55 2.90 9.91Singapore 1 100.00 6.26 2.87 11.45South Korea 1 19.00 5.89 2.70 10.64Israel 1 10.75 7.14 2.65 10.53United Arab Emirates 1 35.82 6.82 2.23 11.17Saudi Arabia 1 19.49 6.37 2.06 10.95Oman 1 32.66 6.85 2.05 10.73Kuwait 1 12.79 6.02 1.91 11.22Qatar 1 39.77 6.37 1.23 11.82Bahrain 1 38.47 6.20 0.54 10.86
Table S2.
Data used in the study. Countries are ordered by UGS. We aggregate city-level data to coverat least 10% of total population.
Section S2. Robustness of the regression
The result of the regression is robust for any green space measure. In table S3, all nine green spacemeasures explain happiness along with GDP, while logarithmic NDVI per capita in the model (5) displaysthe most considerable value of adjusted R compared to other models.3 )( )( )( )( )( )( )( )( ) G D P L N . *** . *** . *** . *** . *** . *** . *** . *** . *** ( . )( . )( . )( . )( . )( . )( . )( . )( . ) G r ee n R a t i o0 . ** -------- ( . ) G r ee np e r C a p i t a - . ** ------- ( . ) N D V I M e a n -- . ** ------ ( . ) N D V I p e r C a p i t a --- . *** ----- ( . ) N D V I L N ---- . *** ---- ( . ) S AV I M e a n ----- . ** --- ( . ) S AV I p e r C a p i t a ------ . *** -- ( . ) E V I M e a n ------- . ** - ( . ) E V I p e r C a p i t a -------- . *** ( . ) C o n s t - . - . - . - . - . - . - . - . - . ( . )( . )( . )( . )( . )( . )( . )( . )( . ) A d j u s t e d R . . . . . . . . . O b s e r v a t i o n s T a b l e S . R e g r e ss i o n a n a l y s i s o f h a pp i n e ss w i t hd i ff e r e n t g r ee n s p a ce m e a s u r e s , *** p ¡ . ; ** p ¡ . ; * p ¡ . . ection S3. Regional influence Regional characteristics affect the level of green space. Figure S1 describes the change of USG by latitude.Countries with a tropical climate such as Southeastern Asia, the Caribbean, and Eastern Africa show arelatively high UGS score. In contrast, Western Asian countries show a relatively low UGS score sincethey are in a dry climate. The UGS score further increases in higher latitudes.
Fig. S1.
Scatter plot of UGS and latitude with country (left) and continent (right) marked. Gray arearepresent the dry climate region.In Table S4, model(3) includes the latitude of the most populated city, model (4-5) includes dummyvariables that tell whether the countries in Western Asia or the dry climate region. These models showthat including regional factors does not improve the model. (1) (2) (3) (4) (5)GDP 1.0120 *** *** *** *** *** (0.6603) (0.6234) (0.6413) (0.6433) (0.6297)UGS - 0.2249 *** ** *** ** (0.2643) (0.3302) (0.3770) (0.3641)Latitude - - 0.0009 - -(0.0250)Western Asia - - - 0.1595 -(1.2679)Dry Climate - - - - -0.0885(1.1301)Const -4.2945 ** -6.4709 *** -6.4326 *** -6.4422 *** -6.4081 *** (6.9672) (6.8998) (7.0474) (6.9518) (7.0037)Adjusted R Table S4.
Regression analysis of happiness with region variables. *** p¡0.01; ** p¡0.05; * p¡0.1.5 ection S4. Distribution of green space Figure S2 describes the distribution of three green space measures. NDVIavg (average NDVI) is calculatedby taking the mean NDVI values over the built-up area, representing how much greenery cities have.NDVIpc (average NDVI per capita) is obtained by dividing the total NDVI by the total population.NDVI per capita describes how much green space is provided to a population. However, NDVI per capitashows a skewed distribution, which is not appropriate for regression analysis. Therefore, we log on toNDVI per capita to get a normal-like distribution of the green space measure.
NDVI mean
NDVI per capita ln(NDVIpc)
Fig. S2.
Distribution plot of NDVI mean, NDVI per capita, logarithmic NDVI per capita
Section S5. Residual analysis
We perform a residual analysis of the regression model in Table 1 to check whether the model is reasonable.First, we need to check the autocorrelation of the residuals by using Durbin-Watson statistics. TheDurbin-Watson statistics show a value of 1.918, which indicates there are no autocorrelations betweenthe residuals. Second, we check for the normality of the residuals. The distribution and Q-Q plot of theresiduals shows that the residuals satisfy the normality condition. Finally, we check for the equality ofvariance by finding outliers using Cook’s distance. The figure shows that every point has a value of lessthan 1, indicating acceptable values.
Residual D e n s i t y f un c t i o n Theoretical quatiles O r d e r e d v a l u e s Country C oo k ' s d i s t a n c e Fig. S3.
Residual analysis of the regression model. (left) The distribution, (middle) Q-Q plot, and(right) cook’s distance of residuals. 6 ection S6. The effect of GDP on green-happiness relation
We can check for a similar result of Fig. 3(c) in the manuscript by calculating the Pearson correlationinstead of the regression coefficient. Figure S4 shows a similar diminishing effect of green space as thegroup contains lower GDP countries. In contrast, log-GDP shows the most strong correlations for theentire dataset containing lower GDP groups.
10 20 30 40 50
GDP Rank P e a r s o n Green spaceGDP
Fig. S4.
Changes of the Pearson correlation between urban green space and happiness for different setsof GDP rank with increasing window size from top 10 to 60.
Section S7. Happiness Report variables
World Happiness Report describes happiness with six main variables:
GDP, social support, life ex-pectancy, freedom, generosity , and corruption perceptions . Social support and freedom are based onbinary responses (yes or no) to World Gallup Poll (WGP) questions; “If you were in trouble, do youhave relatives or friends you can count on to help you whenever you need them, or not?”, and “Are yousatisfied or dissatisfied with your freedom to choose what you do with your life?”, respectively. generosity is the residual of regression for responses for a WGP question “Have you donated money to a charity inthe past month?” on GDP per capita.
Corruption perceptions is based on the response to WGP question,“Is corruption widespread throughout the government or not?” and “Is corruption widespread withinbusinesses or not?”
Life expectancy is based on the Global Health Observatory data from World HealthOrganization (WHO).Here, we checked how our analyses fit into these six variables. The data of 6 variables are retrievedfrom the World Happiness Report, and we took a 3-year average. Figure S5 shows the scatter plotsbetween UGS and six variables in the World Happiness Report. Note that the scatter plot between UGSand social support presents a relatively strong Pearson correlation of 0.4329, while other variables showno correlation with UGS. Therefore, we can suspect that the UGS is connected with the social supportvariable, which should be considered while constructing regression models.Since the data for corruption perceptions is missing for six countries, and it seems to fail to explainhappiness well for developed countries’ data set, we checked the regression with and without corruptionperception. The regression model (1) shows that UGS can explain happiness in place of social support,even although the adjusted R-square value is smaller compared to model (2), which includes socialsupport. Furthermore, model (3), which includes both UGS and social support, shows that UGS losesits explainability while social support. The same result can be found in the model (4-6).7 UGS G D P = -0.2260p = 0.0825 UGS S o c i a l S u pp o r t = 0.4329p = 0.0006 UGS L i f e E x p e c t a n c y = 0.1892p = 0.1478 UGS F r ee d o m = -0.0381p = 0.7723 UGS G e n e r o s i t y = -0.0620p = 0.6380 UGS C o rr u p t i o n P e r c e p t i o n s = -0.1145p = 0.4096 Fig. S5.
Scatter plot between Green space(UGS) and variables in World Happiness Report. ρ indicatesthe Pearson correlations. Section S8. Moderated mediation model for regression
The moderation and mediation technique provides a more complicated regression model, describing moredetailed mechanisms behind regression.The mediation model describes the indirect effect of mediation variables described by the two-stagedregression model. We applied the moderation model for log-GDP since we checked that the regressionanalyses for social support differed depending on the GDP value, which can be described with the crossterm. We can set up the regression model as follows: H = β + β M + β S + β SM (S1) S = β + β G (S2)Now, we can validate the model with regression. The mediation model can be validated by comparingthe multilinear regression model with its explanation of power. We will check whether green space affectshappiness via social support.In Table S6, model (1-3) describes the effect of UGS and social support on happiness. UGS and socialsupport can explain happiness and GDP in the model (1) and (2). However, UGS loses its explainabilitywhen we include both UGS and social support in the model (3), which implies that UGS only indirectlyaffects happiness compared to social support. Note that our mediation model was valid for GDP, so themoderated mediation model would be more appropriate.The moderation of the model can be validated by calculating the regression model with a cross-term.We check for moderation models in our consideration: moderation for green-social, social-happiness, andgreen-happiness. We find that the moderation effect emerges on the social-happiness relation with higheradjusted R-square and significantly low p-value (Table S6 model (4)). Therefore, we can conclude that thegreen space affects happiness through social support, and GDP moderates social support on happiness.8 ithout corruption perceptions With corruption perceptions(1) (2) (3) (4) (5) (6)GDP 0.5187 *** * * ** *** - 0.0339 0.1729 ** - 0.0442(0.2263) (0.2290) (0.3255) (0.3018)Social Support - 5.1863 *** *** - 5.2457 *** *** (3.6667) (4.3514) (4.2056) (4.6136)Life Expectancy 0.0606 *** *** *** ** ** ** (0.0872) (0.0733) (0.0751) (0.1115) (0.0929) (0.0950)Freedom 2.4609 *** *** *** *** ** ** (2.6342) (2.3277) (2.3463) (2.9002) (2.5320) 2.5549)Generosity 0.6584 1.0520 ** ** ** ** (2.0066) (1.7401) (1.7555) (2.2309) (1.9306) (1.4819)Corruption Perceptions - - - -0.3589 -0.3109 -0.3039(1.7389) (8.6572) (8.7725)Const -6.0323 *** -6.0873 *** -6.2005 *** -4.7584 * -4.8097 ** -4.9458 ** (7.1740) (6.0731) (6.1610) (10.2984) (8.5672) (8.7725)Adjusted R Table S5.
Regression analysis of happiness with (1-3) 5 variables and (4-6) 6 variables in the WorldHappiness Report. We separated the models with corruption perceptions since few countries are missingdata: Oman is excluded from the model (1-3), and Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, andthe United Arab Emirates are excluded from the model (4-6). *** p¡0.01; ** p¡0.05; * p¡0.1. (1) (2) (3) (4)log-GDP 1.1321 *** *** *** -4.1830 ** (0.6193) (0.5687) (0.6336) (6.566)UGS 0.2457 *** - 0.0782 -(0.2697) (0.2863)Social Support - 6.3899 *** *** -50.1512 ** (4.4264) (5.3363) (75.625)log-GDP:Social Support - - - 5.5583 *** (7.423)Const -6.5695 *** -6.8962 *** -7.2953 *** ** (6.8599) (5.9001) (6.0703) (66.687)Adjusted R Table S6.
Regression analysis for the moderated mediation model. Coefficient of GDP-Social Supportrepresent cross term of GDP and social support. *** p¡0.01; ** p¡0.05; * p¡0.1. Section S9. Derivation of happiness equation
How much we need green space to increase our happiness? Since our analyses are based on regressionmodels, we can provide a numerical estimation of the required green space to increment happiness.Consider our final regression model: H = β + β M + β S + β SM (S3) S = β + β G (S4)where H is the happiness score, M is GDP per capita, S is social support, and G is UGS. If wesubstitute social support into the equation, we obtain the following equation. H = β + ( β + β ln M ) ln G + β ln M (S5)9f we assume that the value of GDP per capita stays the same, we can solve a fraction of green spacechange to increase a certain amount of happiness. We set the happiness score change to 0.0546, which isan average value for upgrading one rank. G f G i = exp (cid:18) ∆ Hβ + β ln M (cid:19) (S6) Rank G D P p e r c a p i t a N D V I p e r c a p i t a Fig. S6.
Required GDP per capita (yellow) and NDVI per capita (green) to increase average amount ofhappiness for rank up.Country Green Space [%] NDVI per capita GDP per capita [dollar]Qatar 14.50 0.4981 7556Luxembourg 16.00 14.3032 6004Singapore 17.10 3.0292 5199Ireland 19.04 14.9560 4205Kuwait 19.26 1.2950 4115Norway 19.49 28.2059 4026United Arab Emirates 19.79 1.8334 3914Switzerland 21.24 30.8228 3461United States 21.41 36.1908 3416Saudi Arabia 22.50 1.7651 3149Netherlands 23.05 21.1280 3032Sweden 23.52 35.0743 2941Iceland 23.68 56.2871 2909Bahrain 23.75 0.4071 2896Australia 23.80 25.8874 2888Germany 23.99 28.9519 2853Austria 24.31 14.7517 2797Denmark 24.53 11.9521 2761Canada 24.81 35.3155 2716Belgium 25.48 12.6587 2615Oman 26.06 2.0204 2535France 26.29 10.8164 2504United Kingdom 26.37 18.9955 2495Finland 26.68 82.4975 2455Malta 27.81 6.5911 2325South Korea 27.90 4.1332 2315Japan 28.12 5.8425 229210ew Zealand 28.75 59.4447 2229Spain 29.00 18.4593 2205Italy 29.80 52.6365 2133Israel 30.58 4.3248 2069Czech Republic 31.29 36.0511 2015Trinidad and Tobago 32.34 22.7089 1941Slovakia 32.44 56.3721 1934Slovenia 32.72 66.8141 1916Lithuania 33.22 78.0579 1885Estonia 34.51 59.4736 1810Poland 36.40 48.1184 1715Portugal 37.03 41.6695 1686Malaysia 37.40 14.1884 1670Hungary 37.63 45.0125 1660Greece 37.85 10.0445 1652Latvia 38.86 60.5196 1611Russia 40.60 25.0656 1549Kazakhstan 43.12 24.8977 1470Panama 44.84 26.0437 1424Chile 44.84 9.5004 1424Romania 46.17 22.2308 1391Croatia 47.31 88.2509 1365Uruguay 50.84 32.4731 1295Mauritius 51.53 26.5892 1283Turkey 51.91 13.8637 1276Bulgaria 54.05 59.3985 1241Argentina 57.43 15.9950 1193Iran 64.68 11.7746 1112Montenegro 72.63 132.6388 1045Belarus 75.81 46.5536 1023Costa Rica 80.85 41.1266 993Serbia 111.22 154.5325 876Albania 213.00 122.9685 744