Science of the Total Environment 806 (2022) 150308 Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv Environmental heterogeneity in human health studies. A compositional methodology for Land Use and Land cover data Quim Zaldo-Aubanell a,b,⁎, Isabel Serra c,d, Albert Bach b,h, Pablo Knobel b, Ferran Campillo i López b,e, Jordina Belmonte a,f, Pepus Daunis-i-Estadella g, Roser Maneja b,h,i a Institute of Environmental Science and Technology (ICTA), Autonomous University of Barcelona (UAB), Z building, ICTA-ICP, Carrer de les columnes, UAB Campus, 08193, Bellaterra, Barcelona, Spain b Environment and Human Health Laboratory (EH2 Lab), Forest Science and Technology Center of Catalonia, Ctra. de St. Llorenç de Morunys, km 2, 25280 Solsona, Spain c Centre de Recerca Matemàtica, Edifici C, 08193 Bellaterra, Barcelona, Spain d Barcelona Supercomputing Center, 08034 Barcelona, Spain e Paediatric Environmental Health Specialty Unit, Paediatric Team of the Garrotxa Region, Olot and Garrotxa Region Hospital Foundation, Avinguda Països Catalans 86, 17800 Olot, Catalonia, Spain f Department of Animal Biology, Plant Biology and Ecology, Autonomous University of Barcelona (UAB), C building, UAB Campus 08193, Bellaterra, Barcelona, Spain g Department of Computer Science, Applied Mathematics and Statistics, Universitat de Girona, Carrer Universitat de Girona, 6, 17003 Girona, Spain h Forest Science and Technology Center of Catalonia, Ctra. de St. Llorenç de Morunys, km 2, 25280 Solsona, Spain i Geography Department, Autonomous University of Barcelona (UAB), B building, UAB Campus 08193, Bellaterra, Barcelona, Spain H I G H L I G H T S G R A P H I C A L A B S T R A C T • Compositional methodology spotting independent effect of LULC categories • Important effect measure modification of socioeconomic status, age group and sex • Environmental heterogeneity as a key factor in environment-health studies • Compositional methodology as a suit- able tool assessing environmental het- erogeneity a r t i c l e i n f o a b s t r a c t Article history: The use of Land use and Land cover (LULC) data is gradually becoming more widely spread in studies relating the Received 2 June 2021 environment to human health. However, little research has acknowledged the compositional nature of these Received in revised form 8 September 2021 data. The goal of the present study is to explore, for the first time, the independent effect of eight LULC categories Accepted 8 September 2021 (agricultural land, bare land, coniferous forest, broad-leaved forest, sclerophyll forest, grassland and shrubs, Available online 17 September 2021 urban areas, and waterbodies) on three selected common health conditions: type 2 diabetes mellitus (T2DM), Editor: Scott Sheridan asthma and anxiety, using a compositional methodological approach and leveraging observational health data of Catalonia (Spain) at area level. We fixed the risk exposure scenario using three covariates (socioeconomic status, age group, and sex). Then, we Keywords: assessed the independent effect of the eight LULC categories on each health condition. Our results show that each Land use and Land cover LULC category has a distinctive effect on the three health conditions and that the three covariates clearly modify Environmental heterogeneity this effect. Compositional analysis ⁎ Corresponding author at: Environment and Human Health Laboratory (EH2 Lab), Forest Science and Technology Center of Catalonia, Ctra. de St. Llorenç de Morunys, km 2, 25280 Solsona, Spain. E-mail address: quim.zaldo@uab.cat (Q. Zaldo-Aubanell). https://doi.org/10.1016/j.scitotenv.2021.150308 0048-9697/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 Type 2 diabetes mellitus This compositional approach has yielded plausible results supported by the existing literature, highlighting the Asthma relevance of environmental heterogeneity in health studies. In this sense, we argue that different types of envi- Anxiety ronment possess exclusive biotic and abiotic elements affecting distinctively on human health. We believe our contribution might help researchers approach the environment in a more multidimensional man- ner integrating environmental heterogeneity in the analysis. © 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction to explain the geographical distribution of both COVID-19 incidence and mortality in Catalonia (Spain) (Zaldo-Aubanell et al., 2021a). In the last decades, the study of the relationship between the The aim of this study is to explore, for the first time, the inde- environment and human health has increasingly captured the sci- pendent effect of eight LULC categories (agricultural areas, bare entific community's attention worldwide. As a result, growing evi- land, coniferous forest, broad-leaved forest, sclerophyll forest, dence is starting to highlight the potential benefits that the grassland and shrubs, urban areas, and waterbodies) on three environment might provide to human health (Barnes et al., 2019; selected common health conditions (type 2 diabetes mellitus Sandifer et al., 2015; Stier-Jarmer et al., 2021; Taylor and Hochuli, (T2DM), asthma and anxiety) using a compositional methodolog- 2015). Until now, different metrics assessing the environment ical approach. This study leverages observational health data of have been showcased (Bach et al., 2020); from measurements of Catalonia (Spain) at area level (the Basic Health Areas; BHAs). the landscape proportion, the Normalized Difference Vegetation Index (NDVI), to the Simpson's patch diversity and percentages of 2. Materials and methods tree canopy and green spaces (Cushman et al., 2008; Donovan et al., 2013; Frank et al., 2006; Hanski et al., 2012; Sarkar et al., 2.1. Environmental heterogeneity: Land use and Land cover (LULC) dataset 2013). Among these metrics, Land use and Land cover (LULC) data are gradually becoming more widely spread (Zaldo-Aubanell We described the environmental heterogeneity of each Basic et al., 2021b). Health Area (BHA) according to relevant literature (Astell-Burt LULC data is usually a low-cost and high resolution resource, which is and Feng, 2019; Hanski et al., 2012; Wheeler et al., 2015). First, regularly updated (Grekousis et al., 2015). Unlike other environmental we reclassified the prior 23 LULC categories of the Land Use and data, LULC data encompasses both biophysical (e.g., temperature, humid- Land Cover map of Catalonia (Spain) from 2012 into eight major ity, biodiversity, and soil features) and socioeconomic (e.g., political, eco- categories: agricultural areas, bare land, coniferous forest, broad- nomic, and cultural) environmental information (Boada and Gomez, leaved forest, sclerophyll forest, grassland and shrubs, urban 2012). Moreover, LULC data is versatile and can distinguish between dif- areas, and water bodies (see Table S1 and Fig. S2 in Supplementary ferent types of environments, which is particularly relevant since different materials). Then, we calculated the vector of proportions of each types of environments are suggested to have specific elements that might reclassified LULC category for each BHA. promote distinct benefits to human health (Astell-Burt and Feng, 2019; de The 2012 Land Use and Land Cover map of Catalonia is a tool Vries, 2019; Wheeler et al., 2015; White et al., 2013; Zaldo-Aubanell et al., generated with automated image classification of a 30-m resolu- 2021b). tion (minimum area representing 30x30m). Images are obtained Many researchers have used LULC data to describe the environment thought Landsat satellite (Landsat-5, Landsat-7, Landsat-8, and (Groenewegen et al., 2012; Maas et al., 2009a; Mitchell and Popham, Sentinel-2) using both their sensors (Thematic Mapper (TM), En- 2007; van den Berg et al., 2010; Wheeler et al., 2012). However, little re- hanced Thematic Mapper Plus (ETM+), Operational Land Imager search has acknowledged the compositional nature of such data (OLI) and Multispectral Imager (MSI)), and complementary infor- (Aitchison, 2009; Pearson, 1896). mation such as the Urbanistic Map of Catalonia and the graph of The compositional nature of LULC data can be observed in the analysis the Catalonia infrastructures network. The map also incorporates of proportions of specific LULC categories composing geographical regions the cartographic database of forest fires from the Ministry of (Leininger et al., 2013). Each observation is a vector of proportions which Agriculture, Livestock, Fisheries and Food of Catalonia, and has the peculiarity of being constrained, so the sum of all its parts is a con- the LIDAR database from the Institut Cartogràfic i Geològic de stant (Aitchison and Egozcue, 2005). Thus, the vector of proportions is a Catalunya (ICGC) (https://territori.gencat.cat/ca/01_departament/ D-part composition, with D components, and the “sample space” is not 12_cartografia_i_toponimia/bases_cartografiques/medi_ambient_ the real Euclidean space associated with unconstrained data (Aitchison, i_sostenibilitat/usos-del-sol/). 2009). Therefore, D-part compositions only provide information about the relative magnitudes of the compositional components (Aitchison and Egozcue, 2005; Hron et al., 2012), and this information is completely 2.2. Health data gathered in D-1 ratios between the components. Some researchers have avoided the singularity constraint of LULC The prevalence rates of T2DM, asthma, and anxiety for each BHA data by classifying the percentages of LULC categories according to were obtained from the Catalan Health Department and the Catalan quantile division (Bixby et al., 2015; Lachowycz and Jones, 2014; Agency for Health Quality and Evaluation (AQuAS). The health data Mitchell and Popham, 2008; Mytton et al., 2012; Wu et al., 2015). corresponded to 2014, they were aggregated by age group and sex, These methods represent a step forward to the proper use of LULC and were provided for each Basic Health Area (BHA), the fundamental data. However, consideration of the compositional nature of LULC data territorial unit through which the Catalan Healthcare System is articu- might be a critical factor as this has been demonstrated to respect lated (see Fig. S2 in Supplementary materials). scale invariance and the relative scale issues that are completely ignored The data did not distinguish between type 1 diabetes and type 2 diabe- when raw data (e.g., proportions or percentages) is used (Müller et al., tes. However, almost 95% of all diagnosed cases of diabetes are estimated 2018). To our knowledge, no other study relating LULC data to human to be type 2 in adults (Centers for Disease Control and Prevention, health has yet used a compositional approach for the analysis, except 2011). Therefore, we assumed a well representation of individuals with for one study assessing the potential contribution of LULC categories T2DM in our data as previously described (Astell-Burt et al., 2014a). 2 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 2.3. Covariates resolution of our LULC data was 30 × 30 m, we opted for simple imputation of these zeroes. Thus, we replaced the zeroes by the mini- Three covariates (age group, sex, and socioeconomic status (SES)) mum value of 900 m2 to create a vector of positive components w ∈ were used for segmentation to fix the risk exposure scenario and assess SD that then was closured r = ∁ (w). We assumed that it was plausible the independent effect of the eight LULC categories. to consider that each LULC category could be represented in at least one We considered six age groups: paediatric (named 1 in the figures area of 30 × 30 m within each BHA. For instance, agriculture areas might and the tables; <15 years), teenagers and young adults (2; 15–44 take the form of urban agriculture in heavily urbanised BHA. Forested years), adult (3; 45–64 years), senior (4; 65–74 years), old (5; 75–84 areas might also be present, shaping street trees. Even waterbodies years) and very old (6; >84 years). Besides, sex was categorised as might be represented in small pools of water. male or female. Furthermore, SES information was obtained using the 2015 Com- 2.4.2. Statistical analysis posed Socioeconomic Index (CSI) (Colls et al., 2020) from the Catalan Before the analyses, and to avoid potential interaction problems de- Health Observatory. The CSI is a deprivation index calculated for each rived from possible associations, we used a Chi-square testing for signif- BHA and used in assessing resources for Primary Health. The CSI is a icant differences between socioeconomic levels and the presence of continuous variable (0 to 100) that includes the following information: LULC categories. economic income, level of education, professional occupation, life ex- We fitted four different regressions to our data: Binomial, Poisson, pectancy, premature death rate, and evitable hospitalizations rate Negative Binomial, and Beta regression. Similar results were found for (Colls et al., 2020). We used a quintile division of this index, creating all regressions, although Negative Binomial and Beta regression showed five categories: very high (CSI ≤ 34.75), high (34.75 > CSI ≤ 42.60), me- higher confidence intervals. As expected, the best-fit model according to dium (42.60 > CSI ≤ 48.99), low (48.99 > CSI ≤ 56.37) and very low SES AIC was for Negative Binomial regression. However, we noted no im- (56.37 > CSI ≤ 100). portant differences in the estimates derived from the four regressions. Thus, we finally modelled our data using the Binomial regression with 2.4. Data analysis logit link, as it best represented the nature of the assessed health condi- tions (see Eq. (4)). Furthermore, this model was the most feasible in 2.4.1. Compositional ilr-transformation terms of simplicity and interpretability. The population size of each The use of raw compositional data to directly conduct common sta- BHA was used as weights when fitting the model. tistical analyses raises the problem of singularity, which was already warned as problematic over a century ago (Pearson, 1896). Instead, Yi~Bernoulli(pi) for i = 1, …, n. we propose a proper transformation of compositional data, moving   pi n the compositions isometrically from the simplex with the Aitchison ge- Logit ðμ i Þ ¼ log ¼ β 0 þ ∑ βi  X i ∗ ð4Þ ometry to the standard real space with the Euclidean geometry 1−pi i¼1 (Egozcue and Pawlowsky-Glahn, 2016; Hron et al., 2012). The transfor- mation creates new coordinates in the Euclidean geometry and thus where Yi was the binary (Bernoulli) response variable; pi was the allow for popular statistical methods to be applied (Müller et al., probability of successes P(Yi = 1), in this case, 1 stands for a diagnosed 2018). In this case, regression modelling. case; μi was the expected value of each Yi which was equal to the Since regression models are meaningful only when the composi- probability of successes pi; β0 was the intercept, and βi denoted the tional covariates are expressed on an orthonormal basis, a feasible alter- logistic regression coefficients for the design ilr-matrix X ∗ of covariables i. native to raw data (e.g., proportions and percentages) is to use the The role of the covariates ‘age group’, ‘sex’, and ‘socioeconomic sta- isometric logratio (ilr) transformation (Hron et al., 2012). In particular, tus’ has been extensively used to describe health status (Beyer et al., this could be the set of pivot logratios (PLRs), which are a succession of 2018; Frank et al., 2004; Mobley et al., 2006; Richardson and Mitchell, ilr-coordinates where the numerator in the ratio is always a single com- 2010; Van den Berg et al., 2016). We used these covariates for segmen- ponent and the denominator all those other components “to the right” tation to fix the risk exposure scenario. In total, sixty segmentations in the ordered list of components (Greenacre and Grunsky, 2019). were performed. Then, we assessed the independent effect of the For a better interpretation of the regression coefficients, Müller et al. eight ilr-transformed LULC categories on each health condition, as pre- (2018) suggest moving from the orthonormal to the orthogonal coordi- viously described in Müller et al. (2018). nates (see Eq. (1)). So, in a regression with non-compositional response Additionally, we also show the estimated coefficients of all and compositional regressors, an additive increment of one unit in the explanatory variables (sex, age group, socioeconomic status, and ilr- ilr-orthogonal variable (z(l)∗) is equal to a two-fold multiplicative in- transformed LULC categories) for each selected health condition derived crease in the relative dominance of the original composition variable from ordinary general models using non-segmented data (see Table S3 x, if the base-2 logarithm is used (Müller et al., 2018) (see Eq. (2)). in Supplementary materials). This transformation follows: We conducted the statistical analyses using the R language environ- ment for statistical computing, R version 3.6.2 (12 December 2019) (R ðlÞ ðlÞ∗ xi Core Team, 2019). zi ¼ log 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi , i ¼ 1, . . . , D−1 ð1Þ D ðlÞ D−i ∏ j¼iþ1 xj 3. Results 0 1 ðlÞ ðlÞ ðlÞ ðlÞ∗ B xi C xi xi Before conducting any regression analysis, and to avoid potential in- Δzi ¼ log 2 @2  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiA ¼ log 2 ð2Þ þ log 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi ¼1þ log 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi D ðlÞ D ðlÞ D ðlÞ D−i ∏ j¼iþ1 xj D−i ∏ j¼iþ1 xj D−i ∏ j¼iþ1 xj teraction problems derived from possible associations, we used a Chi- i ¼ 1, . . . , D−1 square to test for significant differences between socioeconomical levels ð2Þ and LULC categories. The Chi-squared test showed no significant evi- dence to reject the hypothesis that SES was not related with LULC cate- In our study, we used the transformation suggested by Müller et al. gories: X2(28,N = 369) = 28.403, p = 0.443. Therefore, we assumed no (2018) (Eq. (1)) to transform the vector of proportions of the eight LULC significant relationship between socioeconomic levels of BHAs and the categories describing the environmental heterogeneity for each BHA. presence of any particular LULC category. We observed some zeroes derived from the lack of some LULC cate- Hereunder, we show the independent effect of each ilr-transformed gories in specific BHAs. To simplify and given that the minimum LULC category on the segmented health conditions (see Figs. 2, 3, and 4; 3 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 Fig. 2. Odds ratios (in colour) and 95% CI (grey bands). Associations between the prevalence of T2DM and ilr-LULC categories. 4 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 Fig. 3. Odds ratios (in colour) and 95% CI (grey bands). Associations between the prevalence of asthma and ilr-LULC categories. 5 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 Fig. 4. Odds ratios (in colour) and 95% CI (grey bands). Associations between the prevalence of anxiety and ilr-LULC categories. 6 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 coloured lines). Since we used data representing the entire population Results for ilr-sclerophyll forest showed an increased risk for lower of Catalonia, inference information gathered in the p-values has little SES groups, although the results for very low SES groups were some- relevance. However, we also provide the 95% Wald confidence intervals what inconclusive. For ilr-grassland and shrubs, inconclusive results (95% CI), reporting the precision of the estimations (Figs. 2, 3, and 4; were observed for very high SES groups and an increased risk was ob- grey bands). served for lower SES groups. Likewise, ilr-urban areas showed an in- creased risk for lower SES groups. 3.1. T2DM Finally, ilr-waterbodies showed a reduced risk for most of groups. However, we found some exceptions for medium SES groups and As shown in Fig. 2, most groups for ilr-agricultural areas showed an some age groups of low SES. increased risk of T2DM, with some exceptions on paediatric (1), some young-adults (2), and very low SES groups. 4. Discussion For ilr-bare land areas, very high and very low SES groups were asso- ciated with an increased risk of T2DM, while all the rest were mainly as- Our study has stepped forward to detect the independent contribu- sociated with a reduced risk. tion of eight LULC categories to the prevalence of three assessed health For ilr-coniferous forest, younger groups showed a reduced risk, al- conditions (T2DM, asthma and anxiety). We propose a compositional though the majority of other groups showed an increased risk. approach that shows the estimated odds ratio after segmenting the In the case of ilr-broad-leaved and ilr-sclerophyll forest, the majority data according to the SES status, age group, and sex. Moreover, this of the estimates showed a reduced risk while some age groups, espe- methodology has allowed us to discuss the possible set of particular el- cially paediatric (1), showed the opposite effect. ements of each LULC category related to the studied health conditions. As to the effect of ilr-grassland and shrubs areas, higher average risk was observed for medium SES groups, although low SES paediatric 4.1. Human health conditions (1) male group showed the highest risk. For ilr-urban areas, we found an increasing risk for more For this study, we used the major pathways framework suggested by impoverished areas, especially for older groups. Markevych et al. (2017), and hypothesised that each selected health Finally, results for ilr-waterbodies suggested a reduced risk for more condition was related to the environment through specific pathways. impoverished areas with some exceptions in high and very high SES paediatric (1) and young adult (2) groups and low and very low SES 4.1.1. Type 2 diabetes mellitus young-adult (2) male groups. T2DM has been associated with obesity and sedentary behaviour (low physical activity) (Colagiuri et al., 2010). Thus, the relationship be- tween the environment and T2DM may partly occur through the 3.2. Asthma instoration pathway (Markevych et al., 2017). In particular, the capacity of LULC categories to promote physical activity. In addition to this, re- For ilr-agricultural areas, our results show a reduced risk for lower search is also associating long-term air pollution exposure (Eze et al., SES groups (see Fig. 3). 2015, 2014), as well as traffic noise (Shin et al., 2020) with increased For ilr-bare land areas, most groups showed an increased risk, espe- risk of T2DM. Thus, T2DM might arguably be related to the environment cially females. However, high SES groups and other age groups within through the mitigation pathway (Markevych et al., 2017). each SES group showed the opposite effect. Urban areas showed an increased risk of T2DM for medium to lower Results for ilr-coniferous forest differed between sex groups. In gen- SES groups. This effect could be explained due to the fact that higher SES eral, males showed a predominant reduced risk (except for very high groups, which tend to be more physically active during leisure time SES groups). On the contrary, females showed a predominant increased (Marielle et al., 2012), are also associated with a higher presence of risk for very high, high, and low SES groups, while medium and very low greenspaces in their neighbourhoods (Astell-Burt et al., 2014b). This groups were associated with a reduced risk. might lead to higher levels of physical activity in higher SES groups For ilr-broad-leaved forest, results suggest an increased risk for (Frank et al., 2007; Richardson et al., 2013). Contrarily, lower SES lower SES groups. Furthermore, for ilr-sclerophyll forest, results show groups, which are less able to afford gym fees (Giles-Corti and Donovan, a predominant increased risk for all groups, except for older groups of 2002) and have little access to open public spaces (Koohsari, 2011), very high and very low SES, and younger groups of low and very low tend to walk in more unsupportive build environments (Adkins et al., SES. 2017), which lack certain land-use types such as green spaces and rec- Results for ilr-grassland and shrubs and ilr-urban areas suggest a reation centres (Zandieh et al., 2017). These factors might arguably progressively increased risk across SES variable. lead to less opportunities for physical activity in deprived Lastly, for ilr-waterbodies, most groups showed a predominant re- neighbourhoods, increasing the risk of T2DM in lower SES groups. In duced risk of asthma, especially very low SES groups. fact, higher risk of T2DM and many lifestyle-related risk factors have been reported for people living in deprived neighbourhoods (Astell- 3.3. Anxiety Burt et al., 2014a; Feng and Astell-Burt, 2013; Williams et al., 2012). Moreover, deprived neighbourhoods are more associated with polluted As shown in Fig. 4, we found a predominant decreased risk for most areas (Bolte et al., 2010), higher perceived noise and lower perceived groups for ilr-agricultural areas. safety, cleanliness and aesthetic quality of their neighbourhoods For ilr-bare land, very high and very low SES groups showed a pre- (Mouratidis, 2020). In this sense, the increased risk of T2DM for poorer dominant increased risk of anxiety, while medium and low SES showed groups could also be explained through the mitigation pathway the opposite trend. In contrast, results for high SES groups differed be- (Markevych et al., 2017), since both air pollution exposure and traffic tween sexes. Males showed a predominant decreased risk and females noise have been related to increased risk of T2DM (Eze et al., 2015; showed a predominant increased risk. Shin et al., 2020). As to the effect of ilr-coniferous forest, we observed a predominant Broad-leaved forest, which generally has low undergrowth, cool reduced risk for most groups. However, some groups of high, medium temperatures, and smooth slopes, was predominantly associated with and low SES showed an increased risk in both sexes. Following the lower levels of T2DM. Likewise, sclerophyll forest, which tend to be same trend, we found a predominant reduced risk for ilr-broad-leaved close to urban settlements, showed a generally decreased risk of forest except for some very high and very low SES groups. T2DM. This suggests that, despite its undergrowth densely populated 7 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 with shrubs and some lianas, people might be attracted to perform state that complexity in the prevalence of asthma symptoms calls for a physical activity on the sclerophyll forest principal routes and paths. Re- more comprehensive framework for its understanding. garding the coniferous forest, it might not always be accessible to people to perform physical activity (at least the high mountain forest, which is 4.1.3. Anxiety established on higher altitudes of the mountains). This might explain Anxiety could be mainly linked to the environment through the res- the predominant increased risk of T2DM observed for most groups. toration pathway (Markevych et al., 2017). The restorative role of natu- Regarding waterbodies, we found a heterogeneous effect, suggesting ral environments has been associated to the visual (Franco et al., 2017; that effect measure modification of SES and age group appeared to be Jacobs and Suess, 1975) and auditory stimuli (Coensel et al., 2011), to especially important. the openness of natural spaces (Han, 2007) and to species richness In agricultural areas, results showing increased risk of T2DM could (Aerts et al., 2018). Many studies have highlighted the potential of the be explained through the pesticides exposure pathway (Lee et al., natural environment to increase feelings of restoration (White et al., 2011; Turyk et al., 2009). In a previous study, researchers found five pes- 2013), and reduce stress levels (Morita et al., 2007) and heart rate ticide types positively associated with incident diabetes (Starling et al., (Lee et al., 2014). 2014). In this regard, epidemiological evidence suggests an association Our results suggest that agricultural areas seem foster mental resto- between exposure to organochlorine pesticides and T2DM (Evangelou ration, leading to a main decreased risk of anxiety. Agricultural areas et al., 2016). may represent more diverse landscape configuration, such as the forest-agriculture mosaic, which can increase bird richness (Atauri 4.1.2. Asthma and De Lucio, 2001). Concerning this, recent research has highlighted The relationship between the environment and asthma may occur the positive association between bird diversity and people's wellbeing mainly through the mitigation and instoration pathways (Markevych (Methorst et al., 2020). Moreover, the openness of agricultural scenarios et al., 2017). Air quality is crucial (Toure et al., 2019), not only regarding might also promote restoration for individuals (Han, 2007). the allergenic pollen levels that might aggravate this respiratory disease Likewise, we found that environments with more waterbodies were (Cariñanos and Casares-Porcel, 2011), but also concerning air pollution associated with decreased anxiety levels in the majority of groups. Some (Guarnieri and Balmes, 2014). In addition, other studies have explored authors have underscored the importance of water sounds for mental the connection between air quality and heat events (Soneja et al., health (White et al., 2010). Others have hypothesised the independent 2016) and microbial diversity (Ege et al., 2011). beneficial effect on health for waterbodies (Nutsford et al., 2016), A higher prevalence of asthma is directly related to the higher pres- even without a full exposure but with only observation of water ele- ence of allergenic elements in the air (D'Amato et al., 2007). We found ments photographs (White et al., 2010). that waterbodies show a generally reduced risk of asthma for all SES The beneficial effect of waterbodies might also be linked to evolu- groups. This could be because waterbodies might arguably be associated tionary and cultural theories such as the Biophilia effect (Browning with little presence of allergenic elements due to the lack of vegetation, et al., 2014), which highlights the significance of water for the biological, and because the allergenic elements might be trapped in the wellbeing and survival needs (Han, 2007; Orians and Heerwagen, 1992; waterbodies. In fact, Spanish researchers found that living next to the Ulrich, 2016, 2014). In addition, some research points out to the role of coast may protect against sensitization to pollens and Alternaria negative ions that are present in water environments, especially by the (Moral Gil et al., 2008). sea (Jiang et al., 2018), in lowering depression scores (Perez et al., 2013). Although a set of allergenic species might be distinguished for each Regarding forested LULC categories, only broad-leaved forests, argu- LULC category, the differences across SES groups for LULC categories as- ably the most walkable of the forest types, showed decreased risk of sociated with vegetation suggest that the presence of allergenic species anxiety for most of the SES groups. On the other hand, we found in- might not be the only relevant predictor. For broad-leaved forest and creased anxiety levels for lower SES groups in urban areas. As described grassland and shrubs, we found a higher risk in lower SES groups. On above, important differences between deprived and non-deprived the other hand, sclerophyll forest showed a predominant increased neighbourhoods seem to be essential elements modifying the effect of risk, bare land showed higher predominant risk for females, and conif- urban areas on anxiety levels. These differences include the built envi- erous forest showed a reduced predominant risk for males. ronment, presence of greenspaces (Astell-Burt et al., 2014b), accessibil- For urban areas, we found increased levels of asthma for lower SES ity to public open spaces (Koohsari, 2011), and other important factors groups. Some research has suggested that air polluted environments mentioned above such as noise, light and air pollution. Likewise, results are associated with a higher prevalence of respiratory illnesses such as showing a higher risk for sclerophyll forest and grassland and shrubs asthma (Annesi-Maesano et al., 2007; D'Amato et al., 2002). Further- might be explained by differences in quality of spaces across SES groups. more, lower SES groups are known to be in worse health status and as- sociated with more polluted areas (Bolte et al., 2010; de Vries et al., 4.2. The role of covariates: SES, age group, and sex 2003; Su et al., 2011). Contrarily, the increased levels of greenspaces in richer SES neighbourhoods might lower air pollutant concentrations The three covariates studied (SES, age group, and sex) have widely (Kroeger et al., 2014). been described as important modifiers of nature effects (Markevych In addition, the higher risk of asthma found in lower SES urban areas et al., 2017). Considering this, we have segmented our data accordingly. would also be explained since lower SES groups have been linked with Thus, the three covariates did not play a role in the residuals of our increased susceptibility to heat-associated health effects (Gronlund, models. Nevertheless, we can compare the effects of the covariates 2014). In addition, extreme heat events have been linked with increas- when comparing the estimated odds ratios across different groups ing risk of hospitalization for asthma (Soneja et al., 2016). (Figs. 2, 3, and 4). We found agricultural areas to be generally associated with a reduced We found SES to be the most important covariate as to the effect of risk of asthma for all SES groups except for very high SES groups. Some au- the LULC categories on the selected health conditions. As previously re- thors highlight microbial diversity and specific microbial exposure as pro- ported (Knobel et al., 2021a), we found a dissimilar risk between the tective factors against asthma and atopy (Bach, 2018; Jatzlauk et al., highest and lowest SES groups. This effect measure modification could 2017). In this regard, farm environments that promote higher microbial be explained because lower SES groups might be more likely to benefit exposure would be associated with decreased risk of asthma (Ege et al., from a health promotion intervention (Bolte et al., 2010; de Vries et al., 2011). As previously highlighted (Farfel et al., 2010; Shankardass et al., 2003; Markevych et al., 2017; Su et al., 2011), such as being exposed to 2007), the results for very high SES groups are compatible with the so- particular LULC categories. Lower SES groups have been described to be called “hygiene hypothesis”. However, other studies (Eum et al., 2019) less mobile (Maas et al., 2006; Schwanen et al., 2002), to have less 8 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 health status (Markevych et al., 2017), and to live in more polluted areas might have remained unstudied in these general classifications. One ex- (Bolte et al., 2010). In this sense, lower SES groups show stronger asso- ample is the coniferous forest category, which gathered three forest ciations with their environment (Dadvand et al., 2014; Maas et al., types: low land pine tree forest, the montane pine tree forest, and 2009b). On the contrary, higher SES groups might be more capable of high mountain forest. Each of them possessing individual and excep- changing the environment they live in in order to be less exposed to tional features. In this sense, we argue that using a combination of particular harmful conditions (Bell et al., 2010; Markevych et al., other existing datasets and GIS techniques to construct a more exhaus- 2017). In addition, other factors such as better social environments, tive classification of the different ecosystems or types of environment health care accessibility and use, residential environment preferences, would result in better analyses. In the same direction, authors could and even better general behaviour and lifestyle might also be key factors make use of the present framework to conduct further research testing linking higher SES groups with better general health (Adler and for differences within agricultural areas (irrigated, non-irrigated), urban Newman, 2002). (residential, sprawl, industrialized, urban greenspaces) or waterbodies Regarding the age group, many of the estimates showed a more in- (inland, marine), among other examples. tense effect on very old (6) age group than on young-adult (2), adult (3), senior (4) and old (5). Likewise, we found a recurrent intensified ef- 4.4. Limitations fect for the paediatric (1) group compared to the rest of groups. This ef- fect was found to go either in the same direction as the rest of the age 4.4.1. Unit of analysis groups (e.g., prevalence of T2DM for high SES groups for ilr-broad- We have used the Basic Health Area (BHA) as our unit of analy- leaved forest; Fig. 3), or in the opposite direction (e.g., prevalence of sis to describe the living environment for individuals. However, anxiety for low SES groups for ilr-bare land; Fig. 4). other research encompassing LULC data and human health has On the one hand, elderly people are undoubtedly the groups with used different unit of analysis, from census areas to different buffer less moving capacity. This makes elders more strongly associated with radii (Zaldo-Aubanell et al., 2021b). In this sense, further research their surroundings (Maas et al., 2006). On the other hand, paediatric is needed to test our results and approach using different units of people have unique characteristics that differ from adults (Ortega- analysis. garcía et al., 2019). Additionally, the paediatric group might be under- represented in at least two of the health conditions considered; T2DM 4.4.2. Cross-sectional design and anxiety. Type 1 diabetes, more associated with paediatric groups We followed a cross-sectional design. Thus, limitations derived from (Soltész, 2003), is an autoimmune disorder genetically mediated, this methodology must be carefully considered. Although cross- while type 2 is more of a life style induced disorder (Joshi and sectional designs leverage population-based data and are useful to de- Shrestha, 2010). Regarding anxiety, the anxiety diagnostic criteria in tect differences between areas, they do not allow for causal inference children might also differ from adults. Children have particular features to be established (Wu et al., 2020). (for instance, difficulties in communication, cognition and emotions) Furthermore, we did not consider possible spatial autocorrelations that create unique challenges when distinguishing between normal derived from data. In this sense, robust spatio-temporal methodologies and pathological anxiety (Beesdo et al., 2021). should be needed allowing for the establishment of causal inference. Lastly, in line with other studies (Richardson and Mitchell, 2010), we found sex to be an important variable modifying the effect of LULC cat- egories on the three selected health conditions. 5. Conclusions 4.3. Future research There is a recurrent call for new methodologies that detect the inde- pendent effect of different types of environment on human health. We The compositional methodology used in this study has allowed us to have proposed an innovative methodology using a compositional ap- raise many hypotheses linking LULC categories to the three assessed proach. We have defined eight types of environment using a classifica- human health conditions. Although these hypotheses have been sup- tion of Land use and Land cover data and have tested their individual ported by previous studies, they should be tested further with specific contribution on explaining three selected health conditions using obser- study designs in future studies. vational data. In this regard, the proposed methodology has shown to be an acceptable and a feasible way to address the compositional nature of 4.3.1. Complex pathways linking the environment to human health LULC data, facilitating the interpretation of the estimates through the To simplify the results interpretations, we have assumed that anxi- Log2 ilr-transformation. ety was related to the environment mainly through the restoration Our approach has led us to plausible results supported by the pathway. In contrast, we assumed that T2DM and asthma were related existing literature while has enabled us to push forward the debate on to the environment mainly through two pathways (instoration and mit- the relevance of environmental heterogeneity in health studies. We igation pathways). However, it is possible that, in reality, many other have proposed a detailed conception of the environment that goes be- pathways intertwine (Hartig et al., 2014; Markevych et al., 2017). For in- yond green and natural. In addition, we have discussed how different stance, some of the results found for T2DM needed a broad scope to be types of environment possess exclusive elements (humidity, tempera- interpreted, such as the pesticide exposure for T2DM. ture, type of flora and fauna, accessibility, walkability, openness, pres- Moreover, differences in preferences (Han, 2007; Kiley et al., 2017; ence of water, sounds, air compounds and air quality, heat, and noise, Lyons, 1983) between the different tested groups, the amount of knowl- light contamination, and even chemical exposure) affecting distinc- edge about the environment (Aerts et al., 2018), and even the environ- tively on human health. Furthermore, we have found the relationship mental quality of the LULC categories (Knobel et al., 2021b; Wheeler between the environment and human health to be clearly modified by et al., 2015) might be important factors modifying the LULC – human socioeconomic status, age group, and sex. Lastly, we have highlighted health relationship. Therefore, more studies with accurate information that other important ideas such as the preferences (or agreeableness) about the variables mentioned above should be conducted to relate spe- for specific types of the environment of certain groups and quality of cific health conditions with specific LULC categories. the environments might be important factors and should be considered in future research. 4.3.2. Further exploration of LULC categories We believe that our contribution might help researchers approach We aimed to test the independent effect of broad LULC categories on the environment in a more multidimensional scope, allowing environ- human health. However, we believe that important characteristics mental heterogeneity to be brought into the analysis. 9 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 CRediT authorship contribution statement Atauri, J.A., De Lucio, J.V., 2001. The role of landscape structure in species richness distri- bution of birds, amphibians, reptiles and lepidopterans in Mediterranean landscapes. Landsc. Ecol. 16, 147–159. https://doi.org/10.1023/A:1011115921050. Quim Zaldo-Aubanell: Conceptualization, Formal analysis, Investi- Bach, J.-F., 2018. The hygiene hypothesis in autoimmunity: the role of pathogens and gation, Data curation, Writing – original draft. Isabel Serra: Conceptual- commensals. Nat. Rev. Immunol. 18, 105–120. https://doi.org/10.1038/nri.2017.111. Bach, A., Peñuelas, J., Clarà, J., Llusià, J., López, F.C.I., Maneja, R., 2020. How should forests ization, Methodology, Formal analysis, Data curation, Writing – original be characterized in regard to human health? Evidence from existing literature. Int. J. draft, Supervision. Albert Bach: Writing – review & editing, Visualiza- Environ. Res. Public Health 17, 1–15. https://doi.org/10.3390/ijerph17031027. tion. Pablo Knobel: Writing – review & editing. Ferran Campillo i Barnes, M.R., Donahue, M.L., Keeler, B.L., Shorb, C.M., Mohtadi, T.Z., Shelby, L.J., 2019. Char- López: Resources, Writing – review & editing. Jordina Belmonte: acterizing nature and participant experience in studies of nature exposure for posi- tive mental health: an integrative review. Front. Psychol. https://doi.org/10.3389/ Writing – review & editing. Pepus Daunis-i-Estadella: Methodology, fpsyg.2018.02617. Validation. Roser Maneja: Resources, Supervision, Project administration, Beesdo, K., Knappe, S., Pine, D.S., 2021. Anxiety and Anxiety Disorder s in Children and Ad- Funding acquisition. olescents: Developmental Issues and I mplic ations for DSM - V 32. , pp. 483–524 https://doi.org/10.1016/j.psc.2009.06.002. Bell, M.L., Belanger, K., Ebisu, K., Gent, J.F., Lee, H.J., Koutrakis, P., Leaderer, B.P., 2010. Pre- Declaration of competing interest natal exposure to fine particulate matter and birth weight. Epidemiology 21, 884–891. https://doi.org/10.1097/EDE.0b013e3181f2f405. The authors declare that they have no known competing financial Beyer, K.M.M., Kaltenbach, A., Szabo, A., Bogar, S., Nieto, F.J., Malecki, K.M., 2018. Exposure to neighborhood green space and sleep: evidence from the Survey of the Health of interests or personal relationships that could have appeared to influ- Wisconsin. Sleep Heal. 4, 413–419. https://doi.org/10.1016/j.sleh.2018.08.001. ence the work reported in this paper. Bixby, H., Hodgson, S., Fortunato, L., Hansell, A., Fecht, D., 2015. Associations between green space and health in english cities: an ecological, cross-sectional study. PLoS One 10, e0119495. https://doi.org/10.1371/journal.pone.0119495. Acknowledgments Boada, M., Gomez, F.J., 2012. Boscos de catalunya (Barcelona). Bolte, G., Tamburlini, G., Kohlhuber, M., 2010. Environmental inequalities among children Quim Zaldo-Aubanell was supported by AGAUR FI fellowship (DOGC in Europe - evaluation of scientific evidence and policy implications. Eur. J. Pub. num. 7720, of 5.10.2018). Isabel Serra acknowledges support from Health 20, 14–20. https://doi.org/10.1093/eurpub/ckp213. Browning, W., Ryan, C.O., Clancy, J., 2014. 14 Patterns of Biophilic Design: Improving FIS2015-71851-P and PGC-FIS2018-099629-B-I00 from Spanish Health and Well-Being in the Built Environment. Terrapin Bright Green LLC, New MINECO and MICINN, and was partially funded by the grant RTI2018- York, p. 2014. 096072-B-I00 from the Spanish Ministry of Science, Innovation and Uni- Cariñanos, P., Casares-Porcel, M., 2011. Urban green zones and related pollen allergy: a re- view. Some guidelines for designing spaces with low allergy impact. Landsc. Urban versities. Jordina Belmonte was supported by the Spanish Ministry of Plan. 101, 205–214. https://doi.org/10.1016/j.landurbplan.2011.03.006. Science and Technology through the project CTM2017-86565-C2-1-O Centers for Disease Control and Prevention, 2011. National Diabetes Fact Sheet: National and by the Catalan Government AGAUR through 2017SGR1692. Pepus Estimates and General Information on Diabetes and Prediabetes in the United States, Daunis-i-Estadella acknowledges support from the project RTI2018- 2011. U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, Atlanta, GA. 095518-B-C21 Methods for Compositional analysis of Data (CODAMET), Coensel, B. De, Vanwetswinkel, S., Botteldooren, D., 2011. Effects of natural sounds on the Ministerio de Ciencia, Innovación y Universidades, Spain. The funding perception of road traffic noise. J. Acoust. Soc. Am. 129, EL148–EL153. https://doi.org/ sources did not participate in the design or conduct of the study, the col- 10.1121/1.3567073. Colagiuri, S., Vita, P., Cardona-Morrell, M., Singh, M.F., Farrell, L., Milat, A., Haas, M., Bauman, lection, management, analysis, or interpretation of the data, or the prep- A., 2010. The Sydney Diabetes Prevention Program: a community-based translational aration, review, or approval of the manuscript. study. BMC Public Health, 10 https://doi.org/10.1186/1471-2458-10-328. We want to thank Marcel Bach-Pagès for his insights in the manu- Colls, C., Mias, M., García-Altés, A., 2020. A deprivation index to reform the financing script. model of primary care in Catalonia (Spain). Gac. Sanit. 34, 44–50. https://doi.org/ 10.1016/j.gaceta.2018.07.015. Core Team, R., 2019. R: A Language and Environment for Statistical Computing. R Founda- Appendix A. Supplementary data tion for Statistical Computing. Cushman, S.A., McGarigal, K., Neel, M.C., 2008. Parsimony in landscape metrics: strength, universality, and consistency. Ecol. Indic. 8, 691–703. https://doi.org/10.1016/j. Supplementary data to this article can be found online at https://doi. ecolind.2007.12.002. org/10.1016/j.scitotenv.2021.150308. Dadvand, P., Wright, J., Martinez, D., Basagaña, X., McEachan, R.R.C., Cirach, M., Gidlow, C.J., de Hoogh, K., Gražulevičiene, R., Nieuwenhuijsen, M.J., 2014. Inequality, green spaces, and pregnant women: roles of ethnicity and individual and neighbourhood References socioeconomic status. Environ. Int. 71, 101–108. https://doi.org/10.1016/j.envint. 2014.06.010. Adkins, A., Makarewicz, C., Scanze, M., Ingram, M., Luhr, G., 2017. Contextualizing D'Amato, G., Liccardi, G., D'Amato, M., Cazzola, M., 2002. Outdoor air pollution, climatic walkability: do relationships between built environments and walking vary by socio- changes and allergic bronchial asthma. Eur. Respir. J. 20, 763–776. https://doi.org/ economic context? J. Am. Plan. Assoc. 83, 296–314. https://doi.org/10.1080/ 10.1183/09031936.02.00401402. 01944363.2017.1322527. D'Amato, G., Cecchi, L., Bonini, S., Nunes, C., Annesi-Maesano, I., Behrendt, H., Liccardi, G., Adler, N.E., Newman, K., 2002. Socioeconomic disparities in health: pathways and policies. Popov, T., Van Cauwenberge, P., 2007. Allergenic pollen and pollen allergy in Europe. Health Aff. 21, 60–76. https://doi.org/10.1377/hlthaff.21.2.60. Allergy Eur. J. Allergy Clin. Immunol. 62, 976–990. https://doi.org/10.1111/j.1398- Aerts, R., Honnay, O., Van Nieuwenhuyse, A., 2018. Biodiversity and human health: mech- 9995.2007.01393.x. anisms and evidence of the positive health effects of diversity in nature and green de Vries, S., 2019. Comparison of the health benefits of different types of nature. JAMA spaces. Br. Med. Bull. https://doi.org/10.1093/bmb/ldy021. Netw. Open 2, e198215. https://doi.org/10.1001/jamanetworkopen.2019.8215. Aitchison, J., 2009. A concise guide to compositional data analysis. Ment. Heal. Subst. Use de Vries, S., Verheij, R.A., Groenewegen, P.P., Spreeuwenberg, P., 2003. Natural environ- Dual Diagnosis 2, 101–110. https://doi.org/10.1080/17523280902932649. ments - healthy environments? An exploratory analysis of the relationship between Aitchison, J., Egozcue, J.J., 2005. Compositional data analysis: where are we and where greenspace and health. Environ. Plan. A 35, 1717–1731. https://doi.org/10.1068/ should we be heading? Math. Geol. 37, 829–850. https://doi.org/10.1007/s11004- a35111. 005-7383-7. Donovan, G.H., Butry, D.T., Michael, Y.L., Prestemon, J.P., Liebhold, A.M., Gatziolis, D., Mao, Annesi-Maesano, I., Moreau, D., Caillaud, D., Lavaud, F., Le Moullec, Y., Taytard, A., Pauli, G., M.Y., 2013. The relationship between trees and human health. Am. J. Prev. Med. 44, Charpin, D., 2007. Residential proximity fine particles related to allergic sensitisation 139–145. https://doi.org/10.1016/j.amepre.2012.09.066. and asthma in primary school children. Respir. Med. 101, 1721–1729. https://doi.org/ Ege, M.J., Mayer, M., Normand, A., Genuneit, J., Cookson, W.O.C.M., Phil, D., Braun- 10.1016/j.rmed.2007.02.022. fahrländer, C., Heederik, D., Piarroux, R., von Mutius, E., Transregio, G., Group, S, Astell-Burt, T., Feng, X., 2019. Association of Urban Green Space with mental health and 2011. New England journal 701–709. general health among adults in Australia. JAMA Netw. Open 2, 1–22. https://doi. Egozcue, J.J., Pawlowsky-Glahn, V., 2016. What are compositional data and how should org/10.1001/jamanetworkopen.2019.8209. they be analyzed? Bol. Estad. e Investig. Oper. 32, 5–29. Astell-Burt, T., Feng, X., Kolt, G.S., 2014a. Is neighborhood green space associated with a Eum, Y., Yoo, E.H., Bowen, E., 2019. Socioeconomic determinants of pediatric asthma lower risk of type 2 diabetes evidence from 267,072 australians. Diabetes Care 37, emergency department visits under regional economic development in western 197–201. https://doi.org/10.2337/dc13-1325. New York. Soc. Sci. Med. 222, 133–144. https://doi.org/10.1016/j.socscimed.2019. Astell-Burt, T., Feng, X., Mavoa, S., Badland, H.M., Giles-Corti, B., 2014b. Do low-income 01.001. neighbourhoods have the least green space? A cross-sectional study of Australia's Evangelou, E., Ntritsos, G., Chondrogiorgi, M., Kavvoura, F.K., Hernández, A.F., Ntzani, E.E., most populous cities. BMC Public Health 14, 19–21. https://doi.org/10.1186/1471- Tzoulaki, I., 2016. Exposure to pesticides and diabetes: a systematic review and meta- 2458-14-292. analysis. Environ. Int. 91, 60–68. https://doi.org/10.1016/j.envint.2016.02.013. 10 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 Eze, I.C., Schaffner, E., Fischer, E., Schikowski, T., Adam, M., Imboden, M., Tsai, M., Carballo, Lachowycz, K., Jones, A.P., 2014. Does walking explain associations between access to D., von Eckardstein, A., Künzli, N., Schindler, C., Probst-Hensch, N., 2014. Long-term greenspace andlower mortality? Soc. Sci. Med. 107, 9–17. https://doi.org/10.1016/j. air pollution exposure and diabetes in a population-based Swiss cohort. Environ. socscimed.2014.02.023. Int. 70, 95–105. https://doi.org/10.1016/j.envint.2014.05.014. Lee, D.H., Lind, P.M., Jacobs, D.R., Salihovic, S., Van Bavel, B., Lind, L., 2011. Polychlorinated Eze, I.C., Hemkens, L.G., Bucher, H.C., Hoffmann, B., Schindler, C., Künzli, N., Schikowski, T., biphenyls and organochlorine pesticides in plasma predict development of type 2 di- Probst-Hensch, N.M., 2015. Review association between ambient air pollution and di- abetes in the elderly: the prospective investigation of the vasculature in Uppsala se- abetes mellitus in Europe. Environ. Health Perspect. 123, 381–389. niors (PIVUS) study. Diabetes Care 34, 1778–1784. https://doi.org/10.2337/dc10- Farfel, A., Tirosh, A., Derazne, E., Garty, B.Z., Afek, A., 2010. Association between socioeco- 2116. nomic status and the prevalence of asthma. Ann. Allergy Asthma Immunol. 104, Lee, J., Tsunetsugu, Y., Takayama, N., Park, B., Li, Q., Song, C., Komatsu, M., Ikei, H., 490–495. https://doi.org/10.1016/j.anai.2010.03.019. Tyrväinen, L., Kagawa, T., Miyazaki, Y., 2014. Influence of Forest Therapy on Cardio- Feng, X., Astell-Burt, T., 2013. Neighborhood socioeconomic circumstances and the co-oc- vascular Relaxation in Young Adults. p. 2014. currence of unhealthy lifestyles: evidence from 206,457 Australians in the 45 and up Leininger, T.J., Gelfand, A.E., Allen, J.M., Silander, J.A., 2013. Spatial regression modeling for study. PLoS One 8. https://doi.org/10.1371/journal.pone.0072643. compositional data with many zeros. J. Agric. Biol. Environ. Stat. 18, 314–334. https:// Franco, L.S., Shanahan, D.F., Fuller, R.A., 2017. A review of the benefits of nature experi- doi.org/10.1007/s13253-013-0145-y. ences: more than meets the eye. Int. J. Environ. Res. Public Health https://doi.org/ Lyons, E., 1983. Demographic correlates of landscape preference. Environ. Behav. 15, 10.3390/ijerph14080864. 487–511. https://doi.org/10.1177/0013916583154005. Frank, L.D., Andresen, M.A., Schmid, T.L., 2004. Obesity relationships with community de- Maas, J., Verheij, R.A., Groenewegen, P.P., De Vries, S., Spreeuwenberg, P., 2006. Green sign, physical activity, and time spent in cars. Am. J. Prev. Med. 27, 87–96. https://doi. space, urbanity, and health: how strong is the relation? J. Epidemiol. Community org/10.1016/j.amepre.2004.04.011. Health 60, 587–592. https://doi.org/10.1136/jech.2005.043125. Frank, L.D., Sallis, J.F., Conway, T.L., Chapman, J.E., Saelens, B.E., Bachman, W., 2006. Many Maas, J., van Dillen, S.M.E., Verheij, R.A., Groenewegen, P.P., 2009a. Social contacts as a pathways from land use to health: associations between neighborhood walkability possible mechanism behind the relation between green space and health. Health and active transportation, body mass index, and air quality. J. Am. Plan. Assoc. 72, Place 15, 586–595. https://doi.org/10.1016/j.healthplace.2008.09.006. 75–87. https://doi.org/10.1080/01944360608976725. Maas, J., Verheij, R.A., De Vries, S., Spreeuwenberg, P., Schellevis, F.G., Groenewegen, P.P., Frank, L.D., Saelens, B.E., Powell, K.E., Chapman, J.E., 2007. Stepping towards causation: do 2009b. Morbidity is related to a green living environment. J. Epidemiol. Community built environments or neighborhood and travel preferences explain physical activity, Health 63, 967–973. https://doi.org/10.1136/jech.2008.079038. driving, and obesity? Soc. Sci. Med. 65, 1898–1914. https://doi.org/10.1016/j. Marielle, A.B., Carlijn, B.M., Katrina, K., Johannes, G., B., A.E., K., A., B., F.J., L, van, 2012. Socio- socscimed.2007.05.053. economic inequalities in occupational, leisure-time, and transport related physical activ- Giles-Corti, B., Donovan, R.J., 2002. Socioeconomic status differences in recreational phys- ity among European adults: a systematic review. Int. J. Behav. Nutr. Phys. Act. 9, 1–23. ical activity levels and real and perceived access to a supportive physical environ- Markevych, I., Schoierer, J., Hartig, T., Chudnovsky, A., Hystad, P., Dzhambov, A.M., de ment. Prev. Med. (Baltim). 35, 601–611. https://doi.org/10.1006/pmed.2002.1115. Vries, S., Triguero-Mas, M., Brauer, M., Nieuwenhuijsen, M.J., Lupp, G., Richardson, Greenacre, M., Grunsky, E., 2019. The isometric logratio transformation in compositional E.A., Astell-Burt, T., Dimitrova, D., Feng, X., Sadeh, M., Standl, M., Heinrich, J., data analysis: a practical evaluation. Economics Working Paper Series. Fuertes, E., 2017. Exploring pathways linking greenspace to health: theoretical and methodological guidance. Environ. Res. 158, 301–317. https://doi.org/10.1016/j. Grekousis, G., Mountrakis, G., Kavouras, M., 2015. An overview of 21 global and 43 re- gional land-cover mapping products. Int. J. Remote Sens. https://doi.org/10.1080/ envres.2017.06.028. 01431161.2015.1093195. Methorst, J., Rehdanz, K., Mueller, T., Hansjürgens, B., Bonn, A., Böhning-Gaese, K., 2020. The importance of species diversity for human well-being in Europe. Ecol. Econ. Groenewegen, P.P., van den Berg, A.E., Maas, J., Verheij, R.A., de Vries, S., 2012. Is a green https://doi.org/10.1016/j.ecolecon.2020.106917. residential environment better for health? If so, why? Ann. Assoc. Am. Geogr. 102, Mitchell, R., Popham, F., 2007. Greenspace, urbanity and health: relationships in England. 996–1003. https://doi.org/10.1080/00045608.2012.674899. J. Epidemiol. Community Health 61, 681–683. https://doi.org/10.1136/jech.2006. Gronlund, C.J., 2014. Racial and socioeconomic disparities in heat-related health effects 053553. and their mechanisms: a review. Curr. Epidemiol. Reports 1, 165–173. https://doi. Mitchell, R., Popham, F., 2008. Effect of exposure to natural environment on health in- org/10.1007/s40471-014-0014-4. equalities: an observational population study. Lancet 372, 1655–1660. https://doi. Guarnieri, M., Balmes, J.R., 2014. Outdoor air pollution and asthma. Lancet 383, org/10.1016/S0140-6736(08)61689-X. 1581–1592. https://doi.org/10.1016/S0140-6736(14)60617-6. Mobley, L.R., Root, E.D., Finkelstein, E.A., Khavjou, O., Farris, R.P., Will, J.C., 2006. Environ- Han, K.T., 2007. Responses to six major terrestrial biomes in terms of scenic beauty, pref- ment, obesity, and cardiovascular disease risk in low-income women. Am. J. Prev. erence, and restorativeness. Environ. Behav. 39, 529–556. https://doi.org/10.1177/ Med. 30. https://doi.org/10.1016/j.amepre.2005.12.001 327-332.e1. 0013916506292016. Moral Gil, L., Roig, M., Garde, J., Alós, A., Toral, T., Fuentes, M.J., 2008. Allergen sensitization Hanski, I., von Hertzen, L., Fyhrquist, N., Koskinen, K., Torppa, K., Laatikainen, T., Karisola, in children with asthma and rhinitis: marked variations related to age and P., Auvinen, P., Paulin, L., Mäkelä, M.J., Vartiainen, E., Kosunen, T.U., Alenius, H., microgeographical factors. Allergol Immunopathol (Madr) 36, 128–133. https://doi. Haahtela, T., Haahtelab, T., 2012. Environmental biodiversity, human microbiota, org/10.1157/13124718. and allergy are interrelated. Proc. Natl. Acad. Sci. 109, 8334–8339. https://doi.org/ Morita, E., Fukuda, S., Nagano, J., Hamajima, N., Yamamoto, H., Iwai, Y., Nakashima, T., 10.1073/pnas.1205624109. Ohira, H., Shirakawa, T., 2007. Psychological effects of forest environments on healthy Hartig, T., Mitchell, R., De Vries, S., Frumkin, H., 2014. Nature and health. Annu. Rev. Public adults: Shinrin-yoku (forest-air bathing, walking) as a possible method of stress re- Health 35, 207–228. https://doi.org/10.1146/annurev-publhealth-032013-182443. duction. Public Health 121, 54–63. https://doi.org/10.1016/j.puhe.2006.05.024. Hron, K., Filzmoser, P., Thompson, K., 2012. Linear regression with compositional explan- Mouratidis, K., 2020. Neighborhood characteristics, neighborhood satisfaction, and well- atory variables. J. Appl. Stat. 39, 1115–1128. https://doi.org/10.1080/02664763.2011. being: the links with neighborhood deprivation. Land Use Policy 99, 104886. 644268. https://doi.org/10.1016/j.landusepol.2020.104886. Jacobs, K.W., Suess, J.F., 1975. Effects of four psychological primary colors on anxiety state. Müller, I., Hron, K., Fišerová, E., Šmahaj, J., Cakirpaloglu, P., Vančáková, J., 2018. Interpreta- Percept. Mot. Skills 41, 207–210. https://doi.org/10.2466/pms.1975.41.1.207. tion of compositional regression with application to time budget analysis. Austrian Jatzlauk, G., Bartel, S., Heine, H., Schloter, M., Krauss-Etschmann, S., 2017. Influences of en- J. Stat. 47, 3–19. https://doi.org/10.17713/ajs.v47i2.652. vironmental bacteria and their metabolites on allergies, asthma, and host microbiota. Mytton, O.T., Townsend, N., Rutter, H., Foster, C., 2012. Green space and physical activity: Allergy 72, 1859–1867. https://doi.org/10.1111/all.13220. an observational study using health survey for England data. Health Place 18, Jiang, S.Y., Ma, A., Ramachandran, S., 2018. Negative air ions and their effects on human 1034–1041. https://doi.org/10.1016/j.healthplace.2012.06.003. health and air quality improvement. Int. J. Mol. Sci. 19. https://doi.org/10.3390/ Nutsford, D., Pearson, A.L., Kingham, S., Reitsma, F., 2016. Health & Place Residential expo- ijms19102966. sure to visible blue space ( but not green space ) associated with lower psychological Joshi, S.K., Shrestha, S., 2010. Diabetes mellitus: a review of its associations with different distress in a capital city. Health Place 39, 70–78. https://doi.org/10.1016/j.healthplace. environmental factors. Kathmandu Univ. Med. J. 8, 109–115. https://doi.org/10.3126/ 2016.03.002. kumj.v8i1.3233. Orians, G.H., Heerwagen, J.H., 1992. Evolved responses to landscapes. The Adapted Mind: Kiley, H.M., Ainsworth, G.B., van Dongen, W.F.D., Weston, M.A., 2017. Variation in public Evolutionary Psychology and the Generation of Culture. Oxford University Press, New perceptions and attitudes towards terrestrial ecosystems. Sci. Total Environ. York, NY, US, pp. 555–579. 590–591, 440–451. https://doi.org/10.1016/j.scitotenv.2016.12.179. Ortega-garcía, J.A., Tellerías, L., Ferrís-tortajada, J., Boldo, E., Campillo-lópez, F., Van Den Knobel, P., Kondo, M., Maneja, R., Zhao, Y., Dadvand, P., Schinasi, L.H., 2021a. Associations Hazel, P., Cortes-arancibia, S., Ramis, R., Gaioli, M., Monroy-torres, R., Farias-guardia, of objective and perceived greenness measures with cardiovascular risk factors in C., Borras, M., Cárceles-álvarez, A., Jaimes-vega, D.C., Cordero-rizo, M., López- Philadelphia, PA: a spatial analysis. Environ. Res. 197, 110990. https://doi.org/10. hernández, F., Claudio, L., 2019. Amenazas , desafíos y oportunidades para la salud 1016/j.envres.2021.110990. medioambiental pediátrica en Europa. América Latina y el Caribe. An. Pediatría 90. Knobel, P., Maneja, R., Bartoll, X., Alonso, L., Bauwelinck, M., Valentin, A., Zijlema, W., https://doi.org/10.1016/j.anpedi.2018.11.015. Borrell, C., Nieuwenhuijsen, M., Dadvand, P., 2021b. Quality of urban green spaces in- Pearson, K., 1896. Mathematical contributions to the theory of evolution. Proeedings R. fluences residents' use of these spaces, physical activity, and overweight/obesity. En- Soc. 60, 489–498. viron. Pollut. 271, 116393. https://doi.org/10.1016/j.envpol.2020.116393. Perez, V., Alexander, D.D., Bailey, W.H., 2013. Air ions and mood outcomes: a review and Koohsari, M.J., 2011. Access to public open space: is distribution equitable across different meta-analysis. BMC Psychiatry 13. https://doi.org/10.1186/1471-244X-13-29. socio-economic areas. J. Urban Environ. Eng. 5, 67–72. https://doi.org/10.4090/juee. Richardson, E.A., Mitchell, R., 2010. Gender differences in relationships between urban 2011.v5n2.067072. green space and health in the United Kingdom. Soc. Sci. Med. 71, 568–575. https:// Kroeger, T., Escobedo, F.J., Hernandez, J.L., Varela, S., Delphin, S., Fisher, J.R.B., Waldron, J., doi.org/10.1016/j.socscimed.2010.04.015. 2014. Reforestation as a novel abatement and compliance measure for ground-level Richardson, E.A., Pearce, J., Mitchell, R., Kingham, S., 2013. Role of physical activity in the ozone. Proc. Natl. Acad. Sci. U. S. A. 111, E4204–E4213. https://doi.org/10.1073/ relationship between urban green space and health. Public Health 127, 318–324. pnas.1409785111. https://doi.org/10.1016/j.puhe.2013.01.004. 11 Q. Zaldo-Aubanell, I. Serra, A. Bach et al. Science of the Total Environment 806 (2022) 150308 Sandifer, P.A., Sutton-Grier, A.E., Ward, B.P., 2015. Exploring connections among nature, van den Berg, A.E., Maas, J., Verheij, R.A., Groenewegen, P.P., 2010. Green space as a buffer biodiversity, ecosystem services, and human health and well-being: opportunities between stressful life events and health. Soc. Sci. Med. 70, 1203–1210. https://doi. to enhance health and biodiversity conservation. Ecosyst. Serv. 12, 1–15. https:// org/10.1016/j.socscimed.2010.01.002. doi.org/10.1016/J.ECOSER.2014.12.007. Van den Berg, M., Van Poppel, M., Van Kamp, I., Andrusaityte, S., Balseviciene, B., Cirach, Sarkar, C., Gallacher, J., Webster, C., 2013. Urban built environment configuration and psy- M., Danileviciute, A., Ellis, N., Hurst, G., Masterson, D., Smith, G., Triguero-Mas, M., chological distress in older men: results from the Caerphilly study. BMC Public Health Uzdanaviciute, I., de Wit, P., Van Mechelen, W., Gidlow, C., Grazuleviciene, R., 13, 695. https://doi.org/10.1186/1471-2458-13-695. Nieuwenhuijsen, M.J., Kruize, H., Maas, J., 2016. Visiting green space is associated Schwanen, T., Dijst, M., Dieleman, F.M., 2002. A microlevel analysis of residential context with mental health and vitality: a cross-sectional study in four european cities. Health and travel time. Environ. Plan. A 34, 1487–1507. https://doi.org/10.1068/a34159. Place 38, 8–15. https://doi.org/10.1016/j.healthplace.2016.01.003. Shankardass, K., McConnell, R.S., Milam, J., Berhane, K., Tatalovich, Z., Wilson, J.P., Jerrett, Wheeler, B.W., White, M., Stahl-Timmins, W., Depledge, M.H., 2012. Does living by the M., 2007. The association between contextual socioeconomic factors and prevalent coast improve health and wellbeing. Health Place 18, 1198–1201. https://doi.org/ asthma in a cohort of Southern California school children. Soc. Sci. Med. 65, 10.1016/j.healthplace.2012.06.015. 1792–1806. https://doi.org/10.1016/j.socscimed.2007.05.048. Wheeler, B.W., Lovell, R., Higgins, S.L., White, M.P., Alcock, I., Osborne, N.J., Husk, K., Sabel, Shin, S., Bai, L., Oiamo, T.H., Burnett, R.T., Weichenthal, S., Jerrett, M., Kwong, J.C., Goldberg, C.E., Husk, K., Depledge, M.H., 2015. Beyond greenspace: an ecological study of pop- M.S., Copes, R., Kopp, A., Chen, H., 2020. Association between road traffic noise and ulation general health and indicators of natural environment type and quality. Int. incidence of diabetes mellitus and hypertension in Toronto, Canada: a population- J. Health Geogr., 14–17 https://doi.org/10.1186/s12942-015-0009-5. based cohort study. J. Am. Heart Assoc. 9. https://doi.org/10.1161/JAHA.119.013021. White, M., Smith, A., Humphryes, K., Pahl, S., Snelling, D., Depledge, M., 2010. Blue space: Soltész, G., 2003. Diabetes in the young: a paediatric and epidemiological perspective. the importance of water for preference, affect, and restorativeness ratings of natural Diabetologia 46, 447–454. https://doi.org/10.1007/s00125-003-1101-0. and built scenes. J. Environ. Psychol. 30, 482–493. https://doi.org/10.1016/j.jenvp. Soneja, S., Jiang, C., Fisher, J., Upperman, C.R., Mitchell, C., Sapkota, A., 2016. Exposure to 2010.04.004. extreme heat and precipitation events associated with increased risk of hospitaliza- White, M., Pahl, S., Ashbullby, K., Herbert, S., Depledge, M.H., 2013. Feelings of restoration tion for asthma in Maryland, U.S.A. Environ. Heal. A Glob. Access Sci. Source 15, from recent nature visits. J. Environ. Psychol. 35, 40–51. https://doi.org/10.1016/j. 1–7. https://doi.org/10.1186/s12940-016-0142-z. jenvp.2013.04.002. Starling, A.P., Umbach, D.M., Kamel, F., Long, S., Sandler, D.P., Hoppin, J.A., 2014. Pesticide Williams, E.D., Magliano, D.J., Zimmet, P.Z., Kavanagh, A.M., Stevenson, C.E., Oldenburg, use and incident diabetes among wives of farmers in the Agricultural Health Study. B.F., Shaw, J.E., 2012. Area-level socioeconomic status and incidence of abnormal glu- Occup. Environ. Med. 71, 629–635. https://doi.org/10.1136/oemed-2013-101659. cose metabolism: the Australian Diabetes, Obesityand Lifestyle (AusDiab) study. Dia- Stier-Jarmer, M., Throner, V., Kirschneck, M., Immich, G., Frisch, D., Schuh, A., 2021. The betes Care 35, 1455–1461. https://doi.org/10.2337/dc11-1410. psychological and physical effects of forests on human health: a systematic review Wu, Y.T., Prina, A.M., Jones, A., Matthews, F.E., Brayne, C., 2015. Older people, the natural of systematic reviews and meta-analyses. Int. J. Environ. Res. Public Health 18, environment and common mental disorders: cross-sectional results from the Cogni- 1–39. https://doi.org/10.3390/ijerph18041770. tive Function and Ageing Study. BMJ Open 5, e007936. https://doi.org/10.1136/ Su, J.G., Jerrett, M., de Nazelle, A., Wolch, J., 2011. Does exposure to air pollution in urban bmjopen-2015-007936. parks have socioeconomic, racial or ethnic gradients? Environ. Res. 111, 319–328. Wu, X., Nethery, R.C., Sabath, M.B., Braun, D., Dominici, F., 2020. Air pollution and COVID- https://doi.org/10.1016/j.envres.2011.01.002. 19 mortality in the United States: strengths and limitations of an ecological regres- Taylor, L., Hochuli, D.F., 2015. Creating better cities: how biodiversity and ecosystem func- sion analysis. Sci. Adv. 6, 1–7. https://doi.org/10.1126/SCIADV.ABD4049. tioning enhance urban residents' wellbeing. Urban Ecosyst. 18, 747–762. https://doi. Zaldo-Aubanell, Q., Campillo i López, F., Bach, A., Serra, I., Olivet-Vila, J., Saez, M., Pino, D., org/10.1007/s11252-014-0427-3. Maneja, R., 2021a. Community risk factors in the COVID-19 incidence and mortality Toure, N.O., Gueye, N.R.D., Mbow-Diokhane, A., Jenkins, G.S., Li, M., Drame, M.S., Coker, in Catalonia (Spain). A population-based study. Int. J. Environ. Res. Public Health 18, K.A.R., Thiam, K., 2019. Observed and modeled seasonal air quality and respiratory 3768. https://doi.org/10.3390/ijerph18073768. health in Senegal during 2015 and 2016. GeoHealth 3, 423–442. https://doi.org/10. Zaldo-Aubanell, Q., Serra, I., Sardanyés, J., Alsedà, L., Maneja, R., 2021b. Reviewing the re- 1029/2019GH000214. liability of land use and land cover data in studies relating human health to the envi- Turyk, M., Anderson, H., Knobeloch, L., Imm, P., Persky, V., 2009. Organochlorine exposure ronment. Environ. Res. 194. https://doi.org/10.1016/j.envres.2020.110578. and incidence of diabetes in a cohort of great lakes sport fish consumers. Environ. Zandieh, R., Flacke, J., Martinez, J., Jones, P., Van Maarseveen, M., 2017. Do inequalities in Health Perspect. 117, 1076–1082. https://doi.org/10.1289/ehp.0800281. neighborhood walkability drive disparities in older adults' outdoor walking? Int. Ulrich, R.S., 2014. Aesthetic and Affective Response to Natural Environment. https://doi. J. Environ. Res. Public Health 14. https://doi.org/10.3390/ijerph14070740. org/10.1007/978-1-4613-3539-9. Ulrich, R.S., 2016. Biophilia, biophobia, and natural landscapes. 12