Impact of climate events, pollution, and green spaces on mental health: an umbrella review of meta-analyses

Climate change may affect mental health. We conducted an umbrella review of meta-analyses examining the association between mental health and climate events related to climate change, pollution and green spaces. We searched major bibliographic databases and included meta-analyses with at least five primary studies. Results were summarized narratively. We included 24 meta-analyses on mental health and climate events (n = 13), pollution (n = 11), and green spaces (n = 2) (two meta-analyses provided data on two categories). The quality was suboptimal. According to AMSTAR-2, the overall confidence in the results was high for none of the studies, for three it was moderate, and for the other studies the confidence was low to critically low. The meta-analyses on climate events suggested an increased prevalence of symptoms of post-traumatic stress, depression, and anxiety associated with the exposure to various types of climate events, although the effect sizes differed considerably across study and not all were significant. The meta-analyses on pollution suggested that there may be a small but significant association between PM2.5, PM10, NO2, SO2, CO and mental health, especially depression and suicide, as well as autism spectrum disorders after exposure during pregnancy, but the resulting effect sizes varied considerably. Serious methodological flaws make it difficult to draw credible conclusions. We found reasonable evidence for an association between climate events and mental health and some evidence for an association between pollution and mental disorders. More high-quality research is needed to verify these associations.


Introduction
It is unequivocal that increasing human activities have led to warming the atmosphere, ocean, and land. Widespread and rapid changes in the atmosphere, ocean, cryosphere, and biosphere have occurred (Legg, 2021). The scale of the recent changes across the climate system is unprecedented over many centuries to many thousands of years (Legg, 2021). These changes are suggested to have serious consequences for human health, including renal function loss, dermatological malignancies, tropical infections, pregnancy complications, allergies, and cardiovascular and pulmonary morbidity and mortality (Atwoli et al., 2021;Haines & Ebi, 2019;Rocque et al., 2021).
Over the last few years it has been suggested that mental health may also be affected by these rapidly occurring climate changes. A recent seminal review describes how mental health may be affected by climate change through four different pathways (Clayton, 2021). First, discrete events, such as natural disasters can have a direct impact on mental health. There is suggestive evidence that floods, heatwaves, tornados and hurricanes, wildfires, and earthquakes may be associated with increased rates of post-traumatic stress and depression, substance use disorders, suicidal thoughts, and important risk factors, such as domestic abuse (Cénat, McIntee, & Blais-Rochette, 2020;Chan & Rhodes, 2014;Chen & Liu, 2015;Dai et al., 2016). Second, mental health can be affected by gradual changes, such as rising sea levels and higher temperatures. Although the causal mechanisms are not clear, higher temperatures, for example, have been associated with more aggression and higher suicide rates (Clayton, 2021). Pollution and the 'greenhouse' effect associated with the burning of fossil fuels may also have consequences for mental health. Third, climate change may affect existing physical and social systems, and these changes may have an indirect effect on mental health. For example, the occupational structure and agricultural conditions may change in communities, resulting in economic uncertainties for some groups. Migration may also be the result of areas becoming less inhabitable or disappearing altogether (Clayton, 2021). The fourth pathway refers to the perception of climate change, with anecdotal reports of 'climate anxiety', for example in parents who are worried about their children's future and young adults who are reluctant to procreate because of fear about the future (Clayton, 2021). Solastalgia, the distress caused by the transformation and degradation of one's home environment, is a comparable phenomenon that is somewhat better examined, although it is still mostly unclear what it exactly is and how it affects mental health (Galway, Beery, Jones-Casey, & Tasala, 2019). If climate change does indeed have impact on mental health, it is important to include this in the future projections of global mental health and, if possible, measures should be taken to prevent or treat these increasing numbers of patients with mental disorders.
Over the past decade, the number of meta-analyses of studies examining the impact of climate change on mental health has increased exponentially. That is a positive development, as replication increases knowledge about the impact of climate change. However, the published meta-analyses all have a specific focus, for example on one mental health problem, such as posttraumatic stress disorder (PTSD) (e.g. Chan & Rhodes, 2014;Dai et al., 2016;Rezayat et al., 2020), suicide (e.g. Jahangiri, Yousefi, Mozafari, & Sahebi, 2020), or autism spectrum disorders (ASDs) in children (Chun, Leung, Wen, McDonald, & Shin, 2020;Dutheil et al., 2021). Other meta-analyses focus on specific events (e.g. Chen & Liu, 2015;Dai et al., 2016), on one specific country (e.g. Hosseinnejad et al., 2022;Sepahvand, Hashtjini, Salesi, Sahraei, & Jahromi, 2019), or one polluting substance (Forns et al., 2020). Because of the focus on specific subjects, an overview of the whole field is lacking. Furthermore, an increasing number of meta-analyses is not always positive, because it is not uncommon for meta-analyses on the same research question to reach different conclusions, even when published within the same year (Solmi, Correll, Carvalho, & Ioannidis, 2018). Such discrepancies unavoidably lead to confusion and debate among policy makers and researchers (Solmi et al., 2018). Umbrella reviewsa systematic review of all systematic reviews and/or meta-analyses on a given topicallow a higher-level synthesis of the evidence and a better recognition of the uncertainties, biases, and knowledge gaps, compared to single meta-analyses (Ioannidis, 2009(Ioannidis, , 2017). An umbrella review on environmental and climate-related determinants of mental health is important, because when it is found that mental health is indeed affected, this would be an additional reason to stop the warming of the atmosphere, ocean, and land as soon as possible.
We conducted an umbrella review of meta-analyses examining environmental and climate-related determinants of mental health. In this review, we focused on three major environmental and climate-related determinants that (1) potentially have a major impact on mental health; (2) represent different, non-overlapping domains, and (3) cover as much as possible of the possible environmental and climate-related determinants.
We defined climate events as 'discrete episodes of extreme weather or unusual climate conditions, often associated with deleterious impacts on society or natural systems, defined using some metric to characterize either the meteorological characteristics of the event or the consequent impacts' (Stott et al., 2016). We also used the taxonomy of climate events developed by Stephenson, Diaz, and Murnane (2008), including tropical cyclones, hurricanes, extratropical cyclones, convective phenomena (including tornadoes and severe thunderstorms), mesoscale phenomena (such as polar lows, resulting in e.g. extreme wind speeds and precipitation), floods, drought, heat waves, cold waves, and fog.
In this study we also included natural events that may not be directly influenced by the climate such as landslides. Pollution was defined as the addition of any substance (solid, liquid, or gas) or any form of energy (such as heat, sound, or radioactivity) to the environment at a rate faster than it can be dispersed, diluted, decomposed, recycled, or stored in some harmless form (Nathanson, 2022). The major kinds of pollution, usually classified by environment, are air pollution, water pollution, and land pollution. The impact of green spaces was operationalized as the impact of exposure to the natural environment on mental health problems (Roberts, van Lissa, Hagedoorn, Kellar, & Helbich, 2019).
The domains we choose were also relevant for a separate report by the United Nations Environment Programme (UNEP) (in preparation), that included a modeling study examining the impact of these domains on the costs of global mental health. The three specific subjects that are examined in this umbrella review are (1) the impact of climate events on mental health; (2) the impact of pollution on mental health; and (3) the association between green spaces and mental health, as an indicator of the impact of urbanization on mental health.

Search strategy and selection criteria
We used an umbrella review methodology to systematically collect and review all available meta-analyses examining the potential association between climate change, pollution, green spaces, and mental health. We followed general guidelines for conducting and reporting umbrella reviews (Ioannidis, 2009(Ioannidis, , 2017Papatheodorou, 2019;Solmi et al., 2018). The protocol for this meta-analysis was registered at the Open Science Framework (Cuijpers, 2021).
We conducted systematic searches on 17 June 2021 in three bibliographic databases: PubMed, PsycINFO, and Embase. We first developed a general search string for climate change, pollution, and green spaces, in which we combined text and key words for these subjects with text and key words for mental health and mental disorders. We limited the results to systematic reviews and meta-analyses. Because the impact of climate events (such as tornados, landslides, heatwaves, etc.) may not be captured by key words related to climate change, we conducted separate searches for these events. In these searches we combined text and key words for climate events with text and key words for mental health and mental disorders and again limited the results to meta-analyses and systematic reviews. The full search strings are available in online Supplement S1. All records were read by two independent researchers and we retrieved the full-text of all studies that were selected for retrieval by one or both researchers.
In this umbrella review we included (a) meta-analyses that reported (b) the association between climate events, green spaces, or pollution and mental health or mental disorders, and (c) in which at least one analysis included more than five comparisons (in order to have a reasonable impression of the association). We included any kind of climate event and pollution as defined in the Introduction, including specific substances, and pollution of air, soil, and water. Any mental health problem or mental disorder was included. Because of limited resources, we excluded studies on the association between climate change and intelligence, as well as on dementia and cognitive decline. Only studies in English were included. No other exclusion criteria were applied.
All full text papers were read by two independent researchers and the decision to include or exclude was based on consensus.
Disagreements were solved through discussion. Because no disagreements remained after the discussions, it was not needed to consult a third, senior author.

Quality assessment
The quality of the included meta-analyses was assessed with the AMSTAR 2, a critical appraisal tool for systematic reviews (Shea et al., 2017). AMSTAR-2 critically assesses 16 core characteristics of systematic reviews (online Supplement S2). All assessments of these criteria were conducted by two independent researchers and disagreements were solved by discussion or, when needed, discussed with a third reviewer.

Data extraction
We extracted the following data from the included meta-analyses: the examined climate-related factor (climate event, pollution, specific substance, green spaces, etc.), the design of the included studies in the meta-analysis, the number of included studies, the aggregated number of participants in the primary studies (when reported), the population, and the instrument used to measure the quality of primary studies. We also extracted the mental health outcome, a summary of the pooled outcomes, the significance of the outcomes, the level of heterogeneity [I 2 and its 95% confidence interval (CI)], and (when reported) the outcomes of the analyses examining publication bias. When I 2 or its 95% CI were not reported, we calculated them with the value of χ 2 and degrees of freedom (if available), using the Heterogi module in STATA SE (version 16.1 for Mac). The general characteristics of the meta-analyses were extracted by one reviewer. The outcomes were extracted by one reviewer whose extraction was validated by a second reviewer, who independently extracted 25% of the data. An agreement index of 96.3% between the two reviewers was reached.

Integration of findings
We offer a narrative overview of the identified associations between indicators for climate change, pollution, and green spaces and indicators for mental health and mental disorders.

Selection and inclusion of studies
We examined the abstracts and titles of 575 records (519 after removal of duplicates). The full-texts of 221 studies were retrieved and assessed for eligibility, from which 197 were excluded. A total of 24 meta-analyses met the inclusion criteria, two for green spaces and natural environments, 11 for pollution, and 13 for climate events (with two meta-analyses included in multiple categories). The PRISMA flowchart is presented in Fig. 1.

Characteristics of included meta-analyses
The main characteristics of the 24 included meta-analyses are summarized in Table 1. All identified reviews were relatively recent, published until 2021 and with searches covering from databases' inception to 2021. The total number of primary studies included in the meta-analyses ranged from 5 to 64. The total sample sizes varied among meta-analyses, with the largest analysis including 758 997 participants. The most frequently used design in the included primary studies was cross-sectional, but other common designs included case-control studies, cohorts, casecrossover designs, and time-series analyses. Two of the included studies were not systematic reviews, but performed meta-analyses on outcome data from large European (Forns et al., 2020) and Dutch cohorts (Generaal et al., 2019). More details of the included meta-analyses are reported in online Supplement S3.

Quality of included meta-analyses
The AMSTAR-2 ratings for each of the 24 included meta-analyses are presented in Table 2, and the aggregated ratings across all reviews are reported in Fig. 2. The specification of PICO was judged as adequate for all studies. The majority of the meta-analyses did not register a protocol (n = 16; 67%), did not justify the selection of study designs (n = 17; 71%), did not investigate sources of funding (n = 18; 75%), and did not provide a list of excluded studies with reasons (n = 22; 92%). Comprehensiveness of the literature search was rated as 'partial yes' in the majority of reviews (n = 15; 63%), while only two reviews obtained a complete positive rating. Two reviewers independently selected studies and extracted data in 15 (63%) and 13 (54%) meta-analyses, respectively. Most of the reviews described in great (n = 14; 58%) or sufficient (n = 10; 42%) detail the included studies, and investigated sources for heterogeneity (n = 22; 92%) and publication bias (n = 17; 71%). Above half of them utilized appropriate statistical methods for analyses (n = 15; 63%), particularly those reporting on pollution. Risk of bias was adequately assessed in 19 meta-analyses (79%), but only 7 (29%) of these assessed all relevant criteria. The impact of risk of bias on the effect estimates (or inclusion of only studies at low risk) was examined in 10 (42%) meta-analyses, and 15 (62%) accounted for the risk of bias when interpreting or discussing the results. The vast majority of meta-analysts declared their conflicts of interest (n = 21; 88%).
The overall confidence in the results was high for none of the studies (zero or one non-critical weakness), for three the confidence was moderate (more than one weakness, but no critical flaws), and for the other studies the confidence was low to critically low. The three studies with moderate confidence were on climate events  and on pollution (Hegewald et al., 2020;Lam et al., 2016).

Climate events and mental health
The outcomes reported in the 13 included meta-analyses examining the association between climate events and mental health are summarized in Table 3. We will present the results according to the methodology used, with meta-analyses of pre-post designs first, followed by meta-analyses of estimates of point prevalences of mental disorders (separately for cut-off scores on self-report measures and diagnostic interviews), then meta-analyses of studies with time-series and case-cross-over designs, and finally meta-analyses of correlations between the level of exposure to disasters and mental health outcomes.
One meta-analysis examined the impact of various natural disasters (earthquakes, floods, hurricanes, tsunamis, wildfires, etc.) on psychological distress and mental disorders, and included studies that compared exposed to non-exposed people, as well as studies with measurements before and after the event (Beaglehole et al., 2018). The studies with both pre-test and post-test assessments are especially informative about the actual number of cases triggered by the events. Six included studies that used a pre-post design found a non-significant standardized mean difference (SMD) of 0.32 (95% CI −0.06 to 0.70) on psychological distress. Only two included studies with a pre-post design were available for measuring the impact on mental disorders, which were too few for providing reliable estimates of the impact.
Two meta-analyses reported the prevalence of PTSD according to a diagnostic interview. One found a prevalence of diagnosed PTSD after floods of 16% (95% CI 0.11-0.21) (Chen & Liu, 2015), the other reported a prevalence of PTSD in children and adolescents after earthquakes and floods of 20.8% between 0 and 6 months after the event, 14.6% at 6-12 months, and 16.3% at 12-18 months (Rezayat et al., 2020).
Two meta-analyses of mostly time-series designs and casecrossover studies examined the association between heat exposure and mental health problems. One found a significant risk ratio (RR) of 1.09 (95% CI 1.06-1.13) for the association between daily suicide rates and an increase in temperature of 7.1°C (Heo, Lee, & Bell, 2021). The other  found an overall RR of 1.02 (95% CI 1.01-1.03) for mental health-related mortality for each degree Celsius increase in temperature, and 1.01 (95% CI 1.007-1.015) for mental health-related morbidity. It should be noted that this meta-analysis scored relatively high on AMSTAR-2 (moderate confidence).

Pollution and mental health
The outcomes reported in the 11 included meta-analyses examining the association between pollution and mental health are summarized in Table 4. We will present the results according to the different polluting substances, starting with the most examined substance.
Particulate matter (PM) is a complex mixture that is usually classified into PM 2.5 (<2.5 μm in diameter) and PM 10 (<10 μm) (Chun et al., 2020). One analysis of eight Dutch cohorts found  (Heo et al., 2021) was included in both pollution and climate events, and another study (Generaal et al., 2019) was included in both pollution and green spaces.
Several meta-analyses found significant associations between pollutants and ASD when the mothers were exposed during pregnancy: PM 2.5 (ORs ranging from 1.06 to 2.32), PM 10 (OR = 1.07), ozone (O 3 ; OR = 1.03 and RR = 1.05), NO 2 (OR = 1.02), and solvents (OR = 1.03). One study found a lower risk for ASD and pesticides (OR = 0.83). Other meta-analyses found a significant association between PM 10 and suicide (RR = 1.02) and between NO 2 and suicide (RR = 1.03).
No other significant associations between pollution and mental disorders were found in the included meta-analyses.

Green spaces and mental health
The results of the two meta-analyses examining the association between green spaces and mental health are reported in Table 4. One pooled analysis of eight Dutch cohort studies found a significant negative association between 30% increase of green space in the neighborhood and depression (OR = 0.94), and a significant association between grade of urbanization (mean number of addresses per km 2 within a radius of 1 km) and depression (OR = 1.05) (Generaal et al., 2019). The other meta-analysis included a wide range of designs (randomized, nonrandomized, parallel groups, etc.), and found a significant effect size (SMD = 0.30) for reduction of depressive mood after shortterm exposure to natural environments (Roberts et al., 2019).

Discussion
We conducted an umbrella review of meta-analyses investigating the association between climate events, pollution, and green spaces on the one hand, and any mental health symptoms or disorders on the other. We included 24 meta-analyses, all relatively recent, which analyzed mainly cross-sectional and case-control studies. The 13 meta-analyses examining the impact of climate events on mental health suggested that there may be an association between such events and mental health, although the exact contribution of these events to the prevalence cannot be established based on these studies. Only two meta-analyses used a diagnostic interview to establish the presence of mental disorders, seven pooled different measures for mental health problems incorrectly, and only six primary studies used a pre-post design, which can be used as a good indication for the impact of climate events on mental health problems. Although it seems obvious that traumatic events like climate events result in increased levels of mental health problems, these methodological problems make it impossible to assess the exact size of the impact.
The 10 meta-analyses focusing on the impact of pollution on mental health, found that some substances are associated with mental health problems, especially depression and ASD in children of mothers exposed to substances. However, all associations were very small and in many studies no corrections were applied for other characteristics of the populations, making these associations highly uncertain. PM 2.5 was examined in most studies, but other substances that were examined included PM 10 , NO 2 , O 3 , SO 2 , CO, solvents, styrene exposure, and noise. All associations were small and because unmeasured confounders may further reduce the strength and level of significance of the association, these findings should be considered with caution. We only found two meta-analyses examining green spaces and natural environments (Generaal et al., 2019;Roberts et al., 2019). Both meta-analyses found a small but significant association between green spaces and reduction of mental health symptoms, but the observational nature of many of the primary studies as well as the potential for risk of bias limit confidence in these findings.
Research on the impact of climate events has demonstrated considerable methodological limitations. It is difficult to indicate whether a climate event is caused by climate change or that it is a natural fluctuation. It is also difficult to examine the impact of a climate event on mental health, because this requires at least a measurement before the event and one after, and then the causality is still unclear because other changes may have taken place at the same time that have (partly) caused the changes in mental health. Many studies simply examine correlations between for example pollution and a mental health outcome. Although some studies do adjust for confounders, the number of confounders is usually limited and there is no way to exclude the possibility that an association is in fact caused by a third, unknown variable, especially when the effect size of the identified association is small. Most studies also measured mental health with selfreport measures, while such scales cannot reliably indicate the presence of a mental disorder. Studies based on mental disorders assessed with the gold standard, diagnostic interviews, are hardly done, also because the costs and logistic challenges of such studies are considerable. Based on our umbrella review it is not possible to indicate how much support is available for the four different pathways of how climate can have an impact on mental health.
The results of this study need to be counterbalanced weighing new insights it offers and limitations that remain challenging for generalizability of the findings. The strengths include a rigorous approach to umbrella reviews and the inclusion of a considerable number of studies. However, this study also has some important limitations that have to be taken into consideration. One important limitation is that we only included meta-analyses with at least five studies. That threshold is arbitrary because the strength of evidence does not only depend on the number of studies, but also on the size of the study, the design, and the quality. However, we had to limit the scope of this umbrella review to make the rapid scoping of the evidence manageable and we believe that a threshold of five studies does give good indications for the state of research in a particular area. A second important limitation is that the quality of the included meta-analyses was not     optimal. The overall confidence in the results was high for none of the studies, and moderate for only three of the 24 included meta-analyses. The confidence in the other meta-analyses was low to critically low. This means that the results of all meta-analyses should be considered with caution because of limited quality. Another limitation is that the impact of climate events, pollution, and green spaces change over time, and while most included meta-analyses were conducted over the past few years, it cannot be examined whether the impact of these changes has increased over time. This umbrella review shows that very little good evidence is available on the association between mental health on the one hand and climate events, pollution, and green spaces on the other hand. Because of the devastating impact climate events can have on human lives, there should be no doubt that they result in increased mental health problems. However, the current evidence gives no clear indication of how large that increase is. More research with better designs and assessments are clearly needed to get a better overview of the implication of these developments for mental health. This information is needed to develop strategies to reduce the impact of such events on mental health.
In conclusion, in this umbrella review we found some evidence for an association between climate events and mental health, especially post-traumatic stress, but also depression and anxiety. We also found some evidence for an association between pollution and aspects of mental health, especially depression, suicide and autism spectrum disorders. Finally, we found some indications for an association between green spaces and depression. However, more high-quality research is needed to verify all these associations. Given how rapidly the natural world is changing around us and the policy impetus to reflect and revisit human footprint on natural environment, improving design, measurement, and carrying out evidence informed environmental mental health studies would be important. Financial support. This umbrella review was financially supported by the United Nations Environment Programme, Washington, DC.
Conflict of interest. None of the authors report any conflicts of interest.