Multi-country projections of temperature-related suicide mortality – Nature Mental Health

Climate change is a critical global public health issue, with impacts extending beyond environmental changes, affecting human health, behavior and societal systems in complex ways^1,2,3,4. Notably, rising global temperatures and increasing extreme weather events have been linked to various adverse health outcomes, including mental health disorders^3,4,5,6,7. Epidemiological studies consistently highlight the negative effects of climate change on mental health, showing associations with post-traumatic stress disorder^8,9, anxiety^8,9,10, depression¹⁰, sleep disorders¹¹ and suicide^3,4,5,6.

Among the mental health-related challenges, suicide remains one of the leading causes of death globally, claiming an estimated 720,000 lives annually¹². Climate change has emerged as an important contributor to the existing suicide burden¹³. High temperatures can exacerbate poor mental health outcomes through insomnia, increased stress levels and psychological instability, all of which contribute to suicide risk^{3,4,5,14,15,16}. Numerous studies have identified a positive association between ambient temperature and suicide, suggesting that higher temperatures increase suicide risk^17,18,19,20. The largest-scale multi-city multi-country study demonstrated that higher ambient temperatures generally increased suicide risks with regional variations in the degree and shape of the association²⁰. Nonlinear inverse J-shaped relationships have been reported in East Asian countries such as Japan, the Republic of Korea and Taiwan, where suicide risk tends to decrease at extremely high temperatures after reaching a peak. By contrast, nearly linear patterns appeared in Western countries such as Canada, Spain, Switzerland, the UK and the USA, where suicide risk continued to increase with higher temperatures. Meanwhile, in the Philippines, Brazil, Vietnam and South Africa, the nonlinear curves were unclear or showed large uncertainty²⁰.

Projections of climate-sensitive health responses are essential for anticipating future public health challenges owing to climate change and informing policy. Several studies have predicted the impacts of climate change on human health including morbidity and mortality across diverse countries^{21,22,23,24,25,26,27,28,29}. Projections of temperature-related excess mortality under various climate change scenarios reveal regional disparities, with warmer and low-resource regions expected to bear a greater burden²⁶. Under high-emission scenarios such as representative concentration pathway 8.5 (RCP8.5), which assumes continued high greenhouse gas emissions, warmer regions, including Central and South America, South Europe and particularly Southeast Asia, exhibit sharp increases in heat-related excess mortality (that is, mortality attributable to temperatures above the minimum mortality temperature), which outweigh reductions in cold-related excess mortality (that is, mortality attributable to temperatures below the minimum mortality temperature)^26,27. Consequently, the net change in temperature-related excess mortality is projected to rise substantially in these regions. By contrast, temperate regions such as North Europe, East Asia and Australia, where warming is less intense, show smaller net increases in temperature-related excess mortality, primarily owing to pronounced declines in cold-related mortality^26,27. These findings underscore the importance of policies to mitigate global warming and the disproportionate health risks faced by populations living in warmer and, in some cases, poorer regions.

Despite the importance of projection studies, few have assessed the impact of projected climate change on suicide burden^30,31,32. One study using the high-emission scenario RCP8.5 estimated that a 1 °C monthly average temperature increase would cause a 0.7% rise in suicide rates in US counties and 2.1% rise in Mexican municipalities, potentially resulting in 9,000–40,000 additional suicides up to the year 2050³⁰. A more recent study in Japan projected temperature-related suicide mortality to rise under all climate scenarios³¹. In the 2090s, the projected increases compared with the 2010s were estimated under different shared socioeconomic pathway (SSP) scenarios, 0.6% for SSP1-2.6, 1.3% for SSP2-4.5 and 2.4% for SSP5-8.5, with greater increases under more extreme scenarios. In China, a nationwide case–control study identified a clear positive association between ambient temperature and suicide mortality, estimating that around 15% of suicide deaths were linked to non-optimum temperatures. The study further projected that excess suicide mortality would continue to rise throughout this century under high-emission scenarios, while stabilizing under moderate- and low-emission scenarios³². However, because most studies are country specific, a global comparative analysis is needed to elucidate the future global suicide burden under climate change.

This study aimed to (1) determine the extent of the temperature–suicide association based on observed data, (2) combine these associations with future temperatures under a changing climate and (3) predict future suicide mortality burden associated with temperatures across 751 locations in 26 countries by the 2050s (2050–2059). This work seeks to identify regional variations in the impacts of climate change on suicide risk and inform the development of effective suicide prevention and public health strategies.

Table 1 summarizes the descriptive statistics of suicide and temperature for each of the 26 countries (based on data from 751 locations) as well as the suicide rate obtained from WHO Global Health Estimates³³. The Republic of Korea, Japan, South Africa and Estonia reported high suicide rates of >20/100,000 population, whereas countries such as the Philippines, Mexico, Brazil and Paraguay recorded <5/100,000 population. The total number of suicides in the Multi-Country Multi-City (MCC) dataset was largely contributed by Japan and the Republic of Korea owing to long time coverage and high suicide rate. Average temperatures were relatively high in Southeast Asia (from 24 °C in Taiwan to 28.5 °C in Vietnam), South America (15.4 °C in Chile to 23.9 °C in Brazil) and Central America (18.8 °C in Mexico to 22.7 °C in Costa Rica). By contrast, North Europe recorded much lower average temperatures (3.8 °C in Finland to 10.5 °C in the UK). The average gross domestic product (GDP) for the countries in Central America, South America, Southeast Asia and South Africa was lower than that for other regions. The spatial distribution of mean temperatures across the 751 locations during the observation period is shown in Supplementary Fig. 1.

Table 2 shows the regional averages of modeled temperatures during the 2010s (2010–2019) and projected increases for the 2050s (2050–2059) for each SSP scenario. During the baseline period (2010s), Southeast Asia recorded the highest average temperature of 27.3 °C while North Europe had the lowest at about 5.3 °C. Across all three climate scenarios, North America was projected to show the largest temperature increase (1.25 °C to 2.43 °C), followed by North Europe (1.3 °C to 2.32 °C). By contrast, Southeast Asia (0.75 °C to 1.53 °C) and Australia (0.7 °C to 1.56 °C) were projected to exhibit the smallest rises. Under SSP2-4.5, all regions were projected to experience increases exceeding 1 °C relative to the baseline. Under the highest-emission scenario (SSP5-8.5), North America, North Europe, Central Europe and East Asia saw temperature increases exceeding 2 °C. Supplementary Fig. 2 presents the temporal changes in daily mean temperature relative to the average temperature during the 2010s, showing gradual warming under SSP1-2.6 and steep increases under SSP5-8.5 across all regions.

Figure 1 presents the overall cumulative temperature–suicide relationship by region, showing the relative risk (RR) across temperatures compared with the reference temperature (50th percentile of observed temperature) (see Methods for further details on the choice of reference temperature). The shape of the association varied across regions from linear to nonlinear. In North America, North Europe, Central Europe, South Europe, Southeast Asia and South Africa, the RR increased almost linearly with rising temperatures. Meanwhile, Central America and South America exhibited a J-shaped increase in RR as temperature increased. By contrast, East Asia and Australia demonstrated an inverse J-shaped nonlinear relationship, where RR increased with rising temperatures up to a threshold and then plateaued or slightly declined. Regions with many locations, such as North America, North Europe and East Asia (Table 1), exhibited narrower empirical confidence intervals (eCIs), indicating lower statistical uncertainty, than the other regions. The overall cumulative temperature–suicide relationships by country exhibited similar patterns within the same region, closely resembling the corresponding region-specific association curves (Supplementary Fig. 3).

Figure 2 presents the decadal changes in total attributable fraction (AF; %) of suicide mortality associated with temperatures relative to the 2010s (2010–2019) by region and SSP scenario (the corresponding numerical data are provided in Supplementary Table 1). The total AF represents the proportion of suicide mortality attributable to deviations from the reference temperature (as in Fig. 1) across the entire temperature range. Overall, total AF progressively increased over time and from the best-case (SSP1-2.6) to the worst-case (extreme-warming) scenario (SSP5-8.5), although the magnitude varied by region. In the 2050s (2050–2059), the numerical values of changes in total AF compared with the 2010s show that SSP1-2.6 projected a modest increase ranging from 0.6% in North Europe to 3.2% in Central America, while SSP2-4.5 projected from 0.9% to 4.4%. The largest increase is projected under SSP5-8.5, with projections ranging from 1.2% in North Europe to 6.4% in Central America (Fig. 2 and Supplementary Table 1). Similarly, the numerical projections for the decadal differences in the total attributable number (AN) of suicide deaths are provided in Supplementary Table 2.

In warm climates, such as Central America, South America, Southeast Asia and South Africa, the total AF is projected to increase sharply (Fig. 2). For instance, in Central America under the SSP5-8.5 scenario, the increment in total AF was by 6.4% (3.8, 10.6) from the 2010s to the 2050s (Supplementary Table 1), associated with the J-shaped relationship presented by sharply increasing RR curve over temperatures in this region (Fig. 1). Meanwhile, temperate and colder regions, including North America, North Europe, Central Europe, South Europe, East Asia and Australia, showed smaller increases in total AF across the decades and scenarios compared with other regions. Specifically, in East Asia, the increase in total AF between 2010s and 2050s was minimal, 2% (1.2, 3.5), even under SSP5-8.5 (Fig. 2 and Supplementary Table 1), reflecting the inverse J-shaped relationship in which RR increases and then plateaus with rising temperatures (Fig. 1). At a finer spatial scale, Fig. 3 illustrates location-specific changes in the total AF across 751 locations under SSP5-8.5. These location-specific patterns were largely consistent with the regional-level results. Larger changes were generally observed in locations belonging to regions that exhibited greater increases in Fig. 2. Similar spatial patterns in total AF differences were also observed under SSP1-2.6 and SSP2-4.5 (Supplementary Fig. 4), although the absolute magnitudes of AF changes were smaller in the lower-emission scenarios.

Future suicide burden associated with climate warming was affected jointly by the magnitude of the temperature–suicide relationship and climate change-related temperature increase. Figure 4 shows these factors by region: (a) RRs at extreme temperatures (1st and 99th percentiles) compared with the 50th percentile; (b) temperature changes between the 2010s and 2050s under SSP5-8.5; and (c) the resulting changes in total AF of suicide mortality associated with temperatures between the 2010s and the 2050s under SSP5-8.5 for each region. The regions are ordered by the sum of the two RRs. The five regions exhibiting the largest increases in total AF—Central America, South Africa, South Europe, Southeast Asia and South America—are generally warmer. These regions are projected to show moderate temperature increases of about 1.5–2 °C and larger sums of RRs. By contrast, the remaining regions—North Europe, North America, Central Europe, East Asia and Australia—were colder or temperate, showing smaller increases in total AF mainly because of their smaller sums of the RRs. Supplementary Fig. 5 presents the results for SSP1-2.6 and SSP2-4.5, corresponding to those shown in Fig. 4 for SSP5-8.5. The regional patterns are generally consistent across scenarios, with only the absolute magnitudes. From SSP1-2.6 to SSP5-8.5, both temperature increases and total AF differences become progressively larger across regions.

This study provides the first large-scale multi-country projection of suicide mortality burden associated with climate warming based on daily time series data for temperature and suicide mortality from 751 locations in 26 countries. Applying an advanced statistical methodology, we estimated the relationship between ambient temperature and suicide mortality based on observed data, linked these associations to modeled temperature under three SSP scenarios (SSP1-2.6, SSP2-4.5 and SSP5-8.5), and projected the future change in suicide mortality associated with temperature. We observed increases in future temperature-related suicide mortality from the 2010s to the 2050s, especially under the most extreme scenario SSP5-8.5.

Our findings are consistent with previous research linking rising temperatures to increased suicide risk. For example, our results in North and Central America are consistent with the elevated suicide risks projected under rising temperatures in the USA and Mexico for high-emission scenarios (for example, RCP8.5)³⁰. Similar to our findings in East Asia, a Japanese study identified an inverse J-shaped relationship between temperature and suicide risk, with projected increases in temperature-related suicide mortality across various climate scenarios³¹. Likewise, a nationwide study in China reported that suicide mortality increased with higher temperatures, projecting sustained increases in temperature-related suicide deaths under the high-emission scenario (SSP5-8.5)³². Although previous studies have largely focused on individual countries^30,31,32, their results support our conclusion that, without substantial climate mitigation, temperature-related suicide mortality is likely to rise with warming climate across diverse regions. By providing a global perspective, our study highlights the important worldwide mental health impacts associated with climate change.

We observed regional variations in projection of suicide burden associated with climate change. In warmer regions, such as Central and South America, South Europe, Southeast Asia and South Africa, steep increases were projected by the 2050s in all climate change scenarios. Meanwhile, temperate and colder regions, including North America, North Europe, Central Europe, East Asia and Australia, showed a more gradual increase. These regional differences primarily stem from distinct temperature–suicide relationships and the magnitude of climate change-related temperature increases. Warmer regions exhibited steeper temperature–suicide association curves, indicating that climate change intensifies high-temperature-related suicide risk. By contrast, temperate and colder regions showed gradual association curves. Notably, East Asia and Australia demonstrated inverse J-shaped temperature–suicide relationships, where suicide risk increased with temperature up to a certain point and then plateaued or slightly declined at extreme heat levels. This pattern may reflect behavioral and physiological adaptation to high temperatures among populations regularly exposed to hot conditions (for example, greater use of cooling systems and modified daily activities)³⁴.

Furthermore, the extrapolation range (that is, extent to which future temperatures fall outside the historical temperature range) of this temperature–suicide relationship was influenced by the magnitude of temperature increases. North America and North Europe, having the largest projected temperature increase, showed extensive extrapolation into high-temperature ranges, whereas East Asia and Australia, having a minimal increase, experienced relatively little. Therefore, as temperature distributions shift under climate change, the risks of exposure to low and high temperatures decrease and increase, respectively, while the magnitude of the shifts vary by region.

Another potential explanation for regional variations in future temperature-related suicide mortality may be differences in socioeconomic conditions. Except for South Europe, the warmer regions that showed sharp increases in suicide mortality associated with temperatures over time often overlapped with countries of relatively low socioeconomic status (based on average GDP; Table 1). By contrast, regions where suicide risk increased more gradually tended to comprise countries with higher average GDP. These findings suggest that the impact of climate change on suicide mortality may be mediated indirectly by regional disparities in economic stability and social infrastructure. Climate change can exacerbate existing inequalities, as vulnerable populations and low-income countries face greater economic instability^35,36, escalating conflicts^37,38,39 and declining agricultural productivity/food security⁴⁰. Moreover, in regions with limited access to mental health care and underdeveloped infrastructure, the lack of timely intervention and support may heighten the burden on vulnerable populations³⁶. Therefore, the link between rising temperatures and suicide mortality may reflect, at least in part, the indirect consequences of socioeconomic hardship amplified by climate change.

Our findings underscore the uneven regional distribution of climate change-related mental health risks, highlighting the complex interplay between temperature–suicide associations, climatic shifts and socioeconomic conditions. As global temperatures rise, more frequent and prolonged exposure to extreme heat may worsen mental health outcomes and exacerbate existing disparities. Therefore, considering the projected increases in suicide mortality associated with temperatures, urgent climate mitigation and targeted mental health interventions are necessary to reduce climate-driven public health burdens. Moreover, our findings highlight the importance of region-specific strategies in addressing mental health issues at the global level, considering the local climate conditions, socioeconomic factors, demographic structure and healthcare infrastructure.

The impact of temperature on suicide risk is mediated by a complex interplay of physiological and psychological mechanisms. While the exact pathways are not fully understood, several studies have suggested that rising temperatures exacerbate mental health conditions and stress levels^1,2,33, both of which increase suicide risk. In neuropsychology, heat exposure has been associated with disruptions in regulation of serotonin^41,42—a neurotransmitter critical for mood stabilization. Low serotonin levels have been linked to increased impulsivity and suicidal behavior^43,44,45. In addition, exposure to high temperatures promotes activation of the hypothalamic–pituitary–adrenal axis, leading to elevated blood concentrations of adrenocorticotropic hormone^46,47, which subsequently increases the levels of cortisol—a hormone essential for energy production in response to external stress⁴⁸. Furthermore, higher nighttime temperatures can impair sleep quality and contribute to sleep disturbances^49,50, in turn contributing to emotional dysregulation, cognitive impairment and an increased susceptibility to mental health disorders^5,49,50, including anxiety, depression, aggression and impaired judgment, substantially increasing the risk of suicidal ideation and behaviors^49,50,51,52. Collectively, these mechanisms suggest that the relationship between temperature and suicide is not merely correlational but rather a reflection of how heat exposure disrupts neurobiological and psychological processes, ultimately increasing individual vulnerability to suicidal behavior.

By applying a consistent methodology across diverse countries and regions, this study addresses the limitations of previous single-country/-region studies, offering a more comprehensive and nuanced perspective on the climate–suicide relationship. The two-stage analytical framework addresses the challenges posed by data sparsity in areas with low suicide rates/reports and incorporates hierarchical structures to enhance the precision of location-specific estimates. Moreover, by accounting for within-country heterogeneity and cross-national economic differences, the study improves the accuracy of projections by capturing both local and broader contextual factors. Our approach yielded robust and reliable estimates of future suicide burdens across a range of climate change scenarios, providing an empirical foundation for developing targeted climate adaptation and healthcare strategies that support equitable and effective public health responses.

Several limitations should be acknowledged in this study. To isolate the effects attributable solely to climate change, our projections relied on several simplifying assumptions. First, baseline suicide rates were assumed to remain constant over time, although suicide trends are influenced by evolving socioeconomic conditions, mental health services and prevention policies¹². Second, the exposure–response relationship between temperature and suicide was assumed to remain unchanged, implicitly representing a no-adaptation scenario, whereas behavioral, technological and infrastructural adaptation to heat may modify temperature-related risks²⁶. Third, population size and demographic structure were assumed to remain static, ignoring potential changes owing to population growth, aging and migration^26,27. These assumptions also create a conceptual inconsistency with the use of SSP-based temperature projections, as SSP pathways inherently incorporate assumptions regarding future socioeconomic development, demographic change and adaptation, while our projections of temperature-related suicide burden assumed no changes in population structure and no adaptation. Because all these factors may substantially influence future suicide burdens in complex way, the current projected estimates should be interpreted as scenario-based approximations rather than precise predictions. Future studies should therefore incorporate demographic dynamics, temporal change in suicide trends and potential adaptation when projecting temperature-related suicide burden to produce more realistic estimates^31,32.

Second, the second-stage meta-regression model has limitations in fully capturing heterogeneity at both the location and country levels. In this stage, only a limited set of meta-predictors—including regional indicators, country-level GDP and location-specific temperature—were incorporated. The model may have limited ability to reflect country-specific characteristics among countries within the same region that share similar temperature ranges. Consequently, the modeling framework may tend to integrate suicide risk curves across countries with broadly similar characteristics, potentially overlooking meaningful differences in temperature–suicide relationships across countries and locations.

A further limitation arises from representing thermal exposure using only daily mean temperature. Mean temperature smooths intra-day variability and may fail to capture extreme daytime heat or elevated nighttime temperatures, both of which may influence psychological stress, sleep quality and behavioral responses associated with suicide risk. Because climate change may differentially affect daytime and nighttime temperatures and increase the frequency and persistence of heat extremes, projections based solely on daily mean temperature may underestimate or mischaracterize the future burden of temperature-related suicide. Future studies may benefit from incorporating additional exposure metrics such as daily maximum and minimum temperature, heatwave indicators or composite thermal indices.

Finally, our study is subject to several limitations in data availability. Although our study is conducted on a multiple-countries scale, it still lacks full global coverage owing to data availability limitations, particularly in parts of the Middle East and several areas in Africa. In addition, the observed data primarily represent urban regions, even though suicide behavior may vary between urban and rural communities. Furthermore, Australia and South Africa represent a single country rather than the entire continent, although they are compared individually with other regions. It may not fully reflect regional and climate variations, potentially limiting the broader applicability of the findings. Lastly, the quality of suicide data varies substantially across countries¹². Globally, suicide underreporting—driven by poor certification systems, social stigma and in some settings the illegality of suicidal behavior—is estimated at about 18%, exceeding 30% in low- and middle-income countries^12,53. In addition, the projections were based on suicide count data rather than suicide rates. Because counts are not standardized by the population at risk, this approach may also limit comparability across regions with different population sizes and may obscure the influence of demographic changes such as population growth or aging. Future studies may benefit from incorporating population-standardized suicide rates and demographic projections to better disentangle climate-related risk changes from demographic influences.

This study offers important insights into how warming climates may affect regional mental health burden. Our findings suggest that, without focused climate mitigation efforts, temperature-related suicide mortality could increase substantially in the coming decades owing to rising temperatures, underscoring the urgent need for integrated public health strategies that combine climate action with mental health support. We recommend adopting proactive preventive measures such as expanding community-based mental health programs, strengthening healthcare infrastructure, raising public awareness and addressing climate-driven economic vulnerabilities globally. By aligning climate mitigation efforts with targeted mental health and socioeconomic strategies, we can collectively reduce the psychological impacts of climate change, support vulnerable populations and foster a more sustainable, cooperative and healthier future.

Observed temperature and suicide data

We collected data through the MCC Collaborative Research Network (http://mccstudy.lshtm.ac.uk/), which provides daily time series data of mean temperature and suicide mortality from 751 locations across 26 countries. Suicide cases were identified according to the International Statistical Classification of Diseases and Related Health Problems (ICD) using the codes X60-X84 (ICD-10) and E950-E958 (ICD-8 and ICD-9), which represent intentional self-poisoning and self-harm^54,55,56. Detailed information on data collection is provided in Supplementary Section 1. The dataset covers 1 January 1971 to 31 December 2021, although the years of available data vary by country. The 26 countries were categorized into 10 regions based on geographical and climatic characteristics: North America, Central America, South America, North Europe, Central Europe, South Europe, East Asia, Southeast Asia, Australia and South Africa, with the latter two each representing a single-country region. Table 1 shows the countries included in each region along with their respective observed data periods. This broad geographic coverage provides valuable insights into the impact of climate change on suicide across different climatic zones. Table 1 shows the summary statistics of suicide mortality and temperature for each country. Supplementary Fig. 1 shows the average observed temperatures across 751 locations worldwide.

Modeled temperature and suicide data

We obtained modeled daily mean temperature data for 1971–2059 under three SSP scenarios within Phase 6 of the Coupled Model Intercomparison Project^57,58. The projection periods are defined by decade, starting from the 2010s (2010–2019) and continuing through the 2050s (2050–2059). SSP1-2.6, SSP2-4.5 and SSP5-8.5 scenarios represent combined projections of future socioeconomic development and greenhouse gas emissions, ranging from mitigation (SSP1-2.6) to intermediate stabilization (SSP2-4.5) and high-emission (SSP5-8.5) scenarios⁵⁷. These SSPs are widely used in climate impact assessments to capture varying degrees of future climate change. The Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) framework⁵⁹ provides modeled mean temperature estimates using five general circulation models (GCMs) for each SSP scenario, which are bias-adjusted and downscaled to a spatial resolution of 0.5° × 0.5° (ref. ⁶⁰). The five GCMs from the ISIMIP database (https://www.isimip.org/)⁶¹ include the Geophysical Fluid Dynamics Laboratory Earth System Model Version 4 (GFDL-ESM4), Institut Pierre-Simon Laplace Climate Model 6A-Low Resolution (IPSL-CM6A-LR), Max Planck Institute Earth System Model 1.2-High Resolution (MPI-ESM1-2-HR), Meteorological Research Institute Earth System Model Version 2.0 (MRI-ESM2-0) and UK Earth System Model Version 1.0-LL (UKESM1-0-LL)⁶¹.

For each of the 751 locations, daily mean temperatures were modeled by linking the geographic coordinates of the centroid of each location to the corresponding grid cell in the dataset, using the five GCMs under each SSP scenario. To adjust for differences between the modeled and observed daily mean temperatures, the modeled temperature data were calibrated using additive and multiplicative correction factors to align the monthly mean and daily variability of modeled temperatures with those of the observed temperature data at each location^62,63. Regions showing large median differences (>5 °C) between observed and modeled temperatures after adjustment underwent a secondary bias correction by aligning their future mean temperatures with the national mean. Country-specific root mean square errors, presented in Supplementary Table 3, indicate close alignment between the calibrated modeled and observed temperatures. Table 2 presents the summary statistics of the modeled temperature estimates during the 2010s and projected increase for the 2050s by region and SSP scenario.

Modeled daily suicide mortality were calculated by averaging the observed suicide data for each calendar day and repeating the daily averages in all years during the projection period (2010–2059), assuming no changes in population structure or suicide rate. Hence, the population and suicide rates from the observed data were used as the baseline.

Statistical analysis

To project temperature-related suicide burden, we applied a statistical modeling framework^{26,30,31,62,64}. First, location-specific temperature–suicide associations were estimated based on the observed data using a two-stage approach. Second, the estimated associations were extrapolated to cover the range of the temperature distribution through 2059, and future increases in suicide mortality attributable to temperatures were calculated under three key assumptions: no adaptation, no change in population structure and no change in suicide rate. A detailed explanation is provided in Supplementary Section 2.

Estimation of location-specific temperature–suicide relationship

The temperature–suicide association for each location was estimated using a two-stage approach. First, a conditional Poisson regression model⁶⁵ was fitted for each location according to a time-stratified case-crossover design⁶⁶. The stratum was defined by the three-way interaction of calendar year, month and day of the week, controlling for time-varying confounders such as seasonality and long-term temporal trends. To account for the nonlinear and delayed relationship between temperature and suicide, a distributed lag nonlinear model⁶⁷ was used with a bidimensional cross-basis spline function and up to 3-day lag. The choice of lag was selected based on previous epidemiological studies^31,32, which indicate that suicide risk is primarily influenced by short-term temperature exposure—typically within 2–3 days—rather than by prolonged lag effects. A natural cubic spline was used for the exposure–response association with a single internal knot at the 50th percentile of the location-specific temperature distribution. For the lag–response association, a discrete parameterization was used to define the intervals for lags 0 and 1–3. These specifications were determined through a model selection procedure (Supplementary Section 2.1.1 and Supplementary Table 4) and sensitivity analyses^20,31,32 (Supplementary Fig. 11). Once the model was fit, the cross-basis coefficients were reduced to cumulate the exposure–response associations over the lags, and the reduced coefficients, which represent the cumulative temperature–suicide association, were used as inputs hereafter.

Second, we applied a multivariate mixed-effects meta-regression to pool the location-specific overall cumulative associations⁶⁸. Random effects were included to account for intercept variability at both the country and city levels, assuming a hierarchical structure where cities are nested within countries. Meta-predictors were included as fixed effects, such as region indicator, country-level GDP and location-specific variables including average temperature and daily mean temperature range during the observation period. These specifications were determined through a model selection procedure (Supplementary Section 2.1.2 and Supplementary Table 5). From the model, we derived the best linear unbiased predictions of the overall cumulative associations for each location, expressed as RR curves. This two-stage model improves location-specific estimates by leveraging information from locations with similar characteristics, particularly for low observed daily suicide counts⁶⁸.

Projection of region-specific impact of climate change on suicide

The burden of suicide mortality associated with temperatures (that is, temperatures lower or higher than the median of observed temperatures) was projected to compare the impact of climate change on suicide across regions^{26,31,32,63,65}. The projection assumed (1) no behavioral or physiological adaptation to temperature change, (2) no change in population structure and (3) no change in baseline suicide rates. These assumptions are consistent with standard approaches in climate-health impact projection studies and allow for isolation of the temperature-attributable component of future suicide burden (Supplementary Section 2.2.1). The projection was based on the location-specific temperature–suicide associations, reference temperatures, modeled temperature and modeled suicide data.

Next, a reference temperature needed to be chosen to calculate suicide mortality associated with temperatures. The reference temperature served as the baseline for suicide risk across all temperatures, with ANs or AFs calculated relative to this baseline. To ensure reasonable future projections, the reference temperature was selected within a range of temperatures expected to exist in the future, enabling consistent and meaningful comparisons of suicide risk across different time periods. In this study, the 50th percentile (median) of the observed temperature distribution at each location was selected as the reference, as it is likely to remain within the projected temperature range in the 2050s even under the most extreme scenario (SSP5-8.5). For example, in Fortaleza, Brazil, the lowest projected temperature in the 2050s under SSP5-8.5 is 27.6 °C, which coincides with its observed median temperature (27.6 °C), indicating that future temperatures are expected to begin at levels that historically correspond to the median temperature (Supplementary Table 6 and Supplementary Fig. 6).

We centered the RR curves at the reference temperature (as in Fig. 1), and overall cumulative risks associated with temperature based on observed data were calculated relative to this reference and extrapolated through 2059. The RRs were calculated for future days with temperatures above/below the reference. For each combination of SSPs and GCMs, based on these RRs, daily ANs were calculated for each day from 2010 to 2059 at each location for all combinations of GCMs and SSP scenarios. The location-specific ANs were then aggregated by region, SSP and decade, and averaged across GCMs. The region-specific decadal ANs were divided by the total number of suicide deaths in each region to obtain decadal AF of suicide mortality associated with temperatures^26,31,32,62. Finally, the difference in AF for each decade (2020s–2050s) relative to the 2010s (as presented in Fig. 2) was calculated to assess changes in future suicide burden associated with climate warming.

To put it more concretely, we examined the temporal changes in AF (%) of suicide mortality associated with temperatures above and below the reference. Because the reference temperature corresponds to the median of observed temperature distribution, the total AF can be decomposed into contributions from temperatures below and above this reference. For this decomposition, daily ANs were separately aggregated for days with temperatures above and below the reference at each location and summed by region, SSP and decade. These aggregated values were then divided by the total number of suicide deaths in each region to obtain AFs for warm and cold temperatures, respectively. Accordingly, regional changes in total AF were decomposed into contributions from warm (above-reference) and cold (below-reference) temperatures (Supplementary Fig. 7). While the relative contributions varied by region and location, AFs associated with both warm and cold temperatures increased over time and more intense warming, especially SSP5-8.5. The increase in AF associated with warm temperatures reflects a rise in excess suicide mortality at higher-than-reference temperatures. By contrast, the rise in AF for cold temperatures results from a diminishing protective effect of lower-than-reference temperatures, reducing the contribution of these lower temperatures that would otherwise offset excess suicide mortality. The spatial distribution of these decomposed results for each location is shown in Supplementary Fig. 8 for warm temperatures (above the reference temperature) and Supplementary Fig. 9 for cold temperatures (below the reference temperature).

We quantified the uncertainty in future temperature-related suicide mortality using Monte Carlo simulations at the meta-regression-model level⁶⁴ (Supplementary Section 2.2.2). Specifically, n = 1,000 sets of coefficients were sampled from a multivariate normal distribution defined by the pooled coefficients and their variance–covariance matrix obtained from the second-stage meta-regression model. For each simulation, daily ANs were recalculated and aggregated to derive the empirical distribution of regional AFs. The 2.5th and 97.5th percentiles of this distribution were used as the 95% eCIs. These eCIs accounted for the correlation among location-specific predictions and uncertainty associated with the spatialization of random effects in the meta-regression model. Uncertainty from different climate projections was additionally incorporated by combining temperature estimates from five GCMs under three SSP scenarios, ensuring that the final impact estimates reflect uncertainty across both exposure–response estimation and climate projections. This approach captures the uncertainty in multiple stages of the analysis and reduces the risk of underestimation, particularly in analyses involving small-area data and low heterogeneity⁶⁴.

Using the minimum suicide temperature as reference temperature

We compared our results with those obtained using the reference temperature defined as the temperature corresponding to the lowest suicide risk—the minimum suicide temperature (MinST) in the exposure–response curve (Supplementary Fig. 10). To avoid bias from extreme outliers, the MinST was identified within the 1st to 99th percentile range of observed daily temperatures, which in practice corresponded to the 1st temperature percentile in most locations (top panel in Supplementary Fig. 10). The overall shapes of the temperature–suicide association curves were similar to those obtained using the median temperature as the reference (top panel in Supplementary Fig. 10 and Fig. 1); however, using the MinST as the reference resulted in wider CI for RR estimates at higher temperatures (top panel in Supplementary Fig. 10). Using the MinST as a reference, temporal changes in total AF across future decades remained largely consistent with those obtained using the median-temperature reference (bottom panel in Supplementary Fig. 10 and Fig. 2). Although the absolute values of total AF varied slightly depending on the choice of reference temperature, AF estimates based on the MinST reference exhibited wider uncertainty across regions. Overall, the minimal differences observed between the two reference settings support the robustness of the results and the appropriateness of using the median temperature as the reference in the main analyses.

Sensitivity analysis

To assess the robustness of our results, a sensitivity analysis was conducted, referring to previous projection studies^20,31,32. Owing to data accessibility limitations, the sensitivity analysis was not performed remotely; instead, it was conducted using data from 426 locations included in this study. Since temperature can influence suicide risk over short-term periods and previous epidemiological studies^31,32 have adopted a 3-day lag, we compared only 3-day and 6-day lag structures. Temporal variations in total AF were similar between the two settings, indicating robustness to lag selection (Supplementary Fig. 11).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

1. Watts, N. et al. Health and climate change: policy responses to protect public health. Lancet 386, 1861–1914 (2015).

   Article  PubMed  Google Scholar 

2. Chevance, G. et al. Thinking health-related behaviors in a climate change context: a narrative review. Ann. Behav. Med. 57, 193–204 (2023).

   Article  PubMed  PubMed Central  Google Scholar 

3. McMichael, A. J., Woodruff, R. E. & Hales, S. Climate change and human health: present and future risks. Lancet 367, 859–869 (2006).

   Article  PubMed  Google Scholar 

4. Berry, H. L., Bowen, K. & Kjellstrom, T. Climate change and mental health: a causal pathways framework. Int. J. Public Health 55, 123–132 (2010).

   Article  PubMed  Google Scholar 

5. Mullins, J. T. & White, C. Temperature and mental health: evidence from the spectrum of mental health outcomes. J. Health Econ. 68, 102240 (2019).

   Article  PubMed  Google Scholar 

6. Rataj, E., Kunzweiler, K. & Garthus-Niegel, S. Extreme weather events in developing countries and related injuries and mental health disorders—a systematic review. BMC Public Health 16, 1020 (2016).

   Article  PubMed  PubMed Central  Google Scholar 

7. Tong, M. et al. Heat-attributable hospitalisation costs in Sydney: current estimations and future projections in the context of climate change. Urban Clim. 40, 101028 (2021).

   Article  Google Scholar 

8. Galea, S. et al. Exposure to hurricane-related stressors and mental illness after Hurricane Katrina. Arch. Gen. Psychiatry 64, 1427–1434 (2007).

   Article  PubMed  PubMed Central  Google Scholar 

9. Massazza, A., Ardino, V. & Fioravanzo, R. E. Climate change, trauma and mental health in Italy: a scoping review. Eur. J. Psychotraumatol. 13, 1–16 (2022).

   Article  PubMed  PubMed Central  Google Scholar 

10. Ingle, H. E. & Mikulewicz, M. Mental health and climate change: tackling invisible injustice. Lancet 4, e128–e130 (2020).

    Google Scholar 

11. Rifkin, D. I., Long, M. W. & Perry, M. J. Climate change and sleep: a systematic review of the literature and conceptual framework. Sleep Med. Rev. 42, 3–9 (2018).

    Article  PubMed  Google Scholar 

12. Suicide Worldwide in 2021: Global Health Estimates (World Health Organization, 2021).

13. Protecting Health from Climate Change: Global Research Priorities (World Health Organization, 2009).

14. Liu, J., Varghese, B., Hansen, A. & Bi, P. Effects of high temperatures on poor mental health outcomes: a systematic review and meta-analysis. Eur. J. Public Health 31, ckab164 (2021).

    Article  Google Scholar 

15. Thompson, R. et al. Ambient temperature and mental health: a systematic review and meta-analysis. Lancet 7, e580–e589 (2023).

    Google Scholar 

16. Baecker, L. et al. Impacts of extreme heat on mental health: systematic review and qualitative investigation of the underpinning mechanisms. J. Clim. Change Health 22, 100446 (2025).

    Article  Google Scholar 

17. Heo, S., Lee, W. & Bell, M. L. Suicide and associations with air pollution and ambient temperature: a systematic review and meta-analysis. Int. J. Environ. Res. Public Health 18, 7699 (2021).

    Article  PubMed  PubMed Central  Google Scholar 

18. Kim, S., Kim, Y. & Park, E. Association of ambient temperature with intentional self-harm and suicide death in Seoul: a case-crossover design with a distributed lag nonlinear model. Int. J. Biometeorol. 68, 2321–2331 (2024).

    Article  PubMed  PubMed Central  Google Scholar 

19. Min, H. et al. Ambient temperature, suicide, and urbanicity: a nationwide time-stratified case-crossover study in South Korea. PLoS ONE 20, e0337945 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

20. Kim, Y. et al. Suicide and ambient temperature: a multi-country multi-city study. Environ. Health Perspect. 127, 117007 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

21. Madaniyazi, L., Guo, Y., Yu, W. & Tong, S. Projecting future air pollution-related mortality under a changing climate: progress, uncertainties and research needs. Environ. Int. 75, 21–32 (2015).

    Article  PubMed  Google Scholar 

22. Huber, V. et al. Temperature-related excess mortality in German cities at 2 °C and higher degrees of global warming. Environ. Res. 186, 109447 (2020).

    Article  PubMed  Google Scholar 

23. Hanigan, I. C. & Chaston, T. B. Climate change, drought and rural suicide in New South Wales, Australia: future impact scenario projections to 2099. Int. J. Environ. Res. Public Health 19, 7855 (2022).

    Article  PubMed  PubMed Central  Google Scholar 

24. Liang, C. et al. Projecting the morbidity burden of mental and behavioral disorders associated with increasing humid heat in Shanghai. Nat. Ment. Health 3, 1532–1544 (2025).

    Article  Google Scholar 

25. Madaniyazi, L. et al. Seasonality of mortality under climate change: a multi-country projection study. Lancet 8, e86–e94 (2024).

    Google Scholar 

26. Gasparrini, A. et al. Projections of temperature-related excess mortality under climate change scenarios. Lancet 1, e360–e367 (2017).

    Google Scholar 

27. Masselot, P. et al. Estimating future heat-related and cold-related mortality under climate change, demographic and adaptation scenarios in 854 European cities. Nat. Med. 31, 1294–1302 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

28. Martínez-Solanas, È et al. Projections of temperature-attributable mortality in Europe: a time series analysis of 147 contiguous regions in 16 countries. Lancet Planet. Health 5, e446–e454 (2021).

    Article  PubMed  Google Scholar 

29. Lee, W. et al. Projections of excess mortality related to diurnal temperature range under climate change scenarios: a multi-country modelling study. Lancet 4, e512–e521 (2020).

    Google Scholar 

30. Burke, M. et al. Higher temperatures increase suicide rates in the United States and Mexico. Nat. Clim. Change 8, 723–729 (2018).

    Article  Google Scholar 

31. Thawonmas, R., Hashizume, M. & Kim, Y. Projections of temperature-related suicide under climate change scenarios in Japan. Environ. Health Perspect. 131, 117012 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

32. Zhou, Y. et al. Assessing the burden of suicide death associated with non-optimum temperature in a changing climate. JAMA Psychiatry 80, 488–497 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

33. Global Health Estimates (World Health Organization, 2025); https://www.who.int/data/global-health-estimates

34. Wu, C.-F., Hsieh, Y.-F. & Ou, S.-J. Thermal adaptation methods of urban plaza users in Asia’s hot-humid regions: a Taiwan case study. Int. J. Environ. Res. Public Health 12, 13560–13586 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

35. Barnes, M. C. et al. Understanding vulnerability to self-harm in times of economic hardship and austerity: a qualitative study. BMJ Open 6, e010131 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

36. Sinyor, M. et al. The coming global economic downturn and suicide: a call to action. Nat. Ment. Health 1, 233–235 (2023).

    Article  Google Scholar 

37. Burke, M., Hsiang, S. M. & Miguel, E. Climate and conflict. Annu. Rev. Econ. 7, 577–617 (2015).

    Article  Google Scholar 

38. Hsiang, S. M., Burke, M. & Miguel, E. Quantifying the influence of climate on human conflict. Science 341, 1235367 (2013).

    Article  PubMed  Google Scholar 

39. Extreme Heat Can Take a Toll on Mental Health (American Psychiatric Association, 2023); https://www.psychiatry.org/news-room/apa-blogs/extreme-heat-can-take-a-toll-on-mental-health

40. Kjellstrom, T. & McMichael, A. J. Climate change threats to population health and well-being: the imperative of protective solutions that will last. Glob. Health Action 6, 20816 (2013).

    Article  PubMed  PubMed Central  Google Scholar 

41. Lõhmus, M. Possible biological mechanisms linking mental health and heat—a contemplative review. Int. J. Environ. Res. Public Health 15, 1515 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

42. Gao, J. et al. Ambient temperature, sunlight duration, and suicide: a systematic review and meta-analysis. Sci. Total Environ. 646, 1021–1029 (2019).

    Article  PubMed  Google Scholar 

43. Mann, J. J. The serotonergic system in mood disorders and suicidal behaviour. Phil. Trans. R. Soc. B 368, 20120537 (2013).

    Article  PubMed  PubMed Central  Google Scholar 

44. Korpi, E. R. et al. Serotonin and 5-hydroxyindoleacetic acid in brains of suicide victims. Comparison in chronic schizophrenic patients with suicide as cause of death. Arch. Gen. Psychiatry 43, 594–600 (1986).

    Article  PubMed  Google Scholar 

45. Placidi, G. P. et al. Aggressivity, suicide attempts, and depression: relationship to cerebrospinal fluid monoamine metabolite levels. Biol. Psychiatry 50, 783–791 (2001).

    Article  PubMed  Google Scholar 

46. Popovska-Perčinić, F. et al. A moderate increase in ambient temperature influences the structure and hormonal secretion of adrenal glands in rats. Cell J. 22, 415–424 (2021).

    PubMed  Google Scholar 

47. Park, J. et al. A protective role of urban greenspace on the association between night-time heat and suicide in Seoul, South Korea. Environ. Res. Health 2, 13 (2024).

    Article  Google Scholar 

48. Coryell, W. & Schlesser, M. The dexamethasone suppression test and suicide prediction. Am. J. Psychiatry 158, 748–753 (2001).

    Article  PubMed  Google Scholar 

49. Obradovich, N., Migliorini, R., Mednick, S. C. & Fowler, J. H. Nighttime temperature and human sleep loss in a changing climate. Sci. Adv. 3, e1601555 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

50. Kovats, R. S. & Hajat, S. Heat stress and public health: a critical review. Annu. Rev. Public Health 29, 41–55 (2008).

    Article  PubMed  Google Scholar 

51. Perlis, M. L. et al. Suicide and sleep: is it a bad thing to be awake when reason sleeps? Sleep Med. Rev. 29, 101–107 (2016).

    Article  PubMed  Google Scholar 

52. Shah, A. et al. Suicide rates in five-year age-bands after the age of 60 years: the international landscape. Aging Ment. Health 20, 131–138 (2016).

    Article  PubMed  Google Scholar 

53. Meda, N. et al. How many people die by suicide each year? Not 727,000: a systematic review and meta-analysis of suicide underreporting across 71 countries over 122 years. Front. Psychiatry 16, 1609580 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

54. International Statistical Classification of Diseases and Related Health Problems, 8th Revision (World Health Organization, 1966).

55. International Statistical Classification of Diseases, 9th Revision, Basic Tabulation List with Alphabetic Index (World Health Organization, 1978); https://apps.who.int/iris/handle/10665/39473

56. International Statistical Classification of Diseases and Related Health Problems, 10th Revision (World Health Organization, 2016); https://icd.who.int/browse10/2016/en

57. Riahi, K. et al. The shared socioeconomic pathways and their energy, land use, and greenhouse gas emissions implications: an overview. Glob. Environ. Change 42, 153–168 (2017).

    Article  Google Scholar 

58. O’Neill, B. C. et al. The Scenario Model Intercomparison Project (ScenarioMIP) for CMIP6. Geosci. Model Dev. 9, 3461–3482 (2016).

    Article  Google Scholar 

59. Warszawski, L. et al. The Inter-Sectoral Impact Model Intercomparison Project (ISIMIP): project framework. Proc. Natl Acad. Sci. USA 111, 3228–3232 (2014).

    Article  PubMed  Google Scholar 

60. Lange, S. Trend-preserving bias adjustment and statistical downscaling with ISIMIP3BASD (v1.0). Geosci. Model Dev. 12, 3055–3070 (2019).

    Article  Google Scholar 

61. Lange, S. & Büchner, M. ISIMIP3b bias-adjusted atmospheric climate input data (v1.1). ISIMIP https://doi.org/10.48364/ISIMIP.842396.1 (2021).

62. Vicedo-Cabrera, A. M., Sera, F. & Gasparrini, A. Hands-on tutorial on a modeling framework for projections of climate change impacts on health. Epidemiology 30, 321–329 (2019).

    Article  PubMed  Google Scholar 

63. Hempel, S., Frieler, K., Warszawski, L., Schewe, J. & Piontek, F. A trend preserving bias correction—the ISI-MIP approach. Earth Syst. Dyn. 4, 219–236 (2013).

    Article  Google Scholar 

64. Masselot, P. & Gasparrini, A. Modelling extensions for multi-location studies in environmental epidemiology. Stat. Methods Med. Res. 34, 615–629 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

65. Armstrong, B. G., Gasparrini, A. & Tobias, A. Conditional Poisson models: a flexible alternative to conditional logistic case cross-over analysis. BMC Med. Res. Methodol. 14, 122 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

66. Lu, Y., Symons, J. M., Geyh, A. S. & Zeger, S. L. An approach to checking case-crossover analyses based on equivalence with time-series methods. Epidemiology 19, 169–175 (2008).

    Article  PubMed  Google Scholar 

67. Gasparrini, A., Armstrong, B. & Kenward, M. G. Distributed lag non-linear models. Stat. Med. 29, 2224–2234 (2010).

    Article  PubMed  PubMed Central  Google Scholar 

68. Sera, F., Armstrong, B., Blangiardo, M. & Gasparrini, A. An extended mixed-effects framework for meta-analysis. Stat. Med. 38, 5429–5444 (2019).

    Article  PubMed  Google Scholar 

Download references

Author notes

1. These authors contributed equally: Yoonhee Kim, Yeonseung Chung.

Authors and Affiliations

1. Graduate School of Data Science, Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea

   Hyeyeong Ro

2. Department of Global Environmental Health, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan

   Yoonhee Kim

3. Department of Global Health Policy, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan

   Masahiro Hashizume & Paul Lester Carlos Chua

4. School of Tropical Medicine and Global Health, Nagasaki University, Nagasaki, Japan

   Masahiro Hashizume, Lina Madaniyazi & Xerxes Seposo

5. School of the Environment, Yale University, New Haven, CT, USA

   Michelle L. Bell

6. Center for Climate Change Adaptation, National Institute for Environmental Studies, Tsukuba, Japan

   Yasushi Honda

7. Environment and Health Modelling (EHM) Lab, Department of Public Health Environments and Society, London School of Hygiene and Tropical Medicine, London, UK

   Antonio Gasparrini & Pierre Masselot

8. Department of Statistics, Computer Science and Applications “G. Parenti, ” University of Florence, Florence, Italy

   Francesco Sera

9. Climate, Air Quality Research Unit, School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria, Australia

   Yuming Guo, Shanshan Li & Wenzhong Huang

10. Department of Pathology, School of Medicine, University of São Paulo, São Paulo, Brazil

    Micheline De Sousa Zanotti Stagliorio Coelho

11. INSPER, São Paulo, Brazil

    Paulo Hilario Nascimento Saldiva

12. School of Epidemiology and Public Health, Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada

    Eric Lavigne

13. Environmental Health Science and Research Bureau, Health Canada, Ottawa, Ontario, Canada

    Eric Lavigne

14. Department of Public Health, Universidad de los Andes, Santiago, Chile

    Patricia Matus Correa

15. Centro Latino Americano de Excelencia en Cambio Climático y Salud, Universidad Peruana Cayetano Heredia, Lima, Peru

    Nicolas Valdes-Ortega

16. Department of Environmental Health, School of Public Health, Fudan University, Shanghai, China

    Haidong Kan

17. Biological Mission of Galicia (MBG) – Spanish Council for Scientific Research (CSIC), Pontevedra, Spain

    Dominic Roye

18. CIBERESP, Madrid, Spain

    Dominic Roye & Carmen Íñiguez

19. Institute of Atmospheric Physics, Czech Academy of Sciences, Prague, Czech Republic

    Jan Kyselý & Aleš Urban

20. Faculty of Environmental Sciences, Czech University of Life Sciences, Prague, Czech Republic

    Jan Kyselý & Aleš Urban

21. Institute of Family Medicine and Public Health, University of Tartu, Tartu, Estonia

    Hans Orru & Ene Indermitte

22. Center for Environmental and Respiratory Health Research (CERH), University of Oulu, Oulu, Finland

    Jouni J. K. Jaakkola & Nillo Ryti

23. Medical Research Center Oulu (MRC Oulu), Oulu University Hospital and University of Oulu, Oulu, Finland

    Jouni J. K. Jaakkola & Nillo Ryti

24. Institute of Epidemiology, Helmholtz Zentrum München – German Research Center for Environmental Health (GmbH), Neuherberg, Germany

    Alexandra Schneider, Veronika Huber & Maximilian Schwarz

25. Estación Biológica de Doñana (EBD) – Spanish Council for Scientific Research (CSIC), Seville, Spain

    Veronika Huber

26. Department of Epidemiology, Lazio Regional Health Service – ASL ROMA 1, Rome, Italy

    Paola Michelozzi & Francesca de’Donato

27. Department of Environmental Health, National Institute of Public Health, Cuernavaca, Mexico

    Magali Hurtado Diaz & Eunice Elizabeth Félix Arellano

28. Department of Hygiene, Graduate School of Medicine, Hokkaido University, Sapporo, Japan

    Xerxes Seposo

29. Faculty of Geography, Babes-Bolay University, Cluj-Napoca, Romania

    Iulian Horia Holobaca

30. Department of Environmental Health. Rollins School of Public Health, Emory University, Atlanta, GA, USA

    Noah Scovronick

31. Department of Earth Sciences, University of Torino, Turin, Italy

    Fella Acquaotta

32. Graduate School of Public Health, Seoul National University, Seoul, Republic of Korea

    Ho Kim

33. School of Biomedical Convergence Engineering, College of Information and Biomedical Engineering, Pusan National University, Yangsan, Republic of Korea

    Whanhee Lee

34. Institute of Environmental Assessment and Water Research (IDAEA), Spanish Council for Scientific Research (CSIC), Barcelona, Spain

    Aurelio Tobias

35. Department of Statistics and Computational Research, Universitat de València, València, Spain

    Carmen Íñiguez

36. Institute of Social and Preventive Medicine, University of Bern, Bern, Switzerland

    Ana Maria Vicedo-Cabrera

37. Oeschger Center for Climate Change Research, University of Bern, Bern, Switzerland

    Ana Maria Vicedo-Cabrera

38. Swiss Tropical and Public Health Institute, Allschwill, Switzerland

    Martina S. Ragettli

39. University of Basel, Basel, Switzerland

    Martina S. Ragettli

40. Environmental and Occupational Medicine, National Taiwan University (NTU) College of Medicine and NTU Hospital, Taipei, Taiwan

    Yue Leon Guo

41. National Institute of Environmental Health Science, National Health Research Institutes, Zhunan, Taiwan

    Yue Leon Guo

42. Graduate Institute of Environmental and Occupational Health Sciences, NTU College of Public Health, Taipei, Taiwan

    Yue Leon Guo

43. Research Center for Environmental Changes, Academia Sinica, Taipei, Taiwan

    Shih-Chun Pan

44. Department of Public Health Environments and Society, London School of Hygiene and Tropical Medicine, London, UK

    Ben Armstrong

45. Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA

    Antonella Zanobetti & Joel Schwartz

46. Department of Environmental Health, Faculty of Public Health, University of Medicine and Pharmacy at Ho Chi Minh City, Ho Chi Minh City, Vietnam

    Tran Ngoc Dang & Van Dung Do

47. Department of Mathematical Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea

    Yeonseung Chung

Contributions

H.R., Y.C., M.H. and Y.K. conceived the study. H.R., L.M. and Y.K. wrote the computer code and conducted the analysis. H.R. produced and extracted the data necessary for the analysis. H.R. and Y.C. led the writing, with regular inputs from Y.K. and M.H. All authors (H.R., Y.C., Y.K., M.H., L.M., M.L.B., Y.H., A.G., P. Masselot, F.S., Y.G., S.L., W.H., M.D.S.Z.S.C., P.H.N.S., E.L., P.M.C., N.V.-O., H. Kan, D.R., J.K., A.U., H.O., E.I., J.J.K.J., N.R., A.S., V.H., P. Michelozzi, F.D., M.H.D., E.E.F.A., X.S., P.L.C.C., I.H.H., N.S., F.A., H. Kim, W.L., A.T., C.Í., A.M.V.-C., M.S.R., Y.L.G., S.-C.P., B.A., A.Z., J.S., T.N.D., D.V.D. and M.S.) contributed to the paper and to the interpretation of the results. All authors approved the final version of the paper.

Corresponding authors

Correspondence to Yoonhee Kim or Yeonseung Chung.

Competing interests

The authors declare no competing interests.

Peer review information

Nature Mental Health thanks Brian O’Shea, Rainer Papsdorf, Michael Tong and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions