A health impact assessment of progress towards urban nature targets in the 96 C40 cities

Summary

Background

Urban greenspaces (eg, parks and trees) and blue spaces (eg, rivers and coasts) improve climate regulation and human health. In 2021, the mayors of 31 cities in the C40 Climate Leadership Group set 2030 targets for the percentage of urban greenspace and population with nearby natural (green or blue) space. We quantified annual all-cause mortality reductions from progress towards these targets for C40's 96 member cities.

Methods

We conducted a quantitative health impact assessment, testing three illustrative scenarios to increase urban greenspace: uniformly across space, in areas with the least nature, and in the most populated areas. We converted one percentage point progress towards each target in terms of the Normalized Difference Vegetation Index (NDVI) using previously published associations. We used mortality rate estimates from The Global Burden of Diseases, Injuries, and Risk Factors Study, population data from WorldPop, and a pooled hazard ratio of NDVI and all-cause mortality from an epidemiological meta-analysis.

Findings

Uniformly increasing greenspace by 1% yielded an estimated 96-city median of 1·77 (range 0·65–3·52, IQR 1·46–2·19) fewer annual premature deaths per 100 000 population; increasing the population percentage with nearby natural space yielded an estimated median of 0·56 (range 0·11–1·70, IQR 0·44–0·69) fewer annual premature deaths per 100 000 population. On average, compared with uniform increases, adding greenspace in the least natural areas provided 1·4–1·7 times (depending on the target) the health benefits, and adding greenspace in the most populated areas provided 2·7 times the health benefits.

Interpretation

The geographical distribution of greenspace expansion influences the magnitude of associated health benefits across varied urban contexts. Health benefits are largest when greenspace is added near population centres.

Funding

Wellcome Trust, NASA, and The George Washington University.

Introduction

Theory and empirical evidence support a positive relationship between nature contact and human health. The biophilia hypothesis states that humans have an innate affinity towards nature, including greenspaces (eg, parks and trees) and blue spaces (eg, rivers and lakes), which were once synonymous with food, water, and shelter.¹ Greenspace is positively associated with general health and physical activity and negatively associated with adverse birth outcomes, cardiovascular events, all-cause mortality,² depression and anxiety,³ and lung and prostate cancer.⁴ Although fewer studies address the health benefits of urban blue space compared with urban greenspace, systematic reviews support the association between urban blue space and decreased obesity, decreased all-cause mortality, and improved general and mental health.⁵ Blue space and greenspace are hypothesised to benefit health through common pathways, including increased social interaction and physical activity and decreased stress and environmental harms.^{6, 7}

In 2021, 31 mayors of the C40 Cities Climate Leadership Group signed an Urban Nature Declaration (UND) with two 2030 goals: quality total cover (QTC; ie, 30–40% of total built-up city surface area will consist of green spaces or permeable spaces), and equitable spatial distribution (ESD; ie, 70% of city populations will have access to green or blue public spaces within a 15 min walk or bike ride).⁸ C40 is a global network of 96 cities committed to combating climate change through urban policies to decrease greenhouse gas emissions and climate risk while improving the health of populations. Many cities are now expanding greenspace, with limited characterisation of the resulting public health benefits.

Previous health impact assessments have defined urban greenspace using varied methods. The most commonly used metric is the Normalized Difference Vegetation Index (NDVI).^{9, 10, 11, 12, 13, 14, 15} NDVI is a satellite-derived measure ranging from –1·0 to 1·0, with negative values representing water, values between 0·0 and 0·1 representing barren land or urban area, values between 0·2 and 0·5 indicating sparse vegetation, and values between 0·6 and 1·0 denoting dense vegetation.¹⁶ Each 0·1 increase in NDVI has been associated with 34 000–38 000 annual avoided premature deaths among older adults across 35 US cities.⁹ NDVI has also been used to estimate the health benefits of increasing native plants,¹⁴ expanding urban tree canopy cover,¹⁵ and achieving WHO's recommendation that everyone has access to at least 0·5 hectares of public greenspace within 300 m of their homes across European cities.¹⁰ Beyond NDVI, other greenspace metrics have been used to estimate premature mortality impacts from expanding tree canopy cover in European and American cities^{15, 17, 18} and from greening in a superblock model in Barcelona, Spain.¹⁹ Studies have largely focused on European and US cities and have not explored how geographical distributions of additional greenspace influence estimated health benefits.

Research in context.

Evidence before this study

We searched PubMed for health impact assessments of natural space and all-cause mortality published from Jan 1, 2010, to March 15, 2024, using the search terms (“green space” OR “greenspace” OR “greenness” OR “greenery” OR “natural space” OR “park” OR “tree canopy” OR “green area” OR “NDVI” OR “normalized difference vegetation” OR “blue space” OR “natural space” OR “urban blue-green space” or “UBGS”) AND (“health impact assessment”) AND (“all-cause mortality” OR “mortality”). Most health impact assessments focused on European cities. Four such studies compared existing greenspace to WHO's recommendation that at least 0·5 hectares of public greenspace be accessible within 300 m of all residences. These studies used Normalized Difference Vegetation Index (NDVI) and the percentage of green area within a buffer area to measure greenspace. One study estimated summer deaths from the urban heat island effect that could be avoided by increasing tree canopy coverage to 30% across 93 European cities. Additionally, one study estimated the impact of greening from Barcelona, Spain's superblock model on premature mortality. Several health impact assessments considered US cities. One study of 35 US cities estimated the mortality reduction from a 0·1 increase in NDVI. Two studies quantified avoided premature deaths from meeting city tree canopy targets in Philadelphia, PA; Phoenix, AZ; and Denver, CO. One study estimated the impact of using exclusively native plants to increase green area in Denver, CO.

Added value of this study

We advance upon previous research by considering 96 global cities, estimating the health benefits of incremental progress towards real-world urban nature targets, and exploring the health impact of achieving these targets through different spatial allocations of additional greenspace. We found that the mortality burden that could be avoided from expanding urban greenspace ranged widely across cities and within world regions. Each percentage point increase in greenspace could reduce annual premature deaths by 0·65 (95% CI 0·33–1·00) to 14·72 (7·46–23·26) per 100 000 people, depending on the city and scenario. Each percentage point increase in the proportion of the population with nearby nature yielded 0·11 (0·06–0·16) to 8·33 (4·84–12·29) fewer estimated annual premature deaths per 100 000, depending on the city and scenario. Adding greenspace in areas with the least nature yielded 1·4–1·7 (depending on the target) times greater health benefits than uniform increases; adding greenspace to areas with the densest populations yielded 2·7 times greater health benefits. This work provides practical information for city policy makers and urban planners.

Implications of all the available evidence

Expanding greenspace, particularly near dense populations, benefits health across a wide range of urban contexts. Better understanding the relationship between blue space and mortality, and how greenspace type impacts health benefits, could improve urban policy decisions and the health impact estimates of such policies.

We estimate health benefits associated with incremental progress towards C40's UND targets across its 96 cities, which represent a range of urban contexts in terms of location, climate, income, and population size. Our analysis expands upon the previous literature by considering 96 cities globally; focusing on real-world urban nature targets; estimating health benefits for incremental progress towards urban nature targets, providing flexible results that remain relevant in cases where targets are partly met or exceeded or where targets change; and estimating the influence of the geographical distribution of added greenspace on health benefits. Our approach and results could be used by decision makers to track health benefits of progress towards these targets and urban planners to inform resource allocation to add nature where it most improves human health.

Methods

Overview

We quantified the potential health gains from progress towards the two C40 UND targets across C40's 96 member cities. These cities are in 48 countries across six continents and have 291 million residents (appendix p 1).²⁰ We focused on C40's 96 member cities because of their stated nature goals and commitment to implementing urban changes with climate and health benefits.

Health impact function and inputs

We followed a similar approach to previous multi-national studies estimating avoided premature deaths from changes in environmental exposures, using baseline mortality rates, population counts, and an epidemiologically derived exposure response function.^{10, 13, 21, 22} We estimated avoided mortality among adults 20 years and older associated with increases in greenspace using a linear health impact function:

$$\Delta mortality=\Sigma \left(Y_{oi}*po{p}_i*PAF*-1\right)$$

where mortality represents the annual change in mortality, for a given city, associated with each incremental change in greenspace. Key inputs include national (or subnational where available) baseline mortality for age category i (y_oi), grid cell population in age category i (pop_i), and the population attributable fraction (PAF). We multiplied this quantity by –1 because the hazard ratio (HR) is protective.

Following other health impact assessments of greenspace and all-cause mortality,^{14, 15} and after reviewing multi-national meta-analyses of epidemiological studies examining associations between urban greenspace and all-cause mortality (appendix p 2), we used an HR from Rojas-Rueda and colleagues²³ to estimate the PAF. We identified four meta-analyses and used the pooled HR from this study because it included only low-bias longitudinal studies with consistent greenspace exposure definitions (NDVI) and included adult populations from seven countries. The pooled HR was 0·96 (95% CI 0·94–0·97) for each 0·1 increase in NDVI within 500 m of a person's home. For this reason, we scaled the increase in NDVI by 0·1 and raised the HR to this power.

$$PAF=1-\frac{1}{H{R}^{\frac{\Delta NDVI}{0.1}}}$$

We conducted the health impact assessment at the 100 m pixel level, summing across all pixels within the urban bounds to calculate a city estimate of annual reductions in premature mortality. We used 2019 baseline mortality rates for each 5-year age category from age 20–24 years to 80 years and older from The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2019, the latest version at the time of analysis.²⁴ These data are generally available at the country level, with subnational estimates for some countries. Gridded 100 m 2020 population estimates are from WorldPop for equivalent 5-year age categories.²⁵ We defined city boundaries using the Global Human Settlement Urban Centre Database, which uses a consistent method based on built-up area and population data, capturing greater metropolitan areas rather than administrative boundaries.²⁶

Although evidence suggests that blue space provides similar health benefits to greenspaces,⁷ the association between blue space and all-cause mortality is not well established. One meta-analysis estimated a quantitative relationship between urban blue space and mortality but was limited to three studies using different exposure definitions (appendix p 3).⁵ Therefore, we included blue space in cities’ baseline nature estimates but assumed that progress towards the UND targets was achieved by expanding greenspace. This assumption reflects urban nature policies, which generally aim to increase vegetation as it is often more practical than creating new water bodies.

We assumed a linear health impact function, based on the meta-analysis from which we derived the HR. Some evidence suggests that the exposure–response curve could be non-linear beyond 0·2–0·5, with more uncertain associations outside this NDVI range.²⁷ We analysed the sensitivity of results to restricting the health benefits to NDVI increases within this range. Increasing NDVI within this range, where there is not already dense vegetation or built-up areas, might also be more feasible.

Current and target natural space exposure

We used estimates of 2020 greenspace extent and green and blue space distribution from Martin and colleagues,²⁸ which were derived from European Space Agency (ESA) satellite imagery. We measured nature exposure in two ways: using WorldCover, ESA's 10 m discrete landcover dataset; and using a greenest-pixel composite of all 2020 NDVI images, derived from ESA's Sentinel-2A satellite. Although 29 cities provided data on their cities’ nature levels to C40, these estimates used inconsistent approaches and urban boundary definitions. We relied on satellite data, which provide global consistency and pixel-level estimates needed to assess the geographical distribution of nature. We also compared satellite and city-provided estimates (appendix p 4).

Since the HR uses NDVI, we converted UND targets into the NDVI scale using associations from Martin and colleagues.²⁸ Briefly, Martin and colleagues developed city-specific linear associations at the 100 m grid cell level between landcover-defined and NDVI-defined UND target metrics. For the QTC target, we used the relationship between percentage greenspace and NDVI from Martin and colleagues directly. For the ESD target, we regressed the NDVI-based proportion of a pixel with sufficient green or blue space within a 1000 m buffer from Martin and colleagues (used to assess target progress) on the average NDVI value within the same buffer of the corresponding pixel (to align with HR metric). Buffers were allowed to extend beyond the urban boundary limits for pixels on the city periphery. In this regression, water pixels were identified using the landcover dataset and assigned a value of 1 in both metrics, to align with the ESD target's access to nearby green or blue space objective. Although including water this way increases baseline NDVI estimates for ESD, the health impact assessment considers only incremental increases in NDVI and is not affected by the higher baseline exposure estimates. We used the QTC regression slope to estimate increases in NDVI corresponding to a one percentage point increase in greenspace and the ESD regression slope to estimate increases in NDVI corresponding to one percentage point increase in the population with nearby nature.

Greenspace expansion scenarios

There are many possible approaches to increasing the percentage of greenspace or population with access to nearby natural space. To compare health benefits of alternative approaches, we developed and applied three illustrative scenarios that achieve the same city-average increases in greenspace but distribute additions in different ways. In all scenarios, we restricted NDVI values from exceeding 1, its maximum value.

In the uniform scenario, we increased NDVI in each 100 m land pixel uniformly. For QTC, this increase corresponded to a 1% increase in greenspace. For ESD, this increase was equivalent to a 1% increase in the percentage of the population with nearby nature. In the nature-deprived scenario, we increased NDVI in land pixels with the bottom quintile of nature by five times that of the uniform scenario, resulting in an equivalent city-average increase. In the population-centred scenario, we increased NDVI in land pixels in the top quintile of population by five times that of the uniform scenario, resulting in an equivalent city-average increase.

Realistically, allocation decisions for adding greenspace are driven by a complex multitude of factors; these scenarios are meant to provide a conceptual understanding of how health benefits of alternative approaches might differ. All NDVI increases are assumed to come from expanding greenspace.

Quantifying uncertainty

To estimate percentile intervals for the estimated annual changes in mortality from expanding urban nature, we ran 1000 Monte Carlo simulations of the health impact equation for each of the 96 cities, using error estimates from the meta-analysis,²³ GBD study,²⁴ and Martin and colleagues²⁸ to draw from normal distributions of the inverse HR, baseline mortality rates, and associations of NDVI and the C40 targets. The same draw of the inverse HR was applied across cities in each of the 1000 simulations to enable comparisons.

Role of the funding source

The study funders had no role in data collection, analysis, interpretation, or writing of this manuscript.

Results

Across the 96 cities, one percentage point increases in greenspace (QTC) were associated with a median of 53 (95% CI 45–68) fewer annual all-cause premature deaths in the uniform scenario, 76 (60–100) fewer deaths in the nature-deprived scenario, and 158 (115–212) fewer deaths in the population-centred scenario (figure 1). In general, averted deaths increased with population size (appendix pp 5–6). Where this trend did not hold (eg, Shenzhen–Guangzhou, China and Mexico City, Mexico), lower baseline mortality rates were a key factor. The median number of annual premature deaths averted per 100 000 population from one percentage point increases in greenspace was 1·77 (range 0·65–3·52, IQR 1·46–2·19) in the uniform scenario, 2·49 (range 0·88–6·27, IQR 1·88–3·39) in the nature-deprived scenario, and 4·87 (range 1·59–14·71, IQR 3·70–6·52) in the population-centred scenario (figure 1). Differences in population-standardised avoided mortality were associated with baseline mortality rates, the relationship between greenspace and NDVI, baseline levels of NDVI, and population density (appendix pp 8–9). Across the 96 cities, one percentage point increases in greenspace were associated with a mean of 1·71 (95% CI 0·87–2·67) avoided premature deaths per 100 000 people in the uniform scenario, 2·73 (1·39–4·27) in the nature-deprived scenario, and 4·88 (2·48–7·63) in the population-centred scenario (figure 1). 95th percentile intervals for individual cities and scenarios are shown in the appendix (pp 10–12).

Figure 1.

Quality total cover scenario results by region

Regional box plots of annual reductions in all-cause mortality in absolute (A) and population-standardised (B) terms for each of the scenarios. Results are paralleled for individual cities in the appendix (pp 5–6).

Overall, increasing nature near population centres provided 2·7 times the health benefits of the uniform scenario per 100 000 people, and increasing nature in the least natural areas provided 1·4 times the health benefits of the uniform scenario per 100 000 people. However, there were notable regional differences and city exceptions. North America was the only region where the nature-deprived scenario was associated with fewer median averted deaths than the uniform scenario, and this pattern held for ten of its 17 cities. Generally, central east Asia had the largest intraregional range; Hong Kong had 3·4 times more premature deaths averted per 100 000 than Shenzhen–Guangzhou in the uniform scenario and 5·4 times more in the population-centred scenario.

One percentage point increases in population with nearby nature were associated with a 96-city median of 15 (95% CI 13–20) fewer annual premature deaths in the uniform scenario, 27 (21–37) fewer in the nature-deprived scenario, and 50 (38–60) fewer in the population-centred scenario (figure 2). As with QTC, population was a key driver in estimated absolute deaths avoided. The median reduction in annual premature deaths per 100 000 population associated with a one percent increase in proximity to nature was 0·56 (range 0·11–1·69, IQR 0·44–0·69) in the uniform scenario, 0·97 (range 0·11–5·17, IQR 0·69–1·28) in the nature-deprived scenario, and 1·53 (range 0·16–8·33, IQR 1·08–2·19) in the population-centred scenario (figure 2). One percentage point increases in greenspace across the C40 cities were associated with a mean of 0·60 (95% CI 0·35–0·87) avoided mortality per 100 000 people in the uniform scenario, 1·21 (0·72–1·78) in the nature-deprived scenario, and 1·74 (1·03–2·55) in the population-centred scenario (figure 2). Individual city results and 95th percentile intervals for each city and each scenario are shown in the appendix (pp 7, 13–15).

Figure 2.

Equitable spatial distribution scenario results by region

Regional box plots of annual reductions in all-cause mortality in absolute (A) and population-standardised (B) terms for each of the scenarios. Results are paralleled for individual cities in the appendix (p 7).

Across the cities, the population-centred scenario provided 2·7 times the associated averted mortality of the uniform scenario, and the nature-deprived scenario provided 1·7 times the associated averted mortality of the uniform scenario. Europe generally had the widest intraregional range; Venice, Italy had 8·6 times more premature deaths averted per 100 000 people than Warsaw, Poland in the population-centred scenario, and 6·7 times more in the nature-deprived scenario. North America was the most dissimilar to other regions; three of the four cities that had more avoided deaths in the uniform scenario than in the nature-deprived scenario are North American (Lima, Peru; Houston, TX, USA; Austin, TX, USA; Phoenix, AZ, USA).

In a sensitivity analysis, we separately estimated mortality reductions from NDVI increases in low (0·0–0·2), middle (0·2–0·5), and high (0·5–1·0) ranges to characterise health benefits of adding greenspace within just the middle range, where the greenspace–mortality association is most certain.^{27, 29} For QTC and the uniform scenario, if health benefits are limited to the middle range, 1% increases in greenspace would yield a median of 26 (range 1–408, IQR 11–58) fewer deaths and 0·85 (range 0·16–2·32, IQR 0·64–1·14) fewer deaths per 100 000 per city (figure 3). Half of the estimated health benefits from expanding greenspace came from increases in the middle (0·2–0·5) range of NDVI, ranging from 9% (0·17 deaths per 100 000) in Portland, OR, USA to 83% (1·75 deaths per 100 000) in Phoenix, AZ, USA (figure 3). This contribution is similar for the population-centred scenario and larger (about three-quarters) for the nature-deprived scenario (appendix pp 16–18). If the health benefits of increasing greenspace were limited to increases between 0·2–0·5 NDVI, targeting greenspace interventions where there is currently sparse vegetation (NDVI of 0·2–0·5), rather than built-up areas or dense vegetation, would be most impactful. Greenspace increases to built-up areas would need to be substantial to achieve health benefits, and populations in areas with dense vegetation would not benefit from additional greenspace. Cities with more sparsely vegetated land and greater population density would stand to gain the most from additional greenspace.

Figure 3.

Annual reductions in all-cause mortality in absolute (A) and population-standardised (B) terms for the quality total cover target and uniform scenario

Cities are arranged from highest to lowest adult population size within each world region, indicated by background colour. Each bar is segmented, with colours representing the NDVI value of the contributing pixel. Three pairs of C40 cities (Johannesburg–Ekurhuleni, Shenzhen–Guangzhou, and Tokyo–Yokohama) are shown together because they share urban boundary definitions, which capture greater metropolitan areas. NDVI=Normalized Difference Vegetation Index.

For the ESD target, more of the estimated mortality reductions came from middle (0·2–0·5) and high (0·5–1·0) NDVI ranges, because we included both green and blue space and assigned water pixels a value of 1 (Figure 3, Figure 4). Although water was included in baseline nature exposure estimates, only additional greenspace was considered as an incremental increase to nearby nature. Including just the health gains from the middle range, a 1% increase in the population with nearby access to nature would yield a median of 8 (range 0–147, IQR 1–21) fewer deaths and 0·25 (range 0·00–0·76, IQR 0·10–0·38) fewer deaths per 100 000 people annually per city. NDVI increases between 0·2 and 0·5 contributed a little under half of the health benefits, ranging from 0 in Heidelberg, Germany to 98% (0·17 deaths per 100 000) in Amman, Jordan. Roughly 50% of the estimated mortality reductions from the full range of NDVI could be expected in the population-centred scenario, and 70% in the nature-deprived scenario (appendix pp 19–21).

Figure 4.

Annual reductions in all-cause mortality in absolute (A) and population-standardised (B) terms for the equitable spatial distribution target and uniform scenario

Cities are arranged from highest to lowest adult population size within each world region, indicated by background colour. Each bar is segmented, with colours representing the NDVI value of the contributing pixel. Three pairs of C40 cities (Johannesburg–Ekurhuleni, Shenzhen–Guangzhou, and Tokyo–Yokohama) are shown together because they share urban boundary definitions, which capture greater metropolitan areas. NDVI=Normalized Difference Vegetation Index.

Discussion

We estimated the health impacts of incremental progress towards C40's 2030 urban nature targets: QTC (reaching 30–40% greenspace) and ESD (ensuring 70% of their populations have access to nearby green or blue space). We found that uniformly increasing greenspace by one percentage point (QTC target) yielded an annual city median of 53 (IQR 30–119) fewer all-cause premature deaths, or 1·77 (1·46–2·19) deaths per 100 000 people. Increasing population access to nearby nature (ESD target) by one percentage point was associated with 15 (9–38) fewer all-cause premature deaths, or 0·56 (0·44–0·69) deaths per 100 000. We considered green and blue space in the baseline nature estimates for the ESD target, but considered only increases in greenspace in the health impact assessment. We observed over a five-fold intraregional difference in estimated health gains from adding urban greenspace towards the QTC target and an eight-fold difference in health gains from adding greenspace to the ESD target, driven by population size, baseline mortality rates, and the relationship between the target metrics and NDVI. On average, adding greenspace in areas with the least nature yielded 1·4–1·7 (depending on the target) times greater health benefits than uniform increases; adding greenspace to areas with the densest populations yielded 2·7 times greater health benefits. When we restricted the exposure–response function to NDVI values between 0·2 and 0·5, where the association is most certain, the avoided mortality burden from increasing greenspace was approximately halved, with wide variation across cities.

Our study has several key limitations. Although we included all measured uncertainty in our uncertainty analysis, our estimates ignore unmeasured error. The studies in the meta-analysis from which we draw our exposure–response function controlled for air pollution, factors of the built environment, and socioeconomic status; however, residual confounding from measured and unmeasured covariates probably remains. Many factors (such as mobility, allergies, job location, and social networks) might contribute to how residents interact with nearby residential greenspace. Our application of a single exposure–response function globally does not capture this variability in greenspace exposure, which might be related to an individual's age, sex, and socioeconomic status. Our results are likely to be more accurate for cities with similar demographics and population health status and greenspace contact as those included in the meta-analysis. Additionally, city characteristics such as the built environment, climate, and social and cultural context could influence greenspace use and its health benefits. Residents might face trade-offs in cities with more air pollution and heat and have greater benefits in cities with better pedestrian infrastructure. Furthermore, characteristics not captured by NDVI (eg, community engagement and the type, distribution, and quality of greenspaces) could modify their health and climate benefits. For example, larger, more clustered, more connected, and more asymmetrical greenspaces are associated with stronger health benefits, including reduced all-cause mortality.³⁰ The health and environmental benefits of different greenspace interventions such as planting more trees along major streets, creating a new city park, or adding green roofs are varied. However, increases in NDVI do not provide insight into how greenspace was increased or its subsequent health benefit. Additionally, NDVI is responsive to vegetation greenness, although natural spaces that are not green might provide common pathways to health, such as space to gather and exercise. This area of research could benefit from developing new urban nature metrics that leverage the advantages of NDVI; incorporate more information on greenspace quality, access, and usage; and consider other forms of nature, such as blue space. Considering the effects of urban greenspace more comprehensively (eg, in terms of reductions in noise, traffic, heat, and air pollution) could also improve estimates of its health benefits. Although developing new exposure metrics is beyond the scope of this study, measures that characterise urban nature in terms of possible interventions and by the strength of their health impact could better aid urban planners and decision makers.

As no best practices or guidelines exist on how to measure the C40 targets, estimated health impacts from expanding greenspace might vary considerably using different approaches, data sources, and exposure definitions. For example, we used the Global Human Settlement-Settlement Model dataset to define the urban boundaries, which might exclude natural spaces (due to lack of population and built-up area) that urban dwellers benefit from. Although nature located far from the urban centres might be carved out by this boundary definition, large urban parks (including Stanley Park in Vancouver, BC, Canada; Parque Ibirapuera in São Paulo, Brazil; and Lumpini Park in Bangkok, Thailand) are included. Furthermore, we applied country-level baseline mortality rates from GBD to each city, which could introduce error into our conclusions if the assumption that cities have the same mortality rate as the country as a whole is incorrect.

We explored three illustrative intervention scenarios: increasing nature uniformly across space, increasing nature where the population is most dense, and increasing nature where there is currently the least nature. Adding nature near population centres yielded the greatest benefits, but existing infrastructure might challenge implementation. One study investigated cities with high population density and ample greenspace and found that they had a greater proportion of apartments and multi-family housing units; less area designated for roadways and parking; and land sharing strategies such as tree-lined streets and green roofs, which incorporate nature into the built environment.³¹ These findings could inform implementation strategies for adding greenspace closer to population centres and where there is already built-up area (as in the population-centred and nature-deprived scenarios), where there is greater potential for health benefits.

The feasibility of increasing greenspace or expanding population access to nature by 1% varies across C40 cities. Some cities have climates where maintaining greenspaces is water-intensive and cost-prohibitive, and others have dense urban cores, where adding nature could require extensive urban landscape changes, including building demolition. Some research exists to inform best practices for greenspace development. When residents are involved in the design of urban greenspace, they are subsequently more likely to use these spaces for physical activity and social gatherings.³² We found that expanding urban greenspace, particularly near population centres, can provide important health benefits across a range of urban contexts. Learning from other cities that have successfully increased urban nature, especially in population dense settings, can inform implementation.

Data sharing

Baseline disease rates are available from https://gbd2019.healthdata.org/gbd-compare/. WorldPop datasets are available at https://developers.google.com/earth-engine/datasets/catalog/WorldPop_GP_100m_pop_age_sex_cons_unadj. NDVI datasets are available at https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_SR_HARMONIZED. Other input datasets are available upon request from GKM (gretam@gwu.edu).

Declaration of interests

We declare no competing interests.