Green spaces provide substantial but unequal urban cooling globally

The study addresses how much urban green infrastructure cools cities globally and whether this cooling is unequally distributed between the Global North and Global South. With climate change and urban heat island effects compounding heat exposure—especially in rapidly urbanizing, low-latitude, and often lower-income regions—understanding adaptation capacity via nature-based solutions is crucial. The authors posit that disparities in the quantity and quality of urban green spaces lead to unequal outdoor heat mitigation. They aim to quantify three metrics—cooling efficiency (quality), cooling capacity (city-level cooling), and cooling benefit (population-weighted cooling experienced)—for the ~500 largest cities, evaluate geographic patterns and inequalities (North vs. South), identify natural and socioeconomic drivers, and estimate the potential to enhance cooling and reduce inequality.

Prior work shows cities face combined warming from climate change and urban heat islands, with lethal thresholds increasingly exceeded in low-latitude regions. Urban vegetation buffers heat via shading and transpiration. The 'luxury effect' indicates wealthier neighborhoods and cities generally have more green space and biodiversity, implying unequal access to cooling. Studies have characterized urban heat islands, vegetation patterns, and local cooling effects in select cities and regions, and some global assessments of cooling efficiency or greenspace exposure have emerged. Background climate strongly influences urban heat islands, and risk to urban forests under climate change is higher in the Global South. This study extends the literature by coupling global-scale, city-level assessments of green-space cooling capacity/benefit with inequality analysis and drivers across natural and socioeconomic gradients.

City selection and boundaries: The authors selected 502 cities with populations >1 million (UN 2018), merged contiguous urban agglomerations, excluded those <30 km², and delineated built-up urban extents using GAIA 30 m impervious data (2018). Permanent water bodies were masked (CGLS 10 m, ESA WorldCover, GSHHG). Final sample: 468 urbanized areas. Global North/South classification followed UNDP HDI 2019. Remote sensing data and period: Cooling quantified for each city’s hottest month in 2018 using Landsat 8 Level 2 LST and NDVI (30 m). Population used WorldPop 100 m (2018). Robustness analyses repeated at 1 km using MODIS (MYD11A1 LST, MYD13A2 NDVI), humidity (FLDAS), albedo (MCD43A3), aerosol (MCD19A2), wind (TerraClimate), building height/volume (GHSL/Global 3D), nighttime lights (VIIRS). Local climate zones (LCZ): Within each city, five LCZ-like classes were derived at 100 m using CGLS land cover and GHSL building heights: non-tree vegetation; low-rise; medium-high-rise; open tree cover; closed tree cover. Cooling efficiency (quality): For each LCZ in each city, ordinary least squares regressions related LST (response) to NDVI (predictor), controlling for elevation (MERIT DEM), building height (GHSL), and distance to water (GSHHG). Variables with VIF ≥5 were excluded to limit multicollinearity. Cooling efficiency was the absolute NDVI coefficient (°C per NDVI) for that LCZ. City-scale analyses at 1 km used multi-predictor regressions (LST ~ NDVI + meteorological and urban form covariates) to derive a whole-city efficiency. Cooling capacity (city-average cooling): Computed from grid-cell NDVI relative to the city’s minimum NDVI and the LCZ-specific cooling efficiency, averaged across LCZs weighted by their grid counts. Conceptually, capacity increases with both green space quantity (NDVI) and quality (efficiency). Cooling benefit (population-weighted cooling): Similar to capacity but weighted by local population density relative to the city mean, capturing the average cooling experienced by an urban resident. Alternative neighborhood accessibility windows (300 m, 500 m) were tested. Drivers of inequality: Piecewise structural equation modeling related cooling capacity/benefit to background climate (mean annual temperature, precipitation), topographic variation (elevation range), socioeconomic status (GDP per capita), and city size (area), with direct and indirect effects via green space area (NDVI) and cooling efficiency. Spatial autoregressive models checked robustness to spatial autocorrelation. Inequality metrics: Gini coefficients (population-density weighted; also unweighted and population-size weighted in supplements) measured between-city inequality in capacity and benefit, overall and by North/South. Potential enhancement scenarios: Upper-bound, climate-region-specific targets were defined per LCZ using Köppen classes: raising each city’s grid NDVI to the regional 50th–90th percentile of the best-performing reference city’s NDVI distribution, holding efficiency constant; enhancing efficiency to regional 50th–90th percentiles holding NDVI constant; or enhancing both to matched percentiles. An idealized spatial matching of high NDVI with high population density was also tested. A city-internal reference scenario raised low-NDVI pixels to the city’s own LCZ-specific 50th–90th percentiles.

• Global averages and inequality:
  • Mean city cooling capacity ≈ 2.9 °C (warm season daytime LST). Large variation (CV ~50%), with strong geographic structure: lower capacity in tropical/subtropical, low-latitude cities.
  • Global South vs. Global North gap: capacity 2.5 ± 1.0 °C vs. 3.6 ± 1.7 °C (Wilcoxon p = 2.7e−12); benefit 2.2 ± 0.9 °C vs. 3.4 ± 1.7 °C (p = 3.2e−13). Roughly two-thirds the cooling in Global South.
  • About 85% of the 50 highest-capacity cities are in the Global North; ~80% of the 50 lowest-capacity cities are in the Global South. No Global South city exceeds ~5.5 °C capacity, reflecting lower proportional green-space areas.
• Quantity vs. quality determinants:
  • Cooling capacity correlates strongly with green-space area (mean NDVI) (R² = 0.57) and more modestly with cooling efficiency (R² = 0.22). Thus, quantity dominates over quality in explaining between-city differences.
  • Population distribution reduces benefit slightly relative to capacity because denser areas often have less green space, but this has minor impact on global inequality patterns.
• Drivers (SEM results):
  • Proportional green-space area is positively associated with GDP per capita and city area; negatively with higher temperature and greater topographic variation; positively with precipitation/humidity.
  • Cooling efficiency shows modest associations: warmer cities have higher efficiency; more humid cities have lower efficiency. Model explains little variance in efficiency; green-space area model R² ≈ 0.50; cooling capacity model R² ≈ 0.90.
• Robustness:
  • Results are consistent using MODIS at 1 km, with an even larger North–South gap (~2-fold), and across alternative temporal windows and accessibility assumptions.
• Potential improvements:
  • Raising green-space quantity to regional upper-bound medians could increase capacity by ~2.4 °C globally; to the 90th percentile by ~3.8 °C. Enhancing efficiency alone yields smaller gains (~1.5 °C at 90th percentile) if quantity is fixed.
  • Co-maximizing both quantity and efficiency (to regional 90th percentiles) could yield up to ~10 °C additional cooling and reduce between-city inequality to Gini < 0.1. City-internal percentile targets yield smaller gains (~0.5–1.5 °C) with only slight inequality reduction.

The findings demonstrate a pronounced global inequality in the outdoor heat-mitigation capacity of urban green infrastructure, disadvantaging cities in the Global South where heat exposure is already higher and populations are denser and growing rapidly. This inequality is driven primarily by differences in green-space quantity and, to a lesser extent, quality (cooling efficiency). Wealth and city size correlate with greater green-space provision, reflecting a macro-scale analog of the luxury effect. Background climate and topography also shape green-space quantity, although it remains unclear whether these climatic effects operate directly on vegetation or indirectly through sociopolitical and historical pathways correlated with climate and geography. The study shows that boosting green-space area offers the largest and most practical gains in cooling, while improvements in efficiency provide additional but smaller benefits. Co-enhancing both could substantially increase cooling and drastically reduce global inequality. The implications are significant: without targeted investments and planning to expand and optimize urban green spaces, heat-related health risks and productivity losses may intensify, particularly in the Global South. Strategic greening, including innovative approaches that reconcile high-density development with expansive and effective greenery (e.g., vertical/rooftop greening), could broaden access to cooling benefits, especially if green-space locations are better matched to population concentrations.

This study provides a global, city-scale assessment of how much urban green spaces cool cities and how unequally those benefits are distributed between the Global North and South. By defining and quantifying cooling efficiency, capacity, and benefit, the authors show that the Global South currently realizes only about two-thirds of the cooling experienced in the Global North, largely due to lower green-space quantity. They identify socioeconomic (GDP per capita) and natural (climate, topography) correlates that shape green-space provision and demonstrate substantial potential to both increase cooling (up to ~10 °C under co-maximization scenarios) and reduce inequality (Gini < 0.1) through ambitious, targeted enhancements in green-space quantity and quality. Future research should integrate mechanistic urban biophysical models to predict efficiency from vegetation composition and climate, social theory and urban governance analyses to explain and overcome barriers to greening, economic models to evaluate policy levers and investments, and design/technology innovations to combine high-density living with resilient, accessible green infrastructure. Optimizing the spatial distribution of green spaces relative to population and ensuring climate-resilient species and management will be key to equitable outdoor heat adaptation.

• Causality and omitted variables: Structural equation modeling at global scale captures only a subset of relevant natural and socioeconomic factors; unmeasured sociocultural, historical, and governance variables may drive green-space provision and efficiency.
• Efficiency model fit: Cooling efficiency exhibits low explained variance; linear models may oversimplify local nonlinearities and heterogeneity across LCZs and cities.
• Temperature metric: Land surface temperature was used as the outcome; relationships to near-surface air temperature and human thermal comfort may differ in magnitude.
• Spatial aggregation: City-level averages of capacity/benefit do not fully capture within-city inequities, mobility patterns, or microclimatic effects of landscape configuration beyond LCZ stratification.
• Reference scenarios and assumptions: Potential enhancement estimates depend on assumed regional upper bounds (Köppen classes) and co-maximization feasibility; practical, sociopolitical, and physical constraints may limit attainability.
• Data-year specificity and scale: Main analyses use 2018 hottest-month data; although robustness checks were performed (2017–2019 and 1 km MODIS), interannual variability and data uncertainties remain.