Disproportionate exposure to urban heat island intensity across major US cities

Built environments are typically hotter than nearby rural areas due to the urban heat island (UHI) effect, which is linked to increased mortality and non-fatal health outcomes. Urban form, greenness, and background climate drive both inter- and intra-urban temperature variability, potentially creating unequal heat burdens across sociodemographic groups. Prior evidence suggests marginalized and lower-income communities may face higher heat exposure, but most research has been limited to ad hoc, single or small sets of cities with heterogeneous data and methods, limiting generalizability. Studies on historical redlining show higher summer surface temperatures in formerly redlined areas, but the extent to which these translate to current racial or income disparities remains unclear. A uniform, national-scale assessment is needed to determine whether disparities in SUHI exposure by race/ethnicity, income, and age are systematic across major US cities. This study combines standardized satellite-derived SUHI intensity with consistent census tract-level sociodemographic data, focusing on summer daytime when exposure is highest, to quantify disparities and assess whether income explains racial/ethnic differences.

The UHI effect increases heat-related health risks and reduces productivity and learning. Intracity temperature variation is shaped by greenness, urban form, and city size. Small-scale case studies report disparities in UHI exposure within single cities and across a few global cities, with a tendency for higher-income groups to experience lower heat, though these results are difficult to generalize due to non-uniform definitions and sampling. Historical redlining is associated with higher present-day summer surface temperatures, yet demographic changes complicate direct translation to current racial and income disparities. Existing studies often use varied UHI definitions, non-comparable urban boundaries, and may reflect selection bias. There is also evidence of disproportionately higher heat-related morbidity and mortality among marginalized groups, but national-scale, methodologically consistent evidence linking SUHI exposure disparities to race/ethnicity and income has been lacking.

Study scope and period: The analysis covers the 175 largest US Census-defined urbanized areas (population >250,000 in 2017), encompassing ~65% of the US population and locations where most heat-related deaths occurred in the past 15 years. The focus is on summer months (June–August), when daytime SUHI intensity and mean temperatures are highest.

SUHI intensity data: The study uses a census-tract-consistent SUHI database for the USA, built from NASA MODIS global land surface temperature (LST) products and ESA land cover data. SUHI intensity for an urban census tract is defined using the simplified urban extent method as SUHI = LST_u − LST_r, where LST_u is the tract’s mean LST and LST_r is the mean LST of non-urban, non-water land cover pixels within the tract’s urbanized area (rural reference). When urbanized area and tract boundaries do not coincide, only pixels within the urbanized area are used. Analyses implicitly assume no residents in non-urbanized portions of outlying tracts.

Demographic data and group definitions: Tract-level demographics come from the 2017 5-year American Community Survey (ACS). Group means are computed as population-weighted averages across tracts, weighting by the group’s population in each tract. Race/ethnicity categories include non-Hispanic white, non-Hispanic Black, Hispanic (of any race), and non-Hispanic other (single other races and multi-racial non-Hispanic). A composite People of Color (POC) category includes all Hispanic and all who do not identify as white alone. Income categories are below poverty, 1–2× poverty, and ≥2× poverty. Age vulnerability groups include under 5 and ≥65 years, with race/ethnicity breakdowns where possible (noting Black Hispanics appear in both Black and Hispanic categories for age tables due to ACS limitations).

Climate zone stratification: Disparities are also examined by Köppen–Geiger climate zones (arid, snow, temperate, equatorial) to assess modulation of SUHI by background climate.

Statistical analysis: Population-weighted mean SUHI intensities are computed nationally and by climate zone for each sociodemographic group. City-level comparisons assess differences in group means; statistical significance of differences is indicated at p<0.05. Distributions across cities are visualized with kernel density estimates. To account for nonlinear health damages that increase with temperature, the Kolm–Pollak (KP) inequality index is used to evaluate within-group dispersion (hotspots) and derive equally distributed equivalents (EDEs), presented for a standard range of inequality aversion parameters (κ<0). Mapping highlights the geography of significant disparities across cities.

Software and data access: Statistical analyses were performed in Stata 15 and R 3.6.3 (ggplot2, tmap). The SUHI dataset was developed on Google Earth Engine. SUHI data and an interactive explorer are publicly available; ACS data were retrieved via the Census API.

• National disparities by race/ethnicity and income:

  • Across 175 urbanized areas, the average person of color lives in a census tract with higher summer daytime SUHI intensity than non-Hispanic whites in all but 6 cities (≈97%). Differences are statistically significant in ~75% of cities; non-Hispanic whites have significantly higher exposure in only one city (McAllen, TX).
  • People below poverty have higher SUHI exposure than those at ≥2× poverty in over 70% of cities, with significant differences common and no city where the below-poverty group has significantly lower exposure than the high-income group.
  • Population-weighted mean SUHI (°C), national totals (Table 1a): Total population 2.06. By race/ethnicity: People of Color 2.77; Hispanic 2.70; non-Hispanic Black 3.12; non-Hispanic White 1.47; non-Hispanic Other 2.41. By income: Below poverty 2.77; 1–2× poverty 2.50; ≥2× poverty 1.80.
  • Differences in means (°C), national totals (Table 1b): POC − NH White = 1.30 (p<0.01); Below poverty − ≥2× poverty = 0.97 (p<0.01); POC − Below poverty ≈ −0.00 (ns); NH White − Below poverty = −1.30 (p<0.01).
  • By climate zone, the share of cities where POC have higher exposure ranges from ~42% in arid climates to nearly 90% in snow climates.

• Race vs income disentanglement:

  • Although only about 10% of people of color live below the poverty line, their average SUHI exposure (2.77±2.70 °C) is nearly identical to that of people living below poverty (2.77±2.73 °C), indicating income alone does not explain racial/ethnic disparities.
  • City-level distributions for POC and below-poverty groups are similar; only ~10% of cities show significant differences between these two groups.

• Intragroup variation (hotspots):

  • KP inequality index values are similar across demographic groups within climate zones (national total around 0.26), with small, generally non-significant differences (Table 2). Thus, group mean disparities are not masking systematically greater within-group hotspots for specific groups.

• Age vulnerability:

  • People ≥65 have lower SUHI exposure than those <65 in 86% of cities; the difference is significant in 16%, and no city shows significantly higher exposure for ≥65.
  • National mean SUHI (°C): Under 65 ≈ 2.06; ≥65 = 1.84. For vulnerable subpopulations by race/ethnicity (Table 3a): POC <5 = 2.76; POC ≥65 = 2.88; NH White <5 = 1.45; NH White ≥65 = 1.44.
  • Differences in means (°C), national totals (Table 3b): POC − NH White for <5 = 1.31 (p<0.01); for ≥65 = 1.44 (p<0.01). Racial/ethnic disparities persist within vulnerable age groups.

• Illustrative city cases:

  • Greenville, SC: Black population concentrated in warmest tracts; POC have higher exposure than poor residents, who are more dispersed.
  • Baltimore, MD: Poorest tracts are warmest; POC are more evenly distributed; below-poverty residents have higher exposure than POC.

The study quantitatively demonstrates pervasive inequities in summer daytime SUHI exposure across major US cities: people of color and low-income residents consistently face higher exposures than non-Hispanic white and higher-income populations. This pattern holds nationally and across climate zones, particularly pronounced in the Northeast and upper Midwest. Considering both mean exposure and within-group dispersion using the Kolm–Pollak index shows disparities are not driven by systematically greater intragroup hotspots among disadvantaged groups; rather, the entire distribution is shifted toward higher exposures for these groups.

Age-related vulnerability adds nuance: older adults (≥65) tend to live in greener, cooler suburban areas and have slightly lower average SUHI exposure overall, yet racial/ethnic disparities persist within both the very young and elderly. Older people of color have slightly higher SUHI exposure than all people of color on average, suggesting reduced ability to relocate to cooler areas relative to white peers.

The findings indicate that income differences alone do not account for racial/ethnic disparities, consistent with the legacy of spatial segregation and housing market discrimination. These results have direct relevance for environmental equity and public health planning, given nonlinear health risks from heat and the role of background climate in modulating SUHI. Mapping city-specific disparities enables targeted, equitable interventions (e.g., urban greening, reflective surfaces, cooling infrastructure) aligned with local climate and urban form, ideally co-produced with affected communities.

This work provides the first comprehensive, nationally consistent assessment of disparities in summer daytime SUHI exposure across the 175 largest US urbanized areas, showing that people of color and low-income residents face systematically higher exposures than non-Hispanic white and higher-income populations. These inequities persist across climate zones and within vulnerable age groups. By integrating population-weighted SUHI metrics with inequality indices, the framework enables cities to identify where disparities are greatest and to design locally tailored mitigation strategies that reduce both overall exposure and inequities.

Future research should develop time-series SUHI and sociodemographic panels to evaluate how disparities evolve, especially under climate change and urbanization. Incorporating finer-scale demographic and mobility data, additional vulnerability determinants (e.g., housing quality, access to cooling), and comprehensive thermal stress metrics (including humidity, wind, and radiation) would refine exposure and risk estimates and inform more precise, equitable adaptation policies.

• SUHI as proxy: SUHI reflects land surface temperature and may not capture all components of heat stress (humidity, wind, radiation). Satellite-derived SUHI can overestimate UHI magnitude relative to in situ air temperatures, especially in daytime, due to shading effects not observed from space.
• Seasonal context: While SUHI is harmful in summer, UHI can confer wintertime benefits in very cold climates; however, such benefits may not offset disparities.
• Spatial averaging and exposure assumptions: Assigning a single SUHI to all residents of a census tract ignores within-tract variability and individual behaviors (time spent outdoors/elsewhere, air conditioning). Analyses assume individuals spend the day in their home tract.
• Scale tradeoffs: Using tracts balances demographic precision and satellite uncertainty, but aggregation may underestimate racial disparities (ecological fallacy). Pixel-level LST has higher uncertainty in urban areas; averaging within tracts improves reliability at the cost of spatial detail.
• Cross-sectional design: The study does not assess changes over time; longitudinal analyses are needed to track trends in exposure disparities.
• Data constraints in age-race breakdowns: ACS age data do not separate Black by Hispanic ethnicity, leading to overlap in Table 3 for Black Hispanics.