Widespread race and class disparities in surface urban heat extremes across the United States

Abstract

Here we use 1000-m satellite land surface temperature anomaly measurements to explore the distribution of the United States' urban heating burden, both at high resolution (within cities or counties) and at scale (across the whole contiguous United States). While a rich literature has documented neighborhood-level disparities in urban heat exposures in individual cities, data constraints have precluded comparisons across locations. Here, drawing on extreme summer urban heat measurements from all 1056 U.S. counties with more than 10 developed census tracts, we find that the poorest tracts (and those with lowest average education levels) within a county are significantly hotter than the richest (and more educated) neighborhoods for 76% of these counties (54\% for education); we also find that neighborhoods with higher Black, Hispanic, and Asian population shares are hotter than the more White, non-Hispanic areas in each county. This holds in counties with both large and small spreads in these population shares, and for 71% of all counties the significant racial urban heat disparities persist even when adjusting for income. Although individual locations have different histories that have contributed to race- and class-based geographies, we find that the physical features of the urban environments driving these heat exposure gradients are fairly uniform across the country. Systematically, the disproportionate heat exposures faced by minority communities are due to higher population density, more built-up neighborhoods, and less vegetation.

Lower Urban Humidity Moderates Outdoor Heat Stress

“…Several studies (Benz & Burney, 2021; Chakraborty et al., 2019; Hsu et al., 2021; Hulley et al., 2019; Maimaitiyiming et al., 2014) have examined intra‐urban variability in temperature using satellite‐derived T s . To test whether T s is a useful proxy for urban HI 0 variability within cities, we calculate the intra‐urban gradients in T s , T a , RH, and HI 0 using those clusters (153 for Aqua and 155 for Terra) with enough (>10) CWSs in each decile neighborhood and the rural background (see Section 2; Figure 5).…”

“…Satellite‐derived T s is widely used for urban research (Bechtel et al., 2019; Benz & Burney, 2021; Chakraborty & Lee, 2019; Clinton & Gong, 2013; Manoli et al., 2019; Paschalis et al., 2021; Zhou et al., 2018). For observational studies, this is due to the availability of global and spatially continuous satellite measurements, which enable, among other things, analyses of intra‐urban and inter‐urban variability; difficult using ground‐based measurements.…”

“…Conceptual models of T s are easier to formulate than those for T a or HI 0 , due to strong coupling between T s and the surface energy budget (Li et al., 2019; L. Zhao et al., 2014). Although T s and T a are not strongly correlated over urban areas, especially relevant for public health (Ho et al., 2016; Stone et al., 2019), studies have assumed, either implicitly or explicitly, that Δ T s can still be useful for making decisions about urban heat mitigation (Benz & Burney, 2021; Chakraborty et al., 2020; Hsu et al., 2021; Hulley et al., 2019; L. Zhao et al., 2014; Manoli et al., 2019). We find that for cities in Europe during a heatwave period, the correlations between urban‐scale Δ T s and Δ T a are fairly weak, particularly during daytime ( r 2 = 0.10 for Aqua; 0.09 for Terra; Figure 4a, Figure S6a in Supporting Information ) with only 21% of the variability in Δ T s (slope = 0.21) among cities expected for Δ T a .…”

“…Coarse to medium‐resolution T s from satellites have been used for hotspot analysis within cities (Hulley et al., 2019; Maimaitiyiming et al., 2014; Mentaschi et al., 2022). Several studies have taken advantage of the spatial continuity of satellite observations to map intra‐urban variability of T s across cities with implications for environmental disparities (Benz & Burney, 2021; Chakraborty et al., 2019; Hsu et al., 2021). We find that for the cities considered here, T s is a poor proxy for the intra‐urban variability in HI 0 or other heat indices (including humidex, used in heat warning systems).…”

“…Although the analysis above is done for co‐located pixels, the threshold for the minimum number of stations used (10) is insufficient to represent the mean climatic state of the clusters. Moreover, it is important to also analyze how well T s , which has been extensively used as a proxy for the intra‐urban variability in urban temperatures (Benz & Burney, 2021; Chakraborty et al., 2019, 2020; Hoffman et al., 2020; Hsu et al., 2021; Hulley et al., 2019), represents the within‐city variability in HI 0 . To address this, we estimate the intra‐urban gradients in T s , T a , RH, and HI 0 .…”