Discussion
We used a fine resolution geostatistical prediction model to understand
potential disparities in warm season ambient air temperature exposures
for ethnoracially minoritized groups using measures of residential
segregation in 2003–2019. We found that non-Hispanic Black and Latino
people were consistently exposed to hotter warm seasons than would be
expected if simply using the county average. Asian people also tended to
experience higher average temperatures, but with a notable exception in
Delaware where they experienced cooler summers. Finally, white people
consistently tended to experience cooler warm seasons compared to the
county averages. Spatial segregation analyses suggested that higher
levels of minoritized groups are associated with hotter warm seasons
and, relatedly, higher concentrations of white people are associated
with cooler warm seasons. These results have potential ramifications for
climate, health, and energy research and policy.
Many temperature-related studies have used coarse exposure assessments,
often averaged to the county level or some other large administrative
scale. However, we found systematically different exposure profiles for
populations within counties in this region. This represents potential
differential exposure misclassification, which could be a concern for
temperature exposure and epidemiology studies if not properly addressed.
With regard to policy, energy poverty alleviation programs like LIHEAP
use statewide CDDs among measures of energy demand, implicitly assuming
that all subgroups in the state are exposed to the same magnitude of
season. Yet we found that throughout the study region, minoritized
groups were exposed to higher warm season temperatures. Therefore,
energy demand is likely underestimated for these populations. This is
bolstered by other evidence that Black people are most likely to
experience energy insecurity nationwide year-round (11), but Latino
people experience the highest rates of warm season energy insecurity
(12). Given that we found some of the most prominent warm season
temperatures for Latino people, this may be one possible pathway
explaining this energy insecurity disparity. Finally, other studies have
found that historical redlining is associated with present-day land
cover characteristics associated with the urban heat island effect (16,
17). Those studies were limited to cities with documented redlining, but
our study covers entire states in our region using present-day measures
of segregation and air temperature. Given that we found statewide
temperature disparities beyond just formerly redlined cities, this
suggests that the historical processes associated with land use and land
cover disparities were not restricted to redlined cities. This is an
area for future investigation but speaks to the potential importance of
targeted greening initiatives in urban areas nationwide.
Our study has many strengths. We leveraged fine-resolution air
temperature predictions to understand temperature disparities, while
many other studies have used either coarser models or land surface
temperatures, which we believe more accurately reconstruct
neighborhood-level exposures. We also used metrics and methods to
enhance interpretability and policy relevance—our use of weighted
effect coding provides an intuitive comparison of the ethnoracial
average compared to the county average, and our use of CDDs is relevant
to U.S. energy policy. Our study also covered a large region of more
than 72 million people and more than 17 years. Finally, our data and
programmatic code are freely available to enhance reproducibility of our
findings.
Nonetheless, our study also has limitations. First, we examined
temperature exposures as assigned by residential location, but there may
be exposure disparities due to occupational exposures. Second, all
confidence intervals for Washington D.C. crossed zero, and thus we
concluded there were no apparent disparities. However, it is possible
that our spatial inference methods were overly conservative for a small
spatial area. Third, we examined two measures of residential segregation
based on concentration, but residential segregation is a
multidimensional construct (22). Relatedly, we examined segregation
measures concurrent to temperature exposures. While these and other
research suggest that residential segregation is a pathway towards
temperature exposure disparities, we need to examine the sequence of
segregation and later land development and temperature to make those
conclusions. Further, while we adjusted for year, we did not assess time
trends in these results across the study period. Finally, our
temperature prediction model covered the northeast and mid-Atlantic
regions of the U.S., but future studies should examine other regions or
the entire U.S.
Overall, our findings suggest that ethnoracially minoritized groups
experienced hotter average summers across the 13 states in the
Northeastern U.S. in 2003–2019. This study highlights the importance of
energy poverty alleviation programs, specifically for these subgroups.
These findings are critical to target interventions that enhance the
adaptive capacity of systematically marginalized groups in a warming
climate.