of 9
ENVIR
ONMENT
AL STUDIES
Unequal
exposur
e to
hea
tw
aves
in
Los
Angeles:
Impa
ct
of
unev
en
green
spa
ces
Yi Yin
1
*, Liyin
He
1
, Paul O. Wennberg
1,2
, Chris
tian Frankenberg
1,3
Cities
worldwide
are experiencing
record-br
eaking
summer
temper
atur
es.
Urban
envir
onments
exa
cerba
te
extr
eme
hea
t, resulting
in
not
only
the
urban
hea
t island
but
also
intr
acity
varia
tions
in
hea
t exposur
e.
Under-
standing
these
disparities
is
crucial
to
support
equitable
clima
te
mitiga
tion
and
adapta
tion
efforts.
We found
persis
tent
nega
tiv
e corr
ela
tions
betw
een
da
ytime
land
surfa
ce
temper
atur
e (LST)
and
median
household
income
across
the
Los
Angeles
metr
opolitan
area
based
on
Ecos
ystem
Spa
ceborne
Thermal
Radiometer
Exper-
iment
on
Spa
ce
Sta
tion
observa
tions
from
2018
to
2021.
Lo
wer
evapotr
anspir
ation
resulting
from
the
unequal
dis
tribution
of
vegeta
tion
co
ver
is
a major
factor
leading
to
higher
LST
in
low-income
neighborhoods.
Dispar-
ities
worsen
with
higher
regional
mean
surfa
ce
temper
atur
e,
with
a $10,000
decr
ease
in
income
leading
to
~0.2°C
LST
incr
ease
at 20°C
and
up
to
~0.7°C
at 45°C.
With
mor
e frequent
and
intense
hea
t waves
projected
in
the
futur
e,
equitable
mitiga
tion
measur
es,
such
as
incr
easing
surfa
ce
albedo
and
tree
co
ver
in
low-income
neighborhoods,
are necessary
to
addr
ess
these
disparities.
Copyright
© 2023 The
Authors,
some
rights
reserved;
exclusive licensee
American
Associa
tion
for the Advancement
of Science.
No claim to
original
U.S. Government
Works. Distributed
under
a Creative
Commons
Attribution
NonCommer
cial
License
4.0 (CC BY-NC).
INTRODUCTION
Heat extremes pose a substantial
threat to public
health,
as they in-
crease mortality
and morbidity
(
1
), reduce
physical work capacity
(
2
), and negatively affect mental
health
(
3
). One of the major
factors that exacerbate this risk in cities is the urban
heat island
effect,
which
refers to the phenomenon
that urban
areas are typical-
lywarmer
than theirsurrounding
suburban
andrural areas(
4
). This
phenomenon
results
from several factors,
including
the heat-ab-
sorbing
properties
of building
materials
such as concrete, a lack
of vegetation, heat generated from human
activities,
and the city ge-
ometry
, which
traps heat and reduces
ventilation due to the close
proximity
of tall buildings
(
5
). Characteristics of the built environ-
ment,
such as city geometry
, green space, and surface reflectance,
play a critical
role in not only creating the urban
heat island
effect
compar
ed to rural areas but also resulting
intracity variations (
6
).
The high temper
atures in urban
centers
intensify
energy
consump-
tion, with waste heat further
amplifying
the challenges,
putting
vul-
nerable popula
tions at risk (
7
).
As of 2020, 56% of the world
s
popula
tion lives in urban
areas,
and urbaniza
tion is still a fast-ongoing
process in developing
coun-
tries (United
Nations Population Division).
Projections
sugges
t that
the fraction of the world
popula
tion living
in urban
areas will reach
nearly
70% by 2050.
Accompanying
this rural-to-urban
demo-
graphic
shift is a warming
climate (
8
), with the enhancement
of
heat waves in terms
of frequency
, intensity
, and duration (
5
).
Given the increasing
risks associa
ted with urban
heat extremes,
it
is becoming
increasingly
crucial
for cities to develop strategies
to
reduce
heat exposur
e. Therefore, it is important
to unders
tand
heat exposur
e experienced
by different communities
in urban
areas where the built environment
amplifies
extreme heat.
The impacts of climate change
can exacerbate existing inequali-
ties in society
(
9
), such as uneven distribution
of vegetation and tree
canopy
cover (
10
,
11
), urban
heat islands
(
12
14
),
and environmen-
tal hazards
and pollutants
(
15
,
16
). Disparities
in heat exposur
e have
been observ
ed across major
U.S. cities,
where resource-limited
res-
idents
and communities
of color
tend to live in areas with the
highes
t exposur
e to heat waves (
12
14
).
Hoffman
et al.
(
12
) found
that, across 108 U.S. urban
areas, 94% of studied
urban
areas display
consistent patterns
of elevated land surface temper
atures (LSTs)
in
formerly
redlined
areas relative to their non-redlined
neighbors,
demons
trating the long-las
ting impacts of historical
housing
poli-
cies. Hsu
et al.
(
14
) found
that people
of color and those living
in
poverty experience
greater heat exposur
e in comparison
to non-
Hispanic
whites
and households
above two times the poverty line,
respectiv
ely, across 175 U.S. cities.
These
results
highlight
the dis-
parities
in heat exposur
e among
different racial and socioeconomic
groups in the United
States, underscoring
the urgent
need for de-
tailed
informa
tion about
the sources of these disparities
in current
urban
settings
to support
equitable
adapta
tion and mitiga-
tion efforts.
In this study, we focus on Los Angeles
(LA) County, currently
home
to 9.8 million
residents
with diverse cultural and socioeco-
nomic
backgrounds
(U.S.
Census
Bureau, 2021,
https://www
.
census.go
v/quickfa
cts/fact/table/losangelescountycalifornia,C
A/
PST045221).
Previous
studies
have shown that the case of urban
heat islands
in LA is uniquely
comple
x because
of various
topo-
graphical
features and oceanic
influences
(
17
,
18
). To unders
tand
the tempor
al and spatial variations of urban
heat distribution,
we
use LST measur
ements
from Ecosystem Spaceborne
Thermal
Radi-
ometer
Experiment
on Space Station (ECOSTRESS)
(
18
,
19
). ECO-
STRESS
measur
es high-r
esolution
(~70 m) LST at various
local
times throughout
the year, providing
an opportunity
to unders
tand
the diurnal
and seasonal
variations in heat exposur
e patterns.
We
first analyze
the relationships
between ECOSTRESS
LST and socio-
economic
status across different neighborhoods
from July 2018 to
September
2021, covering four summers
from 2018 to 2021 with a
few notable
heatwaves. Further,
we analyze
the impacts of surface
1
Division
of Geological
and Planetary
Sciences,
California
Institute of Technology
,
Pasadena,
CA, USA.
2
Division
of Engineering
and Applied
Science,
California
Insti-
tute of Technology
, Pasadena,
CA, USA.
3
Jet Propulsion
Laboratory, California
In-
stitute of Technology
, Pasadena,
CA, USA.
*Corresponding
author.
Email:
yinyi11@gmail.com
Pr esent address: Department
of Global
Ecology
, Carnegie
Institution
for Science,
Stanford,
CA, USA.
Yin
et al.
,
Sci. Adv.
9
, eade8501
(2023)
28 April 2023
1 of 8
SCIENCE
ADVANCES
|
RESEARCH
ARTICLE
Downloaded from https://www.science.org at California Institute of Technology on June 05, 2023
albedo,
vegetation coverage, and evapotr
anspir
ation (ET) on the
heat exposur
e pattern to identify
potential
directions
for equitable
climate mitiga
tion and adapta
tion efforts.
RESUL
TS
Nega
tive correlations
betw
een LST and income
An analysis of a typical
summer
mid-da
y observa
tion from the
ECOSTRESS
satellite on 4 June 2021, at 1 p.m. reveals a noteworthy
negative correlation between LST and median
household
income
across the LA County
defined
at the zip code level (
r
=
0.68,
P
<
0.01; Fig. 1A). This negative correlation is also evident
at a finer
spatial resolution
at the block
group level (
r
=
0.63,
P
< 0.01; fig.
S2A),
as well as at the ECOSTRESS
LST pixel level when
we apply
the block-gr
oup level income
data to corresponding
location (
r
=
0.61,
P
< 0.01; figs. S1 and S2B).
The range of neighborhood
average LST spans
from 36° to 54°C along
the income
spectrum,
representing
a difference of almost 20°C (Fig. 1, A and D).
Accompanying
this negative correlation between LST and
income,
we also observ
e a positiv
e correlation between ET and
income
(
r
= 0.62,
P
< 0.01; Fig. 1B). For reference, the downward
solar shortw
ave radiation reaching
the land surface is 963 W m
2
at the time of measur
ement.
Hence,
the highes
t ET rate we
observ
e could
dissipa
te nearly
a third of the incoming
radiation.
The ET and LST retrievals
are based
on ECOSTRESS
observa
tions
and are negatively correlated as expected
(
r
=
0.7,
P
< 0.01; fig.
S3A):
Higher
ET leads to more latent heat loss and, hence,
lower
LST. LA is located in a semiarid
region
with a long dry season
each year, limiting
vegetation growth during
summer
months
due
to limited
water availability
(
20
). High vegetation cover is often
maintained
through irrigation, which
supports
ET and helps
to
cool the land surface (
21
). This is evidenced
by the significant
cor-
relation between normalized
difference vegetation index (NDVI)
derived from Landsa
t reflectance
and ECOSTRESS
ET (
r
= 0.61,
P
< 0.01; fig. S3B).
The strong correlations between income,
NDVI,
ET, and LST
sugges
t that neighborhoods
with higher
income
levels generally
have higher
NDVI,
higher
ET, and, hence,
lower LST. We used
linear
regression
to infer the slopes
of LST change
agains
t income
as the sensitivity
of LST to income
gradient,
which
measur
es the
Fig. 1. Spatial patterns
and correlations
of LST, median
household
income,
and vegeta
tion greenness
across the LA County
.
LST data are from ECOSTRESS
cap-
tured at 1 p.m., 4 June 2021; 2019 median
household
income
data are from U.S. Census
database;
normalized
difference vegetation index (NDVI)
is from Landsa
t 8
averaged over June 2021. (
A
) Correlations between median
household
income
and LST. (
B
) Correlations between median
household
income
and instantaneous
ET.
Each dot in (A) and (B) represents
a ZIP Code Tabulation Area (ZCTA). Corresponding
spatial distributions
of (
C
) ECOSTRESS
LST, (
D
) census
median
household
income,
and (
E
) Landsa
t NDVI at the zip code level. Note that only neighborhoods
with valid observa
tions passing
all the filtering
processes
as documented
in the
Materials
and Methods
are shown in the NDVI and LST distribution
plots.
Yin
et al.
,
Sci. Adv.
9
, eade8501
(2023)
28 April 2023
2 of 8
SCIENCE
ADVANCES
|
RESEARCH
ARTICLE
Downloaded from https://www.science.org at California Institute of Technology on June 05, 2023
degree of disparity
. For example,
in Fig. 1A, we derived a slope of
0.64
(°C/$10,000),
which
indica
tes that for every decrease of
$10,000
in the annual
household
income
across various
neighbor-
hoods,
thereis anincrease of0.64°C
in LST.To assess
the robustness
of the relationships
identified,
we tested different adminis
trative
boundaries
of the LA area here, including
the larger
LA metropol-
itan area, the South
Coast Air Basin
that contains
all of Orange
County, and the nondesert
regions
of LA County, Riverside
County, and San Bernardino
County, as well as the nondesert
regions
of LA County. Similar
negative correlations between LST
and income
are found,
and the results
are not sensitiv
e to the
choice
of adminis
trative boundaries.
Therefore, we present results
using
the boundary
of nondesert
regions
of LA County
hereafter
to reduce
differences
due to natural geographical
settings.
Besides,
we estimated the impacts of geographic settings
in terms
of altitude,
slope aspects,
and distance to the nearest coast; the first two factors
do not appear
to play a significant
role in influencing
the observ
ed
LST pattern,
while
the distance
to the nearest coast has a minor
effect (
r
2
= 0.1; fig. S5).
Tempor
al variations
of the LST and income
relationships
We collect
all available
ECOSTRESS
data that cover more than 30%
of the ZIP Code Tabulation Areas (ZCTAs) in the LA County
from
July 2018 to September
2021 at various
local times
to captur
e
diurnal
and seasonal
variations
in heat exposur
e patterns.
We
have a total of 329 ECOSTRESS
LST images
that passed
all
quality
checks
with a near uniform
distribution
of the diurnal
sampling
(sampling
numbers
are shown in Fig. 2C). We find con-
sistent negative correlations between daytime
LST and median
household
income
(Figs.
2 and 3), which
demons
trates the persis-
tenceoftheobserv
eddisparity
inLSTacrossincome
asshowninthe
case presented
in Fig. 1.
The sensitivity
of LST to income
is greatest around
noon
(Fig. 2A). At night
(defined
as 9 p.m. to 8 a.m.),
the correlation
between LST and median
household
income
is generally insignifi-
cant, although
it is still negative. The correlation becomes
signifi-
cantly
negative from 9 a.m. onward, peaking
at 1 p.m.,
and
thereafter decreases gradually
in the late afternoon
(Fig. 2B). A
strong positiv
e correlation between LST and shortw
ave albedo
is
also observ
ed during
the daytime
(fig. S4), which
sugges
ts that
higher
albedos
are associa
ted with higher
LST
the
opposite
of
what we would
expect if albedo
were the main factor affecting
the
surface energy
budget.
Therefore, we conclude
that ET is the dom-
inant term shaping
the regional
disparities
of LST across income
groups.
The diurnal
variations of the disparity
generally follow
the diurnal
cycle of solar radiation
in
the morning,
as the incom-
ing solar radiation surpasses
the outgoing
longw
ave radiation,
surface temper
atures start to increase. Only areas with vegetation
cover and sufficient
soil water content
can dissipa
te heat through
ETand result in a lower LST. Asthe sun startsto set, photos
ynthesis
and ET rates decline,
and the differences
in LST across income
groups become
smaller.
The daytime
patterns
of the derived sensitivity
of LST to income
show the most negative values
(i.e., the largest disparities)
during
Fig. 2. Diurnal
distribution
of observ
ed disparities.
(
A
) Slopes
of LST agains
t income,
(
B
) correlation coefficients
between LST and income,
and (
C
) number
of ECO-
STRESS
images
analyzed.
Each box shows the quartiles
of the dataset, while the whiskers extend
to show the rest of the distribution,
except
for diamonds
that are
detected
as outliers.
Yin
et al.
,
Sci. Adv.
9
, eade8501
(2023)
28 April 2023
3 of 8
SCIENCE
ADVANCES
|
RESEARCH
ARTICLE
Downloaded from https://www.science.org at California Institute of Technology on June 05, 2023
summer
(June,
July, and Augus
t) and spring
(March, April,
and
May) (Fig. 3). This is consistent with the seasonal
dynamics
of pho-
tosynthesis
in this area, where photos
ynthesis
peaks
from mid-
spring
(~April)
to late summer
(~Augus
t) following the dynamics
of solar radiation and water availability
(
20
). In autumn
and winter,
the magnitude
of disparity
decreases markedly as photos
ynthesis
declines,
confirming
the important
role of ET in driving
the
spatial differences
in LST. The LST and income
correlations and
slopes
are mostly negative, except for a slightly
positiv
e slope and
correlation during
winter
nights.
The correlation coefficient
between LST and median
household
income
is slightly
positiv
ewhen the regional
mean LST is low, and it
gets more negativewith rising
LST from 0° to 30°C and remains
rel-
ativelystable beyond that, sugges
ting persistent patterns
of disparity
in heat exposur
e at high temper
atures (Fig. 4A). However, the mag-
nitude
of observ
ed disparities
across all the ECOSTRESS
images
as measur
ed by the slopes
of the LST, income
regression
contin-
ues to increase with rising
temper
atures, indica
ting a higher
disparity
in heat exposur
e among
income
groups under
global
warming
(Fig. 4B).
Impa
cts of albedo
on LST
Albedo,
which
represents
the fraction ofthe incoming
radiation that
is reflected
by a surface, plays a significant
role in the surface energy
budget.
While
ouranalysis revealed that the spatial pattern of LST is
predominantly
shaped
by ET, we focus on impervious
surfaces of
the built environment
to evalua
te the impacts of albedo
on LST
in the absence
of vegetation impacts. To define
built surfaces, we
used three criteria:
classified
as urban
in the Landsa
t land use/
land coverclassifica
tion, an NDVI
of <0.01 (to remove possible
veg-
etation), and an albedo
of >0.1 (to remove water surface). We used
quantile
regression
to fit the conditional
quantiles
of the LST re-
sponse
to albedo
values
(Fig. 5A). This method
effectiv
ely minimiz-
es the influence
of other
factors
that may affect LST, such as
emissivity
, heat conductance,
and heat capacity. The analysis sug-
gests that higher
albedo
values
are generally associa
ted with lower
Fig. 3. Seasonal
distribution
of slopes
deriv
ed from ECOSTRESS
LST (°C) agains
t annual
household
income
($10,000),
a measur
e for the magnitudes
of dis-
parities
in heat exposur
e among
differ
ent income
groups.
The individual
light grey dot represents one overpassing
image
being
analyzed.
Each box shows the
quartiles
of the dataset, while the whiskers extend
to show the rest of the distribution,
except for diamonds
that are determined
as outliers.
Fig. 4. Enhancing
correlation and magnitudes
of disparity
with
rising
surfa
ce temper
ature.
Two-dimensional
density
distribution
of (
A
) correlation coefficients
between LST with median
household
income
(
y
axis) and regional
mean
LST (
x
axis) and (
B
) slopes
of LST agains
t income
(
y
axis) and the regional
mean
LST (
x
axis).
Yin
et al.
,
Sci. Adv.
9
, eade8501
(2023)
28 April 2023
4 of 8
SCIENCE
ADVANCES
|
RESEARCH
ARTICLE
Downloaded from https://www.science.org at California Institute of Technology on June 05, 2023
LST (
P
< 0.01 for all quantile
regressions
shown in Fig. 5A). The
steeper
slopes
observ
edinhigher
quantiles
providevaluable
insights
into the upper
tails of the LST distribution.
The diurnal
variations
in the LST-albedo
slopes
sugges
t that the variations in LST due to
albedo
are most prominent
around noon when solar radiation isthe
strongest (Fig. 5B), which
further
corroborates our finding.
This analysis provides us with an empirical
reference for the sen-
sitivity
of LST to changes
in albedo.
Specifically
, it implies
that an
increase of 0.1 in albedo
could
potentially
lead to a decrease of
around 2°C in LST at noon.
In July, a typical
daily mean
solar radi-
ation in LA is about
340 W m
2
, and an increase of 0.1 in the surface
albedo
could
lead to a reduction
of absorbed
energy
by 34 W m
2
.
This reduction
is equivalent
to the latent heat lossthrough the ETof
1.23 mm day
1
at a surface temper
ature of 40°C (1.21 mm day
1
at
30°C).
This is a simplified
estimation of the magnitude
of expected
effects
and more sophis
ticated modeling
approaches are clearly
nec-
essary
to fully unders
tand the effects
of albedo
change.
DISCUSSION
Implica
tions
of the observ
ed LST patterns
While
our findings
corroborate earlier
studies
that document
dis-
parities
in surface temper
ature in major
U.S. cities (
12
,
14
), this
analysis advances
upon these by identifying
robust linear
correla-
tions between median
house
income
and daytime
LST across LA
at 70-m
pixel level, block
group level, and zip code level using
spatial analysis based
on high-r
esolution
LST observa
tions from
ECOSTRESS.
We find that vegetation cover, which
correlates with
neighborhood
household
income
due to historical
and contempo-
rary inequities,
is the most significant
driving
factor in explaining
the observ
ed LST variations,
with much
more explaining
power
than topogr
aphic
features such as distance to the nearest coast or
elevation and slope aspects
(fig. S5).
The case of unequal
heat exposur
e in the LA area is particularly
prominent,
as the city locates in a semiarid
environment
with a sig-
nificant
dryseason,
where most urban
vegetation requires irrigation
in summer.
This results
in disparities
in LST that reflect dispropor-
tionately higher
outdoor
water usage in neighborhoods
with higher
income.
Previous
studies
have shown that, across the United
States,
Southeas
t and Western cities exhibit
the greatest intracity differenc-
es in LST between areas with different historical
housing
policies
(
12
). These
spatial differences
could
be related to different back-
ground
climates that have a strong influence
on the intensity
of
urban
heat islands
(
22
). To further
unders
tand regional
patterns
of this important
aspect
of climate and environmental
justice, addi-
tional
studies
are urgently
needed
to expand
the scope
of our
analysis.
One important
limitation of this study is that remotely
sensed
LST is
skin
temper
ature,
which
is only representa
tive of the
surface layer of buildings
or trees and may not reflect the ambient
temper
ature one experiences
within
a building
interior
or under
a
canopy
unders
tory. In particular,
LST does not represent
the
shading
effects
from trees in the canopy
unders
tory; hence,
the
cooling
effect
of tree canopy
is likely under
estimated using
the
LST metrics.
Urban
mobile
platform
observa
tions
have shown
that daytime
air temper
ature decreases nonlinearly
with increasing
canopy
cover (
23
). A case study in Sacramento,
California,
sugges
ts
that shade
trees can yield seasonal
cooling
energy
savings of ~30%
for residential
houses
(
24
).
To complement
the LST analysis, we also used a gridded
daily
temper
ature dataset at 4-km
resolution
that is interpola
ted from
meteor
ological
stations at 4-km resolutions
using
the PRISM
(Pa-
rameter-eleva
tion Relationships
on Independent
Slopes
Model)
product
(
25
). ECOSTRESS
LST at noon is significantly
correlated
with the same day PRISM
daily max air temper
ature at zip code
level, and the range of differences
across neighborhoods
in terms
of 2-m air temper
ature and LST is compar
able, both around
20°C,
although
LST has higher
absolute
values
(fig. S6A).
The re-
gression
of mean
daily
maximum
temper
atures in summer
months
agains
t median
household
income
is not statistically
signif-
icant (fig. S6B). However, there are huge limitations in using inter-
polated air temper
ature at a 4-km
spatial resolution
to represent
spatial variations in urban
heat exposur
e adequa
tely. Hence,
fine-
scaleLSTobserva
tions canstillprovidevaluable
insights
intointrac-
ity temper
ature gradients
and comple
xities at detailed
levelsthat are
difficult
to achieve with meteor
ological
temper
ature data alone.
Moving forward, more urban-scale
in situ observa
tions
of air
Fig. 5. Correlation betw
een LST and shortw
ave albedo
over urban
nonv
egeta
tive built
surfa
ces.
(
A
) Quantile
regression
ofECOSTRESS
LSTagains
t Landsa
t effectiv
e
shortw
ave surface albedo.
The ECOSTRESS
data are the same as shown in Fig. 1. Each dot represents
a colloca
ted pixel. All quantile
regressions
shown are statistically
significant
(
P
< 0.01). (
B
) Diurnal
variations of the slopes
derived from 95% quantile
regression
at different local times in the summer
of 2018. Slopes
that are statistically
significant
(
P
< 0.01) are shown in red in (B).
Yin
et al.
,
Sci. Adv.
9
, eade8501
(2023)
28 April 2023
5 of 8
SCIENCE
ADVANCES
|
RESEARCH
ARTICLE
Downloaded from https://www.science.org at California Institute of Technology on June 05, 2023
temper
ature and other relevant
meteor
ological
variables
are urgent-
ly needed
to better
unders
tand urban
heat exposur
e (
26
).
Possible
mitiga
tion methods
Larger
disparities
in LST are observ
ed as temper
ature increases
(Fig. 4), which
indica
tes that the observ
ed disparities
are likely to
worsen
under
future warming
unless
effectiv
e mitiga
tion efforts
are put into place. Therefore, we propose
the following mitiga
tion
methods
based
on our findings
that underscor
e the importance
of
vegetation cover, evapotr
anspir
ative cooling,
and surface albedo.
Increase
green spaces and greywater recycling
Planting
trees and vegetation are effectiv
e nature-based
solutions
to
mitiga
te urban
heat island
and moder
ate urban
microclima
tes
through shading
and ET (
23
). However, vegetation requires suffi-
cient water to maintain
its health
and cooling
capacity, but water
resources are limited
in the semiarid
environment
of LA. Therefore,
it is crucial
to strategically
locate trees in neighborhoods
with high
vulner
ability
and plan for their sustainable
irrigation to ensure con-
tinued
success.
Previous
studies
found
that turfgrass accounts
for
around 70% of ET from vegetation in LA (
27
), and overwatering
of residential
landscapes
is observ
ed (
28
). In contrast, mature
trees could
access
groundw
ater to complement
shallow water
from irrigation (
29
). Therefore, there is great potential
to support
more tree canopies
in the LA area by shifting
from water-consum-
ing turfgrass into drought-toler
ant trees with thoughtful
planning
and allocating necessary
water resources to the underserv
ed com-
munities.
In the meantime,
promoting
green landscapes
and reduc-
ing impervious
surfaces can increase the ground
s
capacity to
absorb
rainwater. Implementing
rain gardens,
swales, and under-
ground
infiltration trenches
can captur
e, treat, and naturally re-
charge
groundw
ater, thereby augmenting
local water supplies
(
30
). As a cobenefit,
these practices
could
partially
help mitiga
te
flooding
risks that are disproportiona
tely higher
for the disadvan-
taged popula
tions (
31
,
32
).
To reduce
the demand
for additional
irrigation, promoting
effi-
cient use of water already delivered to homes
through greywater re-
cycling
for landscapes
can be a viable
option.
Previous
studies
have
found
that 50 to 80% of household
wastewater can be classified
as
greywater, and greywater recycling
can positiv
ely affect the environ-
ment (
33
35
).
In 2015, LA Department
of Water and Power report-
ed an average residential
water usage
of 338 million
gallons
per day
(
36
). Assuming
that 70% of this water is used indoors
(
37
), if 50% of
the greywater is recycled
for landscape
irrigation and evenly spread
over the service
area of 473 square miles,
then we could
expect an
added
ETofapproxima
tely 0.35 mm day
1
, equivalent
to 30% of the
average daily precipita
tion over this area. However, proper care and
maintenance
are crucial
to ensure environmental
safety.
Solar-r
eflectiv
e cool pavements
and roofs
Asittakestimefortreestomatureandprovideeffectiv
eshading
and
cooling
effects
(
10
), using reflectiv
e surfaces, such as rooftops,
exte-
rior walls, and pavements
could
provide some immedia
te effect to
cool urban
areas (
38
). Pavements
and roofs typically
constitute over
60% of urban
surfaces (roofs, 20 to 25%; pavements,
about
40%)
(
39
). Using
reflectiv
e materials,
roof and pavement
albedos
can be
increased
by about
0.25 and 0.15, respectiv
ely, resulting
in a
net albedo
increase for urban
areas of about
0.1 (
39
). Simula
tions
using
coupled
Earth
System Model
show that altering
the surface
albedo
could
cool not only the LST but also the surface air temper-
ature in many
urbanized
areas across the globe (
40
), with consider-
able cobenefits
to air quality
(
41
). The LA county
has made
cool
roofs manda
tory (https://file.la
county
.gov/SDSInter/bos/supdocs/
128172.pdf
), and a novel approach to adding
a reflectiv
e coating
on pavement
has been applied
in high-vulner
ability
neighborhoods
with projects
such as Cool LA led by the LA Bureau of Street Ser-
vices (https://s
treetsla.la
city.org/cool-la-neighborhoods,
last access
on 16 Augus
t 2022).
While
the effects
of cool roofs have been
well studied,
future studies
are needed
to assess
the overall
thermal
and environmental
effects
of cool pavement
beyond its
effect
on surface temper
ature, as increasing
the surface albedo
for
walkways can reflect
more solar radiation and alter the energy
budget
of the near surface ambient
air in areas of human
activi-
ty (
42
).
Increasing
awareness
and system resilience
It is critical
to identify
the most vulner
able groups and allocate suf-
ficient
resources to adapt
to climate risks,
as climate adapta
tion
competes
with other
priorities
for government
and stakeholders.
Although
our study focuses
on the case of LA, it is essential
to
note that increasing
temper
ate cities,
unaccustomed
to heatwaves,
are also experiencing
rising
heat extremes,
as seen in the deadly
heatwave that occurr
ed in the United
Kingdom
in 2022 (
43
). The
heat exposur
einequity
metric
developed
in thisstudy canhelp iden-
tify vulner
able communities
in other
cities
with different back-
ground
climates, guiding
appropriate mitiga
tion efforts.
Raising
awareness
about
the risks of heatwaves and acknowledging
the
unequal
exposur
e across different socioeconomic
groups are in-
creasingly
important.
This can encour
age urban
planning
choices
and prompt
homeo
wners
to take action
through city programs
and other incentiv
es to promote
environmental
justice and increase
the resilience
of our cities as summers
get warmer.
MATERIALS
AND
METHODS
ECOSTRESS
LST and ET
The ECOSTRESS
was launched
to the Interna
tional
Space Station
(ISS) on 29 June 2018, which
provides by far the highes
t spatial res-
olution
thermal
infrared data available
from space between ~52°N
and ~52°S
(38-m
along-tr
ack and 69-m across-track). Unlike typical
sun-synchronous
orbits,
the precessing
ISS orbit
allows more
dynamic
sampling,
allowing for an overpass
return
frequency
to
be 1 to 5 days for a given location (depending
on the latitude)
at
different local times,
with revisits
around the same local time ap-
proxima
tely every 60 days (
19
). Such a unique
diurnal
sampling
strategy of ECOSTRESS
allows us to gain more insights
into the
nature of tempor
al variation in urban
heat island
effects
at fine
spatial scales.
We use ECOSTRESS
level 2 LST retrievals
(ECO2LSTE
v001,
available
at https://lpdaa
c.usgs.go
v/products/eco2ls
tev001/)
from
July 2018 to September
2021. For each scene,
we apply correspond-
ing cloud
masks
and quality
flags to ensure that only
normal
quality
pixels
are used in further
analysis. Clear-sky
ECOSTRESS
LST retrievals
have been evalua
ted with ground-based
measur
e-
ments
with an average root mean
square error of 1.07 K, a mean
absolute
error of 0.4 K, and
R
2
> 0.988 across 14 sites with different
landco
ver types (
44
). A systematic evalua
tion of different satellite
LST products
agains
t ground
observa
tions
for agricultur
al
Yin
et al.
,
Sci. Adv.
9
, eade8501
(2023)
28 April 2023
6 of 8
SCIENCE
ADVANCES
|
RESEARCH
ARTICLE
Downloaded from https://www.science.org at California Institute of Technology on June 05, 2023