Landscape Ecol (2016) 31:745–760
DOI 10.1007/s10980-015-0284-3

RESEARCH ARTICLE

Micro-scale urban surface temperatures are related to
land-cover features and residential heat related health
impacts in Phoenix, AZ USA
G. Darrel Jenerette . Sharon L. Harlan . Alexander Buyantuev .
William L. Stefanov . Juan Declet-Barreto . Benjamin L. Ruddell .
Soe Win Myint . Shai Kaplan . Xiaoxiao Li

Received: 10 September 2014 / Accepted: 28 September 2015 / Published online: 19 October 2015
Ó Springer Science+Business Media Dordrecht 2015

Abstract
Context With rapidly expanding urban regions, the
effects of land cover changes on urban surface
temperatures and the consequences of these changes
for human health are becoming progressively larger
problems.
Objectives We investigated residential parcel and
neighborhood scale variations in urban land surface
temperature, land cover, and residents’ perceptions of
landscapes and heat illnesses in the subtropical desert
city of Phoenix, AZ USA.

Methods We conducted an airborne imaging campaign that acquired high resolution urban land surface
temperature data (7 m/pixel) during the day and night.
We performed a geographic overlay of these data with
high resolution land cover maps, parcel boundaries,
neighborhood boundaries, and a household survey.
Results Land cover composition, including percentages of vegetated, building, and road areas, and
values for NDVI, and albedo, was correlated with
residential parcel surface temperatures and the effects
differed between day and night. Vegetation was more
effective at cooling hotter neighborhoods. We found

G. D. Jenerette (&)
Department of Botany and Plant Sciences, University of
California Riverside, Riverside, CA, USA
e-mail: darrel.jenerette@ucr.edu

B. L. Ruddell
Fulton Schools of Engineering, Arizona State University,
Tempe, AZ, USA

S. L. Harlan  J. Declet-Barreto
School of Human Evolution and Social Change, Arizona
State University, Tempe, AZ, USA
A. Buyantuev
Department of Geography and Planning, University at
Albany, State University of New York, Albany, NY, USA
W. L. Stefanov
Astromaterials Research and Exploration Science
Division, Exploration Integration and Science Directorate,
NASA Lyndon B. Johnson Space Center, Houston, TX,
USA

S. W. Myint
School of Geographical Sciences and Urban Planning,
Arizona State University, Tempe, AZ, USA
S. Kaplan
The Jacob Blaustein Institutes for Desert Research, BenGurion University of the Negev, Beersheba, Israel
X. Li
Julie Ann Wrigley Global Institute of Sustainability,
Arizona State University, Tempe, AZ, USA

J. Declet-Barreto
Natural Resources Defense Council, Washington, DC,
USA

123

746

consistencies between heat risk factors in neighborhood environments and residents’ perceptions of these
factors. Symptoms of heat-related illness were correlated with parcel scale surface temperature patterns
during the daytime but no corresponding relationship
was observed with nighttime surface temperatures.
Conclusions Residents’ experiences of heat vulnerability were related to the daytime land surface thermal
environment, which is influenced by micro-scale variation in land cover composition. These results provide a
first look at parcel-scale causes and consequences of
urban surface temperature variation and provide a
critically needed perspective on heat vulnerability
assessment studies conducted at much coarser scales.
Keywords Urban heat island  Parcel  MASTER 
Land surface temperature  Social surveys 
Vulnerability

Introduction
Urbanization is associated with increasing air temperatures in cities (Oke 1973), and it also leads to land
surface warming and increasing land surface temperature (LST) heterogeneity (Voogt and Oke 2003;
Jenerette et al. 2007; Cao et al. 2010; Imhoff et al.
2010; Peng et al. 2012; Zhou et al. 2014). Studies of
urban LST have primarily evaluated midday spatial
variation at neighborhood or coarser scales, such as the
city, zip code, and census tract boundaries. However,
extensive scale-dependent variation in LST is
expected throughout any given day due to different
speeds of warming within the complex spatial structure of urban environments (Li et al. 2011; Weng et al.
2011). Furthermore, the human consequences of urban
LST warming are not well characterized at very fine
spatial resolutions, such as individual residential
parcels, although several studies have found LST
variation affects spatial variability in the risk of heatrelated mortality at neighborhood scales (Johnson
et al. 2009; Buscail et al. 2012; Hondula et al. 2012;
Johnson et al. 2012; Harlan et al. 2013).
LST spatial variation is influenced by material
properties such as heat capacity, thermal conductivity,
and thermal inertia (Elachi 1987), and several studies
have proposed that local land cover composition is a
primary determinant of LST patterns (Li et al. 2011;

123

Landscape Ecol (2016) 31:745–760

Connors et al. 2013; Zhou et al. 2014). Increasing
vegetated land cover has been repeatedly shown to
lower LST through radiation interception and shading
and increases in latent heat fluxes associated with
transpiration (Jenerette et al. 2007; Kalma et al. 2008;
Kustas and Anderson 2009; Jenerette et al. 2011;
Xiang et al. 2014; Zhou et al. 2014). In contrast, built
surfaces, including buildings and roads, are generally
warmer than natural surfaces (Buyantuyev and Wu
2010; Zhou et al. 2014). Several studies have shown
that measurement scale affects the amount of observed
variability in land cover—LST relationships (Liang
and Weng 2008; Weng et al. 2011; Song et al. 2014),
but relationships between urban land cover and LST
have almost exclusively been evaluated using satellite-based thermal sensors on Landsat ETM? (60 m
pixels) or ASTER (90 m pixels) instruments (e.g. Li
et al. 2011; Myint et al. 2013; Song et al. 2014). At
these resolutions, extensive mixing of land covers may
obscure stronger or different relationships at finer
resolutions where more urban land cover variability
occurs (Zhou et al. 2011). An evaluation of urban LST
variation at micro scales, such as \10 m, has been
noted as an important research goal for improved
understanding of the biophysical causes urban heat
vulnerability (Small 2003; Deng and Wu 2013; Zhan
et al. 2013; Song et al. 2014).
In addition to improved spatial resolution, extending
LST measurements to both daytime and nighttime
allows for analysis of the differences in causes and
consequences of spatial patterns of daily maximum and
minimum urban LST (Buyantuyev and Wu 2010; Myint
et al. 2013). Because vegetation and high albedo
surfaces are cooler during the day and built surfaces
are warmer, these trends may continue at night due to
differences in stored heat (Buyantuyev and Wu 2010;
Myint et al. 2013). Alternatively, trees may trap infrared
radiation at night and lead to increased surface warming.
Although the upward facing built materials are warmer
during the day, they may more effectively radiate heat at
night, leading to lower nighttime LST while still
radiating heat horizontally. Albedo, as a property
associated with an illuminated surface, may be unrelated
to nighttime LST. Improved understanding of the
biophysical drivers of urban LST will require reconciling the differences in day and night LST variation.
Characterizing the biophysical basis of LST spatial
and temporal variations and connecting LST with
residents’ lived experiences should help us to better

Landscape Ecol (2016) 31:745–760

understand the determinants of variability in heat
vulnerability and implement better climate adaptation strategies for cities. In the large literature that
examines air temperature and heat-related mortality,
there is considerable discussion and debate about the
relative importance of daily and lagged minimum,
mean, and maximum temperature effects on health
outcome (Kalkstein 1991; Basu and Samet 2002;
Hondula and Barnett 2014; Petitti et al. 2015).
Studies in different places and weather conditions
have reached different conclusions. The important
question about the effects of minimum and maximum
temperatures has rarely been evaluated using urban
LST because there have been many fewer studies that
consider LST and most of them exclusively used
either daytime (e.g. Harlan et al. 2013) or nighttime
data (Uejio et al. 2011). One exception was Laaidi
et al. (2012), who examined minimum, mean, and
maximum surface temperatures, and found that
minimum temperature was associated with heat
deaths among the elderly during a heat wave in
Paris. As these authors pointed out, variability in
surface temperatures measured over the extent of the
city more closely approximates the actual living
conditions of urban residents than air temperature
measured at weather stations that are not in close
proximity to people (Laaidi et al. 2012).
In this study, we present new results from high
resolution (7 m) airborne thermal imagery acquired
over the Phoenix, AZ USA metropolitan region. We
link this high resolution thermal data with social
survey data on perceptions of landscapes and heat
illnesses that were collected during the same year. We
use the geographic intersection of these data in
conjunction with a sub-meter land cover classification
and residential parcel boundaries to answer two
questions: (1) How does landscape composition
influence LST variability during the day and night at
parcel and neighborhood scales? (2) Do parcel and
neighborhood surface characteristics correspond to
residents’ perceptions of their local landscapes and
self-reported symptoms of heat-related illnesses?
Answering these questions will improve understanding of how land-cover patterns influence LST in
residential environments and the potential consequences of LST for residents. These are essential
pieces of the puzzle for designing neighborhoods that
reduce human vulnerability to increasing urban
temperatures.

747

Methods
Site description
The Phoenix, AZ USA metropolitan region (population 4.3 million) is a well-studied model system that
has been useful for learning about many aspects of
urban ecology and climate (Grimm and Redman 2004;
Chow et al. 2012). This metropolis has a hot,
subtropical desert climate (Koppen classification
BWh) and features substantial variation in vegetation
and built surfaces associated with residential dwellings. Phoenix is an extreme endpoint of the urban heat
spectrum within the United States (Stone 2012). The
region has a pronounced nighttime urban heat island
(UHI) (Chow et al. 2012), measured in air temperature
and LST, and is home to one of the fastest growing
UHIs within the United States (Stone 2012). Overall,
and consistent with cities throughout the southwestern
United States, the city has more vegetation than the
outlying desert and higher-income neighborhoods
(census tracts) have more vegetation than lowerincome neighborhoods (Jenerette et al. 2013). Vegetation, in turn, is correlated with LST and higherincome neighborhoods are cooler (Jenerette et al.
2007; Buyantuyev and Wu 2010; Jenerette et al.
2011).
Microscale LST: MASTER flight and data
processing
We conducted an airborne LST data collection
campaign over the Phoenix metropolitan area during
a 5-day period in 2011: July 12–13 (daytime flights
collected from 10 am to 1 pm local time; -7 from
UTC) and July 15–16 (nighttime flights collected from
12:30–3 am local time). These times were selected to
be representative of mid-day and mid-night periods
outside the main transitions associated with sunrise or
sunset. For this campaign the MODIS/ASTER Simulator, or MASTER, instrument was mounted on a
Beechcraft B-200 aircraft and flown to obtain high
spatial resolution data. The MASTER sensor acquires
data over the visible through mid-infrared wavelengths (0.46–12.817 lm) in 50 spectral bands (Hook
et al. 2001), which were used to derive LST, normalized difference vegetation index (NDVI), and albedo
at approximately 7 m/pixel. These imagery data
measure a combination of the top of urban canopy

123

748

and ground data, which may be distinct from temperature patterns only at the ground (Goldreich 2006).
Although the original intent was to collect a true
diurnal dataset, this was not possible due to the large
area of data collection and pilot flight rules that
constrain daily flying time and mandate stand-down
periods for flight safety. Fortuitously, a stable weather
pattern prevailed during the week of July 12 that
enabled collection of an approximate diurnal dataset.
During this period, mean daily maximum air temperature was 40.1 °C and differed less than 1.6 °C
between sampling days.
We performed atmospheric correction to obtain
apparent surface reflectance, temperature, and emissivity, which allows for comparison of the flight line
data collected on four different days. The ENVI/IDL
image processing environment was used to perform all
processing of the MASTER data. Atmospheric correction of the mid-infrared wavelength data was
accomplished using an in-scene atmospheric compensation technique (Johnson and Young 1998). We used
an emissivity normalization approach (Kealy and
Hook 1993) to obtain LST from the MASTER
data—an emissivity value of 0.98 was used to
calculate the temperature in each band of the
MASTER thermal IR data for each pixel, and then
the highest temperature obtained was used to create
the LST layer. That temperature was then used to
invert the Planck function to calculate emissivity in
each of the bands. Atmospheric correction of the
visible through shortwave infrared wavelength data
was accomplished using the quick atmospheric correction (QUAC) algorithm (Bernstein et al. 2005)
within ENVI/IDL. Reasonableness of both LST and
apparent reflectance results were assessed qualitatively by random examination of returned pixel values
(surface temperature) and image spectra (vegetation
and soil pixel spectra were compared to laboratory
spectra of similar materials).
In addition to LST we also processed MASTER
data to obtain the NDVI, a measure of vegetation
(Tucker 1979), and an estimate of surface albedo.
NDVI was calculated as follows:
NDVI ¼ ðBand9  Band5Þ=ðBand9 þ Band5Þ
where Band9 and Band5 correspond to MASTER
spectral channels 9 (0.87 lm) and 5 (0.66 lm).
Albedo, the ratio of upwelling to downwelling radiative flux at the surface, was estimated by converting

123

Landscape Ecol (2016) 31:745–760

narrowband to broadband albedo using an empirical
formula previously developed for the MODIS sensor
(Liang 2001). Specifically, we derived the total visible
broadband albedo (avis) as follows
avis ¼ 0:331a5 þ 0:424a1 þ 0:246a3
where a1, a3, and a5 are reflectances in MASTER
spectral channels 1 (0.46 lm), 3 (0.54 lm), and 5
(0.66 lm), respectively.

Phoenix Area Social Survey
The 2011 Phoenix Area Social Survey (PASS) was the
second wave of a household survey conducted every
5 years in the metropolitan area to study humanenvironment interactions in an urban setting. It is part
of the long-term monitoring conducted by the Central
Arizona—Phoenix Long-Term Ecological Research
(CAP LTER) project (http://caplter.asu.edu/research/
long-term-monitoring/). A two-stage sampling design
was used for PASS. First, the sample of PASS
neighborhoods was drawn mostly from long-term
ecological monitoring sites that are part of the CAP
LTER project (Grimm and Redman 2004). U.S. Census Bureau block groups (i.e., relatively homogeneous
populations living within an area of approximately
0.65 km2) were used to define boundaries of PASS
neighborhoods. We selected a sample of 45 PASS
neighborhoods (39 from long-term monitoring sites
and six other CAP LTER research sites) stratified by
location (urban, suburban, fringe) and median household income. Percent retired and percent minority
residents were also considered in the selection to
ensure that different types of residential situations
were represented (Table 1).
Second, the sampling frame for households within
each neighborhood was a list of eligible residential
addresses created from maps and the tax assessor’s
parcel numbers (APN), which included single family
and multi-family houses, apartment buildings, and
mobile home parks. A random sample of 40 households was selected within each neighborhood.
Respondents were offered a small financial incentive
for participation and they could complete the survey
online, by telephone, or in person with an interviewer.
All materials were available in English and Spanish
and repeated follow-up efforts with non-respondents
were on-going through the study period. A total of 806

Landscape Ecol (2016) 31:745–760

749

Table 1 Stratification of the 2011 Phoenix Area Social Survey
neighborhoods and survey respondents

Table 2 Responses of participants to questions in the 2011
Phoenix Area Social Survey (n = 806)

Neighborhood
classification

Survey question

Neighborhoods
(n)

Respondents
(n)

Low income, urban core

6

97

Low income, suburban

6

110

Low income, fringe

2

34

Subtotal low income

14

241

Middle income, urban core

6

105

Middle income, suburban

6

111

Middle income, fringe

7

136

Subtotal middle income

19

352

High income, urban core

1

20

High income, suburban

5

94

High income, fringe
Subtotal high income

6
12

99
213

Grand total

45

806

Median annual household income for each neighborhood was
obtained from 2000 U.S. Census block group data and
classified as follows: low income (\$35,000); middle income
($35,000–$70,000); high income ([$70,000). Urban core
neighborhoods were within 8 km of downtown Phoenix or
within 2.4 km of the 7 other large-city downtowns. Fringe
areas had a moderate amount of undeveloped land within 1.6
km of the neighborhood in 2005. All other neighborhoods were
classified as suburban. At least 30 % of residents in five
neighborhoods were 65 years or older in 2010 (retirement
communities) and at least 50 % of residents in ten
neighborhoods were non-Hispanic white (Hispanic, African
American, Native American, or Asian/Pacific Islander) in 2010

respondents in 45 neighborhoods, ages 18 or older
(one per household), completed the survey for a
minimum response rate of 43.4 % using the standard
definitions of the American Association for Public
Opinion Research (2008). All surveys were completed
between May 26 and December 31, 2011.
The survey asked respondents’ opinions about
quality of life and condition of the natural environment
in their neighborhoods and in the region. Questions
about land use and climate were two areas of
emphasis. We selected three questions from PASS as
indicators of residents’ perceptions of environmental
conditions in their neighborhoods and one question on
self-reported heat illnesses (Table 2). The question
about experiencing heat-related symptoms was asked
in the survey context of 10 other questions about heat
and the symptoms we included in the question
(Table 2) were excerpted from the medical conditions
directly attributable to excessive heat exposure

Response
%

n

During the summer of 2010, do you think your neighborhood was
(_) than most other neighborhoods in the Valley?
A lot cooler
0.9
7
A little cooler
14
117
About the same temperature
22
174
A little hotter
9.4
76
A lot hotter
45
366
Are you (_) with trees that provide shade in your neighborhood?
Very satisfied
27
215
Somewhat satisfied
36
292
Somewhat dissatisfied
22
178
Very dissatisfied
13
106
Please indicate whether having too many paved surfaces such as
roads or parking lots is a (_) in your neighborhood
Big problem
5.8
47
Little problem
18
142
Not a problem at all
74
594
During last summer, did anyone in your household have symptoms
related to heat or high temperatures such as leg cramps, dry
mouth, dizziness, fatigue, fainting, rapid heartbeat or
hallucinations?
Yes
30
245
No
60
486
Do any of the following prevent you from using your airconditioning when the weather is hot? Please select all that apply
Cost of electricity
36
291
Unit doesn’t work
5.3
43
Cost of repairs
4.1
33
Noise
1.0
8
Do not have an air conditioner
3.6
29
Other
1.9
15
Summary: had at least one impediment
45.3
365

Missing and ‘‘don’t know’’ answers were omitted from
calculations on an item-by-item basis and 100 % was based
on number of valid answers for each question, or in the case of
the question on impediments to using air conditioning, valid
answers for each item. (For the question on use of air
conditioning, respondents could select multiple answers).
Responses to the question on neighborhood temperature were
coded so that higher scores indicate hotter neighborhood
relative to others. Responses to the question on satisfaction
with trees were coded so that higher scores indicate higher
satisfaction. Responses to the question on too much pavement
were coded so that higher scores indicate a smaller problem
(higher satisfaction). Subsamples of respondents in
neighborhoods that correspond to useable imagery from the
daytime and nighttime MASTER land cover dataset (n = 640
and 695, respectively) were used in the analyses (see
‘‘Geoprocessing: MASTER georectification and data
overlays’’ section)

123

750

published by the United States Environmental Protection Agency (2006, p. 11). Self-rated health questions
are widely used in the health sciences and other fields
because, over decades of research, they have been
found to be strong correlates of clinical conditions
(Lee 2015) and have been used in assessments of heat
vulnerability (Belanger et al. 2015).
In evaluating the effect of LST on respondents’
reported experiences with heat illness, we used a
question about economic impediments to home air
conditioning. Regulating indoor temperatures with air
conditioning is an important method of coping with
extreme outdoor temperatures in Phoenix and inability
to access this coping mechanism could increase the
likelihood of heat illness. We tested for a potentially
confounding effect of ambient outdoor temperature on
respondents’ answers to the heat illness question by
evaluating the influence of maximum air temperature
(obtained from Sky Harbor Airport) on the survey
response date. The Arizona State University Institutional Review Board approved the use of PASS data
for this study.
Geoprocessing: MASTER georectification
and data overlays
For this study, MASTER data were extracted for 41
(out of the total 45) PASS neighborhoods within the
flight boundary. The atmospherically corrected MASTER data were processed to at-sensor radiance but did
not include geolocation data. We georeferenced each
neighborhood to align with the corresponding
National Agriculture Imagery Program (NAIP) data.
This process included visually locating and matching
feature points in the MASTER image and the base
NAIP image. These feature points were used to create
a polynomial geometric transformation model that
aligned the MASTER image to the same coordinate
system of the base image. In addition we incorporated
a nearest neighbor resampling so that corresponding
pixels represent the same objects. For each neighborhood, a different number of matching points was used
depending on its size and level of distortion in the
MASTER data. All neighborhoods were georeferenced to a target registration error of no more than 0.5
pixels. In cases where severe distortions of the original
MASTER data did not allow registration to this level
of accuracy, registration to the lowest possible error
was performed. In four neighborhoods the daytime

123

Landscape Ecol (2016) 31:745–760

thermal images were too distorted for acceptable georeferencing; these highly distorted images were not
included in further analyses. Registration of the higher
resolution visible reflectance data precluded the use of
two more neighborhoods for NDVI or albedo data.
The final processed MASTER dataset included high
spatial resolution LST, NDVI, albedo, and direct
estimates of emissivity that were atmospherically
corrected, and georeferencing (Fig. 1).
We linked the MASTER and PASS data with a high
resolution (1 m) land cover classification for the
Phoenix metropolitan region, which had a classification accuracy of 92 % and a kappa statistic of 0.91 (Li
et al. 2014). This classification was conducted using
NAIP aerial photography and following an objectoriented expert rule algorithm (Baatz et al. 2008),
which is becoming increasingly common for urban
regions (Myint et al. 2011; Grove et al. 2014).
Preliminary analyses showed that 2.8 % of parcels
had no observed built land cover. The presence of trees
above the built surface may have obscured buildings
and there may also have been errors in the NAIP data
or land cover classification. We overlaid the land
cover classification and MASTER data with parcel
boundaries obtained from the Maricopa County Tax
Assessors office and extracted all residential plots
(Fig. 2). A complete geographic overlay of land cover,
LST, and parcel boundaries for 37 PASS neighborhoods during the day (n parcels = 21,589 and n
survey respondents = 640) and 41 neighborhoods at
night (n parcels = 25,011 and n survey respondents = 695) was available for our analyses.
Image acquisition occurred over approximately a
3-h flight period and therefore we evaluated the
potential for LST to increase during daytime sampling
and decrease during nighttime sampling. We compared means across all parcels in maximum, minimum, and mean LST. During the day we only
observed increasing LST in the minimum parcel
temperatures and, therefore, no correction was needed.
At night, we observed a significant cooling in the
maximum, minimum, and mean parcel temperatures
over the flight periods. We used linear regression to
identify the rate of cooling and then corrected all
parcel temperatures to the median sampling time
(approximately 1 am local time). Subsequent analyses
conducted using both time-corrected and uncorrected
data had consistent results for correlation analyses and
significance tests.

Landscape Ecol (2016) 31:745–760

751

Fig. 1 Representative MASTER LST measurements for a highly vegetated neighborhood (upper) and minimally vegetated
neighborhood (lower). MASTER bands 9, 5, and 3 were used for RGB display shown as false-color composites on the left

Analysis
We used a suite of standard statistical tools including
correlations, linear regressions, and generalized liner
models (GLM) for data analysis; all statistical analyses were conducted in Matlab 2013 (The Mathworks,
Natick, MA). For each individual residential parcel we
determined the means of LST, NDVI, and albedo and
proportions of each land cover type. We analyzed the
relationships between LST and the proportions of tree,
grass, built, and road area at the parcel scale using a
non-spatial linear correlation analysis. Because there
are many thousands of parcels, all results at the parcel
scale were statistically significant, consistent with
most ‘‘big-data’’ analyses and, therefore, we focused
on effect sizes in these analyses (Nuzzo 2014).
We also conducted analyses that compared
parcel variation in LST between different neighborhoods. We computed regressions between mean
parcel LST and NDVI within each single neighborhood in order to compare differences in vegetated cooling between neighborhoods. We
compared expected bare surface temperatures (i.e.
the intercepts of the linear regressions) with the
LST-cooling effectiveness of vegetation (i.e. the

slopes of the NDVI-LST regressions). This analysis
showed the relative sensitivity of LST to vegetation
abundance across neighborhoods.
We examined survey responses at the parcel scale
and compared mean parcel LST for the group of
respondents who reported symptoms of heat illness
with the group reporting no symptoms. To account for
differences in capacity to cope with heat, we restricted
some analyses to the subsample of respondents who
reported an impediment to using home air conditioning. These analyses were conducted using GLM for
binomial distributions. At the neighborhood scale we
used regression to evaluate the effects of mean
neighborhood LST and land cover characteristics on
aggregate measures (means) of respondents’ perceptions of the local landscape and experiences of heat
illness symptoms.

Results
Parcel land cover and LST relationships
Mean LST of individual residential parcels varied
tremendously: daytime parcels ranged between 35.1

123

752

Landscape Ecol (2016) 31:745–760

Fig. 2 Representative overlay of land cover data, parcel
boundaries, and MASTER LST measurements for a highly
vegetated neighborhood. The upper left and upper right show
day and night time LST (°C) and parcel boundaries. The lower

left shows land-cover classification and parcel boundaries. The
lower right shows a detail subset of the nighttime LST and
parcel delineations

and 67.4 °C and nighttime parcels ranged between
15.1 and 32.1 °C.
At the parcel scale, some measures of residential
land cover were correlated with LST. During the
daytime (Fig. 3), parcel-scale mean LST was inversely correlated with vegetation density. This finding
was supported across three different measures of
vegetation density: mean NDVI and the proportions of
tree or grass cover. Tree cover had the highest

correlation with LST (r = -0.44) and grass cover
had the lowest correlation (r = -0.27). Built surface
area, including roads, and albedo were weakly positively associated with mean daytime parcel LST. At
night (Fig. 4) many of the relationships were reversed.
Vegetation, quantified as NDVI, tree area, or grass
area, was weakly positively associated with nighttime
LST. As in the daytime, tree area had the highest of
these correlations at night, although the relationship

123

Landscape Ecol (2016) 31:745–760

753

65

65

35

r = -0.34

55

55

45

45

35

r = 0.05

r = -0.09

25

0.4

0.8

35
0

0.3

0.6

Albedo

NDVI
65

65

r = -0.27

r = -0.44

55

55

45

45

35
0

50

100

35
0

50

100

Grass Area (%)

Tree Area (%)
65

65

55

45

45

15
0

0.4

0.8

r = 0.21

50

Built Area (%)

100

35
0

50

0.3

35

35

r = 0.07

r = 0.20

25

25

15
0

0.6

Albedo

50

100

15
0

50

100

Grass Area (%)

Tree Area (%)
35

r = 0.06

r = -0.36

25

25

35
0

15
0

NDVI

35

r = 0.10

55

Night Land Surface Temperature (C)

Day Land Surface Temperature (C)

25

35
0

r = -0.23

100

Road (%)

Fig. 3 Parcel scale relationships between land cover characteristics and daytime land surface temperature. Pearson correlation coefficients are shown for all relationships

with LST was much weaker at night than in the day
(r = 0.20). Larger areas of built surfaces and higher
albedo were associated with cooler residential parcel
LST at night (r = -0.36 and -0.23, respectively).
Roads were minimally positively associated with
nighttime LST. For all of these daytime and nighttime
relationships there were large amounts of unexplained
variation in mean LST at the parcel scale.
Regressions of parcel daytime LST on NDVI
calculated independently for each neighborhood
revealed that vegetation had a significant cooling
effect within all neighborhoods. The magnitude of
these effects, however, varied tremendously between
neighborhoods. The slope magnitude between LST
and NDVI in Fig. 5 is a quantification of the
cooling effectiveness for differences in NDVI
between the neighborhoods. Notably, the slope was
strongly related to the regression intercept for each
neighborhood, such that vegetation was more effective at cooling hotter neighborhoods (p \ 0.001;
R2 = 0.54).

15
0

50

Built Area (%)

100

15
0

50

100

Road Area(%)

Fig. 4 Parcel scale relationships between land cover characteristics and nighttime land surface temperature. Pearson
correlation coefficients are shown for all relationships

Neighborhood land cover, LST and household
income
Consistent with previous findings in the Phoenix
region, more impoverished neighborhoods were hotter
(r = 0.47, p \ 0.01; proportion of neighborhood
households living below the federal poverty line and
mean residential parcel daytime LST) and had less
vegetation (r = -0.45; proportion of households in
poverty and proportional cover of both grass and trees
in parcels). Mean neighborhood daytime LST was
inversely correlated with vegetation (r = -0.37,
p \ 0.01; proportion of neighborhood tree and grass
area with mean daytime LST).
Micro-scale environments, residents’ perceptions,
and symptoms of heat illness
Mean daytime LST in the PASS neighborhoods
ranged from 40 to 55 °C, tree area varied from 1.2 to
26.8 %, and road area varied from 3.7 to 53 %. Survey

123

754

Landscape Ecol (2016) 31:745–760

Vegetation Cooling Effect (C)

5

-5

-15

-25

-35
46

50

54

58

62

Bare Temperature (C)

Fig. 5 Neighborhood variation in average parcel-scale daytime
vegetative cooling (quantified as the regression slope between
NDVI and LST) effect in relation to projected bare surface
temperature (regression p \ 0.001, R2 = 0.54). Each datum
represents a single neighborhood

respondents also varied widely in their perceptions of
neighborhood environmental conditions and they used
the full spectrum of available answers for the survey
questions (Table 2). Residents’ perceptions of temperature and land cover correlated with empirical
observations of these characteristics (Fig. 6). Residents’ mean perception of neighborhood temperature
relative to other neighborhoods on a five-point scale
(higher values indicate a hotter neighborhood) was
higher in neighborhoods with higher daytime LST
(p = 0.002, R2 = 0.23). No relationship was detected
between perception of relative neighborhood temperature and nighttime LST (p [ 0.3). Perceptions of
drivers of LST were also consistent with observed
data. Residents’ mean satisfaction with shade trees in
their neighborhood on a four-point scale (higher
values indicate more satisfaction) was positively
related to percentage of neighborhood area covered
by trees (p = 0.006, R2 = 0.18). Residents’ mean
perception of the magnitude of paved surfaces as a
neighborhood problem on a three-point scale (higher
values indicate smaller problem, thus higher satisfaction) was negatively related to road area (p = 0.002,
R2 = 0.23).
The frequency of self-reported heat- related symptoms was correlated with daytime LST at both the
parcel and neighborhood scales. The mean parcel
daytime LST was significantly higher (0.93 °C,
p \ 0.001) for respondents who reported a heat-

123

Fig. 6 PASS respondents’ perceptions of their neighborhoods
and neighborhood conditions observed by MASTER. Upper
panel shows mean perception of relative neighborhood temperature and observed daytime neighborhood mean temperature
(p = 0.002; R2 = 0.23). Middle panel shows mean satisfaction
with neighborhood trees and neighborhood tree area (p \ 0.006;
R2 = 0.18). Lower panel shows mean satisfaction with amount
of roads and paved area (p = 0.002; R2 = 0.23). For all panels
the scale of y-axis represents range of possible survey answers.
Each datum represents a single neighborhood

related illness in their household than for those
reporting no heat related illnesses (Table 3). Furthermore, within the subgroup of respondents who had
impediments to using home air conditioning, the
parcel daytime LST difference between those who
reported heat illness and those who did not was more
than 50 % higher (1.52 °C, p \ 0.001) than in the
whole sample. This means the effect of parcel daytime
surface temperature on heat illness was greater for
people with reduced access to home air conditioning.
The relationship between parcel nighttime LST and
reported heat illness was not statistically significant
for the whole sample or for the subgroup with
restricted access to air conditioning. Daytime maximum air temperature on the day of survey completion
did not have a statistically significant effect on the
likelihood of reporting heat symptoms, indicating that
it did not affect the results of our study.
At the neighborhood scale, the incidence of experiencing heat illness varied between 5 and 53 % of
residents. An exponential increase in the rate of

Landscape Ecol (2016) 31:745–760

755

Table 3 Differences in parcel land surface and air temperature (°C) between groups of PASS respondents who reported yes or no for
experiencing household heat-related illnesses in the summer of 2010
Temperature metric (°C)

All respondents

Reduced access to air
conditioning

No heat illness
symptoms

Heat illness
symptoms

GLM p
value

Daytime LST mean

49.05 (3.71)

49.98 (3.83)

p \ 0.001

Nighttime LST mean

23.95 (1.66)

23.84 (1.71)

p = 0.107

Maximum Tair day of
survey

35.98 (9.23)

35.12 (9.85)

p = 0.258

Daytime LST mean

48.60 (3.67)

50.12 (4.15)

p \ 0.001

Nighttime LST mean

23.92 (1.77)

23.92 (1.67)

p = 0.197

Maximum Tair day of
survey

35.75 (8.91)

37.28 (8.25)

p = 0.491

Values in cells represent means (standard deviations) of temperature for the individual parcels of respondents during daytime, means
(standard deviations) of temperature for the individual parcels of respondents during the nighttime, and the maximum air temperature
at Sky Harbor Airport on the day of survey response. The upper panel represents the 2011 PASS sample. The lower panel represents
PASS respondents who reported an impediment to using air conditioning in their home (too expensive to operate, broken unit, etc.).
Statistical significance was assessed through a GLM for binomial distribution

evaluation of land cover composition and LST variation, our results showed large differences in day and
night LST accompanied by moderate couplings of
LST with some land cover characteristics that differed
between day and night. There was consistency
between residents’ perceptions of their neighborhood
environments and empirically measured temperature
and land covers. Using self-reported symptoms of heat
illnesses, we found heat-related health impacts in the
Phoenix region were positively associated with daytime LST of the respondent’s residential parcel and
with neighborhood daytime LST.
Fig. 7 Proportion of PASS respondents reporting someone in
their household had suffered from symptoms related to heat
illness during summer and corresponding neighborhood mean
daytime LST (exponential regression; p = 0.011; R2 = 0.42).
Each datum represents a single neighborhood

reported heat-related illness was associated with
neighborhood daytime LST (Fig. 7, p = 0.011;
R2 = 0.42). Variation in nighttime LST was not
correlated with reported symptoms of heat-related
illness at the neighborhood scale (p [ 0.1).

Discussion
Fine resolution LST data coupled with comprehensive
land cover maps, household surveys, and residential
parcel delineations provide a robust description of the
biophysical environment of homes and experiences
with heat of the occupants. In this first parcel scale

Improving understanding of parcel
and neighborhood LST variation
Both daytime and nighttime surface temperatures
were influenced by characteristics of the vegetated and
built land covers at the parcel scale, which is
consistent with previous findings at neighborhood
scales (Weng et al. 2011; Myint et al. 2013). Of
importance for designing heat mitigation strategies,
the magnitude and direction of the relationships
differed between day and night. Strategies to reduce
day and night LST may differ in their effectiveness or
even be in opposition. The hypothesis that vegetationbased LST cooling during the day is dominated by
evaporative energy partitioning was supported by
consistent results for all the greenness metrics—NDVI
mean and percentages of tree cover or grass cover.
Among these different vegetation variables, tree cover

123

756

was the strongest correlate, which supports previous
findings that grass is less effective than tree canopy for
LST cooling (Myint et al. 2013), likely because trees
with deeper roots have more consistent access to water
and thereby maintain more cooling capacity. Our
finding of a small positive nighttime parcel vegetationLST relationship contrasts with previous results from
Phoenix conducted at much larger (90 m pixels) scales
(Buyantuyev and Wu 2010; Myint et al. 2013), which
showed negative relationships between vegetation and
LST during both the day and night. Also in contrast to
regional UHI studies in this hot semi-arid environment
(Myint et al. 2015), parcel scale built land covers were
negatively related to nighttime LST. These reversals
of daytime patterns at night suggests that at the very
local scale, vegetation traps heat and building roofs
more rapidly lose heat at night. Resolving these scale
differences are an important research need. The
negative relationship between albedo and nighttime
LST and a smaller cooling effect of albedo throughout
the day is consistent with an explanation that higher
albedo surfaces have reduced energy storage and are
therefore cooler early in the night. For parcel-scale
land surface cooling, our results support both the use
of vegetation and high albedo (e.g. white roofing)
building materials (Georgescu et al. 2014). Because of
generally stronger land cover—LST relationships
during the day, peak temperatures may be more easily
managed than nighttime minimum temperatures
through landscape modification.
Notably, the effectiveness of vegetation (measured
either as NDVI or vegetated area) on moderating
temperatures varied substantially across neighborhoods. Although prior work has shown a cooling
capacity for vegetation, it is generally believed that
this capacity is regulated primarily by the area of
vegetation (e.g. Jenerette et al. 2011; Li et al. 2011).
Our findings suggest that, in addition to the area of
vegetation, other factors influence the cooling capacity
of vegetation, which we quantified as slope of LSTNDVI relationship within a neighborhood. We showed
that vegetation cooling effectiveness at the parcel
scale was correlated with expected bare surface
temperature, which is consistent with prior neighborhood analyses examining patterns across seasons
(Jenerette et al. 2011; Zhou et al. 2014). These
findings suggest increased transpiration rates associated with hotter environments may lead to enhanced
cooling. However, the effectiveness of vegetation may

123

Landscape Ecol (2016) 31:745–760

also be regulated by other factors we did not examine,
including irrigation (Jenerette et al. 2011), landscape
configuration (Connors et al. 2013; Zheng et al. 2014;
Zhou et al. 2014), and plant species characteristics
(Lundholm et al. 2010; McCarthy et al. 2011).
Strategies for reducing LST through increased planting should be targeted toward areas with the highest
temperatures because in these locations mitigation is
likely to be more effective than planting in areas with
lower LST.
Connecting environment with heat vulnerability
Daytime LST at neighborhood and parcel scales were
correlated with residents’ perceptions of the heatrelated risk factors and heat-related symptoms of
illness. Residents recognized regional variation in
temperature and were similarly aware of important
drivers of local LST including tree and road cover area
in their surroundings. Such consistencies between
perceptions and environment were previously suggested by comparisons with modeled air temperature
data (Ruddell et al. 2012), although we present the first
data-based comparison of perception-environment linkages for LST urban microclimates. The consistency
between perceptions of factors associated with heatrelated health risks and observed patterns of these
drivers suggest there is value in consulting residents
about their views of effective heat mitigation strategies.
Critical for understanding residents’ heat vulnerability, we showed that both parcel and neighborhood
daytime LSTs were correlated with self-reported
symptoms of heat-related illnesses. Symptoms are
markers of specific heat-health effects that provide
more information about the broader health impacts of
heat than can be gleaned from retrospective analysis of
mortality and morbidity records. The incidence of selfreported heat illness symptoms in our sample greatly
exceeded the frequencies of deaths and hospitalizations attributed to heat in the Phoenix region (Petitti
et al. 2015).
Heat-related illnesses are associated with a suite of
individual, social, and environmental factors that are
themselves strongly interrelated (Reid et al. 2009;
Hondula et al. 2012; Harlan et al. 2013; Petitti et al.
2015). The causal mechanisms leading to heat illnesses and deaths are not fully understood, although
research is making progress in explaining linkages
between spatial variations in living conditions and

Landscape Ecol (2016) 31:745–760

individual risks. For example, poverty is associated
with chronic health problems that are exacerbated by
heat (Balbus and Malina 2009), reduced coping
capacities that include access to air conditioning (Reid
et al. 2009), and elevated risks from working in higher
temperature conditions (Petitti et al. 2013).
At present the correlations between LST and heat
symptoms do not imply direct causation of heat illness
but our study clearly shows that daytime LST can be
one spatial indicator of heat risk. This was borne out in
our analysis showing positive relationships between
reported heat symptoms and daytime parcel and
neighborhood LST, as well as a much stronger positive
relationship between daytime parcel LST and symptoms of heat illness in the subsample of residents
without consistent access to air conditioning (the most
vulnerable subpopulation) compared to the entire
population. Perhaps surprisingly, no similar correlations were observed between heat stress and nighttime
LST in this setting.
Comparisons between LST and survey respondents’ lived experiences were consistent with several
of the few existing neighborhood scale studies,
including Harlan et al. (2013) analysis of Phoenix,
which showed an association between spatial variability in daytime LST and heat-related deaths. The
present study also examined, but did not find, a
relationship in this setting between mean parcel
nighttime LST in mid-summer and the likelihood that
survey respondents reported experiences with symptoms of heat illness. Our finding differs from a coarserscale (1 km LST data) study in Paris, which found
nighttime (minimum) LST was significantly associated with elderly heat deaths but there was no
relationship between deaths and daytime (maximum)
LST (Laaidi et al. 2012). More comparative research
on spatial and temporal variability in surface temperatures and heat-related health outcomes in urban areas
is needed.
Resolving the influence of daytime and nighttime
temperature variations on health impacts at appropriate spatial and temporal scales is important for
identifying effective landscape interventions that
reduce health vulnerability. Yet a limitation of all
temperature-related health studies, including our own,
is the inability to make direct mechanistic links
between the temperature metric, the time, and the site
of where a heat-related health incident occurred. Some
of the symptoms reported in our study undoubtedly

757

occurred outside the home and neighborhood. Future
work directed to individual microclimate monitoring
(Kuras et al. 2015) will greatly improve the ability to
make such temporal and spatial linkages.
Looking towards more heat resilient cities
Our findings highlight the potential value for microscale landscape interventions to reduce the heat
exposure risk and vulnerability of urban residents.
To meet current and future challenges of reducing
urban heat vulnerability, cities are using four main
strategies to manage heat risks: trees and vegetation,
green roofs, cool roofs, and cool pavement (United
States Environmental Protection Agency 2008).
Georgescu et al. (2014) concluded from their model
of the regional climate in Phoenix that more cool
roofing materials would have a greater impact than
more vegetation on decreasing regional air temperatures. However, for reducing health risks associated
with higher LST, mitigating daytime LST in residential areas may be more effective than lowering
nighttime air temperature in hot climates like Phoenix.
At microscales, increasing greening in pedestrian
pathways and outdoor play spaces may be especially
useful (Vanos 2015). In comparing the relative cooling
effectiveness of vegetation and albedo, our findings
support increasing urban greening, especially by
planting trees around homes with a large fraction of
human-built land surfaces. Because the cooling
effectiveness of increasing albedo was observed
primarily at night, this strategy may not be as valuable
for reducing heat-related illnesses. However, as trees
in hot arid cities require irrigation, balancing the
cooling benefits with water use is an essential
sustainability challenge (Gober et al. 2010; Jenerette
et al. 2011; Pataki et al. 2011). Increasing albedo does
not have an associated water consumption trade-off
and could more readily be implemented anywhere.
Reconciling the effectiveness of alternate mitigation strategies is an important challenge for reducing
urban vulnerability to climate changes (Liang and
Weng 2008; Georgescu et al. 2014). Current global
demographic trends projecting 2.5 billion more urban
residents by 2050, a 64 % increase from current
distributions (United Nations Department of Economic and Social Affairs Population Division 2014),
will place more people’s health at risk from extreme
heat due to combinations of global warming

123

758

associated with greenhouse gas emissions and regional
warming associated with UHIs (Intergovernmental
Panel on Climate Change 2014). Interdisciplinary
approaches such as this one, which geographically
overlay data describing high resolution environmental
and human variation at the scale of individual homes,
provide a powerful lens for understanding and designing the coupled natural and human systems that
contribute to reducing heat vulnerability.
Acknowledgments This work was supported by National
Science Foundation Grants GEO-0816168, GEO-0814692,
BCS-1026865, EF-1049251, EF-1049224, and DEB-0919006.
We thank Anthony Brazel and Chris Martin for their advice on
the MASTER data collection effort and David Hondoula for
insightful discussions. All data are available from CAP-LTER
(caplter.asu.edu).

References
American Association for Public Opinion Research (2008)
Standard definitions: final dispositions of case codes and
outcome rates for surveys (revised). http://www.aapor.org/
AAPORKentico/Education-Resources/For-Researchers/
Poll-Survey-FAQ/Response-Rates-An-Overview.aspx
Baatz M, Hoffmann C, Willhauck G (2008) Progressing from
object-based to object-oriented image analysis. In: Blaschke T, Lang S, Hay GJ (eds) Object-based image analysis:
spatial concepts for knowledge-driven remote sensing
applications. Springer, Berlin, pp 29–42
Balbus JM, Malina C (2009) Identifying vulnerable subpopulations for climate change health effects in the United
States. J Occup Environ Med 51:33–37
Basu R, Samet JM (2002) Relation between elevated ambient
temperature and mortality: a review of the epidemiologic
evidence. Epidemiol Rev 24:190–202
Belanger D, Gosselin P, Valois P, Abdous B (2015) Neighbourhood and dwelling characteristics associated with the
self-reported adverse health effects of heat in most
deprived urban areas: a cross-sectional study in 9 cities.
Health Place 32:8–18
Bernstein LS, Adler-Golden SM, Sundberg RL, Levine RY,
Perkins TC, Berk A, Ratkowski AJ, Felde G, Hoke ML
(2005) Validation of the quick atmospheric correction
(QUAC) algorithm for VNIR-SWIR multi- and hyperspectral imagery. In: SPIE proceedings, algorithms and
technologies for multispectral, hyperspectral, and ultraspectral imagery XI, pp 668–678
Buscail C, Upegui E, Viel JF (2012) Mapping heatwave health
risk at the community level for public health action. Int J
Health Geogr. doi:10.1186/1476-1072x-1111-1138
Buyantuyev A, Wu JG (2010) Urban heat islands and landscape
heterogeneity: linking spatiotemporal variations in surface
temperatures to land-cover and socioeconomic patterns.
Landscape Ecol 25:17–33

123

Landscape Ecol (2016) 31:745–760
Cao X, Onishi A, Chen J, Imura H (2010) Quantifying the cool
island intensity of urban parks using ASTER and IKONOS
data. Landsc Urban Plan 96:224–231
Chow WTL, Brennan D, Brazel AJ (2012) Urban heat island
research in Phoenix, Arizona: theoretical Contributions and
policy applications. Bull Am Meteorol Soc 93:517–530
Connors JP, Galletti CS, Chow WTL (2013) Landscape configuration and urban heat island effects: assessing the
relationship between landscape characteristics and land
surface temperature in Phoenix, Arizona. Landscape Ecol
28:271–283
Deng CB, Wu CS (2013) Examining the impacts of urban biophysical compositions on surface urban heat island: a
spectral unmixing and thermal mixing approach. Remote
Sens Environ 131:262–274
Elachi C (1987) Introduction to the physics and techniques of
remote sensing. Wiley, New York
Georgescu M, Morefield PE, Bierwagen BG, Weaver CP (2014)
Urban adaptation can roll back warming of emerging
megapolitan regions. Proc Natl Acad Sci USA
111:2909–2914
Gober P, Brazel A, Quay R, Myint S, Grossman-Clarke S, Miller
A, Rossi S (2010) Using watered landscapes to manipulate
urban heat island effects: how much water will it take to
cool Phoenix? J Am Plan Assoc 76:109–121
Goldreich Y (2006) Ground and top of canopy layer urban heat
island partitioning on an airborne image. Remote Sens
Environ 104:247–255
Grimm NB, Redman CL (2004) Approaches to the study of
urban ecosystems: the case study of Central Arizona—
Phoenix. Urban Ecosyst 7:199–213
Grove JM, Locke DH, O’Neil-Dunne JPM (2014) An ecology of
prestige in New York City: examining the relationships
among population density, socio-economic status, group
identity, and residential canopy cover. Environ Manag
54:402–419
Harlan SL, Declet-Barreto JH, Stefanov WL, Petitti DB (2013)
Neighborhood effects on heat deaths: social and environmental predictors of vulnerability in Maricopa County,
Arizona. Environ Health Perspect 121:197–204
Hondula DM, Barnett AG (2014) Heat-related morbidity in
Brisbane, Australia: spatial variation and area-level predictors. Environ Health Perspect 122:831–836
Hondula DM, Davis RE, Leisten MJ, Saha MV, Veazey LM,
Wegner CR (2012) Fine-scale spatial variability of heatrelated mortality in Philadelphia County, USA, from
1983-2008: a case-series analysis. Environ Health. doi:10.
1186/1476-1069x-1111-1116
Hook SJ, Myers JEJ, Thome KJ, Fitzgerald M, Kahle AB (2001)
The MODIS/ASTER airborne simulator (MASTER)—a
new instrument for earth science studies. Remote Sens
Environ 76:93–102
Imhoff ML, Zhang P, Wolfe RE, Bounoua L (2010) Remote
sensing of the urban heat island effect across biomes in the
continental USA. Remote Sens Environ 114:504–513
Intergovernmental Panel on Climate Change (2014) Climate
change 2014: impacts, adaptation, and vulnerability.
Cambridge University Press, Cambridge
Jenerette GD, Harlan SL, Brazel A, Jones N, Larsen L, Stefanov
WL (2007) Regional relationships between surface

Landscape Ecol (2016) 31:745–760
temperature, vegetation, and human settlement in a rapidly
urbanizing ecosystem. Landscape Ecol 22:353–365
Jenerette GD, Harlan SL, Stefanov WL, Martin CA (2011)
Ecosystem services and urban heat riskscape moderation:
water, green spaces, and social inequality in Phoenix, USA.
Ecol Appl 21:2637–2651
Jenerette GD, Miller G, Buyantuev A, Pataki DE, Gillespie T,
Pincetl S (2013) Urban vegetation and income segregation
in drylands: a synthesis of seven metropolitan regions in
the southwestern United States. Environ Res Lett 8:044001
Johnson BR, Young SJ (1998) In-scene atmospheric compensation: application to SEBASS data collected at the ARM
site. The Aerospace Corporation
Johnson DP, Wilson JS, Luber GC (2009) Socioeconomic
indicators of heat-related health risk supplemented with
remotely sensed data. Int J Health Geogr. doi:10.1186/
1476-1072x-1188-1157
Johnson DP, Stanforth A, Lulla V, Luber G (2012) Developing
an applied extreme heat vulnerability index utilizing
socioeconomic and environmental data. Appl Geogr
35:23–31
Kalkstein LS (1991) A new approach to evaluate the impact of
climate on human mortality. Environ Health Perspect
96:145–150
Kalma JD, McVicar TR, McCabe MF (2008) Estimating land
surface evaporation: a review of methods using remotely
sensed surface temperature data. Surv Geophys 29:421–469
Kealy PS, Hook SJ (1993) Separating temperature and emissivity in thermal infrared multispectral scanner data—Implications for recovering land-surface temperatures. IEEE
Trans Geosci Remote Sens 31:1155–1164
Kuras ER, Hondula DM, Brown-Saracino J (2015) Heterogeneity in individually experienced temperatures (IETs)
within an urban neighborhood: insights from a new
approach to measuring heat exposure. Int J Biometeorol.
doi:10.1007/s00484-00014-00946-x
Kustas W, Anderson M (2009) Advances in thermal infrared
remote sensing for land surface modeling. Agric For
Meteorol 149:2071–2081
Laaidi K, Zeghnoun A, Dousset B, Bretin P, Vandentorren S,
Giraudet E, Beaudeau P (2012) The impact of heat islands
on mortality in Paris during the August 2003 heat wave.
Environ Health Perspect 120:254–259
Lee S (2015) Self-rated health in health surveys. In: Johnson TP
(ed) Handbook of health survey methods. Wiley, Hoboken
Li JX, Song CH, Cao L, Zhu FG, Meng XL, Wu JG (2011)
Impacts of landscape structure on surface urban heat
islands: a case study of Shanghai, China. Remote Sens
Environ 115:3249–3263
Li X, Myint SW, Zhang Y, Galletti C, Zhang X, Turner Ii BL
(2014) Object-based land-cover classification for
metropolitan Phoenix, Arizona, using aerial photography.
Int J Appl Earth Obs Geoinf 33:321–330
Liang SL (2001) Narrowband to broadband conversions of land
surface albedo I: algorithms. Remote Sens Environ 76:
213–238
Liang BQ, Weng QH (2008) Multiscale analysis of censusbased land surface temperature variations and determinants
in Indianapolis, United States. J Urban Plan Dev 134:
129–139

759
Lundholm J, MacIvor JS, MacDougall Z, Ranalli M (2010)
Plant species and functional group combinations affect
green roof ecosystem functions. PLoS One. doi:10.1371/
journal.pone.0009677
McCarthy HR, Pataki DE, Jenerette GD (2011) Plant water-use
efficiency as a metric of urban ecosystem services. Ecol
Appl 21:3115–3127
Myint SW, Gober P, Brazel A, Grossman-Clarke S, Weng QH
(2011) Per-pixel vs. object-based classification of urban
land cover extraction using high spatial resolution imagery.
Remote Sens Environ 115:1145–1161
Myint SW, Wentz EA, Brazel AJ, Quattrochi DA (2013) The
impact of distinct anthropogenic and vegetation features on
urban warming. Landscape Ecol 28:959–978
Myint SW, Zheng B, Talen E, Fan C, Kaplan S, Middel A, Smith
M, Huang HP, Brazel A (2015) Does the spatial arrangement of urban landscape matter? Examples of urban
warming and cooling in Phoenix and Las Vegas. Ecosystem Health and Sustainability 1:art15
Nuzzo R (2014) Statistical errors. Nature 506:150–152
Oke TR (1973) City size and urban heat island. Atmos Environ
7:769–779
Pataki DE, Boone CG, Hogue TS, Jenerette GD, McFadden JP,
Pincetl S (2011) Ecohydrology bearings-invited commentary socio-ecohydrology and the urban water challenge.
Ecohydrology 4:341–347
Peng SS, Piao SL, Ciais P, Friedlingstein P, Ottle C, Breon FM,
Nan HJ, Zhou LM, Myneni RB (2012) Surface urban heat
island across 419 global big cities. Environ Sci Technol
46:696–703
Petitti DB, Harlan SL, Chowell-Puente G, Ruddell D (2013)
Occupation and environmental heat-associated deaths in
Maricopa County, Arizona: a case-control study. PLoS
One 8:e62596
Petitti DB, Hondula DM, Yang S, Harlan SL, Chowell G (2015)
Multiple trigger points for quantifying heat-health impacts:
new evidence from a hot climate. Environ Health Perspect.
doi:10.1289/ehp.1409119
Reid CE, O’Neill MS, Gronlund CJ, Brines SJ, Brown DG,
Diez-Roux AV, Schwartz J (2009) Mapping community
determinants of heat vulnerability. Environ Health Perspect 117:1730–1736
Ruddell D, Harlan SL, Grossman-Clarke S, Chowell G (2012)
Scales of perception: public awareness of regional and
neighborhood climates. Clim Change 111:581–607
Small C (2003) High spatial resolution spectral mixture analysis
of urban reflectance. Remote Sens Environ 88:170–186
Song J, Du S, Feng X, Guo L (2014) The relationships between
landscape compositions and land surface temperature:
quantifying their resolution sensitivity with spatial
regression models. Landsc Urban Plan 123:145–157
Stone B Jr (2012) The city and the coming climate: climate
change in the places we live. Cambridge University Press,
New York
Tucker CJ (1979) Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens Environ
8:127–150
Uejio CK, Wilhelmi OV, Golden JS, Mills DM, Gulino SP,
Samenow JP (2011) Intra-urban societal vulnerability to
extreme heat: the role of heat exposure and the built

123

760
environment, socioeconomics, and neighborhood stability.
Health Place 17:498–507
United States Environmental Protection Agency (2006) Excessive heat events guidebook. EPA 430-B-06-005 (June).
http://www.epa.gov/heatisland/about/pdf/EHEguide_final.
pdf
United States Environmental Protection Agency (2008) Chapter 1 Heat island basics in reducing urban heat islands:
compendium of strategies. http://www.epa.gov/heatisland/
resources/compendium.htm
United Nations Department of Economic and Social Affairs
Population Division (2014) World urbanization prospects:
the 2014 revision, highlights (ST/ESA/SER.A/352)
Vanos JK (2015) Children’s health and vulnerability in outdoor
microclimates: a comprehensive review. Environ Int
76:1–15
Voogt JA, Oke TR (2003) Thermal remote sensing of urban
climates. Remote Sens Environ 86:370–384
Weng QH, Rajasekar U, Hu XF (2011) Modeling urban heat
islands and their relationship with impervious surface and
vegetation abundance by using ASTER images. IEEE
Trans Geosci Remote Sens 49:4080–4089

123

Landscape Ecol (2016) 31:745–760
Xiang TT, Vivoni ER, Gochis DJ (2014) Seasonal evolution of
ecohydrological controls on land surface temperature over
complex terrain. Water Resour Res 50:3852–3874
Zhan WF, Chen YH, Zhou J, Wang JF, Liu WY, Voogt J, Zhu
XL, Quan JL, Li J (2013) Disaggregation of remotely sensed
land surface temperature: literature survey, taxonomy,
issues, and caveats. Remote Sens Environ 131:119–139
Zheng B, Myint SW, Fan C (2014) Spatial configuration of
anthropogenic land cover impacts on urban warming.
Landsc Urban Plan 103:104–111
Zhou WQ, Huang GL, Cadenasso ML (2011) Does spatial
configuration matter? Understanding the effects of land
cover pattern on land surface temperature in urban landscapes. Landsc Urban Plan 102:54–63
Zhou WQ, Qian YG, Li XM, Li WF, Han LJ (2014) Relationships between land cover and the surface urban heat island:
seasonal variability and effects of spatial and thematic
resolution of land cover data on predicting land surface
temperatures. Landscape Ecol 29:153–167