Towards More Comprehensive Projections of Urban Heat-Related Mortality: Estimates for New York City under Multiple Population, Adaptation, and Climate Scenarios

Abstract

High temperatures have substantial impacts on mortality and, with growing concerns about climate change, numerous studies have developed projections of future heat-related deaths around the world. Projections of temperature-related mortality are often limited by insufficient information necessary to formulate hypotheses about population sensitivity to high temperatures and future demographics.

This study has derived projections of temperature-related mortality in New York City by taking into account future patterns of adaptation or demographic change, both of which can have profound influences on future health burdens.

We adopt a novel approach to modeling heat adaptation by incorporating an analysis of the observed population response to heat in New York City over the course of eight decades. This approach projects heat-related mortality until the end of the 21st century based on observed trends in adaptation over a substantial portion of the 20th century. In addition, we incorporate a range of new scenarios for population change until the end of the 21st century. We then estimate future heat-related deaths in New York City by combining the changing temperature-mortality relationship and population scenarios with downscaled temperature projections from the 33 global climate models (GCMs) and two Representative Concentration Pathways (RCPs).

The median number of projected annual heat-related deaths across the 33 GCMs varied greatly by RCP and adaptation and population change scenario, ranging from 167 to 3331 in the 2080s compared to 638 heat-related deaths annually between 2000 and 2006.

These findings provide a more complete picture of the range of potential future heat-related mortality risks across the 21(st) century in New York, and highlight the importance of both demographic change and adaptation responses in modifying future risks.

Full text

Environmental Health Perspectives
•
volume 125 | number 1 | January 2017
47
Research
A Sectio n 508–conformant H TML version of this arti cle
is available at http://dx.doi.org/10.1289/EHP166.
Introduction
High temperatures have long been recognized
to have substantial impacts on mortality, and
with growing concerns about climate change,
numerous studies have projected future heat-
related mortality due to climate change in
recent years (Baccini et al. 2011; Dessai 2003;
Gosling et al. 2009; Hayhoe et al. 2004, 2010;
Jackson et al. 2010; Knowlton et al. 2007;
Ostro et al. 2012; Peng et al. 2011; Sheridan
et al. 2012). Some studies have characterized
the relationships between temperature and
mortality over the full temperature spectrum
at a given location in order to estimate the
current and future “net impact” of tempera-
ture on mortality (Doyon et al. 2008; Guest
et al. 1999; Li T et al. 2013; Martens 1998;
Martin et al. 2012). We chose to focus here
on heat-related mortality because adapta-
tion responses to cold would likely be quite
different, and to date, adaptation responses
to cold have not been as thoroughly studied
as those to heat. In addition, previous work
in New York City (New York) suggested that
increases in heat-related mortality are likely
to be substantial and may not be offset by
decreases in cold-related mortality (Li T
et al. 2013).
Projections of temperature-related
mortality are, unfortunately, often limited
by insufficient understanding of the popula-
tion adaptation to heat. To date, relatively
few heat-health impact studies have consid-
ered future adaptation. Some studies have
used temperature–mortality curves from
“analogue cities” that currently experience
temperatures similar to those projected to
occur in the future at a location of interest
(Kalkstein and Green 1997; Knowlton et al.
2007) or temperature–mortality curves from
hotter “analogue summers” that have previ-
ously occurred in the same location (Hayhoe
et al. 2004). Other studies have developed
scenarios for acclimatization to specific
increases in temperatures over time (Dessai
2003; Gosling et al. 2009; Kalkstein and
Green 1997). However, to our knowledge,
no previous studies have quantified future
adaptation trends based on historical patterns
of adaptation in the city under study.
An important question to consider is
whether the future population response to
high temperatures should be projected based
on observations from the present and/or the
recent past. Cities are complex adaptive systems
(Holland 1995; Lansing 2003) capable of self-
organizing in order to adapt to environmental
conditions. At the same time, there are limits to
social adaptation (Dow et al. 2013) that are yet
to be well understood and quantified.
In addition to future changes in climate
and population adaptation to heat, future
demographics are important determinants of
health impacts (Huang et al. 2011). Utilizing
multiple population change scenarios is also
important for quantifying the range and
uncertainty of potential temperature-related
health impacts.
We start by developing heat adaptation
models that project the population response
to heat until the year 2100 based on eight
decades of historical daily temperature and
mortality data. The approach is particularly
suitable for New York City, which is among
Address correspondence to E.P. Petkova, National
Center for Disaster Preparedness, Earth Institute,
Columbia University, 215 W. 125th St., New
York, NY 10027 USA. Telephone: (646) 845-2325.
E-mail: epp2109@columbia.edu
Supplemental Material is available online (http://
dx.doi.org/10.1289/EHP166).
is work was supported by the Consortium for
Climate Risk in the Urban Northeast (CCRUN), the
National Center for Disaster Preparedness (NCDP;
E.P.P.), National Institutes of Health/National
Institute of Environmental Health Sciences (NIEHS
Center grant ES009089; E.P.P. and P.L.K.), and
a Methodology Research fellowship from Medical
Research Council (grant MR/M022625/1; A.G.).
e authors declare they have no actual or potential
competing financial interests.
Received: 12 October 2015; Revised: 16 January
2016; Accepted: 13 May 2016; Published: 23 June
2016.
Note to readers with disabilities: EHP strives
to ensure that all journal content is accessible to all
readers. However, some figures and Supplemental
Material published in EHP articles may not conform to
508 standards due to the complexity of the information
being presented. If you need assistance accessing journal
content, please contact ehponline@niehs.nih.gov.
Our staff will work with you to assess and meet your
accessibility needs within 3 working days.
Towards More Comprehensive Projections of Urban Heat-Related Mortality:
Estimates for New York City under Multiple Population, Adaptation, and
Climate Scenarios
Elisaveta P. Petkova,1 Jan K. Vink,2 Radley M. Horton,3 Antonio Gasparrini,4,5 Daniel A. Bader,3 Joe D. Francis,2
and Patrick L. Kinney6
1National Center for Disaster Preparedness, Earth Institute, Columbia University, New York, New York, USA; 2Cornell Program on
Applied Demographics, Cornell University, Ithaca, New York, USA; 3Center for Climate Systems Research, Columbia University, New
York, New York, USA; 4Department of Social and Environmental Health Research, and 5Department of Medical Statistics, London
School of Hygiene & Tropical Medicine, London, UK; 6Department of Environmental Health Sciences, Mailman School of Public Health,
Columbia University, New York, New York, USA
Back gro und : High temperatures have substantial impacts on mortality and, with growing
concerns about climate change, numerous studies have developed projections of future heat-
related deaths around the world. Projections of temperature-related mortality are often limited by
insufficient information to formulate hypotheses about population sensitivity to high temperatures
and future demographics.
oBjectives: e present study derived projections of temperature-related mortality in New York
City by taking into account future patterns of adaptation or demographic change, both of which
can have profound influences on future health burdens.
Methods: We adopted a novel approach to modeling heat adaptation by incorporating an analysis
of the observed population response to heat in New York City over the course of eight decades. is
approach projected heat-related mortality until the end of the 21st century based on observed trends
in adaptation over a substantial portion of the 20th century. In addition, we incorporated a range
of new scenarios for population change until the end of the 21st century. We then estimated future
heat-related deaths in New York City by combining the changing temperature–mortality relation-
ship and population scenarios with downscaled temperature projections from the 33 global climate
models (GCMs) and two Representative Concentration Pathways (RCPs).
results: e median number of projected annual heat-related deaths across the 33 GCMs varied
greatly by RCP and adaptation and population change scenario, ranging from 167 to 3,331 in the
2080s compared with 638 heat-related deaths annually between 2000 and 2006.
conclusions: ese findings provide a more complete picture of the range of potential future
heat-related mortality risks across the 21st century in New York City, and they highlight the
importance of both demographic change and adaptation responses in modifying future risks.
citation: Petkova EP, Vink JK, Horton RM, Gasparrini A, Bader DA, Francis JD, Kinney PL.
2017. Towards more comprehensive projections of urban heat-related mortality: estimates for New
York City under multiple population, adaptation, and climate scenarios. Environ Health Perspect
125:47–55; http://dx.doi.org/10.1289/EHP166

Petkova et al.
48
volume 125 | number 1 | January 2017
•
Environmental Health Perspectives
the largest cities in the world and has retained
a relatively consistent urban shape over the
entire historical period covered by this study.
We then develop demographic scenarios that
characterize potential changes in the city
population during the study period. Finally,
we calculate future heat-related deaths by
combining the derived temperature–mortality
relationships and population scenarios with
the downscaled temperature projections from
the 33 global climate models (GCMs) and the
two Representative Concentration Pathways
(RCPs), RCP4.5 and RCP8.5, developed in
support of the Intergovernmental Panel on
Climate Change (IPCC)’s Fifth Assessment
Report (AR5) (IPCC 2013).
Methods
Daily Mortality Data
e process of the historical daily mortality
data preparation and validation has been
discussed in detail previously (Petkova et al.
2014). Death records prior to 1949 are stored
at the New York City Department of Records
and Information Services (DORIS 2016).
Death indexes for all years between 1900 and
1949, including each documented death in
the five New York City boroughs (Bronx,
Brooklyn, Manhattan, Queens, and Staten
Island) from 1900 to 1948, were scanned
by the Genealogy Federation of Long Island
(http://freepages.genealogy.rootsweb.ancestry.
com/~gfli/). Annual numbers of deaths calcu-
lated from these records were compared with
the numbers published in the New York City
Department of Health’s annual Summary
of Vital Statistics reports. Annual calculated
numbers of deaths were between 0.02% and
4.94% (median 0.95%) higher than those
reported in the annual Summary of Vital
Statistics reports (Petkova et al. 2014).
Death records for the years after 1950 are
stored at the New York City Department of
Health and Mental Hygiene (NYC DOH
MH 2016) and were not directly accessible or
available in digital format for this study.
Daily multiple-cause-of-death mortality
data for all five New York City boroughs for
1973–2006 were obtained in collaboration
with Joel Schwartz and colleagues at Harvard
University School of Public Health from the
U.S. National Center for Health Statistics
(NCHS 2016).
Temperature Data
Daily temperature data before 1949 were
obtained for New York Central Park from
the United States Historical Climatology
Network (USHCN) [National Oceanic
and Atmospheric Administration (NOAA)
National Centers for Environmental
Information 2016]. There were five missing
records in the data prior to 1949 that were
substituted with the averages of the previous
and following day temperatures. Daily
temperature data, also from the New York
Central Park station from 1973 onwards, were
obtained from the National Climatic Data
Center (NCDC 2016).
Historical Heat–Mortality
Relationships
We used the distributed lag nonlinear model
(DLNM) module in R (Gasparrini 2011) to
characterize the temperature–mortality rela-
tionships for each time period. Distributed lag
nonlinear models allow simultaneous charac-
terization of the nonlinear and lagged effects
of temperature on mortality (Armstrong 2006;
Gasparrini et al. 2010). Decadal models for
1900–1909 (1900s), 1910–1919 (1910s),
1920–1929 (1920s), 1930–1939 (1930s),
1940–1948 (1940s), 1973–1979 (1970s),
1980–1989 (1980s), 1990–1999 (1990s),
and 2000–2006 (2000s) were developed
using the mean daily temperature, and 22°C
(corresponding to approximately the 80th
percentile of annual temperature) was used as
a reference temperature for calculating relative
risk. e temperature-mortality analysis was
restricted to the summer months (June to
September) in order to focus on heat-related
mortality. The model is represented by the
following equation:
log[E(yt)] = α + f(xt; β) + s(t; γ) + g(jt; η)
+ Σ6
p = 1 δpIp(dt) [1]
where
• E(yt) is the expected number of deaths at day t
• f is the function modeling the association
with x, a moving average of temperature over
a lag of 3 days (lag 0–3), with parameters β
• s is the function of time t modeling the
long-term trend with parameters γ
• g is the function of the day of the year j
modeling the seasonal trend with parameters η
• Ip is a series of indicators modeling the
association with day of the week d with
parameters δp.
Although longer lags have been found to
be appropriate in modeling heat-mortality
impacts in the beginning of the 20th century
because of some immediate partial harvesting
following exposure to heat, shorter lags have
been found to adequately capture heat effects
in recent decades (Petkova et al. 2014). us,
a lag of 3 days was selected to focus on the
immediate impact of heat on mortality.
We defined f as a cross-basis composed of a
quadratic spline with 4 degrees of freedom
with 2 knots at equally spaced percentiles of
temperature distribution for the exposure–
response function, and a natural spline with 2
degrees of freedom with 2 knots for the lag–
response function. e functions s and g were
defined as natural cubic splines with 7 degrees
of freedom per decade and with 4 degrees of
freedom for day in year, respectively.
Temperature Projections
e methods used here have been described
by Petkova et al. (2013). Downscaled climate
projections were developed using monthly
Bias Corrected Spatially Disaggregated
(BCSD) data at 1/8° resolution (Maurer et al.
2007). e data are derived from the WCRP
CMIP5 multi-model data set and include 33
GCMs used in the IPCC’s Fifth Assessment
Report. e global climate models along with
their originating institution and atmospheric
resolution are presented in Table 1.
Projections are provided for two RCPs
(Moss et al. 2010). e pathways are the basis
for short- and long-term climate modeling
experiments and make various underlying
assumptions about radiative forcing through
time, which depends upon future global
greenhouse gas and aerosol concentrations,
and land use changes.
e two RCPs used in this analysis were
RCP4.5 and RCP8.5, which are the most
frequently used RCPs among the climate
modeling community. These two scenarios
represent relatively low (4.5) and high (8.5)
greenhouse gas projections/radiative forcing
through the end of the century. Under
RCP4.5, stabilization of greenhouse gas
concentrations occurs shortly after 2100 as
a product of emissions reduction before that
time. RCP8.5 is a scenario with increasing
emissions through the century, associated with
an energy-intensive future and limited use of
green technologies (van Vuuren et al. 2011).
To develop daily temperature projections,
the monthly output from the climate models
for the 1/8° grid box corresponding to New
York City (Central Park) was used to develop
change factors for each calendar month based
on the difference between a 30-year future
average for that calendar month and the same
model’s 30-year baseline average for that same
calendar month (Horton et al. 2011). The
monthly change factors were then applied
to the corresponding observed daily weather
data to create a future projection.
e combination of 33 models and 2 RCPs
yielded 66 synthetic future temperature projec-
tions for daily mean temperature from 2010 to
2099 that are based on three 30-year time slices,
defined as the 2020s (2010–2039), the 2050s
(2040–2069) and the 2080s (2070–2099).
Population Projections
A comprehensive set of population projections
for New York State until 2040 and a detailed
methodology were previously derived by the
Cornell Center for Applied Demographics
(Vink 2009).
Projections were developed for this
study by establishing a range of assumptions

Projections of urban heat-related mortality
Environmental Health Perspectives
•
volume 125 | number 1 | January 2017
49
regarding the components of the basic
demographic equation based on the Cohort
Component model (Smith et al. 2001):
forage 0and 0at>>
,
forage 0and 0
POP
POPD NM
at
>
1
1
1, 1,
1, 1, 1,
t
a
t
a
tt
att
a
tt tt
a
tt
a
=
-+
=
-
-
--
-
--
,BDNM-+
Z
[
\
]
]
]
]
]
]
]
]
]
]
]
]
]
]
[2]
where
• POPa
t is the city population age a in at
point t in time
• POPa
0 is the population age a at the begin-
ning of the projection according to the
Decennial Census 2010. See 2010 Census
Summary File 1 (U.S. Census Bureau 2010)
• Bt – 1,t is the number of births between the
year before point t in time and point t and is
a function of age-specific fertility rates and
the number of females at each age
• Da
t – 1,t is the number of deaths between
the year before point t in time and point t
of people who would otherwise have been
age a at point t. It is a function of age-
specific mortality rates and the number of
people at risk
• NMa
t – 1,t is the net migration between the
year before point t in time and point t of
people who are age a at point t. Net migra-
tion is the difference between the number
of people moving in (a function of an age
profile and the total level of people moving
in) and the number of people moving out
(a function of age-specific rates and the local
population of a certain age).
is set of equations was set up separately
by sex.
We defined five different scenarios for
projecting future New York City popula-
tions by altering the parameters of the
above-mentioned equations. The “baseline”
scenario assumed that all parameters of the
model remained constant; that is, age-specific
fertility and mortality rates and age character-
istics of migration were all held constant, but
the population aged forward. e “decreased
mortality” scenario assumed a decrease in age-
specific mortality rates such that the values
reached 2/3 of the 2010 values in 2100. Life
expectancy at birth would increase by 6 years
over time under this scenario, which is in line
with the mortality assumptions in the recent
Census Bureau projections (U.S. Census
Bureau 2012). e third scenario, “increased
in-migration,” assumed that the growth of
domestic in-migration (from other parts of the
United States to New York City) would be half
of the growth of the U.S. population and that
the growth of international in-migration (from
outside of the United States to New York City)
would be half of the growth of the projected
international in-migration nationwide [from
Table1. IPCC AR5 GCMs used in this study. The models were developed by 22 modeling centers (left column). Some centers support multiple GCMs, and/or
versions of their GCM.
Modeling center Institute ID Model name
Atmospheric
resolution
(lat × lon) References
Commonwealth Scientific and Industrial Research Organization (CSIRO)
and Bureau of Meteorology (BOM), Australia
CSIRO-BOM ACCESS1.0 1.25 × 1.875 Bi etal. 2013
ACCESS1.3 1.25 × 1.875
Beijing Climate Center, China Meteorological Administration BCC BCC-CSM1.1 2.8 × 2.8 Wu 2012
BCC-CSM1.1(m) 1.1 × 1.1
College of Global Change and Earth System Science, Beijing Normal
University
GCESS BNU-ESM 2.8 × 2.8
Canadian Centre for Climate Modelling and Analysis CCCMA CanESM2 2.8 × 2.8 von Salzen etal. 2013
National Center for Atmospheric Research NCAR CCSM4 0.9 × 1.25 Gent etal. 2011; Neale etal. 2013
Community Earth System Model Contributors NSF-DOE-NCAR CESM1(BGC) 0.9 × 1.25 Long etal. 2013; Neale etal. 2013;
Hurrell etal. 2013
CESM1(CAM5) 0.9 × 1.25
Centro Euro-Mediterraneo per l Cambiamenti Climatici CMCC CMCC-CM 0.75 × 0.75 Scoccimarro etal. 2011; Roeckner
etal. 2006
Centre National de Recherches Météorologiques/Centre Européen de
Recherche et Formation Avancée en Calcul Scientifique
CNRM-CEFRACS CNRM-CM5 1.4 × 1.4 Voldoire etal. 2013
Commonwealth Scientific and Industrial Research Organization in
collaboration with Queensland Climate Change Centre of Excellence
CSIRO-QCCE CSIRO-Mk3.6.0 1.9 × 1.9 Rotstayn etal. 2012
LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences
and CESS, Tsinghua University
LASG-CESS FGOALS-g2 2.8 × 2.8 Li L etal. 2013a, 2013b
The First Institute of Oceanography, SOA, China FIO FIO-ESM 2.8 × 2.8 Collins etal. 2006
NOAA Geophysical Fluid Dynamics Laboratory NOAA GFDL GFDL-CM3 2.0 × 2.5 Donner etal. 2011; Dunne etal.
2013; Delworth etal. 2006
GFDL-ESM2G 2.0 × 2.5
GFDL-ESM2M 2.0 × 2.5
NASA Goddard Institute for Space Studies NASA GISS GISS-E2-R 2.0 × 2.5 Schmidt etal. 2006
National Institute of Meteorological Research/Korea Meteorological
Administration
NIMR/KMA HadGEM2-AO 1.25 × 1.875 Collins etal. 2011; Davies etal. 2005
Met Office Hadley Centre (additional HadGEM2-ES realizations
contributed by Instituto Nacional de Pesquisas Espaciais)
MOHC (additional
realizations by INPE)
HadGEM2-CC 1.25 × 1.875 Collins etal. 2011; Davies etal. 2005
HadGEM2-ES 1.25 × 1.875
Institute for Numerical Mathematics INM INM-CM4 1.5 × 2.0 Volodin etal. 2010
Institut Pierre-Simon Laplace IPSL IPSL-CM5A-LR 1.9 × 3.75 Dufresne etal. 2013; Hourdin etal.
2013a, 2013b
IPSL-CM5A-MR 1.3 × 2.5
IPSL-CM5B-LR 1.9 × 3.75
Japan Agency for Marine-Earth Science and Technology, Atmosphere
and Ocean Research Institute (The University of Tokyo), and National
Institute for Environmental Studies
MIROC MIROC-ESM 2.8 × 2.8 Watanabe 2008; Watanabe etal.
2011
MIROC-ESM-CHEM 2.8 × 2.8
Atmosphere and Ocean Research Institute (The University of Tokyo),
National Institute for Environmental Studies, and Japan Agency for
Marine-Earth Science and Technology
MIROC MIROC5 1.4 × 1.4 Watanabe etal. 2010
Max Planck Institute for Meteorology MPI-M MPI-ESM-MR 1.9 × 1.9 Stevens etal. 2013
MPI-ESM-LR 1.9 × 1.9
Meteorological Research Institute MRI MRI-CGCM3 1.1 × 1.1 Yukimoto etal. 2012
Norwegian Climate Centre NCC NorESM1-M 1.9 × 2.5 Iversen etal. 2013; Kirkevåg etal.
2013; Tjiputra etal. 2013
NorESM1-ME 1.9 × 2.5

Petkova et al.
50
volume 125 | number 1 | January 2017
•
Environmental Health Perspectives
the Census 2010 projections (U.S. Census
Bureau 2010)]. e fourth scenario, “increased
out-migration,” assumed that the rate of out-
migration would increase by 25% over the
projection period. The assumptions for the
increased in-migration and increased out-
migration are rather arbitrary, but they aim
to strike a balance between reasonable and
informative. More radical assumptions would
lead to New York City populations that would
introduce various complications because of
overcrowding or high vacancy rates. Finally, we
also used a “constant” no-population change
scenario in which the population and the age
of the population remained constant at the
2010 levels.
Projected Heat-Related Mortality
As previously reported (Petkova et al. 2014),
relative risks (RRs) estimated for heat-related
mortality were relatively constant during
the first part of 20th century, suggesting
little adaptation to heat during this period,
whereas RRs decreased from the 1970s to the
2000s, consistent with substantial adapta-
tion to heat. Specifically, the average relative
risk of mortality associated with a daily mean
temperature of 29°C versus 22°C during
June–September ranged from 1.30 [95%
confidence interval (CI): 1.25, 1.36] in the
1910s to 1.43 (95% CI: 1.37, 1.49) in the
1900s. In contrast, predicted average RRs
for the same exposure contrast fell from 1.38
(95% CI: 1.31, 1.44) during 1900–1948 to
only 1.15 (95% CI: 1.09, 1.20) during 1973–
2006 (p-value < 0.001), suggesting rapid
adaptation since the 1970s (Petkova et al.
2014). We believe that increased access to air
conditioning in recent years was the primary
cause of the apparent increase in adaptation.
A random-effects meta-regression including
a linear term for decade predicted a decrease
of 4.6% (95% CI: 2.4%, 6.7%) per decade
(p-value < 0.001) (Petkova et al. 2014).
Because we did not have mortality data
from the 1950s and 1960s, we could not
verify the precise onset of the adaptation
process (as indicated by the downward shift
in the trend for RRs). However, if we assume
that access to air conditioning was the major
driving force behind heat adaptation, it is
plausible to define three stages in the popula-
tion response to heat: before the introduc-
tion of domestic air conditioning, during
air conditioning penetration, and after air
conditioning penetration levels reach a steady
state. Because 84% of surveyed households in
New York City in 2003 already had air condi-
tioning in their homes (U.S. Census Bureau
2004), compared with only 39% in 1970
(U.S. Census Bureau 1978), we assume that
the prevalence of air conditioning will reach a
steady state level sometime in the near future.
Future heat-related mortality relative risks
at each degree Celsius (°C) were derived for
temperatures ≥ 25°C using the temperature-
specific relative risk estimates from the
historical decades as described above. Decade-
specific temperature curves were linearly
extrapolated for temperatures ≤ 41°C, the
highest projected temperature, using the last
four temperature data points of each curve.
We chose a sigmoid function to model the
decadal change in the heat-mortality response
because it permits an accurate approximation
of the three stages in the adaptation process:
1
RR RR
e
RR
ADAPTMAX YY
RANGE
0
=-
+
#\--
^h
[3]
The initial level of temperature-specific
relative risk (RRMAX) at each temperature was
determined by selecting the mean relative risk
from the first part of the 20th century, corre-
sponding to the preadaptation part of the
sigmoid curve. e RRRANGE was derived as
the difference between the RRMAX and RRMIN,
where RRMIN is the minimum relative risk
for a given temperature or the value to which
the sigmoidal curve converges. We developed
two future adaptation scenarios in addition
to a no-adaptation scenario: a scenario of
high adaptation where the projected RRMIN
in 2100 is 80% lower than the RR observed
at the same temperature during the 2000s,
and a scenario of moderate adaptation where
the projected RRMIN in 2100 is 20% lower
than the corresponding observed RR during
the 2000s. Y represents the year for which
RRADAPT is calculated, and Y0 represents the
half decay point, or the year in which RRMAX
drops by half of the RRRANGE. The steep-
ness of the transition between the periods
of no adaptation and complete adaptation is
determined by the coefficient α. Both α and
Y0 were subjected to nonlinear least squares
optimization using the data points for the last
four decades. We are not proposing a scenario
Figure1. Temperature-specific mortality curves for New York City, 1900–2100. (A) Adaptation model assumes that temperature-specific relative risks will
decrease by an additional 20% (“low adaptation”) between 2010 and 2100 compared with the 2000s. (B) Adaptation model assumes that temperature-specific
relative risks will decrease by an additional 80% (“high adaptation”) between 2010 and 2100 compared with the 2000s. Points represent the relative risks (RRs)
calculated using the distributed lag non-linear model (DLNM) for each temperature for the 1970s (1973–1979), 1980s (1980–1989), 1990s (1990–1999), and 2000s
(2000–2006). RRs were calculated for June–September using a model with a quadratic spline with 4 degrees of freedom and 22°C as a reference temperature.

Projections of urban heat-related mortality
Environmental Health Perspectives
•
volume 125 | number 1 | January 2017
51
assuming 100% adaptation because sub-
populations of vulnerable individuals without
access to air conditioning or other means of
heat relief are likely to continue to exist in the
future; thus, heat-related mortality may not
be completely avoidable.
Future heat-related deaths were calculated
as described by Petkova et al. (2013). In the
present study, population change and heat
adaptation scenarios were also incorporated
into the calculations. e temperature-specific
relative risks derived from the no adapta-
tion, high-adaptation and low-adaptation
scenarios were applied to the daily, downscaled
temperature projections until 2100.
Results
Our previous study of heat adaptation
patterns in New York City that examined
daily temperature and mortality data spanning
more than a century found no evidence of
adaptation during the beginning of the 20th
century, but evidence of rapid adaptation in
subsequent decades was observed (Petkova
et al. 2014). Based on these findings, we
developed a three-stage model of adapta-
tion. We also developed two future adapta-
tion scenarios, of low and high adaptation,
assuming different levels of adaptation
throughout the 21st century. Temperature-
specific mortality curves for New York City
calculated according to the low- and high-
adaptation scenarios are illustrated in Figure 1.
Points represent the relative risks calculated
using the DLNM model for each temperature
for the 1970s through the 2000s.
To characterize possible population change
pathways in New York City throughout the
21st century, we developed four new popula-
tion scenarios, making a range of assumptions
about future mortality, in-migration, and out-
migration. Population projections (Figure 2)
based on the four scenarios developed for this
study were used in addition to a no- population-
change (constant) scenario to derive assessments
of future heat-related mortality. Annual popu-
lation projections according to each scenario
along with the corresponding mortality rates are
provided in Table S1.
Finally, we obtained statistically down-
scaled future mean temperature projections
for New York City from 33 GCMs used in
the IPCC’s Fifth Assessment Report and two
RCPs, RCP4.5 and RCP8.5, representing
relatively low and high greenhouse gas projec-
tions, respectively. Combining these yielded an
ensemble of 66 model/scenario combinations
for future health impact calculations.
Future mortality estimates varied greatly
depending on the choice of demographic
and adaptation scenario. To emphasize the
influence of both population change and heat
adaptation, we used the 33 climate model
median and the two RCPs. Median numbers
of projected heat-related deaths across the 33
GCMs used during the 2020s, 2050s and
2080s are summarized by RCP, adaptation
scenario and population scenario in Figure 3
and Table 2.
The estimated median number of heat-
related deaths across the 33 GCMs is substan-
tially higher under RCP8.5 as the century
progresses, and in many cases, the number of
deaths projected under RCP8.5 is more than
twice the corresponding estimate for RCP4.5
under the same time, population, and adap-
tation scenarios. These findings suggest that
the number of deaths would be substantially
reduced under the lower-emission pathway,
RCP4.5. For example, we estimate that by the
2080s, 1,494 annual heat-related deaths could
be avoided under the increased in-migration/
low adaptation scenario, based on projections
of 2,771 versus 1,277 deaths under RCP8.5
and RCP4.5, respectively (Table 2).
Projected heat-related mortality was
highest for the increased in-migration popu-
lation scenario, followed by the baseline,
increased out-migration, decreased mortality,
and constant population scenarios. As an
example, for the 2080s under the RCP8.5/
high adaptation scenario, we projected 804
deaths under the increased in-migration
scenario, 698 deaths under the baseline
scenario, 624 deaths under both the increased
out-migration and decreased mortality
scenarios, and 379 deaths under the constant
population scenario.
Increasing levels of adaptation reduced
the number of projected deaths substantially.
For example, by the 2080s, 3,331, 2,271, and
804 deaths were projected to occur under
RCP8.5 and the increased in-migration/
no adaptation, increased in-migration/low
adaptation, and increased in-migration/high
adaptation, respectively. As another example,
during the 2020s and under RCP4.5, the
median number of heat-related deaths across
the 33 GCMs was 370 for the constant popu-
lation scenario with no adaptation and 149 for
the same scenario with high adaptation.
Discussion
To our knowledge, this study is the first to
present projections of heat-related mortality
until the end of the 21st century while incor-
porating assumptions of heat adaptation
based on historical mortality data spanning
over a century. Our adaptation model char-
acterized long-term trends in the popula-
tion response to heat and under alternative
assumptions about the limits to future adap-
tation. ere is considerable agreement that
limits to adaptation to climate change exist
Figure2. New York City (NYC) population by 2100 calculated according to the five population scenarios
developed for this study. “Baseline” assumed that all parameters of the model remain constant; that is,
age-specific fertility and mortality rates and age characteristics of migration are all kept constant, but the
population ages forward. “Decreased mortality” assumed a decrease in age-specific mortality rates such
that the values reach 2/3 of the 2010 values by 2100. “Increased in-migration” assumed that the growth of
domestic in-migration (from other parts of the United States to New York City) will be half of the growth of
the U.S. population and that the growth of international in-migration (from outside of the United States to
New York City) will be half of the growth of the projected international in-migration nationwide. “Increased
out-migration”: assumed that the rate of out-migration would increase by 25% over the projection period.
“Constant” assumed that the population and the age of the population remain constant at 2010 levels.

Petkova et al.
52
volume 125 | number 1 | January 2017
•
Environmental Health Perspectives
and are often defined by interactions between
climate change and biophysical and socio-
economic constraints, among other factors
(Klein et al. 2014). Quantifying the potential
limits and obstacles to climate change adapta-
tion as they relate to various health outcomes
is critical for achieving optimal resource
allocation and long-term planning.
Projecting future population adaptation
to heat is among the most important chal-
lenges in assessing the burden of heat-related
mortality under a changing climate. Here, we
Figure3. Median annual projected heat-related
deaths in New York City according to two
Representative Concentration Pathways (RCPs),
(A)RCP4.5 and (B) RCP8.5, and across 33
global climate models (GCMs) during the 2020s
(2010–2039), the 2050s (2040–2069), and the
2080s (2070–2099). The corresponding numeric
data are provided in Table2. Heat adaptation
scenarios are indicated by circle size and
include“high adaptation,” where adaptation,
as measured by the minimal relative risk for a
given temperature to be reached by the year
2100 (RRmin), is projected to reach a value 80%
lower than the RR calculated at each degree
Celsius (°C) during the 2000s; “low adapta-
tion,” where adaptation, as measured by RRmin,
is projected to reach a value 20% lower than
the RR calculated at each degree Celsius (°C)
during the 2000s; and“no adaptation,” wherein
future adaptation does not occur and adapta-
tion, as measured by RRmin, remains the same
as the RR calculated at each degree Celsius
(°C) during the 2000s. Population scenarios are
indicated by color and included “baseline,”
which assumed that all parameters of the
model remain constant; that is, age-specific
fertility and mortality rates and age char-
acteristics of migration are all kept constant,
but the population ages forward; “decreased
mortality,” which assumed a decrease in age-
specific mortality rates such that the values
reach 2/3 of the 2010 values by 2100; “increased
in-migration,” which assumed that the growth
of domestic in-migration (from other parts of
the United States to New York City) will be half
of the growth of the U.S. population and that
the growth of international in-migration (from
outside of the United States to New York City)
will be half of the growth of the projected inter-
national in-migration nationwide; “increased
out-migration,”which assumed that the rate
of out-migration would increase by 25% over
the projection period; and “constant,” which
assumed that the population and the age of the
population remain constant at 2010 levels. For
reference, there were 638 heat-related deaths
annually between 2000 and 2006.

Projections of urban heat-related mortality
Environmental Health Perspectives
•
volume 125 | number 1 | January 2017
53
have proposed a novel approach to modeling
heat adaptation that allows the consideration of
observed trends in adaptation since the begin-
ning of the 20th century. Because our previous
findings suggested that there was no adapta-
tion to heat in New York City during the first
part of the 20th century (Petkova et al. 2014),
we used the mean relative risk estimated for
the early part of the 20th century to anchor
the upper segment of the sigmoidal adaptation
function (Equation 3) for that period. We used
the declining relative risks estimated for recent
decades to characterize adaptation that occurred
as the prevalence of air conditioning increased,
and we extrapolated this decline through 2100
under two different adaptation scenarios repre-
senting both modest and substantial increases
in adaptation from the 2010 level.
Although population change is considered
to be among the most important factors in
estimating future temperature impacts, future
demographics are often not taken into account
because location-specific population projec-
tions are generally not readily available beyond
several decades. To address this issue, we
developed new population change scenarios
to apply to our projections of heat-related
mortality. Finally, we combined the developed
population and heat adaptation scenarios with
temperature projections from multiple GCMs
and two RCPs to derive a comprehensive
assessment of heat-related mortality until the
end of the 21st century.
Annual future mortality estimates varied
greatly by RCP, as well as by population
change and adaptation scenario. For instance,
the constant population/high adaptation
scenario produced the lowest death estimates,
projecting 167 and 379 heat-related deaths
during the 2080s under RCP4.5 and RCP8.5,
respectively. The increased in-migration/no
adaptation scenario produced the highest
mortality estimates under RCP8.5, projecting
555 and 3,331 deaths during the 2020s and
the 2080s, respectively.
Both the heat adaptation and demo-
graphic scenarios have several limitations.
First, our model of heat-related mortality over
time was based on an empirical fit to historical
data and extrapolation using a sigmoidal curve
into the future. We did not identify and incor-
porate causal factors such as air conditioning
use into the projection of future heat response.
Future research that focuses on characterizing
the impact of heat over time among vulner-
able populations would be particularly useful
in improving the utility of the adaptation
models. In addition, studies quantifying the
impact of various public health interventions
such as heat warning systems, cooling centers,
and other preventive measures on heat-related
mortality would be valuable for the further
development of this work. Another important
limitation of the study is that decade-specific
mortality versus temperature curves were
linearly extrapolated to high temperatures
projected to occur under changing climate
(e.g., temperatures of 41°C) for which no
historical mortality data exist. This extrapo-
lation may underestimate mortality impacts
at such very high temperatures, particularly
during the initial exposures of the populations
to temperatures that they have not previously
experienced. Studies of mortality responses in
unacclimatized populations would be particu-
larly useful in characterizing heat impacts at
very high temperatures. Finally, we acknowl-
edge that the assumptions underlying the
two adaptation scenarios developed for this
study were arbitrary, but we believe that they
capture a reasonable range of potential future
adaptation, from modest (20%) to substantial
(80%). More data over a longer time period
will be needed to determine which end of this
range is most realistic.
Although we believe that the assumptions
of the demographic models developed for
this work are reasonable, they are based on
historical trends that may or may not continue.
Population projections are rarely developed
beyond several decades, particularly on a
fine, city-level geographical scale. Given the
increasing importance of projecting popula-
tion health impacts under a changing climate,
additional work focused on developing and
validating long-term population projections
will be of critical importance for improving the
accuracy of projecting heat-related mortality
and other health impacts. Nevertheless, by
including five different population scenarios,
our study is among the first to examine
sensitivity to this important assumption.
Conclusion
e methods and findings of this study may be
particularly relevant to estimating heat-related
mortality in cities currently experiencing heat
impacts and increasing urbanization with or
without population growth. Because the choice
of adaptation scenario substantially affected
the number of projected heat-related deaths,
improved understanding of heat adaptation
is necessary in order to refine projections.
Nonetheless, the substantial reduction of heat-
related mortality, particularly under the high-
adaptation scenario, provides evidence of the
importance of public policy measures leading
to continuous heat adaptation. Finally, the
number of median annual heat-related deaths
calculated across all models under RCP8.5 was
in many instances more than twice as high as
the number of deaths projected under RCP4.5.
This difference highlights the magnitude of
the potential public health benefit associated
with reducing greenhouse gas concentrations
in the atmosphere.
Table2. Median number of projected heat-related deaths in New York City across the 33 GCMs used
in this study for the 2020s (2010–2039), 2050s (2040–2069) and 2080s (2070–2099) by Representative
Concentration Pathway (RCP), adaptation scenario and population scenario.
Period Population scenario
RCP4.5 RCP8.5
No
adaptation
Low
adaptation
High
adaptation
No
adaptation
Low
adaptation
High
adaptation
2020s Baseline 492 412 191 549 460 215
2050s Baseline 1,084 891 267 1,449 1,196 365
2080s Baseline 1,348 1,109 308 2,893 2,407 698
2020s Decreased mortality 472 395 184 527 442 207
2050s Decreased mortality 1,001 823 247 1,339 1,104 338
2080s Decreased mortality 1,205 991 275 2,585 2,151 624
2020s Increased in-migration 497 416 193 555 465 217
2050s Increased in-migration 1,151 946 283 1,539 1,270 387
2080s Increased in-migration 1,552 1,277 354 3,331 2,771 804
2020s Increased out-migration 489 409 190 546 457 214
2050s Increased out-migration 1,040 855 257 1,391 1,147 351
2080s Increased out-migration 1,206 991 275 2,587 2,152 624
2020s Constant 370 311 149 413 347 167
2050s Constant 608 500 150 813 671 205
2080s Constant 733 603 167 1,573 1,309 379
Heat adaptation scenarios include a)“high adaptation”: adaptation, as measured by RRmin or the minimal relative risk
for a given temperature to be reached by the year 2100, projected to reach a value 80% lower than RR calculated at
each degree Celsius (°C) during the 2000s; b)“low adaptation”: adaptation, as measured by RRmin or the minimal relative
risk for a given temperature to be reached by the year 2100, projected to reach a value 20% lower than RR calculated
at each degree Celsius during the 2000s; and c)“no adaptation”: future adaptation does not occur. Adaptation, as
measured by RRmin or the minimal relative risk for a given temperature to be reached by the year 2100, remains the same
as the RR calculated at each degree Celsius during the 2000s. Population scenarios included the following: a)“baseline”
assumed that all parameters of the model remain constant; that is, age-specific fertility and mortality rates and age
characteristics of migration are all kept constant, but the population ages forward; b)“decreased mortality” assumed
a decrease in age-specific mortality rates such that the values reach 2/3 of the 2010 values by 2100; c)“increased
in-migration” assumed that the growth of the domestic in-migration (from other parts of the United States to New York
City) will be half of the growth of the U.S. population and that the growth of the international in-migration (from outside
of the United States to New York City) will be half of the growth of the projected international in-migration nationwide;
d)“increased out-migration” assumed that the rate of out-migration would increase by 25% over the projection period;
and e)“constant” assumed that population and age of the population remain constant at the 2010 levels. For reference,
there were 638 heat-related deaths annually between 2000 and 2006.

Petkova et al.
54
volume 125 | number 1 | January 2017
•
Environmental Health Perspectives
RefeRences
Armstrong B. 2006. Models for the relationship
between ambient temperature and daily mortality.
Epidemiology 17:624–631.
Baccini M, Kosatsky T, Analitis A, Anderson HR,
D’Ovidio M, Menne B, etal. 2011. Impact of heat on
mortality in 15 European cities: attributable deaths
under different weather scenarios. JEpidemiol
Community Health 65:64–70.
Bi D, Dix M, Marsland SJ, O’Farrell S, Rashid H,
UotilaP, et al. 2013. The ACCESS coupled model:
description, control climate and evaluation. Aust
Meteorol OceanogrJ 63:41–64.
Collins WD, Rasch PJ, Boville BA, Hack JJ, McCaaJR,
Williamson DL, etal. 2006. The formulation and atmo-
spheric simulation of the Community Atmosphere
Model version 3 (CAM3). J Clim 19:2144–2161.
Collins WJ, Bellouin N, Doutriaux-Boucher M, GedneyN,
Halloran P, Hinton T, etal. 2011. Development and
evaluation of an Earth-System model – HadGEM2.
Geosci Model Dev 4:1051–1075.
Davies T, Cullen MJ, Malcolm AJ, Mawson MH,
Staniforth A, White AA, etal. 2005. A new dynam-
ical core for the Met Office’s global and regional
modelling of the atmosphere. Q J R Meteorol Soc
131:1759–1782.
Delworth TL, Broccoli AJ, Rosati A, Stouffer RJ,
BalajiV, Beesley JA, et al. 2006. GFDL’s CM2
global coupled climate models. Part I: formulation
and simulation characteristics. J Clim 19:643–674.
Dessai S. 2003. Heat stress and mortality in Lisbon
Part II. An assessment of the potential impacts of
climate change. Int J Biometeorol 48:37–44.
Donner LJ, Wyman BL, Hemler RS, Horowitz LW,
MingY, Zhao M, et al. 2011. The dynamical core,
physical parameterizations, and basic simulation
characteristics of the atmospheric component
AM3 of the GFDL global coupled model CM3. J Clim
24:3484–3519.
DORIS (New York City Department of Records and
Information Services). 2016. The Vital Records
Collection at the Municipal Archives. http://www.
nyc.gov/html/records/html/archives/genealogy.
shtml [accessed 10May 2016].
Dow K, Berkhout F, Preston BL, Klein RJ, Midgley G,
Shaw MR. 2013. Limits to adaptation. Nat Clim
Chang 3:305–307.
Doyon B, Bélanger D, Gosselin P. 2008. The potential
impact of climate change on annual and seasonal
mortality for three cities in Québec, Canada. Int J
Health Geogr 7:23, doi: 10.1186/1476-072X-7-23.
Dufresne JL, Foujols MA, Denvil S, Caubel A, MartiO,
Aumont O, etal. 2013. Climate change projec-
tions using the IPSL-CM5 Earth System Model:
from CMIP3 to CMIP5. Clim Dyn 40:2123–2165, doi:
10.1007/s00382-012-1636-1.
Dunne JP, John JG, Shevliakova E, Stouffer RJ, Krasting
JP, Malyshev SL, etal. 2013. GFDL’s ESM2 global
coupled climate–Carbon Earth System Models. Part
II: carbon system formulation and baseline simula-
tion characteristics. J Clim 2013 26:2247–2267.
Gasparrini A. 2011. Distributed lag linear and non-linear
models in R: the package dlnm. J Stat Softw 43:1–20.
Gasparrini A, Armstrong B, Kenward MG. 2010.
Distributed lag non-linear models. Stat Med
29:2224–2234.
Gent PR, Danabasoglu G, Donner LJ, Holland MM, Hunke
EC, Jayne SR, etal. 2011. The Community Climate
System Model version 4. JClim 24:4973–4991.
Gosling SN, McGregor GR, Lowe JA. 2009. Climate
change and heat-related mortality in six cities Part
2: climate model evaluation and projected impacts
from changes in the mean and variability of
temperature with climate change. IntJ Biometeor
53:31–51.
Guest CS, Wilson K, Woodward AJ, Hennessy K,
Kalkstein LS, Skinner C, etal. 1999. Climate
and mortality in Australia: retrospective study,
1979–1990, and predicted impacts in five major
cities in 2030. Clim Res 13:1–15.
Hayhoe K, Cayan D, Field CB, Frumhoff PC, MaurerEP,
Miller NL, etal. 2004. Emissions pathways, climate
change, and impacts on California. Proc Natl Acad
Sci U S A 101:12422–12427.
Hayhoe K, Sheridan S, Kalkstein L, Greene S. 2010.
Climate change, heat waves, and mortality projec-
tions for Chicago. J Great Lakes Res 36:65–73.
Holland JH. 1995. Hidden Order: How Adaptation
Builds Complexity. Cambridge, MA:Perseus Books.
Horton RM, Gornitz V, Bader DA, Ruane AC, Goldberg
R, Rosenzweig C. 2011: Climate hazard assess-
ment for stakeholder adaptation planning in New
York City. J Appl Meteorol Climatol 50:2247–2266,
doi:10.1175/2011JAMC2521.1.
Hourdin F, Foujols MA, Codron F, Guemas V,
DufresneJL, Bony S, et al. 2013a. Impact of the
LMDZ atmospheric grid configuration on the climate
and sensitivity of the IPSL-CM5A coupled model.
Clim Dyn 40:2167–2192.
Hourdin F, Grandpeix JY, Rio C, Bony S, Jam A,
CheruyF, et al. 2013b. LMDZ5B: the atmospheric
component of the IPSL climate model with revisited
parameterizations for clouds and convection. Clim
Dyn 40:2193–2222.
Huang C, Barnett AG, Wang X, Vaneckova P,
FitzGerald G, Tong S. 2011. Projecting future heat-
related mortality under climate change scenarios:
a systematic review. Environ Health Perspect
119:1681–1690, doi: 10.1289/ehp.1103456.
Hurrell JW, Holland MM, Gent PR, Ghan S, Kay JE,
Kushner PJ, etal. 2013. The Community Earth
System Model: a framework for collaborative
research. Bull Am Meteorol Soc 94:1339–1360, doi:
10.1175/BAMS-D-12-00121.1.
IPCC (Intergovernmental Panel on Climate Change).
2013. Climate Change 2013—The Physical Science
Basis: Working Group I Contribution to the Fifth
Assessment Report of the Intergovernmental Panel
on Climate Change. StockerTF, QinD, PlattnerGK,
TignorM, Allen SK, Boschung J, et al., eds.
Cambridge, UK:Cambridge University Press.
Iversen T, Bentsen M, Bethke I, Debernard JB,
Kirkevåg A, Seland Ø, etal. 2013. The Norwegian
Earth System Model, NorESM1-M—Part 2: climate
response and scenario projections. Geosci Model
Dev 6:389–415.
Jackson JE, Yost MG, Karr C, Fitzpatrick C, LambBK,
Chung SH, etal. 2010. Public health impacts of
climate change in Washington State: projected
mortality risks due to heat events and air pollution.
Clim Change 102:159–186.
Kalkstein LS, Greene JS. 1997. An evaluation of
climate/mortality relationships in large U.S. cities
and the possible impacts of a climate change.
Environ Health Perspect 105:84–93.
Kirkevåg A, Iversen T, Seland Ø, Hoose C,
KristjánssonJE, Struthers H, et al. 2013. Aerosol–
climate interactions in the Norwegian Earth
System Model–NorESM1-M. Geosci Model Dev
6:207–244.
Klein RJT, Midgley GF, Preston BL, Alam M,
BerkhoutFGH, Dow K, etal. 2014. Adaptation oppor-
tunities, constraints, and limits. In: Climate Change
2014: Impacts, Adaptation, and Vulnerability. Part
A: Global and Sectoral Aspects. Contribution of
Working Group II to the Fifth Assessment Report
of the Intergovernmental Panel on Climate Change.
Field CB, Barros VR, Dokken DJ, Mach KJ,
MastrandreaMD, Bilir TE, etal., eds. Cambridge,
UK:Cambridge University Press, 899–943.
Knowlton K, Lynn B, Goldberg RA, RosenzweigC,
Hogrefe C, Rosenthal JK, etal. 2007. Projecting
heat-related mortality impacts under a changing
climate in the New York City region. Am J Public
Health 97:2028–2034.
Lansing JS. 2003. Complex adaptive systems. Annu
Rev Anthropol 32:183–204.
Li L, Lin P, Yu Y, Wang B, Zhou T, Liu L, etal. 2013a. The
Flexible Global Ocean-Atmosphere-Land System
Model, Grid-point Version 2: FGOALS-g2. Adv
Atmos Sci 30:543–560.
Li L, Wang B, Dong L, Liu L, Shen S, Hu N, etal. 2013b.
Evaluation of Grid-point Atmospheric Model of IAP
LASG version 2 (GAMIL2). Adv Atmos Sci 30:855–867.
Li T, Horton RM, Kinney P. 2013. Projections of
seasonal patterns in temperature-related deaths
for Manhattan. Nat Clim Chang 3:717–721.
Long MC, Lindsay K, Peacock S, Moore JK, Doney SC.
2013. Twentieth-century oceanic carbon uptake
and storage in CESM1 (BGC). J Clim 26:6775–6800.
Martens WJ. 1998. Climate change, thermal stress and
mortality changes. Soc Sci Med 46:331–344.
Martin SL, Cakmak S, Hebbern CA, Avramescu ML,
Tremblay N. 2012. Climate change and future
temperature-related mortality in 15 Canadian
cities. Int J Biometeorol 56:605–619.
Maurer EP, Brekke L, Pruitt T, Duffy PB. 2007. Fine-
resolution climate projections enhance regional
climate change impact studies. Eos Trans Am
Geophys Union 88:504, doi: 10.1029/2007EO470006.
Moss RH, Edmonds JA, Hibbard KA, ManningMR, Rose
SK, van Vuuren DP, etal. 2010. The next genera-
tion of scenarios for climate change research and
assessment. Nature 463:747–756.
NCDC (National Climatic Data Center). 2016. Climate
Data Online. https://www.ncdc.noaa.gov/
cdo-web/ [accessed 10May 2016].
NCHS (National Center for Health Statistics). 2016.
Mortality Data. http://www.cdc.gov/nchs/nvss/
deaths.htm [accessed 10May 2016].
Neale RB, Richter J, Park S, Lauritzen PH, VavrusSJ,
Rasch PJ, etal. 2013. The mean climate of the
Community Atmosphere Model (CAM4) in forced SST
and fully coupled experiments. J Clim 26:5150–5168.
NOAA (National Oceanic and Atmospheric
Administration) National Centers for Environmental
Information. 2016. U.S. Historical Climatology
Network (USHCN). https://www.ncdc.noaa.gov/
data-access/land-based-station-data/land-based-
datasets/us-historical-climatology-network-ushcn
[accessed 10May 2016].
NYC DOH MH (New York City Department of Health
and Mental Hygiene). 2016. Death Certificates.
http://www1.nyc.gov/site/doh/services/death-
certificates.page [accessed 10May 2016].
Ostro B, Barrera-Gómez J, Ballester J, BasagañaX,
Sunyer J. 2012. The impact of future summer
temperature on public health in Barcelona and
Catalonia, Spain. Int J Biometeorol 56:1135–1144.
Peng RD, Bobb JF, Tebaldi C, McDaniel L, Bell ML,
Dominici F. 2011. Toward a quantitative estimate
of future heat wave mortality under global climate
change. Environ Health Perspect 119:701–706, doi:
10.1289/ehp.1002430.
Petkova EP, Gasparrini A, Kinney PL. 2014. Heat and
mortality in New York City since the beginning of
the 20th century. Epidemiology 25:554–560.
Petkova EP, Horton RM, Bader DA, Kinney PL. 2013.
Projected heat-related mortality in the U.S.
urban northeast. Int J Environ Res Public Health
10:6734–6747.

Projections of urban heat-related mortality
Environmental Health Perspectives
•
volume 125 | number 1 | January 2017
55
Roeckner E, Brokopf R, Esch M, Giorgetta M,
Hagemann S, Kornblueh L, etal. 2006. Sensitivity
of simulated climate to horizontal and vertical
resolution in the ECHAM5 atmosphere model. J
Climate 19(16):3771–3791.
Rotstayn LD, Jeffrey SJ, Collier MA, Dravitzki SM, Hirst
AC, Syktus JI, etal. 2012. Aerosol- and greenhouse
gas-induced changes in summer rainfall and
circulation in the Australasian region: a study using
single-forcing climate simulations. Atmos Chem
Phys 12:6377–6404, doi:10.5194/acp-12-6377-2012.
Schmidt GA, Ruedy R, Hansen JE, Aleinov I, Bell N,
Bauer M, etal. 2006. Present-day atmospheric simu-
lations using GISS ModelE: comparison to insitu,
satellite, and reanalysis data. J Clim 19:153–192.
Scoccimarro E, Gualdi S, Bellucci A, Sanna A,
Giuseppe Fogli P, Manzini E, etal. 2011. Effects
of tropical cyclones on ocean heat transport in a
high-resolution coupled general circulation model.
J Clim 24:4368–4384.
Sheridan SC, Allen MJ, Lee CC, Kalkstein LS. 2012.
Future heat vulnerability in California, Part II:
projecting future heat-related mortality. Clim
Change 115:311–326.
Smith SK, Tayman J, Swanson DA. 2001. State and Local
Population Projections: Methodology and Analysis.
New York:Kluwer Academic/Plenum Publishers.
Stevens B, Giorgetta M, Esch M, Mauritsen T, CruegerT,
Rast S, etal. 2013. Atmospheric component of the
MPI-M Earth System Model: ECHAM6. JAdv Model
Earth Syst 5:146–172, doi: 10.1002/jame.20015.
Tjiputra JF, Roelandt C, Bentsen M, Lawrence DM,
Lorentzen T, Schwinger J, etal. 2013. Evaluation
of the carbon cycle components in the Norwegian
Earth System Model (NorESM). Geosci Model Dev
6:301–325.
U.S. Census Bureau. 1978. Housing Characteristics
for Selected Metropolitan Areas. Annual Housing
Survey: 1976. Washington, DC:U.S. Census Bureau.
U.S. Census Bureau. 2004. Current Housing Reports,
Series H170/03-53. American Housing Survey for
the New York-Nassau-Suffolk-Orange Metropolitan
Area: 2003. Washington, DC:U.S. Census Bureau.
https://www.census.gov/prod/2004pubs/h170-03-53.
pdf [accessed 12May 2016].
U.S. Census Bureau. 2010. American FactFinder.
New York City, New York, 2010. http://factfinder.
census.gov/bkmk/table/1.0/en/DEC/10_SF1/
PCT12/1600000US3651000 [accessed 12May 2016].
U.S. Census Bureau. 2012. Methodology and
Assumptions for the 2012 National Projections.
Washington DC:U.S. Census Bureau. http://
www.census.gov/population/projections/files/
methodology/methodstatement12.pdf [accessed
10 May 2016].
van Vuuren DP, Edmonds J, Kainuma M, Riahi K,
Thomson A, Hibbard K, etal. 2011. The representa-
tive concentration pathways: an overview. Clim
Change 109:5–31.
Vink J. 2009. New York Population Projection by Age
and Sex: County Projections 2005–2035. Model
Description. New York:Cornell University, Program
on Applied Demographics.
Voldoire A, Sanchez-Gomez E, Salas y Mélia D,
DecharmeB, Cassou C, Sénési S, etal. 2013. The
CNRM-CM5.1 global climate model: description
and basic evaluation. Clim Dyn 40:2091–2121.
Volodin EM, Dianskii NA, Gusev AV. 2010. Simulating
present-day climate with the INMCM4.0 coupled
model of the atmospheric and oceanic general
circulations. Izv An Fiz Atmos OK+ 46:414–431.
von Salzen K, Scinocca JF, McFarlane NA, Li J,
ColeJN, Plummer D, etal. 2013. The Canadian
fourth generation Atmospheric Global Climate
Model (CanAM4). Part I: representation of physical
processes. Atmos Ocean 51:104–125.
Watanabe M. 2008. Two regimes of the equatorial
warm pool. Part I: a simple tropical climate model.
J Clim 21:3533–3544.
Watanabe M, Chikira M, Imada Y, Kimoto M. 2011.
Convective control of ENSO simulated in MIROC.
JClim 24:543–562.
Watanabe M , Suzuki T, O’ishi R, Komuro Y ,
WatanabeS, Emori S, etal. 2010. Improved climate
simulation by MIROC5: mean states, variability,
and climate sensitivity. J Clim 23:6312–6335.
Wu T. 2012. A mass-flux cumulus parameterization
scheme for large-scale models: description and
test with observations. Clim Dyn 38:725–744.
Yukimoto S, Adachi Y, Hosaka M, Sakami T,
YoshimuraH, Hirabara M, etal. 2012. A new global
climate model of the Meteorological Research
Institute: MRI-CGCM3—model description and
basic performance. J Meteorol Soc Jpn 90A:23–64.