Background and objectives
Heat related mortality is of great concern for public health, and estimates of future mortality under a warming climate are important for planning of resources and possible adaptation measures. Papers providing projections of future heat-related mortality were critically reviewed with a focus on the use of climate model data. Some best practice guidelines are proposed for future research.
The electronic databases Web of Science and PubMed/Medline were searched for papers containing a quantitative estimate of future heat-related mortality. The search was limited to papers published in English in peer-reviewed journals up to the end of March 2017. Reference lists of relevant papers and the citing literature were also examined. The wide range of locations studied and climate data used prevented a meta-analysis.
A total of 608 articles were identified after removal of duplicate entries, of which 63 were found to contain a quantitative estimate of future mortality from hot days or heat waves. A wide range of mortality models and climate model data have been used to estimate future mortality. Temperatures in the climate simulations used in these studies were projected to increase. Consequently, all the papers indicated that mortality from high temperatures would increase under a warming climate. The spread in projections of future climate by models adds substantial uncertainty to estimates of future heat-related mortality. However, many studies either did not consider this source of uncertainty, or only used results from a small number of climate models. Other studies showed that uncertainty from changes in populations and demographics, and the methods for adaptation to warmer temperatures were at least as important as climate model uncertainty. Some inconsistencies in the use of climate data (for example, using global mean temperature changes instead of changes for specific locations) and interpretation of the effects on mortality were apparent. Some factors which have not been considered when estimating future mortality are summarised.
Most studies have used climate data generated using scenarios with medium and high emissions of greenhouse gases. More estimates of future mortality using climate information from the mitigation scenario RCP2.6 are needed, as this scenario is the only one under which the Paris Agreement to limit global warming to 2°C or less could be realised. Many of the methods used to combine modelled data with local climate observations are simplistic. Quantile-based methods might offer an improved approach, especially for temperatures at the ends of the distributions. The modelling of adaptation to warmer temperatures in mortality models is generally arbitrary and simplistic, and more research is needed to better quantify adaptation. Only a small number of studies included possible changes in population and demographics in their estimates of future mortality, meaning many estimates of mortality could be biased low. Uncertainty originating from establishing a mortality baseline, climate projections, adaptation and population changes is important and should be considered when estimating future mortality.
Citation: Sanderson M, Arbuthnott K, Kovats S, Hajat S, Falloon P (2017) The use of climate information to estimate future mortality from high ambient temperature: A systematic literature review. PLoS ONE 12(7): e0180369. https://doi.org/10.1371/journal.pone.0180369
Editor: Juan A. Añel, Universidade de Vigo, SPAIN
Received: January 17, 2017; Accepted: June 14, 2017; Published: July 7, 2017
Copyright: © 2017 Sanderson et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: This work was funded under the NIHR Health Protection Research Unit (NIHR HPRU) in Environmental Change and Health, led by the London School of Hygiene and Tropical Medicine in partnership with Public Health England (PHE), and in collaboration with the University of Exeter, University College London, and the Met Office. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Warming of the Earth’s climate is now unequivocal; global average temperatures have risen by 0.85°C between 1880 and 2012 . Increases in temperature over land areas are almost always higher than global average increases but vary between different regions of the Earth [2,3]. The frequency of heat waves has also increased in many continents . Global mean temperature is projected to increase by about 1.6 to 2.6°C above the preindustrial period by the 2050s, depending on the scenario used . Using median values, projected temperature increases for Europe and America are between 2 and 4°C for the 2050s (relative to present-day climate). Higher increases are projected over much of Asia and Australia .
There is increasing concern over the effects of hot weather on public health, including heat-related mortality and morbidity . Deaths from high temperatures and heat waves are greater than deaths from other weather events such as tornados and flooding . These deaths are not only a result of heatstroke. Existing studies generally examine the relationship between short-term fluctuations in temperature and all-cause (or cause specific, e.g. cardiovascular) mortality. The first studies linking mortality to warm temperatures were published in the early twentieth century . Since this time, there have been numerous additional studies of the effects of specific periods of warm and hot weather on mortality, many of which have been reviewed elsewhere [7,8]. More recently, heat-related mortality has gathered increased attention in public health research owing to the acceptance that the Earth’s climate is warming and the large number of deaths caused by extreme heat waves (for example, Europe, 2003, 2015; Russia, 2010; Australia, 2012/2013 and 2016/2017; North America, 2012; India and Pakistan, 2015). Some of these events have led to the implementation of specific policies to reduce heat-related mortality such as the National Heat Wave Plan in France  and the Heatwave Plan for England .
Excess mortality from high temperatures has been reported in the first five assessment reports published by the Intergovernmental Panel on Climate Change (IPCC) to varying degrees. In the first report  the effects of warm temperatures and heat waves on mortality were briefly discussed. It was speculated that mortality from heat waves was likely to increase under a warming climate. The second , third , fourth  and fifth  assessment reports each contain a chapter devoted to human health. Mortality from high temperatures was discussed briefly in each report, but few studies were cited and the methods used to estimate future mortality were not assessed. Fourteen studies of projections of future heat related mortality have been reviewed previously , but these authors did not critically review the use of climate model data.
The aim of the present study is to review the use of climate model data in projections of future heat-related mortality. Morbidity was not considered; there are very few papers projecting future morbidity, and the burdens are very dependent on changes in health care. An important aspect of the present review is the critical appraisal of the selection of climate data and its use, and the methods employed to combine climate model data with observations. The treatment of uncertainty in climate model projections is also assessed. The epidemiologic models used to relate mortality to temperature and other variables have been reviewed elsewhere [5,8,16] and will not be addressed in the present study.
Data sources and search strategy
The peer-reviewed databases Web of Science and PubMed/Medline were searched with a focus on the titles of the articles; trial searches using a more general search of topics identified many hundreds of articles, most of which were irrelevant. Groups of two or three of the following keywords were used in the searches: mortality, future, climate, climate change, impacts, projection, heat, temperature, deaths and scenario. The search was limited to papers published in peer-reviewed journals in English with no restriction on year up to the end of March 2017. It is noted that the Web of Science does not contain articles published before 1981.
Two criteria were used to select articles for further study. The articles had to include at least one quantitative estimate of future heat-related mortality. Studies which only reported changes in morbidity, mortality resulting from air pollution or infectious diseases, or focused on winter and the effects of cold temperatures were not selected. Conference abstracts, books and publications by governments or international organisations were not included. Reference lists in the articles selected, and those studies which cited them were examined to ensure no relevant publications had been missed. Two of the authors independently examined the titles and abstracts of the articles identified in the searches of the databases to assess their relevance. A protocol for this systematic review has not been published.
There is no accepted standard procedure for assessing the quality of climate models and their data, although some recommendations have been made [17,18]. The risk of bias was assessed in two domains (spread in projections amongst climate models and emissions scenarios used). The study quality was assessed by one reviewer.
Initially, 608 articles were identified from the literature searches after duplicate entries had been removed. Thirteen additional articles were selected from examination of reference lists and citing literature. After screening the titles and abstracts, the full texts of 130 articles were examined in detail. Of these articles, 63 were found to contain quantitative estimates of future heat-related mortality and so were selected for the systematic review (Fig 1). The locations studied, time periods, climate models, emissions scenarios and treatment of adaptation are summarised in Table 1, together with the meteorological variables used in each study. Further technical details of each study, specifically the variables used to model mortality, calibration methods, time of year considered and consideration of changes in population are shown in Table 2. The locations of the cities studied are shown in S1 Fig, and the time periods considered in each article are illustrated in S2 Fig.
The first known attempt to estimate future heat-related mortality under a warming climate was published in 1988 . Future mortality was estimated for fifteen cities in the USA using a range of prescribed temperature increases inferred from a single global climate model simulation. A small number of studies were published afterwards in the 1990s and early 2000s (Table 1). The number of studies of future heat-related mortality increased considerably after 2007, for which there are several possible reasons. Improved access to data from both global and regional climate model simulations around the same time could be one reason. Improved epidemiological methods, easier access to and speed of the Internet, and increased computational power of researchers’ workstations might be other factors.
The majority of the locations studied are cities in high income countries including Canada, USA, European countries, South Korea and Australia (Table 1). The locations and numbers of studies which estimated future mortality for each city are shown in S1 Fig. There are no studies specifically of rural areas, although two studies calculated future mortality in different regions of the UK which included both urban and rural populations [25,29]. Two other studies [30,31] considered both rural and urban areas in parts of the USA. There are very few or no projections of future mortality for populations in Africa, the Middle East, South America, and much of northern, central and southern Asia. The only studies of future mortality in tropical and sub-tropical areas are those for Thailand , China , Korea [34,35,36] and India . In some countries, complete mortality records may not be routinely available which limits epidemiological analyses.
Observations of local weather and climate
All of the selected studies have either used local observations of weather variables and mortality records to construct suitable mortality models, or used mortality functions from previous studies. Very few cities contain dense weather observation networks. These networks tend to be organised for specific research projects and only exist for short time periods (for example, ). Many studies of mortality within cities have therefore used observations from nearby airports or parks within the city. Temperatures at these locations may not be representative of conditions in other parts of the city . Estimates of the number of days above a threshold temperature could therefore be over- or underestimated in some city areas.
Some studies averaged observations within a given area to produce mean values for that area [29,27,31]. Given the sparseness of surface observations, these estimates may not represent the “true” area- averaged temperatures. Other studies used gridded temperature data which had been created by use of a regression model to interpolate irregularly spaced weather observations onto a regular grid [37,40,41,42,43]. These gridded datasets provide useful estimates of climate information for locations where surface-based observations are unavailable. Their accuracy depends on the number of weather stations available and the predictors used (for example, altitude, proximity to the coast, local topographical features). Some important climatic effects, such as the urban heat island, may not be included leading to underestimates of temperature in the gridded data at urban locations. Many observations contain measurement and sampling errors, but the magnitude of these errors is not always known. These errors are likely to be small compared with other sources of uncertainty.
A few studies have used climate data from reanalyses to supplement surface observations [37,44,45,46]. Global reanalyses are created by assimilating observations every 3–12 hours within a weather forecast model to provide a dynamically consistent description of the atmosphere. Global reanalyses available at the time of writing have spatial scales of the order of 30–180 km. In one study , climate variables from a global reanalysis were further downscaled and corrected using surface observations. Regional reanalyses, created by driving higher resolution models with climate data from global reanalyses, have resolutions of approximately 10–50 km [47,48].
Projections of future climate
In this section, projections of future climate are discussed. Global climate models are briefly described, followed by the scenarios used to drive them.
Global climate models.
Projections of future climate originate with global climate models (GCMs), which embody the current understanding of the dynamical, physical and biogeochemical processes that control the climate system . The GCMs used for the fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC), published in 2013, had horizontal resolutions between 60 and 150 km [50,51].
In addition to atmospheric processes, current GCMs include representations of the ocean and its circulation, aerosol particles, the land surface, land and sea ice, vegetation and the carbon cycle, and, more recently, atmospheric chemistry [50,51]. The ability of GCMs to simulate observed climate variables and their spatial patterns has continuously improved .
No two GCMs are identical; they contain different but plausible methods for representing climatic processes, numerical methods for solving equations and representations of processes which occur at spatial scales that cannot be resolved directly by the climate model . There are two important consequences of the choices made when constructing climate models. First, a range of changes in temperature, rainfall and other climate variables are produced by different climate models when they are forced with the same estimates of future greenhouse gas emissions. Secondly, systematic errors (or “biases”) are apparent when comparing simulations of present day climate with observations. Correction of these biases is especially important when absolute thresholds are used, as in temperature-mortality models. Methods for correcting biases, often referred to as calibration, are discussed below.
It is impossible to predict future emissions of greenhouse gas emissions and changes in land use. Projections of future climate are created by driving global climate models with greenhouse gas emissions or atmospheric concentrations prescribed in scenarios. Scenarios are neither forecasts nor predictions; they provide descriptions of possible future socioeconomic and technological changes, population growth and land use change, from which emissions of greenhouse gases can be estimated. The earliest scenarios used either a fixed increase in carbon dioxide levels (1% per year, for example) or involved executing a GCM with constant levels of carbon dioxide, at present day and doubled levels . These simulations provided useful information on possible future climatic conditions, but no indication of when those conditions might occur.
The IPCC published a set of six scenarios (“IS92”) in the early 1990s  which were used to assess climate change for the IPCC second and third assessment reports that were published in 1996 and 2001. Increased understanding of the driving forces of emissions and assessment methodologies led to the production of a new set of scenarios which are described in the Special Report on Emissions Scenarios . These “SRES” scenarios were derived from four different socioeconomic storylines based on various assumptions regarding population growth, technological changes, energy sources and land use . None of these scenarios included policies to reduce greenhouse gas emissions. The SRES scenarios most commonly used are high emissions (A1FI, A2), medium emissions (A1B, B2) and low emissions (B1). Climate simulations using a subset of the SRES scenarios were used to inform the third and fourth assessment reports of the IPCC. About half of the studies in Tables 1 and 2 used climate data generated under one or more of the SRES scenarios.
A new set of scenarios based on radiative forcings were developed to replace the SRES scenarios; they were given the label representative concentration pathways, or RCPs . The RCPs, unlike the SRES scenarios, are not based on socioeconomic storylines . Instead, a specific emission scenario, including land use and land cover changes, was identified which would lead to each target radiative forcing trajectory . The four RCPs include a mitigation scenario leading to a very low radiative forcing level (RCP2.6), two medium stabilisation scenarios (RCP4.5 and RCP6.0) and one very high emission scenario (RCP8.5). A comparison of global mean temperature changes and associated carbon dioxide levels shows that the SRES A1FI and RCP8.5 scenarios are similar, and the SRES A2 scenario lies between RCP6.0 and RCP8.5. SRES A1B is close to RCP6.0 and SRES B1 is similar to RCP4.5. The RCP2.6 scenario includes policies which result in net negative emissions of carbon dioxide, and so temperature changes projected with this scenario are notably lower than any projections using the SRES scenarios. Climate projections using four RCPs formed the basis of the Coupled Model Intercomparison Project 5  (CMIP5). These projections were analysed extensively for the IPCC fifth assessment report [1,2,3]. Thirteen studies used climate data generated under one or more of the RCPs to estimate future mortality (Table 1).
Two studies used other scenarios. A study of future mortality in cities in the USA  used two different scenarios (“REF” and “POL3.7”), which are similar to the RCPs but have different radiative forcings. A study of future mortality in the Netherlands  used four scenarios (“KNMI’14”) which were based on a downscaled subset of the CMIP5 projections.
No likelihood is attached to the IS92, SRES or RCP emissions scenarios. They are assumed to be equally plausible representations of future emissions. Ideally, future climates generated under all scenarios within a group (i.e., all SRES or all RCPs) would be used to explore the impacts of different policy options on projected mortality. However, it may not be practical or possible to do so owing to the large volumes of data involved. Some downscaled climate model datasets were created using a single emissions scenario (see next section).
Downscaling climate simulations
GCMs depict the climate using a three dimensional grid over the globe. Current GCMs have resolutions between about 60 km and 150 km . For many impacts studies, climate data are required at higher spatial scales, so downscaling is required. Downscaling refers to a process whereby climate information at large spatial scales is used to create projections at smaller spatial scales. There are two main approaches for downscaling, dynamical and statistical.
Dynamical downscaling involves the use of a regional climate model (RCM) over a smaller domain (e.g., a continent or country) at higher horizontal resolution. RCMs, like their parent GCMs, are based on physical principles. The resolution of RCMs has also increased over time. The RCMs used in earlier studies had resolutions of the order of 50 km (Table 1), whereas some more recent RCM simulations have resolutions between 10 km and 25 km. Occasionally, RCMs with higher horizontal resolutions have been used (250 m , 1 km , 4 km ).
Statistical downscaling is based on relationships between local climate variables, such as temperature and rainfall and large-scale “predictor” variables such as air pressure or temperature. First, these relationships are derived using observations of climate variables. Next, these relationships are applied to projections from GCMs to produce local climate data for the future. Statistical methods have been used to produce monthly, daily and sub-daily climate data at local scales [23,56,57].
Three studies used both downscaling methods [34,35,43]. A regional climate model was used to dynamically downscale global climate model simulations to a higher resolution, and then statistical methods were used to produce data at either specific locations or on a regular grid. A similar combination of methods was used to create the KNMI’14 scenarios .
Dynamical and statistical downscaling methods have advantages and disadvantages. RCMs can produce a wide range of climate variables at high spatial and temporal resolutions, but are computationally expensive to execute. Some ensembles of regional climate model projections have been created using only a single emissions scenario. For example, the ENSEMBLES project  used the SRES A1B scenario, whereas the North American Regional Climate Change Assessment Program  (NARCCAP) used the A2 scenario. The CORDEX initiative  has produced downscaled climate data using multiple global and regional climate models under the RCP scenarios for many land regions of the world. Climate data from RCMs are likely to require calibration, as any errors in the driving GCM climatology will also be present in the RCM climate. Calibration of climate data is discussed below.
Statistical methods require less computational resources than RCMs, but need a long series of observations of the climate variables of interest in order to establish robust relationships with large scale predictors. The relationships can vary temporally and spatially. Additionally, statistical methods implicitly assume that the relationship between local and large scale variables does not change over time, which may not be true. Statistical downscaling methods generally incorporate a calibration step.
Four studies [43,61,62,63] used a weather generator to create daily series of climate variables for specific locations. A weather generator is not, strictly speaking, a downscaling method, but can be used with other downscaling techniques to produce local climate information. A weather generator is a statistical model designed to generate synthetic but realistic series of climate variables of an arbitrary length. Weather generators incorporate a stochastic rainfall model which simulates rainfall sequences. Other climate variables (such as daily maximum and minimum temperatures) are then calculated from regression relationships with the rainfall amounts and current state (i.e., wet or dry) . Most weather generators operate on daily time scales, although some also produce hourly values of climate variables . Data for future time periods can be created by either adding climate change factors to a present-day series, or modifying the relationships between the weather variables. A long standing issue with weather generators is their inability to reproduce long periods of persistent weather such as warm temperatures and droughts .
The very first study of future heat-related mortality used prescribed temperature increases which were based on a single GCM simulation . Six other studies also used prescribed increases in temperature [32,33,65,66,67,68]. Of these studies, two [33,65] added temperature changes based on global mean changes to observed daily summertime temperatures of the cities under study. Analyses of regional temperature changes projected by global models shows that simulated temperature changes over individual land points are almost always greater than global mean changes [2,3]. The exact change in temperature varies considerably within a given region . Modelled temperature changes for the location of interest should therefore be used.
Most studies have used climate data from the grid box that encloses the location of interest. Several studies [37,69,70,71,72] interpolated modelled data to the point(s) of interest, whereas two others [73,74] used the average of the values from the central and eight surrounding boxes.
Several different climate variables have been used to project future mortality (Table 2), although justification for the choice of variable is rarely given. Two studies [32,75] related monthly mortality to monthly means of daily temperatures. Most other studies used daily data, of which the most common variables were daily maximum and daily mean temperatures (Table 2). Two studies used daily minimum temperatures [31,51], and so relate mortality to hot nights instead of hot days.
Some studies modelled mortality using variables calculated from temperature and humidity which are thought to be physiologically relevant (S1 Appendix). Apparent temperature, a function of air and dew point temperature was used by a number of studies [30,40,46,65,76,77,78]. One study  used the Humidex and another  the heat index . A study of mortality in Vienna  used physiological equivalent temperature (PET), which is calculated with a heat balance model of the human body.
Many of the climate variables (daily temperatures, apparent temperatures, etc) used to construct mortality models are correlated , so that the choice of variable may not be important. However, another study compared the numbers of days identified as being important for heat-related deaths using four different heat-health warning systems based on different climate variables . The numbers of hot days were dependent on the variable chosen, even though some of the variables were highly correlated. A study of mortality in seven US cities found that the correlations between different variables (daily minimum, mean and maximum temperatures) were weaker for the extremes than for the entire distribution . Additionally, daily maximum or mean temperatures were more strongly associated with mortality than minimum temperatures . Some climate models project larger increases in daily maximum temperatures during the warm season than daily mean or minimum temperatures . The use of mortality models based on daily minimum or mean temperatures may therefore produce lower estimates of future mortality than models based on daily maximum temperatures.
A small number of studies [44,56,67,76,85,86,87] classified air masses into different weather types based on temperature, humidity and other characteristics. Models linking mortality with metrics such as apparent temperature (Table 2; S1 Appendix) were then built separately for oppressive weather types (those associated with high temperatures and/or high humidity likely to cause large increases in mortality) and other weather types. This approach has the advantage of not requiring any downscaling of global climate model data to local levels. However, the ability of climate models to simulate the correct numbers and seasonality of the air mass types was not always assessed. Errors in the modelled circulation could result in over- or under-estimation of oppressive air mass types which would impact upon projected changes in mortality.
Calibration of climate data
Climate models have improved considerably since they were first developed in the late 1960s . Despite the continuous developments, systematic errors or biases (for example, over- or underestimation of summer temperatures, simulation of too many wet days) are apparent when simulations of present day climate are compared with observations, although the magnitudes of these biases have steadily decreased . Correction of biases in the modelled climate data (a procedure referred to as calibration in the present study) is therefore required. Any biases in the modelled climate would affect the estimated baseline mortality and projected changes in mortality, which can be significant [42,88]. Despite this issue, four studies appear to have used raw (i.e. uncalibrated) climate model data to estimate future mortality [63,71,77,89] which would introduce errors into their estimates.
Almost all of the studies in Table 1 have combined data from climate models with local observations of climate to reduce biases in the modelled data. Regardless of the method used, the climate change information from a GCM or RCM is often at a coarser resolution than the local climate data. Hence, the local scale characteristics of the calibrated data are dependent on the observations whereas climate change effects are controlled by coarse-scale data from climate models .
The calibration methods used by the studies are summarised in Table 2 and most fall into one of two groups, “delta” and “shift”. Under the “delta” method, differences in modelled climate between a baseline and future period (called change factors) are calculated and added to an observed time series . In many cases, the time scales of the change factors were different to the observed time series. Monthly or annual mean changes in temperature were added to observed daily temperatures in most studies (Table 2). One study calculated monthly temperature changes from climate model simulations and added them to observed monthly mean temperatures . Two studies calculated average daily temperature changes and added them to observed daily temperatures [51,91].
Seven studies used the shift method to calibrate their climate data (Table 2). Modelled and observed data over a common period were used to calculate daily or monthly correction factors which were then added to the modelled data over all time periods . Seven other studies used the percentile approach to calibrate their modelled data (Table 2). The percentile of the threshold temperature from observations (above which excess mortality occurs) is applied to modelled temperatures in the baseline period. The new threshold is then used with the modelled data. The percentile approach is equivalent to using the shift method with a single value, as the same value would effectively be added to the entire modelled temperature distribution.
Other methods have been used to calibrate climate data, including quantile mapping and fitting of functions to the distributions of the data. Four studies used quantile mapping methods [37,40,62,92]. Empirical cumulative distribution functions of the modelled and observed data are used to calculate a correction factor for each pair of data values. This method can be extended to include modelled changes between a baseline and future periods.
Two studies fitted logistic distribution functions to the distributions of the modelled and observed temperatures [42,88]. The logistic function is defined by two parameters analogous to the mean and standard deviation of a normal distribution. Changes in the parameters between the functions fitted to the modelled future and baseline temperature distributions were then added to the respective parameters estimated from the observational distribution. Finally, the resulting distribution was sampled to produce a daily series of calibrated temperatures for the future periods.
The ability of various calibration methods, including the “delta”, “shift” and quantile mapping, to reduce errors and reproduce high and low extremes of temperature distributions has been assessed . It was found that the delta and shift methods performed the worst in reproducing the higher and lower temperatures in the distribution compared with quantile mapping. The delta method performed the worst overall, whereas the shift was the worst method for reproducing temperatures in the upper half of the distribution . The shift method implicitly assumes that the biases are stationary, so that the correction factors calculated for the present day are applicable to future periods. This assumption of constant biases may not be correct, and could be invalidated seasonally, geographically, and also with the amount of global warming .
Most of the studies reviewed have considered the well established effect of general summertime temperatures on mortality . Twelve of the studies considered the effects of heat waves on future mortality and are summarised in Table 3. Heat waves (a period of consecutive anomalously hot days and/or hot nights) are comparatively rare events, whereas warm and hot days occur in most if not all summers. Hence, mortality attributable to hot days is generally much larger than mortality from heat waves. A wide range of heat wave definitions have been used (Table 3), which makes comparisons between studies difficult. Five studies modelled mortality as a function of the lengths of the heat waves [31,35,40,94,95], whereas five others used temperatures during the heat waves [29,55,70,92,96]. One study calculated a separate mortality risk for each day in sequence of the heat wave , and another modelled mortality as a function of the number of heat wave days in the hot season .
Only three studies calculated mortality from both individual hot days and heat waves. The importance of heat waves differed considerably. A study of future mortality from hot days and heat waves in three cities in Alabama found that deaths from heat waves were at most a few percent of the deaths from high temperatures . In contrast, deaths from heat waves were estimated to be of similar magnitude to deaths from individual hot days for four cities in Spain . In a study of heat-related mortality in UK regions during the twenty-first century, the extra mortality from heat waves was only important for London .
A systematic review of heat wave definitions and associated mortality  concluded that the impact of heat waves on mortality was important, but the magnitude of the effect varied under different heat wave definitions. Generally, the higher the temperature threshold used, the higher the impact on mortality. The intensity of the heat wave appeared to be more important for mortality than the duration. However, it was unclear whether the effects of the intensity and duration of heat waves were independent or interactive .
There is considerable evidence to show that populations in some areas have adapted to warmer temperatures over past few decades [98,99,100,101,102,103]. Adaptation can occur via physiological acclimatisation and behavioural changes. Other mechanisms include improved health care, provision of heat-health warning systems  and increased installation of air conditioning systems . Adaptation to warmer temperatures can occur within a season and over many years. Only two studies included within-season adaptation [40,85], where modelled mortality at the end of the warm season would be lower than at the beginning given the same climatic conditions.
Most of the methods by which longer-term adaptation in its various forms has been included in quantitative estimates of future mortality are simplistic. In many studies, the mortality threshold was increased by an arbitrary amount with no justification or reference to epidemiological evidence (Table 1). Some studies reduced the gradient of the temperature-mortality relationship [30,46,53,75,85,105]. Other studies estimated the effects of adaptation by extrapolating mortality-temperature trends into the future , or simply ignoring mortality in the first few days of a heat wave . A small number used the “analogue city” approach, where the mortality model for a city with a warmer climate is applied to the city under study [28,69,86]. None of the studies have considered “negative adaptation”, where communities become less well adapted to warmer temperatures, owing to failure of power generation or transmission grids in populations accustomed to using air-conditioning, for example .
Future changes in population and demographics (specifically aging) are important when calculating heat-related mortality. The numbers of deaths would be expected to increase owing to larger populations and projected higher proportions of older people who would be more vulnerable to the effects of high temperatures. Many studies did not include estimates of population growth in their projections (Table 1) and so would underestimate the numbers of heat-related deaths in the future. Several studies have shown that projected numbers of deaths were considerably higher when population growth and changes in demographics were included, compared with results using a static population [29,55,70].
Sixty three papers estimating future mortality from high temperatures and heat waves have been reviewed. These studies used a wide range of surface observations and climate model projections (Tables 1 and 2). All the studies indicate that heat-related mortality would increase under a warming climate. The projected impacts of climate change on mortality are highly dependent on the future scenarios and climate models chosen. The majority of studies have used a small number of climate simulations without considering where they lie within the range of future projections. The difficulty in obtaining and processing data from climate models is likely to have been one factor in older studies, although access to climate model data has greatly improved in recent years. The use of climate information from a small number of models means future mortality estimates could be biased low or high. Ideally, all available simulations would be used to estimate future mortality. Alternatively, a subset of the climate models could be selected which captures key regional climate processes and the range of possible changes in climate [18,53]. Calibration of data from climate models is required to reduce the impacts of any biases. It should be noted that no calibration method will remove all deficiencies in the modelled climate, such as over- or under-prediction of the persistence of periods of hot and cold weather.
Uncertainty in estimates of future mortality originates from several sources, including climate models and emissions scenarios , the calibration method, the mortality model , treatment of adaptation [44,46] and future population changes . Consequently, estimates of future mortality for the same city can vary considerably between different studies. As an example, mortality estimates for Chicago from five studies are compared in S3 Fig. These estimates vary by a factor 4 or more. In areas where surface observations are sparse and the terrain is complex, uncertainty in interpolated or gridded data derived from observations can be large . This issue has not been considered in projections of climate impacts on health.
The relative importance of these various sources of uncertainty is likely to change temporally. In the near future, the choice of mortality model might be one of the larger sources of uncertainty [70,84], whereas over longer time periods climate model uncertainty, the choice of scenario and treatment of adaptation would become more important [40,44,70,73,75,91,107,108]. A study of mortality in 14 European cities for the end of the twenty-first century using six different adaptation methods showed that the uncertainty originating from the adaptation methods was mostly larger than uncertainty from climate models and emissions scenarios .
Two studies [27,78] used median or mean changes in temperature from an ensemble of climate model simulations and did not consider the range of projections. Two others [109,110] estimated mortality using projections from multiple climate models but only reported median changes in mortality. There were large differences in projected mortality in all of the studies which used data from two or more climate models driven by same emissions scenario (e.g., [76,88,111,112]).
A few studies quoted future mortality estimates as averages or ranges across different emissions scenarios [31,62,84,107]. Another study appears to have used a weighted average of projections from two scenarios which were created with two different models to estimate future mortality . These mortality averages and ranges are very dependent on the scenarios and models used and are therefore potentially misleading. Results should be presented separately for each scenario.
About half of the studies reviewed here used climate data generated with medium or high emissions scenarios (SRES B2, A1B, A1FI and A2; RCP6.0 and RCP8.5). A smaller number used lower emission scenarios (SRES B1; RCP4.5). The Paris Agreement to limit global warming to less than 2°C and pursue efforts to limit the temperature increase to 1.5°C was ratified in November 2016. The RCP2.6 scenario is the only one consistent with the aims of the Paris Agreement, but just three of the studies [36,37,46] considered it. Further studies of future mortality using the RCP2.6 scenario are therefore required. However, even if global warming was limited to 2°C, increases in numbers of hot days and lengths and intensities of heat waves are still likely .
The use of short time lengths for a baseline and future periods (for example, 3–5 years; [31,33,51,69,77]) should be avoided. Regional climate in most areas of the world is highly variable, and the climate in a short period could be anomalously warm or cold relative to a longer-term average. Very different projections of mortality could be obtained if different years had been chosen for the baseline and future periods. One study  showed that projected mortality using 3-year and 30-year future periods were very different, by about a factor of 2.
Some studies have only considered mortality in the summer months or warm season (for example, June–August or May–September in the Northern Hemisphere). This approach would exclude unusually warm months outside of these periods and bias mortality estimates low. For example, in 2003 and 2011, temperatures in April in London exceeded the mortality threshold used in  for several days. The length of the warm season would be expected to expand in the future , increasing the chance that some warm days and the associated mortality would not be included if a fixed time period was used.
The variable chosen to model heat-related mortality can also affect estimates of future mortality. A study of mortality in 107 cities in the USA  used several different climate variables (such as daily maximum temperature and apparent temperature). The best variable for modelling mortality varied between the different cities. Another study  estimated future mortality in Europe using two different variables and obtained similar (but not identical) results. In contrast, a study of mortality in the UK  found that mortality was best modelled using daily maximum temperatures. Some climate models project larger increases in daily maximum temperatures during the warm season than daily means and minimums . Estimates of future mortality will be partly dependent on the variable chosen, but the importance of variable choice is likely to vary with location.
The urban population is growing and is expected to continue increasing in the future . Urban areas have their own climates which are different to surrounding rural areas. They are generally warmer than rural areas, especially at night, owing to absorption and release of heat by buildings, waste heat from energy use, and a lack of surface moisture [55,116]. The temperature differences between towns and cities and rural areas are referred to as the urban heat island (UHI). The UHI can reach values of up to 10°C in large cities . Urban populations are therefore exposed to higher temperatures than rural populations. Urban temperatures in the future could increase from expansion of urban centres as well as the warming climate .
Many climate models do not simulate urban climates, so that future heat-related mortality within cities is likely to be underestimated. Three of the studies in Tables 1 and 2 used models which explicitly simulated urban climates [54,55,61]. One study  used a modified weather generator to simulate the climate of London. The effects of climate change as well as increased urbanisation and anthropogenic heat emissions on mortality in London were examined. Future mortality increased as a result of climate change, and the inclusion of increased urban land use and anthropogenic heat release resulted in a further increase in mortality of about 10–15%.
Another study calculated the effect of the urban climate on mortality in a large city in the UK during a severe heat wave . Calculations of heat-related mortality over the same period in which urban areas had been replaced by a rural land type were about 50% smaller. Future mortality, estimated by adding temperature changes from a regional climate model to the modelled present-day temperatures, was notably higher when urban temperatures were used compared with the rural values.
The impacts of increased urbanisation on the climate of a city in a semi-arid area have been studied . Daily minimum temperatures were increased by a larger amount than daily mean temperatures under greater urbanisation; little effect was simulated on daily maximum temperatures. Changes in heat-related mortality based on daily minimum temperatures were notably higher than estimates based on daily mean temperatures . Changes in mortality based on changes in daily maximum temperatures were small and negative, so that the increased urbanisation acted to reduce mortality slightly. Similar changes in temperatures and impacts were found in a study of heat stress (but not mortality) in Sydney resulting from urban expansion and climate change .
These studies mates highlight the fact that changes in urban climates are different to those in rural areas. The extra heat from anthropogenic activities further raises urban temperatures but was only considered in one study . Many global and regional climate models do not explicitly simulate urban climates. High resolution model simulations of urban areas are few in number, probably due to the high computational cost of running such models. If climate data from these high resolution climate simulations are used for estimating future mortality, the variable used to model mortality needs to be chosen with care.
One issue with heat-mortality models occurs when they are used with temperatures higher than those used to construct them. Projections based on simple linear models could underestimate mortality, especially when extreme temperatures are experienced. Some non-linear models have very steep gradients for high temperatures [42,70,78]. A small increase in temperature would produce a very large increase in mortality which might be unrealistic. It could be insightful to compare the projected temperature changes with the calibration range to understand how much extrapolation is occurring.
When presenting estimates of future mortality, results using no change in population and demographics should be given alongside results with population and demographic changes. The separate effects of changes in climate and changes in population on mortality can then clearly be seen. For example, in a study of heat-related mortality in the UK , mortality was estimated to increase by 66%, 257% and 535% by the 2020s, 2050s and 2080s respectively when population size and aging was included, but only by 46%, 169% and 329% if population size and ages were held constant. Large differences in projected mortality were calculated in other studies using different scenarios of population growth and aging [70,118].
Adaptation or acclimatisation of the population to warmer temperatures is important when estimating future mortality . There is evidence from the epidemiological literature that in some locations, heat related mortality has decreased over time [98,99,100,101,102,103]. These results might suggest that existing measures are keeping pace with warming so far, but it is unclear if and how such measures could continue to succeed in the future. Some degree of adaptation to heat is likely to have occurred naturally . Evidence of short-term adaptation to heat is also supported by physiological studies . It is not possible to say how much of the decreased sensitivity to heat demonstrated in these studies is due to physiological, behavioural or adaptive structural changes in the environment (e.g., increased availability of air conditioning, planting of trees to provide shade, etc).
In brief, there are two aspects which can be considered: adaptation within the warm season and longer term adaptation to warmer temperatures. Two studies accounted for possible in-season adaptation, so that mortality at the end of a warm season would be lower than at the beginning, given the same climatic conditions [40,65]. Longer term adaptation has been estimated using several different methods. Three studies used the differences in mortality between warm and cool summers [40,85,107]. Many other studies have estimated the possible effects of adaptation by increasing the mortality threshold temperature by a fixed amount, typically between 0.5°C and 4.0°C (e.g., [42,46,61]). Other studies changed the slope of the exposure-response function [30,46,53,75,85,105]. These changes to the threshold temperatures or slopes are often arbitrary and are rarely supported by epidemiological evidence .
The differences in future mortality estimates incorporating six different adaptation models has been studied for selected European cities . This study showed that uncertainty in future heat-related mortality resulting from different adaptation methods was larger than uncertainty associated with emissions scenarios and climate models. A study of mortality in Beijing reached similar conclusions . There is a need for greater evaluation of intervention methods to improve modelling of adaptation within epidemiological models .
Heat waves are rare events, meaning there are few examples to study. It is therefore difficult to assess by how much mortality could be elevated by the persistence of the hot conditions during heat waves. The effects of heat waves on future mortality have only been assessed by twelve studies (Table 3), and of those only three modelled mortality from both hot days and heat waves. The importance of heat waves for excess mortality varied considerably. Heat waves are projected to become more frequent, hotter and longer as the climate warms . Whether the importance of heat waves compared with individual hot days for mortality would also increase in the future is unclear.
The timing of hot days and heat waves may also be important. Those that occurred early in the warm season in temperate zones might have a larger effect on mortality than those which happened later [7,85,121,122]. High temperatures have the largest effect on older people, whereas prolonged heat waves can impact on the entire population. A modified mortality relationship may be needed for heat waves than for the general effect of high temperatures. It is unclear whether the lengths and intensities of heat waves act independently or synergistically on mortality . Further research is needed to fully understand the effects of heat waves on mortality.
Factors that have not generally been considered
There are several other factors which could be important when estimating future heat-related mortality, but have not been included in the studies reviewed here. Summer mortality from high temperatures may be moderated by mortality in the previous winter [123,124,125]. If mortality during winter was low, mortality in the following summer could be elevated, owing to a larger number of vulnerable people. Similarly, a winter with high mortality could mean mortality in the following summer would be reduced. However, aside from a few studies, the epidemiological evidence for linkages between winter and summer mortality are not well established. Any linkage may reduce in importance as the climate warms.
The importance of the socioeconomic status of the population when calculating future mortality is unclear. Two studies [126,127] found some evidence to show that populations in deprived areas of Chicago and Paris respectively were more vulnerable to heat related mortality than those in affluent areas. In contrast, a study of mortality in Australian cities  saw little or no evidence for modification of the temperature-mortality relationships by socioeconomic status.
A rapid change in temperature within a day (the diurnal temperature range, DTR) could be a risk to human health . Those with cardiovascular and respiratory diseases are most at risk from large changes in DTR. The elderly and children appear to be more susceptible to large changes in the DTR than other age groups. Further research is needed to confirm and understand any effects of DTR on health and mortality . So far, DTR has not been included as an explanatory factor in models used to estimate future mortality.
Strengths and limitations
This review is the first to synthesise and critically assess the use of information from climate models to estimate future heat-related mortality. Additionally, three factors which have not been considered so far in projections of future mortality were identified.
There are several limitations of the current review. Many different combinations of keywords were used to identify relevant articles, but some may have been missed. Only articles published in English were searched for, so any relevant studies in other languages will have been omitted. The databases searched did not index journals in fields such as economics and social sciences. Two relevant articles published in economics journals were found in reference lists of other papers, and there may be other papers in similar journals which would be of interest. Some studies may not have been published, especially if they had negative results, so some degree of publication bias cannot be ruled out. All articles in which an estimate of future heat-related mortality was reported were included regardless of the quality of the study. The assessment of quality is subjective, and was not included in the eligibility criteria. However, inclusion of those papers regarded as low quality would not alter the conclusions or recommendations of this review.
Heat-related mortality is now recognised as a serious issue which is likely to increase in severity as the climate warms. Studies of future mortality would benefit from more interdisciplinary collaborations to improve the quality of research and results and also to broaden the interest and readership of work that is likely to be important for public policy across a number of sectors. There are very few studies of the effects of warm temperatures on mortality outside of high-income countries. More studies are needed in middle and low income countries, and in sub-tropical and tropical areas. Additionally, there are very few studies specifically estimating future mortality in rural populations.
Adaptation will play a key role in reducing the effects of a warmer climate on heat-related mortality. In some locations, heat-related mortality has fallen over the past few decades, suggesting existing measures are keeping pace with the rate of warming. It is unclear whether these existing measures will continue to succeed in the future; some newer measures are likely to be needed. Methods by which adaptation has been included in mortality models are often simplistic and are not linked to epidemiological evidence. More research is needed to improve the representation of adaptation within mortality models.
Uncertainty in projections of future mortality originates from several sources, but only a small number of studies have partially or fully addressed this issue. In particular, uncertainty from the choice of climate model simulation(s) is not often considered. Ideally, climate projections from multiple models under different emissions scenarios with a range of greenhouse gas emissions would be used to estimate future heat-related mortality. Alternatively, a subset of the climate model projections which captures the range of climate change over the area of interest could be selected.
Estimates of future heat-related mortality are partially controlled by two competing effects: population growth and aging would act to increase mortality, whereas adaptation would reduce mortality. It would be of interest to understand how the magnitudes of these two effects change temporally within epidemiological models and with the amount of warming.
The recognition of and interest in heat-related mortality may provide opportunities for more interdisciplinary studies involving both epidemiologists and climate scientists to better estimate the impacts of high temperatures on health and properly include uncertainty in the projections of future climate. The outcomes of such studies should be directed toward answering policy-relevant questions and contributing toward the design of suitable adaptation measures.
There are many different factors to be considered when estimating future heat-related mortality. Much depends on what data are available for any given area and the purpose of the research, for example, exploring worst case scenarios and impacts of different policy options. One important recommendation from this review is transparency in reporting, ensuring the data used fit the purpose of each study and any limitations are reported.
An example ‘checklist’ for authors to consider when publishing results using climate model projections is suggested in Table 4.
S1 Fig. Cities for which future mortality has been estimated.
The symbols indicate the number of studies of mortality for that city.
S2 Fig. Time periods considered in each study of future mortality.
Magenta lines indicate the periods of observations used. Green and black lines show the model baseline and future time periods. Dashed lines and open symbols indicate time periods implied but not stated by the authors, or where prescribed temperature increases are assumed to represent the indicated time period. Single years are shown by solid or open circles.
S3 Fig. Present day and estimated future mortality rates for Chicago.
Mortality rates are in units of deaths per 100,000 of population. Mortality shown in magenta and green were calculated using high (SRES A1FI, A2) and low (SRES B1) emissions scenarios respectively. Mortality rates in grey were estimated using other scenarios. Error bars (where shown) represent the lowest and highest estimates using data from multiple climate models with the same mortality model. The estimates from Kalkstein and Greene (1997) were made using three different GCMs, and assume full adaptation of the population to the future temperatures. The mortality estimates for Chicago were normalised to deaths per 100,000 of population using census data for the specified year: Kalkstein and Smoyer (1993) - 1970 census, population 3366957. Kalkstein and Greene (1997) - 1980 census, population 3005072. Greene et al. (2011) - 2000 census, population 2896000. Petkova et al. (2014) - 2010 census, population 2707120. Hayhoe et al. (2011) quoted mortality as deaths per 100,000 and so their results are shown without any modification.
- 1. IPCC, 2013a: Summary for Policymakers. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change Change [Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 1–30.
- 2. Collins M, Knutti R, Arblaster J, Dufresne JL, Fichefet T, Friedlingstein P, et al. 2013. Long-term Climate Change: Projections, Commitments and Irreversibility. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 1029–1136
- 3. IPCC, 2013b: Annex I: Atlas of Global and Regional Climate Projections Supplementary Material RCP8.5 [van Oldenborgh GJ, Collins M, Arblaster J, Christensen JH, Marotzke J, Power SB, Rummukainen M, Zhou T (eds.)]. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.
- 4. Bassil KL, Cole DC. 2010. Effectiveness of public health interventions in reducing morbidity and mortality during heat episodes: A structured review. Int J Environ Res Pub Health 7:991–1001.
- 5. 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 pmid:21816703
- 6. Gover M. 1938. Mortality during periods of excessive temperature. Public Health Reports 53(27):1122–1143.
- 7. Basu R, Samet JM. 2002. Relation between elevated ambient temperature and mortality: A review of the epidemiologic evidence. Epidem Rev 24:190–202.
- 8. Gosling SN, Lowe JA, McGregor GR, Pelling M, Malamud BD. 2009a. Associations between elevated atmospheric temperature and human mortality: a critical review of the literature. Clim Change 92:299–341.
- 9. Pascal M, Le Tertre A, Saoudi A. 2012. Quantification of the heat wave effect on mortality in nine French cities during summer 2006. PLoS Curr 4:RRN1307, pmid:22453900
- 10. Public Health England. 2015. Heatwave Plan for England. Protecting health and reducing harm from severe heat and heatwaves. Public Health England, London, UK.
- 11. Rouvière C, Williams T, Ball R, Shinyak Y, Topping J Nishioka S, et al. 1990. Human settlement; the energy, transport and industrial sectors; human health; air quality; and changes in ultraviolet-B radiation, in Climate Change: The IPCC Impacts Assessment. Report prepared for Intergovernmental Panel on Climate Change by Working Group II. Tegart WJMcG, Sheldon GW, Griffiths DC (eds). Australian Government Publishing Service, Canberra, Australia.
- 12. Ando M, Carcavallo R, Epstein P, Jendritzky G, Kalkstein L, Odongo R, et al. 1996. Human Population Health, in: Climate Change 1995. Impacts, Adaptations and Mitigation of Climate Change: Scientific-Technical Analyses. Contribution of Working Group II to the Second Assessment Report of the Intergovernmental Panel on Climate Change, Watson RT, Zinyowera MC, Moss RH, Dokken DJ (Eds), Cambridge University Press, Cambridge, UK.
- 13. McMichael A, Githeko A, Akhtar R, Carcavallo R, Gubler D, Haines A, et al. 2001. Human Health. Climate Change 2001: Impacts, adaptation and vulnerability. Contribution of Working Group II to the Third Assessment Report of the Intergovernmental Panel on Climate Change, Parry ML, Canziani OF, Palutikof JP, van der Linden PJ, Hanson CE (eds), Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 451–485.
- 14. Confalonieri U, Menne B, Akhtar R, Ebi KI, Hauengue M, Kovats RS, et al. 2007. Human Health. Climate Change 2007: Impacts, adaptation and vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Parry ML, Canziani OF, Palutikof JP, van der linden PJ, Hanson CE, eds., Cambridge University Press, Cambridge, UK, 391–431.
- 15. Smith KR, Woodward A, Campbell-Lendrum D, Chadee DD, Honda Y, Liu Q, et al. 2014. Human Health: Impacts, adaptation, and co-benefits. 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 University Press, Cambridge, United Kingdom and New York, NY, USA, 709–754.
- 16. Armstrong B. 2006. Models for the relationship between ambient temperature and daily mortality. Epidemiology 17:624–31. pmid:17028505
- 17. IPCC 2010. Meeting report of the Intergovernmental Panel on Climate Change Expert Meeting on assessing and combining multi model climate projections [Stocker TF, Qin D, Plattner G-K, Tignor M, Midgley PM (eds.)]. IPCC Working Group I Technical Support Unit, University of Bern, Bern, Switzerland, 117pp.
- 18. McSweeney CF, Jones RG, Lee RW and Rowell DP. 2015. Sub-selecting CMIP5 GCMs for downscaling over multiple regions. Clim Dyn 44:3237–3260.
- 19. Leggett J, Pepper WJ, Swart RJ, Edmonds J, Meira Filho LG, Mintzer I, et al. 1992. Emissions scenarios for the IPCC: an update. Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, Cambridge University Press, UK.
- 20. IPCC, 2000. Special Report on Emissions Scenarios [Nakicenovic N, Swart R (Eds)], Cambridge University Press, Cambridge, UK.
- 21. Moss RH, Edmonds JA, Hibbard KA, Manning MR, Rose SK, van Vuuren DP, et al. 2010. The next generation of scenarios for climate change research and assessment. Nature 463:747–756. pmid:20148028
- 22. Mills D, Schwartz J, Lee M, Sarofim M, Jones R, Lawson M. 2015. Climate change impacts on extreme temperature mortality in select metropolitan areas in the United States. Clim Change 131:83–95.
- 23. Brekke L, Thrasher BL, Maurer EP, Pruitt T. 2013. Downscaled CMIP3 and CMIP5 Climate and Hydrology Projections: Release of Downscaled CMIP5 Climate Projections, Comparison with preceding Information, and Summary of User Needs, prepared by the U.S. Department of the Interior, Bureau of Reclamation, Technical Services Center, Denver, Colorado. 47 pp.
- 24. Murphy JM, Sexton DMH, Jenkins GJ, Boorman PM, Booth BBB, Brown CC, et al. 2009. UK Climate Projections Science Report: Climate change projections. Met Office Hadley Centre, Exeter, UK.
- 25. Vardoulakis S, Dear K, Hajat S, Heaviside C, Eggen B, McMichael AJ. 2014. Comparative assessment of the effects of climate change on heat- and cold-related mortality in the United Kingdom and Australia. Environ Health Perspect 122:1285–1292. pmid:25222967
- 26. Weedon GP, Gomes SS, Viterbo PP, Shuttleworth WJ, Blyth EE, Österle HH, et al. 2011. Creation of the WATCH forcing data and its use to assess global and regional reference crop evaporation over land during the twentieth century. J Hydrometeorol 12:823–848.
- 27. Räisänen J, Räty O. 2013. Projections of daily mean temperature variability in the future: cross-validation tests with ENSEMBLES regional climate simulations. Clim Dyn 41:1553–1568.
- 28. Kalkstein LS. 1988. The impacts of predicted climate changes upon human mortality. Publications in Climatology, 41(1), 120 pp.
- 29. Hajat S, Vardoulakis S, Heaviside C, Eggen B. 2014. Climate change effects on human health: projections of temperature-related mortality for the UK during the 2020s, 2050s and 2080s. J Epidemiol Community Health 68:641–648. pmid:24493740
- 30. Ostro B, Rauch S, Green S. 2011. Quantifying the health impacts of future changes in temperature in California. Environ Res 111:1258–1264. pmid:21975126
- 31. Wu J, Zhou Y, Gao Y, Fu JS, Johnson BA, Huang C, et al. 2014. Estimation and uncertainty analysis of impacts of future heat waves on mortality in the eastern United States. Environ Health Perspect 122:10–16. pmid:24192064
- 32. Tawatsupa B, Dear K, Kjellstrom K, Sleigh A. 2014. The association between temperature and mortality in tropical middle income Thailand from 1999 to 2008. Int J Biometeorol 58:203–215. pmid:23100101
- 33. Zhang J, Tian TL, Guo TJ, Huang CR, Dong KH. 2014. Impact of temperature on mortality in three major Chinese cities. Biomed Env Sci 27:485–496.
- 34. Kim Y-M, Kim S, Liu Y. 2014. The impact of climate change on heat-related mortality in six major cities, South Korea, under representative concentration pathways (RCPs). Front Environ Sci 2,
- 35. Kim D-W, Deo RC, Chung J-H, Lee JS. 2016. Projection of heat wave mortality related to climate change in Korea. Nat Hazards 80:623–637.
- 36. Lee JY, Kim H. 2016. Projection of future temperature-related mortality due to climate and demographic changes. Environment International 94:489–494. pmid:27316627
- 37. Murari KK, Ghosh S, Patwardhan A, Daly E, Salvi K 2015. Intensification of future severe heat waves in India and their effect on heat stress and mortality. Reg Environ Change 15:569–579.
- 38. Watkins R, Palmer J, Kolokotroni M, Littlefai P. 2002. The London heat island: results from summertime monitoring. Building Serv Eng Res Technol 23:97–106.
- 39. Dousset B, Gourmelon F, Laaidi K, Zeghnoun A, Giraudet E, Bretin P, et al. 2011. Satellite monitoring of summer heat waves in the Paris metropolitan area. Int J Climatol 31:313–323.
- 40. 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 USA 101:12422–12427. pmid:15314227
- 41. Takahashi K, Honda Y, Emori S. 2007. Assessing mortality risk from heat stress due to global warming. J Risk Res 10:339–354.
- 42. Gosling SN, McGregor GR, Lowe JA. 2009b. 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 Biometeorol 53:31–51.
- 43. Morabito M, Crisci A, Moriondo M, Profili F, Francesconi P, Trombi G, et al. 2012. Air temperature-related human health outcomes: Current impact and estimations of future risks in Central Italy. Sci Tot Env 441:28–40.
- 44. 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.
- 45. Honda Y, Kondo M, McGregor G, Kim H, Guo Y-L, Hijioka Y, et al. 2014. Heat-related mortality risk model for climate change impact projection. Environ Health Prev Med 19:56–63. pmid:23928946
- 46. Gosling SN, Hondula DM, Bunker A, Ibarreta D, Liu J, Zhang X, et al. 2016. Adaptation to climate change: a comparative analysis of modelling methods for heat-related mortality. Environ Health Perspect, (in press).
- 47. Mesinger F, DiMego G, Kalnay E, Mitchell K, Shafran PC, Ebisuzaki W, et al. 2006. North American Regional Reanalysis. Bull Amer Meteorol Soc 87:343–360.
- 48. Vautard R, Gobiet A, Jacob D, Belda M, Colette A, Déqué M, et al. 2013. The simulation of European heat waves from an ensemble of regional climate models within the EURO-CORDEX project. Clim Dyn 41:2555–2575.
- 49. Sexton DMH, Murphy JM, Collins M, Webb ML. 2012. Multivariate probabilistic projections using imperfect climate models part I: outline of methodology. Clim Dyn 38:2513–2542.
- 50. Cubasch U, Wuebbles D, Chen D, Facchini MC, Frame D, Mahowald N, et al. 2013. Introduction. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 119–158.
- 51. Flato G, Marotzke J, Abiodun B, Braconnot P, Chou SC, Collins W, et al. 2013: Evaluation of Climate Models. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker T.F., Qin D., Plattner G.-K., Tignor M.,Allen S.K., Boschung J., Nauels A., Xia Y., Bex V. and Midgley P.M. (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.
- 52. Taylor KE, Stouffer RJ, Meehl GA. 2012. An overview of CMIP5 and the experiment design. Bull Amer Meteor Soc 93:485–498.
- 53. Huynen MMTE Martens P. 2015. Climate change effects on heat- and cold-related mortality in the Netherlands: A scenario-based integrated environmental health impact assessment. Int J Environ Res Pub Health 12:13295–13320.
- 54. Martinez GM, Baccini M, De Ridder K, Hooyberghs H, Lefebvre W, Kendrovski V, et al. 2016. Projected heat-related mortality under climate change in the metropolitan area of Skopje. BMC Public Health 16:407, pmid:27183821
- 55. Heaviside C, Vardoulakis S, Cai X-M. 2016b. Attribution of mortality to the urban heat island during heatwaves in the West Midlands, UK. Environ Health 15(Suppl 1):S27, pmid:26961286
- 56. Greene S, Kalkstein LS, Mills DM, Samenow J. 2011. An examination of climate change on extreme heat events and climate–mortality relationships in large U.S. cities. Wea Clim Soc 3:281–292.
- 57. Kingsley SL, Eliot MN, Gold J, Vanderslice RR, Wellenius GA. 2016. Current and projected heat-related morbidity and mortality in Rhode Island. Environ Health Perspect 124:460–467. pmid:26251954
- 58. van der Linden P, Mitchell J. (Eds.) 2009. ENSEMBLES: Climate change and its impacts: Summary of research and results from the ENSEMBLES project, 160 pp, Met Office Hadley Centre, Exeter, UK.
- 59. Mearns LO, Arritt R, Biner S, Bukovsky MS, McGinnis S, Sain S, et al. 2013. Climate change projections of the North American Regional Climate Change Assessment Program (NARCCAP). Clim Change 120:965–975.
- 60. Giorgi F, Gutowski WJ 2015. Regional dynamical downscaling and the CORDEX initiative. Ann Rev Environ Res 40:467–490.
- 61. Jenkins K, Hall J, Glenis V, Kilsby C, McCarthy M, Goodess C, et al. 2014. Probabilistic spatial risk assessment of heat impacts and adaptations for London. Clim Change 124:105–117.
- 62. Guo Y, Li S, Liu DL, Chen D, Williams G, Tong S. 2016. Projecting future temperature-related mortality in three largest Australian cities. Environ Pollut 208:66–73. pmid:26475058
- 63. Baaghideh M, Mayvaneh F. 2017. Climate change and simulation of cardiovascular disease mortality: A case study of Mashhad, Iran. Iran J Pub Health 46:396–407.
- 64. Ailliot P, Allard D, Monbet V, Naveau P. 2015. Stochastic weather generators: An overview of weather type models. http://perso.univ-rennes1.fr/valerie.monbet/doc/papiers_pdf/SWGEN_review.pdf. Accessed on 2nd December 2016.
- 65. 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. pmid:19858539
- 66. Bobb JF, Peng RD, Bell ML, Dominici F. 2014. Heat-related mortality and adaptation to heat in the United States. Environ Health Perspect 122:811–816. pmid:24780880
- 67. Kalkstein LS. 1993. Health and climate change: direct impacts in cities. The Lancet 342:1397–1399.
- 68. Heaviside C, Tsangari H, Paschalidou A, Vardoulakis S, Kassomenos P, Georgiou KE, et al. 2016a. Heat-related mortality in Cyprus for current and future climate scenarios. Sci Tot Env 569–570:627–633.
- 69. 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. pmid:17901433
- 70. 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. pmid:22370738
- 71. Barreca A. 2012. Climate change, humidity, and mortality in the United States. J Environ Econ Manage 63:19–34. pmid:25328254
- 72. Deschênes O, Greenstone M. 2011. Climate change, mortality, and adaptation: Evidence from annual fluctuations in weather in the US. Amer Econ J: Appl Econ 3: 152–185.
- 73. Muthers S, Matzarakis A, Koch E. 2010. Climate change and mortality in Vienna—A human biometeorological analysis based on regional climate modeling. Int J Environ Res Pub Health 7:2965–2977.
- 74. Martin SL, Cakmak S, Hebbern CA, Avramescu M-L, Tremblay N. 2012. Climate change and future temperature-related mortality in 15 Canadian cities, Int J Biometeorol 56:605–619. pmid:21597936
- 75. Martens WJM. 1998. Climate change, thermal stress and mortality changes. Soc Sci Med 46:331–344. pmid:9460815
- 76. Hayhoe K, Sheridan S, Kalkstein L, Greene S. 2010. Climate change, heat waves, and mortality projections for Chicago. J. Great Lakes Res 36(suppl. 2):65–73.
- 77. Voorhees AS, Fann N, Fulcher C, Dolwick P, Hubbell B, Bierwagen B, et al. 2011. Climate change-related temperature impacts on warm season heat mortality: A proof-of-concept methodology using BenMAP. Environ Sci Technol 45:1450–1457. pmid:21247099
- 78. Ballester J, Robine J-M, Herrmann FR, Rodó X. 2011. Long-term projections and acclimatization scenarios of temperature-related mortality in Europe. Nature Comm 2:358, pmid:21694706
- 79. Jackson JE, Yost MG, Karr C, Fitzpatrick C, Lamb BK, Chung S, 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.
- 80. Robinson P. 2001. On the definition of a heat wave. J Appl Meteor 40:762–775.
- 81. Barnett AG, Tong S, Clements A. 2010. What measure of temperature is the best predictor of mortality? Environ Res, 110:604–611. pmid:20519131
- 82. Hajat S, Sheridan SC, Allen MJ, Pascal M, Laaidi K, Yagouti A, et al. 2010. Heat–Health Warning Systems: A comparison of the predictive capacity of different approaches to identifying dangerously hot days. Amer J Pub Health 100:1137–1144.
- 83. Davis RE, Hondula DM, Patel AP. 2016. Temperature observation time and type influence estimates of heat-related mortality in seven U.S. cities. Environ Health Perspect 124:795–804. pmid:26636734
- 84. Benmarhnia T, Sottile M-F, Plante C, Brand A, Casati B, Fournier M, et al. 2014. Variability in temperature-related mortality projections under climate change. Environ Health Perspect 122:1293–1298. pmid:25036003
- 85. Kalkstein LS, Smoyer KE. 1993. The impact of climate-change on human health: Some international implications. Experientia 49:969–979. pmid:8253171
- 86. Kalkstein L, Greene J. 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. pmid:9074886
- 87. Guest CS, Willson 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.
- 88. Gosling SN, McGregor GR, Lowe JA. 2012. The benefits of quantifying climate model uncertainty in climate change impacts assessment: an example with heat-related mortality change estimates. Clim Change 112:217–231.
- 89. El-Fadel M, Ghanimeh S. 2013. Climate change and temperature rise in the Greater Beirut Area: implications on heat-related premature mortality. Reg Environ Change 13:1059–1067.
- 90. Baker DJ, Hartley AJ, Butchart SHM, Willis SG. 2016. Choice of baseline climate data impacts projected species' responses to climate change. Glob Change Biol 22: 2392–2404.
- 91. 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. pmid:12750971
- 92. Zhou J, Chang HH, Fuentes M. 2012. Estimating the health impact of climate change with calibrated climate model output. J Agric Biol Environ Stat 17:377–394. pmid:24039385
- 93. Dixon KW, Lanzante JR, Nath MJ, Hayhoe K, Stoner A, Radhakrishnan A, et al. 2016. Evaluating the stationarity assumption in statistically downscaled climate projections: is past performance an indicator of future results? Clim Change, 135:395–408.
- 94. Peng RD, Bobb JR, 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. pmid:21193384
- 95. Zacharias S, Koppe C, Mücke H-G. 2015. Climate change effects on heat waves and future heat wave-associated IHD mortality in Germany. Climate 3:100–117.
- 96. Roldán E, Gómez M, Pino MR, Pórtoles J, Linares C, Díaz J. 2016. The effect of climate-change-related heat waves on mortality in Spain: uncertainties in health on a local scale. Stoch Environ Res Risk Assess 30:831–839.
- 97. Xu Z, FitzGerald G, Guo Y, Jalaludin B, Tong S. 2016. Impact of heatwave on mortality under different heatwave definitions: A systematic review and meta-analysis. Environ Int 89–90:193–203.
- 98. Davis RE, Knappenberger PC, Michaels PJ, Novicoff WM. 2003. Changing heat-related mortality in the United States. Environ Health Perspect 111:1712–1718. pmid:14594620
- 99. Carson C, Hajat S, Armstrong B, Wilkinson P. 2006. Declining vulnerability to temperature-related mortality in London over the 20th century. Am J Epidemiol 164:77–84. pmid:16624968
- 100. de’Donato FK, Leone M, Scortichini M, De Sario M, Katsouyanni K, Lanki T, et al. 2015. Changes in the effect of heat on mortality in the last 20 years in nine European cities. Results from the PHASE project. Int J Environ Res Public Health 12:15567–15583. pmid:26670239
- 101. Schwartz JD, Lee M, Kinney PL, Yang S, Mills D, Sarofim MC, et al. 2015. Projections of temperature-attributable premature deaths in 209 U.S. cities using a cluster-based Poisson approach. Environ Health 14:85, pmid:26537962
- 102. Gasparrini A, Guo Y, Hashizume M, Kinney PL, Petkova EP, Lavigne E, et al. 2015. Temporal variation in heat–mortality associations: a multicountry study. Environ Health Perspect 123:1200–1207. pmid:25933359
- 103. Arbuthnott KA, Hajat S, Heaviside C, Vardoulakis S. 2016. Changes in population susceptibility to heat and cold over time: assessing adaptation to climate change. Environ Health 15:(Suppl 1), S33, 73–93.
- 104. Lowe D, Ebi KL, Forsberg B. 2011. Heatwave early warning systems and adaptation advice to reduce human health consequences of heatwaves. Int J Environ Res Public Health 8:4623–4648. pmid:22408593
- 105. Li T, Ban J, Horton RM, Bader DA, Zhou M, Liang X, et al. 2016. Aging will amplify the heat-related mortality risk under a changing climate: projection for the elderly in Beijing, China. Sci Reps 6:21861, pmid:27320724
- 106. Anderson GB, Oleson KW, Jones B, Peng RD. 2016. Projected trends in high-mortality heatwaves under different scenarios of climate, population, and adaptation in 82 US communities. Clim Change,
- 107. Cheng CS, Campbell M, Li Q, Li G, Auld H, Day N, et al. 2008. Differential and combined impacts of extreme temperatures and air pollution on human mortality in south-central Canada. Part II: future estimates. Air Qual Atmos Health 1:223–235.
- 108. Watkiss P, Hunt A. 2012. Projection of economic impacts of climate change in sectors of Europe based on bottom up analysis: human health. Clim Change 112:101–126.
- 109. 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. pmid:24300074
- 110. Petkova EP, Bader DA, Anderson BG, Horton RM, Knowlton K, Kinney PL. 2014. Heat-related mortality in a warming climate: Projections for 12 U.S. cities. Int J Environ Res Public Health 11:11371–11383. pmid:25365060
- 111. Li T, Horton RM, Kinney PL. 2013. Projections of seasonal patterns in temperature- related deaths for Manhattan, New York. Nature Clim Change 3:717–721.
- 112. Li T, Ban J, Horton RM, Bader DA, Huang G, Sun Q, et al. 2015. Heat-related mortality projections for cardiovascular and respiratory disease under the changing climate in Beijing, China. Sci Rep 5:11441; pmid:26247438
- 113. Clark RT, Murphy JM, Brown SJ 2010. Do global warming targets limit heatwave risk? Geophys Res Lett 37:L17703,
- 114. Armstrong BG, Chalabi Z, Fenn B, Hajat S, Kovats S, Milojevic A, et al. 2011. Association of mortality with high temperatures in a temperate climate: England and Wales. J Epidemiol Community Health 65:340–345. pmid:20439353
- 115. United Nations 2014. World Urbanization Prospects: The 2014 Revision, Highlights. United Nations Department of Economic and Social Affairs, Population Division (ST/ESA/SER.A/352).
- 116. Hondula DM, Georgescu M, Balling RC Jr. 2014. Challenges associated with projecting urbanization-induced heat-related mortality. Sci Tot Env 490:538–544.
- 117. Argüeso D, Evans JP, Pitman AJ, Di Luca A. 2015. Effects of city expansion on heat stress under climate change conditions. PLoS ONE 10(2):e0117066, pmid:25668390
- 118. Petkova EP, Vink JK, Horton RM, Gasparrini A, Bader DA, Francis JD, et al. 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. pmid:27337737
- 119. Linares C, Mirón IJ, Montero JC, Criado-Álvarez JJ, Tobías A, Díaz J. 2014. The time trend temperature–mortality as a factor of uncertainty analysis of impacts of future heat waves. Environ Health Perspect 122:A118; pmid:24784020
- 120. Garrett AT, Creasy R, Rehrer NJ, Patterson MJ, Cotter JD. 2012. Effectiveness of short-term heat acclimation for highly trained athletes. Eur J Appl Physiol 112:1827–1837. pmid:21915701
- 121. Hajat S, Kovats RS, Atkinson RW, Haines A. 2002. Impact of hot temperatures on death in London: a time series approach. J Epidemiol Community Health 56:367–372. pmid:11964434
- 122. Anderson GB, Bell ML. 2011. Heat waves in the United States: mortality risk during heat waves and effect modification by heat wave characteristics in 43 U.S. communities. Environ Health Perspect 119:210–218. pmid:21084239
- 123. Rocklöv J, Forsberg B, Meister K. 2009. Winter mortality modifies the heat-mortality association the following summer. Eur Respir J 33:245–251. pmid:18799511
- 124. Stafoggia M, Forastiere F, Michelozzi P, Perucci CA. 2009. Summer temperature-related mortality: effect modification by previous winter mortality. Epidemiology, 20:575–583. pmid:19295435
- 125. Ha J, Kim H, Hajat S. 2011. Effect of previous winter mortality on the association between summer temperature and mortality in South Korea. Environ Health Perspect 119:542–546. pmid:21233056
- 126. Curriero FC, Heiner KS, Samet JM, Zeger SL, Strug L, Patz JA. 2002. Temperature and mortality in 11 cities of the eastern United States. Amer J Epidemiol, 155:80–87.
- 127. Semenza JC, Rubin CH, Falter KH, Selanikio JD, Flanders D, Howe HL, et al. 1996. Heat related deaths during the July 1995 heat wave in Chicago. N Engl J Med 335:84–90. pmid:8649494
- 128. Cheng J, Xu Z, Zhu R, Wang X, Jin L, Song J, et al. 2014. Impact of diurnal temperature range on human health: a systematic review. Int J Biometeorol 58:2011–2024. pmid:24535132
- 129. 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, pmid:18498630