Population ageing determines changes in heat vulnerability to future warming

Population ageing, an increase in the older age group’s portion of the total population, worsens the heat tolerance of a society. However, impacts of ageing on the social exposure to projected unprecedented hot summers (UHSs) are uncertain. We show that a shifting of the population distribution towards older ages amplifies the vulnerability of a country to the increasing frequency of UHSs as a result of warming during 2040–2070, especially in most populated regions such as China, India, and sub-Saharan countries. The warming scenarios from Representative Concentration Pathway (RCP) 8.5 are combined with population scenarios from three Shared Socio-economic Pathways (SSPs) SSP2, SSP3, and SSP5 together to estimate the exposure to UHSs. The ageing-driven increase in the exposure of elderly to UHSs ranges 51–198, 91–261, and 47–156 million in China, India, and sub-Saharan countries, respectively, between population scenarios. In China, with decreasing total population, the exposure to UHSs will be increased by rapid population ageing. In India and sub-Saharan countries, the potential of ageing to raise the exposure to UHSs will be even larger than that of warming. In contrast, in aged societies with slow ageing trend, e.g. United States and Europe, the warming mainly increases the exposure to UHSs. Our results suggest the changing age structure could exacerbate a country’s heat vulnerability despite limiting warming to a certain level in the future.


Introduction
Population ageing is one of the most significant contributors to world population change (Baccini et al 2011, Basu 2009, World Health Organization (WHO) 2015a, Li et al 2016, United Nations 2019a. Developed countries are entering the aged society due to the decrease in mortality and fertility rates (United Nations 2019b). The population aged 65 years and above (P65+) in the United States (US) in 2014 was 46 million, which is twice that in any year in the 1960s (Mather et al 2015). As a result, the share of P65+ to total population reaches 15% in 2014 in 6 Author to whom any correspondence should be addressed. the US, which is larger than that in 1960s by 6% (Mather et al 2015). In Germany, the share of the P65+ reached 21.1% of the total population in 2016, which was larger than that in 1996 by 5.5% (Statistical Office of the European communities 1990). In the future, the number of elderly will rapidly increase in highly populated countries with middle-and lowincomes (World Health Organization (WHO) 2011, United Nations 2019b). In China, the population aged 60 years and beyond (P60+) is projected to account for 28% of total population in 2040, which is more than twice the fraction in 2010 (World Health Organization (WHO) 2015b). The percentage of P60+ in India will exceed 10% of the total population in the 2020s, which will further increase by 10% in the 2060s (World Health Organization (WHO) 2015a). As a result of worldwide ageing, one of six people in the world will be aged 65 years or above in 2050 (United Nations 2019b).
The projected growth in the population of the elderly poses a serious danger to public health under a given warming scenario (World Health Organization (WHO) 2015a, Arnell et al 2016). Elderly people are much more vulnerable to heat stress in conjunction with their low adaptability to hot weather through factors, such as reduced sweat gland output, reduced skin blood flow, and smaller increase in cardiac output (Kenney andMunce 2003, Guo et al 2017). The high possibility of living alone, physical inactivity, having chronic diseases, and using of medications also increases heat-related mortality of the elderly (Hajat et al 2010, Guo et al 2017. The susceptibility of the elderly to heat stress has been reported when historically extreme heat events occur during previous record-breaking hot summers (Larrieu et al 2008, Dole et al 2011, Yiou et al 2020. During the 2003 summer for example-one of the hottest summers in Europe-the heat wave-induced deaths in France were mostly among the elderly (Larrieu et al 2008). According to the ageing trend in recent decades, more elderly persons have been exposed to extreme heat waves occurred in the record-breaking hot summers (Urban et al 2017(Urban et al , 2020. Further, the accelerated ageing trend in future will increase the exposure of elderly to projected unprecedented hot summers (UHSs), defined as hotter summer than the hottest summer in historical record, accompanied by strong heat waves (Jones et al 2015, King 2017, Harrington and Friederike 2018. Thus, there is a need to investigate the ageing-driven change in the elderly population exposed to UHSs for future climates to aid decisionmakers in reducing hot-weather-induced damages on public health. This study aims to quantify the importance of ageing with reference to the increase in the exposure of older persons to unprecedented future heat in six regions with large populations, namely, China, India, sub-Saharan countries with low-incomes (SSA-L), Brazil, US, and the 15 countries that joined the European Union prior to 2004 (EU15) during the mid-21st century. The change in P65+ UHS is computed based on the combination climate change projections of 27 global climate models (GCMs) under the Representative Concentration Pathway 8.5 (RCP8.5) scenario in combination with three population scenarios following Shared Socio-economic Pathways (SSPs) SSP2, SSP3, and SSP5. Individual impacts of ageing on the change in P65+ UHS are separated and are compared to those of warming and total population change. The results can provide relative importance of aforementioned phenomena for the change in regional exposure to UHSs, which can be a guideline for planning mitigation policies.

Definition of summer and the hottest summer
We used the monthly mean temperature of highresolution gridded datasets from the Climate Research Unit Time Series version 4.03 (CRUTS4) for 1901-2018 as an observation (Harris et al 2014). A summer is defined as the hottest season according to records of three consecutive months derived from the 20-year monthly climatology for 1986-2005 at individual grid cells (figure S1 (available online at stacks.iop.org/ERL/15/114043/mmedia)). Further, the summer with the highest temperature during 1901-2018 is defined as the hottest summer. Figure  S2 shows the global distribution of the seasonal mean temperature and year of occurrence of the hottest summer over the land surface.

Projection of UHSs
Seasonal mean temperature during summer is projected using the monthly mean temperature for 1861-2005 and the RCP8.5 scenario for 2006-2100 obtained from 27 GCMs that participated in the fifth phase of the Climate Model Intercomparison Project (CMIP5) (Taylor et al 2011;table S1). The first ensemble member of the individual GCMs is interpolated to the half-degree in latitude and longitude and is adjusted to the observation using the biascorrection procedures and observation data from the Climate Prediction Center (see Feng and Fu 2013 for details). The occurrence of UHSs is determined every year for the period of 2006-2100 when the summer mean temperature exceeds the hottest summer temperature based on CRUTS4. The probability of UHSs occurring within 20-year moving windows is computed as the ratio of the number of UHS occurrences within each 20-year period to 20. Each 20-year period is indexed as its final year; thus, the 20-year moving windows for 2006-2100 are from 1987-2006 to 2081-2100. The probability of UHS occurrence is projected at the times when the global mean temperature crosses 1.5, 2, and 3 • C relative to the pre-industrial level (t 1.5 , t 2 , and t 3 ). Here, t 1.5 , t 2 , and t 3 are calculated as the times whereby the 20-year moving averaged global temperature initially reach 0.9, 1.4, and 2.4 • C above 1986-2005, respectively, because this period was probably, at least, 0.6 • C warmer than the pre-industrial level (Schleussner et al 2016, Hawkins et al 2017. The mean, 84th and 16th percentile values of the UHS probability of occurrence are calculated based on the spread of 27 GCM projections. Also, the 16%-84% range of the probability of UHS occurrence is computed as the difference between the 84th and 16th percentile values. Note that extreme heat waves, which can induce numerous damages on public health, generally occurred during the recorded hottest summer (Dole et al 2011, Steffen et al 2017, Yiou et al 2020, therefore, the use of UHSs to project exposure of older people in future can be appropriate.

Projection of POP UHS and P65+ UHS
We estimated the total population exposed to UHSs (POP UHS ) in six regions, China, India, sub-Sahara countries with low income (SSA-L), Brazil, the US, and the 15 countries that joined European Union prior to 2004 (EU15), at the year of 2040, 2050, and 2070, which are the nearest years to the ensemble median of t 1.5 , t 2 , and t 3 , across 27 GCM simulations (2039, 2052, and 2075), respectively. The gridded dataset of the total population under SSP2, SSP3, and SSP5 for 2010-2100 with a 10-year temporal resolution is considered and interpolated to the halfdegree in latitude and longitude (Jones and O'Neil 2016, Fricko et al 2017, Fujimori et al 2017, Gao 2017, Kriegler et al 2017. Both SSP3 and SSP5 can be relevant to the RCP8.5 scenario because these two scenarios are characterised by large emissions from fossil-fuel usage (Riahi 2017). The SSP2 scenario is selected as a middle-of-the-road condition (Riahi 2017). Thus, three SSPs can provide a suitable range of population situations relevant to climate change under the RCP8.5 scenarios. Among the three SSP scenarios, SSP3 and SSP5 show the largest and smallest increase in population with slow and rapid ageing, respectively, and SSP2 represents the middle-of-theroad between SSP3 and SSP5 (table S2;  The projected regional-level population structure dataset under SSP2, SSP3, and SSP5 for 2005-2100 with 5-year discrete temporal resolutions obtained from the SSP database are used to estimate the P65+ UHS for 2040, 2050, and 2070 (Kc and Lutz 2017). Initially, the ratio of the P65+ to the total population count (R P65+ ) is calculated in individual regions. Under an assumption that R P65+ is the same for all grid cells within individual regions, P65+ UHS is computed as the product of POP UHS and R P65+ . Note that this assumption is used to match the gridded population count and regional-level composition of population data.

Separating individual impacts of warming, population change, and ageing
We calculated the change in P65+ UHS in 2070 relative to that in 2040, wherein the global mean temperature exceeded approximately 1.5 • C above the pre-industrial level under the RCP8.5 scenario. The change in the P65+ UHS relies on changes in three factors, namely, the UHS occurrence, total population count, and the ratio of P65+ to total population. The impact of individual factors is designated by impacts of warming (WARM), population change (POP), and ageing (AGEING). Here, we estimated the impact of each factor on the change in the P65+ UHS by subtracting the change in the P65+ UHS , considering two factors different from all the factors. For example, the impact of POP on the change in the P65+ UHS is computed as the difference between the P65+ UHS change due to the combined impacts and that due to WARM and AGEING. The estimated impacts of individual factors on P65+ UHS can explain which factor causes the change in P65+ UHS , as well as the associated potential risks.

Projection of UHSs
The multi-model ensemble of probability of UHS occurrence derived from the 27 GCM projections increases with the magnitude of global warming as the summer mean temperature rises (figure 1(a)-(c) and S4). In the 1.5 • C warming scenario, the probability of UHS occurrence shows a distinct regional variation ( figure 1(a)). UHSs are likely to occur in many tropical regions, indicating that the people in these regions are at risk of being exposed to UHSs and the associated heat stresses. In contrast, the probability of UHS occurrence is much lower in mid-and high latitudes, except for some regions in North America. This spatial UHS probability of occurrence pattern is connected to the low and high internal variabilities of temperature in the tropics and high latitudes (Harrington et al 2016, Lehner et al 2018). The probability increases under the 2 • C global warming condition relative to that under the 1.5 • C scenario (figure 1(b)). In particular, the likelihood of UHS occurrence in North America, East Asia, and Southeast Asia shows an increase of more than 30%, whereas the probability is low in Europe, Siberia, and southeastern Australia. When the global mean temperature reaches 3 • C, the summer mean temperature is likely to exceed the observed record of the hottest summer over almost all of the land surface ( figure 1(c)). The regions with high UHS probabilities of occurrence are generally consistent with the areas where the 16%-84% range of UHS probability of occurrence is small, thereby suggesting that the possibility of UHSs occurring is high in those regions (figure 1).

Total population and P65+ exposed to UHSs
Projecting POP UHS is important for estimating the potential impacts of UHSs on society. In general, more people will experience UHSs in the future due to world population growth, as well as the increase of the probability of UHSs in many areas (Lehner et al 2018, Harrington andFriederike 2018). However, the POP UHS can vary by regions and population scenarios. Figure 2 shows the estimated The population structure evolves differently according to regions and population scenarios (Kc and Lutz 2017); thus, temporal changes in P65+ UHS are different to those of POP UHS (figures 2 and 3). For all regions, the increasing rate of P65+ UHS is larger than that of POP UHS under all SSP scenarios. China shows a large increase in P65+ UHS , which contrasts the slight change in POP UHS (figures 2(a) and 3(a)). In India, under SSP2, the ensemble median of P65+ UHS in 2070 is three times larger than that in 2040, while POP UHS increases by 1.5 times for the same periods (figures 2(b) and 3(b); tables S3 and S4). The P65+ UHS in the other four regions are also projected to increase drastically (figure 3(c)-(f)). In addition, the changes in P65+ UHS according to individual SSPs are different from those in the POP UHS values in China, India, SSA-L, and Brazil. In these regions, the increase in P65+ UHS is the largest and smallest under the SSP5 and SSP3 scenarios, respectively, which is contrary to those in POP UHS (figures 2(a)-(d) and 3(a)-(d)). For example, in India, the increase in the ensemble median of POP UHS in 2070 relative to that in 2040 is 876 million under SSP3, which is three times larger than that under SSP5 (tables S3 and S4). However, for the same period, the increase in the ensemble median of P65+ UHS in India reaches 173 million under SSP3, which is less than two-thirds of that under SSP5 (tables S3 and S4). Both the US and EU15 show consistent changes in POP UHS and P65+ UHS for the three SSPs (figures 2(e), (f), (e), and 3(f)). The contrasting responses of POP UHS and P65+ UHS to the population scenarios in China, India, SSA-L, and Brazil are consequent of the distinct evolution of the elderly population to that of the total population (inner boxes in figures 2 and 3), thereby indicating that ageing should be considered as an important factor for estimating potential risks from heat stresses.

Driving cause of P65+ UHS increase
The change in P65+ UHS depends on three factors: (1) increase in UHS occurrence according to warming (WARM), (2) change in total population (POP),  4(a)). AGEING is the largest contributor to the increase in P65+ UHS in India, SSA-L, and Brazil under population scenarios following SSP2 and SSP5, which represent moderate-and fast-ageing scenarios, respectively (figure 4(b)-(d)). When The wide blue, orange, and purple bars indicate the 16%-84% percentile ranges of the projections due to the combined impacts (COMBINE; C) under SSP2, SSP3, and SSP5, respectively. The white dots denote the ensemble median of individual projections. The thin blue bars with circles, squares, and cross markers indicates the 16%-84% percentile ranges and median of projected change under the SSP2 scenarios due to the impact of warming (WARM; W), population change (POP; P), and ageing (AGEING; A), respectively. The thin orange and purple bars with markers are same, but for projections under SSP3 and SSP5 scenarios, respectively. population change follows the slow ageing scenario SSP3, the combined impacts of the three factors increase P65+ UHS in China, India, SSA-L, and Brazil (figure 4(a)-(d)). In the US and EU15, the WARM is the primary reason for the increase in P65+ UHS rather than the other considered parameters under all population scenarios (figures 4(e) and (f)). This can be explained by the remarkable increase in the probability of UHS occurrence around 2070, as well as the relatively small change in the total population and P65+ in the US and EU15 relative to that in 2040 (figures 1, 2(e), (f), 3(e), and (f)).

Discussions
Our results predict that more elderly people will be exposed to heat stresses occasioned by UHSs due to the combined impacts of WARM, POP, and AGE-ING. Furthermore, the individual contributions of the aforementioned phenomena towards increasing P65+ UHS vary by region. In China, India, SSA-L, and Brazil, the ageing-driven increase in P65+ UHS in 2070 relative to that in 2040 is larger than the impact of warming and population change, except in the SSP3 scenario (figure 4(a)-(d)). This indicates that the potential impacts of UHSs on public health will increase due to ageing in those regions even if we limit global warming to the level in 2040, comparable to 1.5 • C above the pre-industrial level. In addition, the health risks resulting from warming and ageing may be amplified in the aforementioned regions due to the relatively deficient economic capabilities (Harrington et al 2016). Thus, the economic support of developed countries is necessary for developing regions to establish age-friendly health and social services to mitigate the threat of UHSs to the aged societies of the future. In contrast, in both the US and EU15, the growth in P65+ UHS according to warming and the associated increase in the probability of UHS occurrence is much larger than that due to the change in other factors under all population scenarios (figures 4(e) and (f)). In addition, the regulated strength of UHSs under a target warming level, such as a 1.5 • C warming relative to the pre-industrial period, can suppress heat stresses in other regions that experience rapid ageing (King et al 2018, Lehner et al 2018. Efforts on traditional mitigation strategies that focus on reducing the emission of greenhouse gases and the relevant global warming effects are essential to reduce health problems associated with unprecedented summer heat in an aged society.