Depopulation, super aging, and extreme heat events in South Korea

Abstract

South Korea’s population is declining and its composition changing, associated with lowest-low fertility rates and rapid aging (super aging). When estimating changes in future exposure to extreme heat events (EHE), events that are predicted to be intensified due to climate change, it is important to incorporate demographic dynamics. We analyze business-as-usual (BAU) population and climate scenarios—where BAU refers to no significant change in current processes and trends in either domain—from 2010 to 2060 for South Korea. Data for both BAU scenarios are spatially linked and used to measure and identify national and sub-national and age-group specific EHE exposure. The results reveal an increasing exposure to EHE over time at the national level, but this varies widely within the country, measured at the municipal level. The most intensive exposure levels will be in the decade ending in 2040 driven by high estimated severe EHE. Sub-nationally, Seoul will be the most vulnerable municipality associated with super aging, while severe EHE not demographic factors will be relevant in Daegu, the second-most vulnerable metropolitan area. By 2060, national estimates suggest the older population will be up to four times more exposed to EHE than today. While the population of South Korea will decline, the rapid aging of the population ensures that specific regions of the country will become exceedingly vulnerable to EHE.

1. Introduction

South Korea has good reason to be concerned about climate change. The nation’s mean temperature is getting hotter more rapidly than global and East Asian averages (Korea Meteorological Administration, 2018, IPCC, 2013); over the last century (1912–2018) the average mean temperature in South Korea has increased by 1.9 °C which is 0.85 °C higher than the global average (National Institute of Meteorological Sciences, 2018). In 2018, excess mortality and the number of patients with compromised health due to extreme heat was estimated at 929 and 44,060 people, respectively (Park and Chae, 2020). In the last decade increased concern over population health risks has led to a series of national climate change adaptation plans. In 2020, the 3rd version was announced which included a comprehensive list of climate change risks and adaptation measures for all ministries and municipalities (both urban and rural) in the nation as climate change risks, including but not limited to extreme heat, will likely become more intense and more frequent in the near future (Ministries of Republic of Korea, 2020).

Meanwhile, South Korea is losing population related to its lowest-low fertility rate and super aging (Kim and Kim, 2020). In 2020, the national level total fertility rate (TFR) was 0.84, the lowest in world history (Statistics Korea, 2021). In order for South Korea to sustain its population size, a TFR equivalent to 2.1 is necessary or significant in-migration, both of which are highly unlikely.

A series of important questions arise when we bring together projections related to climate change and demographic change: What is the population-level exposure risk to extreme heat events (EHE), when changes in population size and composition are considered? Will more people be exposed to more frequently occurring EHE? How will different age-groups be impacted by EHE? How do the geographic patterns of estimated EHE and age-composition changes vary sub-nationally (i.e., by municipality)?

Exposure is distinct from vulnerability (Adger, 2006, IPCC, 2014a). In this paper we use definitions by the Intergovernmental Panel on Climate Change (IPCC). In the context of this study, vulnerability to EHE would differ for different age-groups of the population due to general differences in health and wellbeing (Kenney et al., 2014). We argue that higher levels of exposure will amplify vulnerability and risk (both mortality and morbidity). Thus, understanding the exposure to EHE is fundamental in terms of comprehending the associated vulnerability and risks; i.e., which populations and areas are the most vulnerable.

Literature regarding future population exposure to EHE in a spatially explicit fashion exists but are limited in geographical foci and demographic context (Jones et al., 2015, Forzieri et al., 2017). In the continental United States, the exposure to EHE is likely to increase by four- to sixfold in the late twenty first century by at least 10 billion person-days per year, where 29 % of the exposure was related to population growth and geographical redistribution of residents (Jones et al., 2015). In Europe, EHE are expected to be the most fatal weather-related disaster as the research estimates that 80–380 million people will be exposed in an annual basis by 2100 (Forzieri et al., 2017). Both of these studies, motivational studies for this paper, consider population and climate change scenarios and conclude that the future population will be more exposed to the climate change threats, but do so mainly because in the aggregate there will be more people than there are now, and the threats will become more severe than they are now. In other words, these prior studies have not been applied to demographic contexts of population decline and rapid changes in the age composition, which can influence the size of the vulnerable population. Thus, a study focusing on South Korea is an important contribution to the current literature.

Unlike the US and Europe, very few East Asian case studies explore future exposure to EHE in a spatially explicit fashion. Most literature focuses on analyzing relative risks with respect to EHE using time-series data, and their research outcomes are rarely presented in a map form or able to generate sub-national estimates. Instead, the studies account for socio-economic variables, especially aging and older population, to shape vulnerability and the associated risks (Kim et al., 2020, Kang et al., 2020, Son et al., 2019, Lee et al., 2018a, Kim et al., 2016a, Kim et al., 2014, Kyselý and Kim, 2009, Kim and Joh, 2006, Li et al., 2016, Cheng et al., 2018). Only a few studies employ population and climate change scenarios in tandem to depict the future situation, but none are spatially explicit (Li et al., 2016, Kim et al., 2016a, Shim et al., 2017). Thus, our second contribution to the literature is to leverage sub-national spatio-temporal data to help unpack the within-county heterogeneity in the patterns and changes in the patterns of EHE exposure.

Aging can exaggerate vulnerability to EHE because physically, the older population’s ability to control body heat diminishes (Kenney et al., 2014), and the older population is likely to have chronic disease or preexisting conditions and/or live alone (Cheng et al., 2018, Oudin Åström et al., 2011). Mentally, EHE would pose stress to the older population affecting their mental health, e.g., dementia (Berry et al., 2010, Hansen et al., 2008, Nitschke et al., 2013). We also note that the impact of EHE on individuals and regions with high concentrations of vulnerable populations is complex, with differential temporal impacts based on whether the focus is on immediate health outcomes or temporal lags examining short-term, long-term, or even cumulative health-related outcomes. Regardless of the inherent complexity, we believe it is important to take population aging into account when quantifying exposure, vulnerability, and risk due to EHE, particularly for countries in East Asia.

East Asia overall has been going through a rapid aging transition (Basten, 2013) and a recent Chinese case study demonstrates how an aging population is at risk of EHE in Beijing, China (Li et al., 2016). According to the authors, it is expected that Beijing would have 8.6–10.6 thousand heat-related deaths by the 2080s. In the case of South Korea, it is also anticipated that the number of deaths will increase by at least fivefold by 2060, where projection of the growing size of older population was incorporated (Kim et al., 2016). The expected deaths in Beijing and South Korea would be even higher if based on the adoption of a more radical climate change scenario (Li et al., 2016, Kim et al., 2016a).

In summary, this study fills the gap in the literature as to the best of our knowledge this is the first study of EHE in a country where population totals are declining and changing rapidly in age composition. Further, we designed our research to incorporate the future population exposure to EHE in a spatially explicit fashion, leveraging the best available data on population projections in South Korea. Our work has been motivated by Jones et al. (2015), but as their work focused on the United States they were unable to consider population aging or depopulation processes. Even the global assessment comparing historical urban warming (1983–2016) did not thoroughly discuss areas with declining populations and did not consider age compositions (Tuholske et al., 2021). However, in East Asia, one of the most populous regions on Earth including the two Koreas, China, and Japan, the aging transition has progressed rapidly in comparison to either the US or Europe (Parsons and Gilmour, 2018, Kishida and Nishiura, 2018, Basten, 2013, Muramatsu and Akiyama, 2011, Yu et al., 2020). Without considering super aging in East Asia, no research would be able to accurately estimate the vulnerability and risks associated with EHE exposure.

We aim to quantify the exposure to future EHE in South Korea by employing two business-as-usual (BAU) scenarios estimated for the period from 2011 to 2060: a BAU population scenario and a BAU climate scenario. A business-as-usual scenario in lay terms assumes that there is no significant change in current processes and trends over time. The spatio-temporal population and climate projections are mapped and overlaid to estimate the number of people exposed to EHE at the sub-national level. We follow the government’s definition of EHE—a daily maximum temperature above 33 °C—because the temperature threshold was defined by considering national and regional health impacts in South Korea (Korea Meteorological Administration, 2020). The definition does not consider the duration of extreme heat; hence, we prefer the term “extreme heat events,” instead of heatwaves. The three primary research questions are:

1. In the future, will there be a larger or smaller population exposed to future EHE in South Korea given the demographic changes (lowest-low fertility and super aging)?

2. Will the patterns of exposure to EHE vary within-country? Which municipal areas in South Korea will be most exposed to future EHE and why? Is the main driver more EHE or changes in the size and proportion of the older population?

3. Given that the older population is more vulnerable to EHE, how do estimates vary when different age thresholds are applied?

2. Material and Methods

The research employs population and climate data published in the most recent literature (see below). We generate spatio-temporal datasets matching population and climate data so they can be spatially overlaid and allowing us to link the frequency and intensity of EHE and population exposure. Our measure of exposure to EHE is based on aggregate number of person-days per year.

In terms of EHE exposure, we also quantify demography- and climate-induced change by holding either population or climate constant and then calculate the exposures. With sub-national data we compare EHE exposures in the four large metropolitan areas, namely Seoul, Incheon, Daegu, and Busan, with respect to their estimated future climate and demography. Lastly, we calculate the change in EHE exposure using established age thresholds that define the older population. The data and methods are explained in the following paragraphs in more detail.

2.1. Population Data & Methods

2.1.1. Population data

A sub-national population projection of South Korea (Kim and Kim, 2020) and administrative records provided by the Korean Statistical Information Services are used in this study. The administrative records report the most recent demographic structure, such as age and sex composition for different sub-national or small area geographies for every year through mid-century. These demographic data are linked to a series of geographic information system (GIS) layers. In brief, the demographic data are suitable to be used as a baseline when spatially downscaling sub-national population projections. Kim and Kim (2020) employed a cohort component method, where each demographic component was estimated using the following models: a generalized log gamma distribution model for fertility (Kaneko, 2003), Heligman and Pollard model for mortality (Heligman and Pollard, 1980), and a migrant pool model for gross migration (Kim and Kim, 2020).

In this research, a small area refers to an administrative level that covers rural and urban municipalities. South Korea’s land area was divided into 162 small areas, approximately 620 square kilometers in size on average (i.e., approximately 25 by 25 km).

2.1.2. Spatial downscaling of population projection

Spatial downscaling proportionally allocates the 37 sub-national population projections (K. W. Kim and Kim 2020) to 162 small-areas by preserving the pycnophylactic property (Tobler 1979). This is based on the assumption that the most up-to-date age and sex ratios for each area would remain the same for the entire projection horizon. The following equations are used for the allocation:

$${\text{Proportion}}_{i,g,a,s}=\frac{{\text{Population}}_{i,g,a,s}}{\sum _{g,a,s}^{n_i}{\text{Population}}_{i,g,a,s}}$$

$$\text{Total Population in 2020}=\sum _{i=1}^{37}\sum _{g=1}^2\sum _{a=1}^{101}\sum _{s=1}^{n_i}{\text{Population}}_{i,g,a,s}$$

where $i$ indicates one of the 37 sub-national divisions of South Korea, $g$ specifies sex, $a$ denotes each age cohort, $s$ refers to each small area, and $n_i$ indicates the number of small areas in sub-national division $i$. Once a proportion is calculated then the value is multiplied to the corresponding population projection (and rounded up to an integer). We repeated this procedure for 2025, 2030, 2035, 2040, 2045, 2050, 2055, and 2060 to result in a series of single-year small-area population projections.

2.1.3. Age-specific population change

After tabulating the small-area population projections, we estimate total population and the older populations based on the three often used age thresholds (i.e., aged ≥ 65, aged ≥ 75, and aged ≥ 85). We have used the age 65 as the threshold identifying the older population in South Korea because it is a threshold used by the American Geriatric Society and also the legal definition of older population by the Ministry of Health and Welfare in South Korea (Vaughan et al., 2019, Ministry of Health and Welfare, 2021).

Regarding population data we can estimate and map population changes from the year 2020 through to 2060 and examine spatio-temporal trends in the total and age-specific threshold populations.

2.2. Climate Data & Methods

2.2.1. Climate change scenario data

We used a representative concentration pathway 4.5 (RCP 4.5) as our climate change scenario because this scenario is considered BAU in South Korea (Shim et al., 2017). RCPs are the climate change scenarios used in the IPCC’s 5th assessment report (IPCC, 2014b, Moss et al., 2010). RCP 4.5 indicates that the future radiative forcing is stabilized at 4.5 W per square meter (hence, RCP 4.5), and accordingly, the mean temperature will rise by 2.4 °C and 2.9 °C at the global and Korean peninsula levels, respectively, by 2071–2100 (Korea Meteorological Administration, 2018). This scenario assumes that global and national climate change mitigation policies are implemented and working effectively to some extent and that the carbon dioxide level in the atmosphere does not increase more than 538.4 ppm by 2100 (IPCC, 2013). It is important to note that IPCC regards RCP 8.5 as BAU, the scenario with no mitigation activities by the global community, resulting in the most radical climate change. We used the highest spatial resolution for RCP 4.5 (i.e., 1 km × 1 km) provided by the Korea Meteorological Administration and the National Institute of Meteorological Sciences. These data were processed as follows: The Coupled Model Intercomparison Project 5 (CMIP5), which is a collaborative global climate modeling process coordinated by the World Climate Research Programme, generated global RCP scenarios, including the Hadley Global Environment Model Version 2 – Atmosphere and Ocean (HadGEM2-AO) by the Met Office in the UK. The global RCPs (HadGEM2-AO) were spatially downscaled to produce the regional RCPs, entitled Hadley Global Environment Model 3 Regional Atmosphere (HadGEM3-RA) (Suh et al., 2016). By employing the Modified Korea – Parameter-elevation Regressions on Independent Slopes Model (MK-PRISM), the regional RCPs (HadGEM3-RA) were spatially downscaled to the highest spatial resolution scenarios (1 by 1 km) used in this research (Kim et al., 2016).

2.2.2. Frequency and intensity of extreme heat events

We have accounted the number of days of EHE as follows. The summer in South Korea usually ranges from June to August, but recently, EHE are becoming more frequent in both May and September (Moon et al. 2020). Therefore, and similar to Perkins-Kirkpatrick and Gibson (2017) we use EHE occurrence between May and September. The following equation shows how an annual EHE frequency was accounted for every decade:

$$EHE\mathrm{ }in\mathrm{ }decade\mathrm{ }ending\mathrm{ }in\mathrm{ }2020=\mathrm{ }\frac{{\sum }_{y=1}^{10}{\sum }_{m=1}^{5}EHE\mathrm{ }}{10\mathrm{ }years}$$

where $m$ denotes each month, i.e., from May to September, $y$ means each year, i.e., from 2011 to 2020. We repeated this procedure for the decades ending in 2030 (2021–2030), 2040 (2031–2040), 2050 (2041–2050), and 2060 (2051–2060) and mapped the outcomes. Also, we calculate differences in the number of EHE across four time periods (specifically 2020 and 2030, 2020 and 2040, 2020 and 2050, and 2020 and 2060) and map them to examine change in the geographical distribution of EHE over time.

The intensity regarding area and frequency (IAF) of EHE is calculated for each month through the decade ending in 2060. According to Shim et al. (2017), IAF is defined as follows:

$$IAF\left(\mathrm{\%}\right)=\mathrm{ }\frac{{\sum }_{i=1}^h{A}_i{d}_i\times 100}{{\sum }_{i=1}^t{A}_i{D}_i}$$

where $A$ is the area of pixel $i$ with an EHE; $h$denotes the total number of pixels with EHE; $d$ shows the number of days of EHE in pixel $i;\mathrm{ }t$ is the total number of pixels in the study area; and $D$ indicates the total number of days for each month. By design, IAF ranges from 0 to 100. If an IAF is 100% in August, that means the entire South Korea’s land is exposed to EHE every day that month. If an IAF is 50%, the half of the nation is exposed to EHE every day that month, or the entire nation is exposed to EHE for half of the month. Comparing IAFs in different months and decades enables readers to understand the fluctuating intensities of EHE.

2.3. Exposure to extreme heat events

The main purpose of this research is to estimate the number of people exposed to future EHE in South Korea at a small-area (i.e., sub-national) scale. When estimating exposure, it is mandatory to match the unit of analysis so that the population projection and climate change scenario can be tabulated in tandem. Then, the population projections and the EHE frequencies are overlaid, multiplied, and normalized by each municipal area to estimate the populations exposed to future EHE using number of person-days per year (Jones et al., 2015, Tuholske et al., 2021).

Once completed the analytical process proceeded as follows. First, the exposures for the total population and the older populations (defined by different age thresholds) are calculated at the national level and compared over time. This step incorporates attention to how sensitive the population’s exposure to EHE would be with respect to the aging transition. Second, the exposure of the total population is visualized in a spatially explicit way to portray its geographical distribution for different points in time. When tabulating the EHE exposure we have multiplied EHE for the decade ending in 2030 (2021–2030) with the population in 2025. Similarly, the exposures of the older populations with three age thresholds are mapped. Third, we decompose the aggregate exposure to quantify the relative effects of population, climate, and their interaction. By fixing the population constant but letting EHE to fluctuate, we measure the climate effect. Similarly, we calculate the population effect by fixing the EHE frequency constant but allowing population to decrease. The interaction effect is tabulated by taking the difference between the total exposure change and the sum of the population and climate effects (Jones et al., 2015). Fourth, a series of comparison are conducted for four metropolitan areas—Seoul, Incheon, Daegu, and Busan—to analyze how future climate and demographic change will contribute to EHE exposure with respect to older populations in these areas. Lastly, we calculate the percent increase of exposures to EHE from the decade ending in 2030 through to the decade ending in 2060 with respect to the current condition (the decade ending in 2020) for the older populations, defined using three age thresholds.

3. Results

The purpose of the range of model runs was to assess population exposure to future EHE and to analyze the spatio-temporal trends of this exposure.

3.1. Business-as-usual depopulation and aging

At the national level, the decline in the total population is evident, and the speed of decline is rapid. In 2030, the total population of South Korea is smaller than 50 million and by the year 2060 is projected to be under 38 million (Fig. 1). Over the next four decades the proportion of the older population will become larger. The older population aged ≥ 65 is estimated to steadily increase through 2050, will plateau and begin to decline again around 2060. In 2020, the older population aged ≥ 65 represented 16.6 % of the total population but will rapidly increase to almost 48 %. Indeed, by 2060, more than half of all South Koreans (52.7 %) will be aged ≥ 65 under the BAU population scenario. Relatedly, the proportion of the population aged ≥ 75 and ≥ 85 will increase to 36.4 % and 17.6 %, respectively, of the total population by 2060.

Figure 1.

National population change from 2030 to 2060 for the total population and the older populations aged ≥65, aged ≥75, and aged ≥85.

Only a handful of municipalities will gain total population between 2020 and 2060 (Fig. 2a). However, such gain will not be substantial as those areas are mostly distributed in rural regions, except Sejong city. The remainder of South Korea will lose population between 2020 and 2060. The two areas indicated with the darkest blue are the most populous regions in South Korea: Seoul, the capital city, and Busan, the largest harbor and the second largest city in the nation (Fig. 2a). In general, the larger the population in 2020, the larger the population loss by 2060.

Figure 2.

Small-area population differences between 2020 and 2060 for (a) the total population, (b) the older populations aged ≥65, (c) aged ≥75, and (d) aged ≥85.

The geographical distribution of the older population aged ≥ 65 is more clustered than that of the total population (Fig. 2b). The areas showing negative values are well known rural areas in South Korea. The four clusters color-coded in light blue are located either deep in the forest regions of the country or close to the ocean, although some distance from the metropolitan areas of Seoul and Busan. In these rural areas, the younger population often migrate to the larger urban centers (for social and economic opportunities such as jobs and better family infrastructure such as schools), leaving the older population behind and thereby changing the local age composition. The rest of the nation will numerically gain older population aged ≥ 65 between 2020 and 2060 (Fig. 2b).

Only a few municipal areas (light blue) are expected to experience a decreasing proportion of the population aged ≥ 75 by 2060 (Fig. 2c). Seoul (the darkest red) shows that its level of super aging will be the highest in South Korea. Aging in Busan and Incheon (shown in bright red) is not at the level observed in Seoul but nevertheless these metropolitan areas are higher than the rest of the nation. The lighter red indicates other large cities in South Korea, while the lightest red portrays small cities and rural areas (Fig. 2c). In short, the larger the population, the larger the proportion that are older and where places experience super aging. This pattern becomes more apparent for the older population aged ≥ 85 (Fig. 2d).

3.2. Business-as-usual extreme heat events

The spatial distribution of EHE frequency (the number of days per year) for the decade ending in 2020 is mapped (Fig. 3a). Differences compared to the decade ending in 2020 are shown in Fig. 3b–e, and the series of maps illustrates that the increasing and decreasing trends of EHE frequency and their geographic patterning. EHE frequency will not increase universally over regions, and it is important to note that some regions may experience fewer EHE in the future than in the present.

Figure 3.

Geographical distribution of EHE for the decade (a) 2020 and the difference maps for (b) 2020–2030, (c) 2020–2040, (d) 2020–2050, and (e) 2020–2060.

Both increasing and decreasing trends in the short term, through the decade ending in 2030 are relatively moderate. The pixel (or grid) count indicating the largest increase in terms of frequency is only in the single-digits. Further, the pixel indicating the largest decrease shows the largest negative value, compared to other time periods (Fig. 3b). The change between the decades ending in 2020 and in 2040 is more dramatic. The pixel value with the largest increase is 23.8 days per year (i.e., over three weeks), whereas the largest decrease is shorter than one day. Briefly, compared to the decade ending in 2020, EHE will become more common and frequent by 2040 (Fig. 3c). The 2020–2050 change in number of days per year is less dramatic than for the 2020–2040 comparison, but nevertheless higher than in 2020–2030 (Fig. 3b–d). The 2020–2060 difference parallels the earlier period through to 2040 (Fig. 3c and 3e).

The areas with little change or a decreased EHE frequency are mostly mountainous regions with much vegetation, high altitude, and steep slope, and the eastern coastal regions. Jeju island, located off the southern coast, will likely have fewer EHE in the future. In contrast, the large urban areas in the northwest, and coastal and agricultural regions in the southwestern part of the Korean peninsula are estimated to experience more frequent EHE compared to the past (Fig. 3a–e).

IAF efficiently summarizes such dynamics in terms of frequency and area. These results are shown in Fig. 4. Although Fig. 4 does not portray any spatially explicit information, it shows the EHE intensity in a monthly basis, averaged to show per annum estimate for every decade. Generally, EHE of July and August will become more intensive. From the decade ending in 2040 through 2060, all IAFs in July and August are higher than 15, with July 2040 (i.e., 2031–2040) estimated to have the highest IAF (27). The result implies that the entire South Korea will either be under the EHE for more than a week or 27 % of South Korea will experience the EHE every day for July 2040. August 2060 (i.e., 2051–2060) illustrates the second highest IAF (26.7), followed by August 2040 (25.7).

Figure 4.

National trend of decennial average EHE intensity for each month from the decade 2030 to 2060.

3.3. Future exposure to extreme heat events

The number of people exposed to EHE are summarized in Fig. 5. An increasing trend is identified when all ages are considered; the spike in 2040 is due to the fact that the EHE are unusually intensive compared to the other decades. The pace of population decline (Fig. 1) is relatively slow compared to the pace of change in EHE, hence, generating an increase in exposure to the latter at the national level. The trends of exposure to EHE among the older populations aged ≥ 65 and aged ≥ 75 show a similar yet more moderate results, whereas the heat exposure on the older population aged ≥ 85 reveals an increasing trend (Fig. 5). All maps portraying the spatial distribution of populations exposed to EHE have similar legend schemes, to aid comparison over time. The four major metropolises, namely, Seoul, Incheon, Daegu, and Busan are always the highest top four (Fig. 6, Fig. 7, Fig. 8, Fig. 9).

Figure 5.

National trend of exposure to EHE for the total population and the older populations aged ≥65, aged ≥75, and aged ≥85 from the decade 2030 to 2060.

Figure 6.

Geographical distribution of exposure to EHE for the total population for the decade (a) 2030, (b) 2040, (c) 2050, and (d) 2060.

Figure 7.

Geographical distribution of exposure to EHE for the older population aged ≥65 for the decade (a) 2030, (b) 2040, (c) 2050, and (d) 2060.

Figure 8.

Geographical distribution of exposure to EHE for the older population aged ≥75 for the decade (a) 2030, (b) 2040, (c) 2050, and (d) 2060.

Figure 9.

Geographical distribution of exposure to EHE for the older population aged ≥85 for the decade (a) 2030, (b) 2040, (c) 2050, and (d) 2060.

The exposure of the total population demonstrates the similar increasing trend through 2060 as identified in Fig. 5 as the spatial coverage of EHE exposure expands over time (Fig. 6). The exposure to EHE for the decade 2040 is most severe (Fig. 6b); the number of person-days per year in Seoul, for instance, is estimated at more than 274.7 million, followed by Daegu (68.4 million) and Busan (46.4 million). By 2035 Busan’s population (3.1 million) will be larger than Daegu’s (2.2 million) but Daegu’s exposure to EHE is expected higher, implying that EHE in Daegu will be more intensive than in Busan (Fig. 6b).

As for the exposure of the population aged ≥ 65, the decade ending in 2040 may not be most severe for Seoul and Incheon (Fig. 7). At the aggregate level, it is true that the majority of the population, including Daegu and Busan, will be most exposed to EHE for the decade ending in 2040 (Fig. 5, Fig. 7b). However, in terms of the largest number of person-days per year, Seoul is expected to experience the most severe exposure for the decade ending in 2060. Seoul’s and Incheon’s exposures will be 124.9 and 13.1 million, respectively (Fig. 7d), and these figures account for 45.5 % and 32.2 % of Seoul’s and Incheon’s exposure of the total population considering all ages in the same time period, respectively (Fig. 6d).

It is estimated that the decade ending in 2060 will be the most severe in terms of EHE exposure for the population aged ≥ 75, not only at the national level (Fig. 5), but also at the small-area scale (Fig. 8d). Note that Seoul’s exposure will be more severe in the decade ending in 2050 than the earlier decade (Fig. 8b and 8c), a pattern not found in the prior results (Fig. 5, Fig. 6, Fig. 7). In Daegu, on the other hand, the most severe exposure is estimated in the decade ending in 2040 (Fig. 8b), but this trend will not be maintained for the population aged ≥ 85 (Fig. 9b). The exposure in Seoul and Daegu will get more intensive over time (Fig. 9).

It is evident that the exposure is negatively affected by the population effect when the change in total aggregate exposure is decomposed (Fig. 10). In 2060, the exposure to total population to EHE will increase by 524 million person-days relative to 2020 by increasing EHE frequency only. However, the population and interaction effects offset the climate effect resulting in a lower exposure (336 million person-days). It is important to note that solely considering the total population decline, without referring to the corresponding age compositions, may result in an unrealistic underestimation in assessing EHE exposure and vulnerability.

Figure 10.

Decomposition of aggregate projected change in exposure into climate, population, and interaction effects.

As for the older population aged ≥ 65, the contributions of EHE frequency and population size to EHE exposure in the four metropolitan areas are shown in Table 1. In cases of Seoul and Incheon, it is evident the growing size of the older population is the main contributor for increasing EHE exposure in these metropolitan areas. In cases of Daegu and Busan, however, the contribution of the growing older population to increase in EHE exposure diminishes in the latter decades implying no change in population size but a change in EHE frequency (Table 1).

Table 1.

Population aged ≥65 exposure to EHE in four metropolitans.

		Population (persons)	Δ Population	EHE (days)	Exposure (person-days)
Seoul	2030	1,968,125		17.4	34,168,524
	2040	3,045,217	1,077,092	30.9	94,177,831
	2050	3,929,865	884,648	22.7	89,289,526
	2060	4,284,734	354,869	29.1	124,881,632
Incheon	2030	208,880		7.7	4,330,412
	2040	412,518	203,638	14.6	13,397,620
	2050	675,035	262,517	11.0	13,096,601
	2060	830,037	155,002	15.6	20,238,277
Busan	2030	849,319		4.3	3,636,083
	2040	1,182,313	332,994	14.8	17,527,968
	2050	1,383,880	201,567	10.8	14,988,212
	2060	1,384,756	876	13.7	18,989,300
Daegu	2030	504,782		14.1	7,100,362
	2040	726,568	221,786	30.6	22,253,806
	2050	864,439	137,871	21.7	18,746,104
	2060	860,483	−3,956	27.9	24,040,324

Fig. 11 demonstrates how the increases in the age-specific populations are estimated to become more exposed to EHE over time. The exposure to EHE of population aged ≥ 85 will increase more rapidly than those of the populations aged ≥ 75 and those ≥ 65. The age-specific exposures are driven, not surprisingly, by demographic change rather than by EHE.

Figure 11.

Percent increase of EHE exposure from the decade 2030 to 2060 with respect to the decade 2020 for the older populations aged ≥65, aged ≥75, and aged ≥85.

4. Discussion

Our South Korean case study started with a simple question: What will happen to population-level exposures to EHE when a country’s population begins to decline in an era where there is an anticipated increase in the number of EHE? And, in particular, what will be the exposure to EHE among different older populations, due to rapidly changing age composition of the country (i.e., super aging). One might anticipate that population-level exposures will decline as population declines but if the frequency of EHE increases rapidly then this can offset the population decline. Our results show that the overall population exposure to EHE in South Korea is expected to increase in the future even with the anticipated population decline. A similar story is anticipated for Europe (Forzieri et al., 2017).

What about population exposures EHE at a sub-national scale? Shim et al. (2017) tabulated the potential population exposure to EHE in South Korea and argued that by 2050 the older population aged ≥ 65 would become more exposed to EHE compared to the total population. The authors were only able to present aggregate national findings. Our work is spatially disaggregated and we find, as expected, high exposures are consistently found in metropolitan areas, especially Seoul, Incheon, Daegu, and Busan, due to their larger population. An exception to this is Daegu, where EHE are expected to increase dramatically (Fig. 3a–e). Indeed, future mortality rates due to extreme heat stress has identified Daegu as an area of concern (Kim et al., 2014).

The population in rural areas are expected to be relatively less exposed to EHE, according to our maps (Fig. 6, Fig. 7, Fig. 8, Fig. 9). However, it is important to stress that this does not necessarily mean the rural population would be less vulnerable to EHE or are at minimal risk. As noted earlier, risk is a function of exposure and vulnerability (IPCC 2014a). The rural population may be at more risk. Reid et al. (2009) argued that factors—namely, age, poverty, education, living alone, ethnicity, air conditioning, diabetes, and green space—shape heat-related vulnerability. And, recent South Korean case studies identify the major factors that contribute to the heat-related deaths (Kim et al., 2017) identifying rural areas as more vulnerable due to super aging, a high proportion of older population living alone, and agricultural employment. Heat-related mortality and risk in the rural areas of South Korea is estimated to be about 5.6 times higher than in the metropolitan areas (Kim et al., 2017).

What does the higher exposure in the metropolitan areas imply? Our finding indicates that the super aging in these areas would amplify vulnerability to EHE in the future. South Korea is a highly urbanized nation (Kim and Kim, 2020, Park et al., 2021), and this is linked to high levels of social isolation (i.e., living alone). In 2017, 28.5 % of households in South Korea were one-person households, and it is estimated this will rise to 37.3 % by 2047 (Statistics Korea 2019). In turn, as with the rural areas, the higher social isolation is associated with the higher heat-related mortality and risk (Yong-ook Kim et al., 2020). In short, it is likely that high mortality and risk in the metropolitan areas will be exacerbated by super aging, smaller kin networks, and social isolation. There is a bright side. At this time, it is estimated that 87 % of South Korean population has access to air-conditioning in their homes (Gallup Korea 2018). As Reid et al. (2009) argued home air-conditioning can be a strong protective factor against heat-related deaths, but only if it is supported with stable electricity supply (Santamouris et al., 2015).

The most up-to-date national climate change adaptation plan of South Korea does not take such population dynamics and composition into account when designing adaptation measures with respect to future EHE (Ministries of Republic of Korea 2020). Among 84 major climate change risks complied by the adaptation plan, 10 of them are about EHE and health impacts; the ratio (10/84) implies that EHE already is an important concern in South Korea. The associated adaptation measures are geared towards installing additional air-conditioning and other cooling units for vulnerable population (Ministries of Republic of Korea 2020). However, the measures may be unrealistic, if not too costly, given the anticipated pace of growth in the vulnerable older population over the next 40 years. The adaptation plan is science-based with the supports of three decision making systems developed by Korea Adaptation Center for Climate Change. However, future population dynamics and composition are ignored (Korea Adaptation Center for Climate Change, 2016, Korea Adaptation Center for Climate Change, 2020a, Korea Adaptation Center for Climate Change, 2020b). For example, our study illustrates why it may be strategic for Seoul to prepare future EHE adaptations by focusing on the needs of the growing older population, but for Daegu, the local government will need to consider different strategies to reduce the impacts of extreme heat.

In a study such as this it is important to discuss modeling uncertainties. For the population projection, it is worth noting the simplicity of downscaling the sub-national population projection to the small area scale. To quantify a population exposure to EHE at the municipal level, it is necessary to have spatially explicit population data match the spatial resolution of downscaled RCP scenario (1 km × 1 km). It is also important that the population projections should cover the entire nation, not solely focusing on a small area or two, so that our work could address future EHE in both urban and rural regions. However, such data or the relevant literature regarding South Korea’s downscaled future population has not yet been made available. Among numerous methods for small area population forecasts, we have chosen to use the downscaling and disaggregation approach (Wilson et al., 2021), an approach frequently used by climate change researchers (Jones et al., 2015, Jones and O’Neill, 2013, O’Neill, 2005, O’Neill et al., 2001). Although the approach is sometimes criticized for oversimplification and its accuracy not comprehensively assessed (Wilson et al., 2021), the existing climate change literature confirms and justifies that our population data are suitable for quantifying EHE exposure. In the literature, the downscaling and disaggregation approach is applied at a pixel (grid) level because this spatial scale is meaningful in climate change research. Most applications using gridded population data do not consider sex or age but only use total population (Jones et al., 2015, Jones and O’Neill, 2013, O’Neill, 2005, O’Neill et al., 2001). In our research, as the age information is critical if we are to better understand exposure and vulnerability to EHE in South Korea. We have downscaled the sub-national population projections to the municipal level, where the most recent population composition information is provided by Statistics Korea. In summary, our population downscaling may show a coarser spatial resolution than the previous research (small area vs pixel) but we have included population age composition.

The modeling uncertainty for the climate scenario data (RCP 4.5) must be addressed as well. The combination of HadGEM3-RA and HadGEM2-AO could be limited in estimating EHE without any systematic assessment and validation of modeling performance. According to Suh et al. (2016), HadGEM2-AO performed well for East Asia in general, but the performance was limited or biased in the South Korea case because lower spatial resolution (approximately 135 km × 135 km) was incapable of dealing with complex local geography. South Korea has many mountainous areas, and the model results were biased. On the other hand, HadGEM3-RA was able to correct the bias in such high-altitude mountainous areas, and this justifies the application of this model in our work. In our previous research, we compared the performance of RCP 4.5 and RCP 8.5 ensemble climate scenarios at the municipal level, and the uncertainties varied over regions. For example, more uncertainties were observed in Daegu (July) compared to Seoul (July); hence, it is important to take such uncertainty into consideration when utilizing our research outcome for municipal EHE adaptation policies (Han et al., 2018).

In our future work, we plan to study not only the vulnerability associated with EHE and super aging in South Korea, but also to map the associated risks and potential excess deaths at the municipal level. It is important to note that estimating excess deaths and excess mortality are distinct. To estimate the number of deaths due to aggravated EHE, the data of population exposure is mandatory. Numerous South Korean case studies, however, are mostly about quantifying relative risks due to future EHE (Park et al., 2019, Chae and Park, 2021, Lee et al., 2018b), and such research cannot estimate a number of prospective deaths at a regional or local scale because of the lack the small area population forecasts. In short, at the present time, a relative increase in mortality due to EHE cannot be converted into economic losses or costs due to morbidity-related health outcomes, nor tease out short-term, long-term or cumulative effects. Unfortunately, the most up-to-date national climate change adaptation plan of South Korea does not include such perspectives (Ministries of Republic of Korea 2020), and we believe that our present and future research on EHE exposure can contribute to the national policy.

5. Conclusions

Many South Koreans will face EHE and other heat-related risks in the future. However, the population-level exposure surface is not even within a country and nor will it be only determined by the increase in EHE. The rapidly changing demographics of South Korea, particularly the emergence of super aging populations will elevate the health and mortality risks. Coupling climate change scenarios with population dynamics and paying close attention to composition in a spatially explicit manner can reveal areas and populations at high (and lower) risk. Our strategy looks beyond just climate change-related weather events (Meehl and Tebaldi 2004). To date, the national adaptation plan regarding EHE and other climate change risks has not fully incorporated demographic futures, and this may hinder designing and implementing an effective adaptation measure in terms of coping with climate change risks in South Korea.

Highlights:

• The rapidly growing older population in South Korea will become increasingly vulnerable to extreme heat events (EHE).

• Over time the large metropolitan areas are expected to become more exposed to EHE compared to other parts of South Korea.

• The national population decline will offset the climate effect, hence resulting in lower exposure to EHE when the age composition is not considered.

• The impact on the older population was assessed by different age cut-points, which reveal that as the population aged ≥ 85 is expected grow rapidly this group’s exposure over time will be more affected by population change than EHE.