Impact of Extreme Heatwaves on Population Exposure in China Due to Additional Warming

Abstract

Extreme heatwaves are among the most important climate-related disasters affecting public health. Assessing heatwave-related population exposures under different warming scenarios is critical for climate change adaptation. Here, the Coupled Model Intercomparison Project phase 6 (CMIP6) multi-model ensemble output results are applied over several warming periods in the Intergovernmental Panel on Climate Change AR6 report, to estimate China’s future heatwave population exposure under 1.5 °C and 2.0 °C warming scenarios. Our results show a significant increase in projected future annual heatwave days (HD) under both scenarios. With an additional temperature increase of 0.5 °C to 2.0 °C of warming, by mid-century an additional 20.15 percent increase in annual HD would occur, over 1.5 °C warming. If the climate warmed from 1.5 °C to 2.0 °C by mid-century, population exposure would increase by an additional 40.6 percent. Among the three influencing elements that cause the changes in population exposure related to heatwaves in China–climate, population, and interaction (e.g., as urbanization affects population redistribution)–climate plays the dominant role in different warming scenarios (relative contribution exceeds 70 percent). Therefore, considering the future heat risks, humanity benefits from a 0.5 °C reduction in warming, particularly in eastern China. This conclusion may provide helpful insights for developing mitigation strategies for climate change.

1. Introduction

Meteorological measurements and general circulation model (GCM) simulations yield strong evidence for warming globally over the last several decades. Continued warming will likely cause increases in the frequency of extreme climatic and weather occurrences, challenging human systems. For instance, occurrences such as the 2013 heatwave over China, the 2011–2014 persistent drought over California, and the devastating 2013 flood in India resulted in significant large economic losses and left many people wounded or homeless [1]. Climate extremes include a broad range of physical effects that have been recorded extensively across the globe, including a significant danger to natural resources, the environment, and the social economy [2].

As is revealed by the newest climate change scenario design—called ‘shared socio-economic pathway’ (SSP)—identifying vulnerabilities in climate change risks is crucial for the public and governments [3]. In general, vulnerability is characterized as a result of a society’s exposure and sensitivity to hazards, as well as its ability to adapt. Therefore, assessing future climate change risks, particularly the vulnerability of population populations to a certain disasters, is critical in developing social risk management measures [4].

Global-warming-affected extreme events that alter risks include many aspects, such as heatwaves, droughts, hurricanes, floods, mudslides, and many other hazards [5,6,7,8,9,10,11]. Heatwaves represent one of the most serious meteorological disasters impacting public health, and their frequency and severity are on the rise, with a tendency to worsen [12,13,14]. For example, the 1995 Chicago heatwave claimed the lives of 739 people [15]. At least 70,000 people died in the 2003 heatwave in central Europe, in addition to the catastrophic socioeconomic effects [16,17]. In 2010, a catastrophic heatwave in eastern Europe and western Russia killed over 55,000 people in Russia alone [18]. This was followed by a severe drought, which reduced food production by 25 percent and cost the local economy almost USD 15 billion [19]. In 2019, two heatwaves in western Europe caused deaths that exceeded normal rates by 50 percent, with temperatures in the Netherlands, Belgium, France, Germany, and the United Kingdom setting new highs in meteorological history [20]. In China, heatstroke (HS) during the summer has also become a serious public health issue in recent years [1]. The 2013 record-breaking high-temperature event in Shanghai caused about 160 deaths in the Pudong New District [21].

China is the most populated country in the world, and the overwhelming majority of the population is concentrated in the heatwave-prone eastern monsoon region [22], where air temperatures will continue to rise under current emissions scenarios. Population exposure to extreme drought is projected to increase by 17% if the global temperature increase is limited to 2.0 °C rather than 1.5 °C in the future [23], increasing the threat to the living conditions of most people living in China [24,25]. Like other disasters, heatwave-related disasters are impactful due to two major factors–the exposure to the phenomenon and the consequence of its occurrence. On the one hand, climate change scenarios including the magnitude of warming affect the hazard exposure [26,27]; while on the other hand, consequence is influenced by population and the exposure and socioeconomic vulnerability of that population [28,29]. Thus, heatwave risk can be described as a function of climate and population [28], both of which change over time and geographic area [4]. According to Jones et al., climate change is a stronger determinant of exposure than demographic change [30]. The determinants of population exposure, in general, are mostly determined by geographical disparities, taking into consideration regional population policies. The overall change in exposure could also result predominantly from the interactive effect of both factors, such as in the case of climate migrations across areas such as Africa [31,32].

Previous studies on heatwaves in China have mostly concentrated on the traits of intense heatwave episodes and their fluctuations [33,34,35,36,37,38,39,40]. Recent work has also provided a clearer understanding of the dangers of heatwaves in future temperature increase scenarios. Owing to recognition of the importance of public health management techniques, population exposure to intense heatwaves has received more attention recently [41]. Evidence suggests that climate change may be a major driver of changes in population exposure in China [42,43], and some studies show that the interaction between climate and population determines the population exposure to future extreme heatwaves in China [44]. However, there is currently little understanding of the relative contributions of heatwaves and demographic changes on population exposure changes on a regional scale. It is important to fill this gap as a critical next step for risk assessment of extreme heatwaves.

This study explores changes in heatwave and population exposure under two temperature increasing scenarios of 1.5 °C and 2.0 °C in China based on a daily maximum and minimum surface air temperature (T_max and T_min) dataset from CMIP6 and gridded daily datasets over China, along with population data that consider local population policies, from Oak Ridge National Laboratory’s LandScan program and the Shared Socioeconomic Pathways (SSPs) from Tsinghua University. In addition, differences in population exposure under the different warming scenarios are analyzed, with explanations of possible reasons for the differences. It is important to separately examine the differences in impact of the one-half degree of difference in increases from 1.5 °C to 2.0 °C, because the Intergovernmental Panel on Climate Change (IPCC) [45] recognizes these two thresholds, with the former being deemed likely to occur—with high confidence—between 2030 and 2052 if current rates are extrapolated. The 2 °C threshold is selected because IPCC (2018) notes, with medium confidence, that it would result in an additional 420 million people being frequently exposed to extreme heatwaves beyond those exposed to 1.5 °C of warming.

2. Materials and Methods

2.1. Dataset

For comparison and bias correction, daily T_max and T_min across China from a high-resolution (0.25° $\times$ 0.25°) gridded daily dataset, namely CN05.1, generated from more than 2400 stations affiliated with China Meteorological Administration [46] are taken into considered. Daily T_max and T_min output from 20 CMIP6 model simulations are employed to project the spatial distribution of future heatwaves under 1.5 °C and 2.0 °C warming scenarios. Information about these CMIP6 models is provided in Table S1. The historical scenario used is the “all-forcing” simulations (1995–2014). Projections of GCMs under three scenarios based on the combination of the shared socioeconomic pathways (SSPs) [47] and the representative concentration pathways (RCPs) [48]—including SSP1–2.6, SSP2–4.5, and SSP5–8.5—are included. The time frame of 1.5 °C and 2.0 °C warming globally is shown in

Table 1. Earliest 20-year averaging period that displays 1.5 °C and 2.0 °C of global warming, by SSP threshold. The change is displayed in °C relative to the 1850–1900 reference period for the selected time periods.

	SSPs	SSP1-2.6	SSP2-4.5	SSP5-8.5
Warming Scenarios
1.5 °C		2023–2042	2021–2040	2018–2037
2.0 °C		post–2100	2043–2062	2032–2051

, which is consistent with the IPCC AR6 [49]. Output datasets from the CMIP6 simulations are downscaled into 0.25° $\times$ 0.25° grid cells using the bilinear interpolation method via xarray (version 2022.3.0) [50].

The historical population data come from LandScan, which contains the global population for the period 2000–2020 and is widely used in natural and human-caused disaster risk assessments [51]. The projected future population data come from the gridded dataset of Tsinghua University from 2010 to 2100 [52]. Both historical and future population data have a resolution of 30 arc-seconds (~1 km) and are up-scaled to 0.25° $\times$ 0.25° using the grid area scaling method. The population SSPs scenarios included in this dataset are consistent with CMIP6, including data from three scenarios: SSP1–2.6, SSP2–4.5, and SSP5–8.5.

2.2. Definition of Present, 1.5 °C, and 2.0 °C Warming

In this study, the historical baseline period of 1995–2014 is used, based on the recommendation of IPCC AR6. The historical climate or historical scenario in this work refers to the climate of present. The selected time periods of 1.5 °C and 2.0 °C of warming relative to 1850–1900 for 20-year averaged global surface air temperature (GSAT) changes are based on multiple lines of evidence (the time span of each SSP as shown in Table 1). Therefore, the meteorological and population data for 1.5 °C and 2.0 °C warming are averaged using multi-model ensembles in the corresponding warming scenarios, respectively. Historical climate and population data (1995–2014) are compared with the projected data for 1.5 °C and 2.0 °C of warming. Because the population of LandScan is as early as 2000, the historical population is represented by 16 years of data (2000–2014).

2.3. Quantile Mapping Bias Correction

Before using GCM data, it is critical to correct for biases in the GCM simulations against observed data. After downscaling the climate data into 0.25° $\times$ 0.25° grid cells, a quartile mapping technique is used to bias correct the historical simulation and future projection of the GCM models, using cumulative density function matching methods [53]. The key to quantile deviation correction is to establish a transfer function, which generally includes a theoretical probability distribution function [54,55,56] and empirical probability function [57], using theoretical cumulative distribution function and cumulative distribution function of observed data, respectively. In comparison, the empirical probability function has wider utility because it does not require any assumptions about the statistical distribution of the original data [58]. For each grid cell, given month, and variable (T_max and T_min), the transform function is fitted by the qq-plot of GCM historical simulations and observations [58]. Then, bias correction using transfer functions for historical simulation and warming scenarios results. In this work, the quantile mapping bias correction of the GCM output is implemented with xclim [59].

2.4. Heatwave Events and Population Exposure

Early research on heatwaves only considered whether the highest temperature or lowest temperature on a single day exceeded a certain threshold. However, exceedance thresholds that correspond optimally to population exposure are likely to be affected by antecedent and subsequent temperature [60,61]. More recently, compound hot extremes have considered both daily T_max and T_min based on a bivariate definition framework [62,63,64]. Recently, studies on heatwaves in the Northern Hemisphere have used relative temperature thresholds and diurnal temperature range (DTR) to identify heatwaves. This method has been shown to offer improvements by defining heatwave events to be defined more strictly and requiring both T_max and T_min to exceed the 95th quantile [65].

In this study, a heatwave is considered to occur if all of the following criteria are met: (1) daily T_max ≥ 95th percentile of observed daily T_max in May, June, July, August, and September during 1985–2014); (2) daily T_max ≥ 35 °C [66]; (3) daily DTR in those same months ≤ 10 °C, to ensure that uncomfortable conditions persisted throughout the day and in recognition that both sensible and latent heating are key factors in the warming and stress effects [65,67]. The number of days spent experiencing a heatwave in a year is referred to as heatwave days (HD). When a heatwave event straddles two calendar years, it is counted in the year in which the heatwave began. Expressing “risk” as the product of hazard intensity and hazard exposure [68] population exposure (E) is defined as the product of HD and the total population (Pop) [69] at the grid point, and the unit is person-days, shown as

$$E=HD\times Pop$$

(1)

2.5. Amplified Impacts and Relative Contributions

In this study, heatwave effect is defined as E of compound heatwave events. Li et al. [70] and Jones et al. [28] showed that the absolute amplified impact ($A{I}_a$) due to an 0.5 °C additional warming is defined as

$$A{I}_a=E_{2.0}-E_{1.5}$$

(2)

where $E_{1.5}$ and $E_{2.0}$ represent the population exposure under the 1.5 and 2.0 °C warming scenarios, respectively. Relative amplified impact ($A{l}_r$) is defined as the ratio of the additional exposure to the baseline exposure, thereby indicating the proportion of the population that will be additionally exposed to heatwaves by increasing the delta temperature relative to the baseline temperature rise scenario. The equation is

$$A{I}_r=\frac{E_{2.0}-E_{1.5}}{E_{1.5}}\times 100\%$$

(3)

According to the definition of population exposure, the increases in population exposure relative to historical simulations can be calculated as

$$\begin{array}{c}\Delta E=E_f-E_h\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}=H{D}_f\times Po{p}_f-H{D}_h\times Po{p}_h\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}=\Delta HD\times Po{p}_h+\Delta Pop\times H{D}_h+\Delta HD\times \Delta Pop\end{array}$$

(4)

where $\Delta E$ is relative to the historical baseline period, $E_h$ and $E_f$ depict population exposure under historical simulation and future global warming scenarios, respectively; $H{D}_f$ and $H{D}_h$ represent HD under future global warming scenarios and historical simulation, respectively; $Po{p}_f$ and $Po{p}_h$ correspond to the population numbers under the future global warming scenarios and historical simulation, respectively; and $\Delta HD$ and $\Delta Pop$ represent the changes of HD and population in the future global warming scenarios and historical simulations, respectively. Furthermore, $\Delta HD\times Po{p}_h$ represents the change in population exposure due to climate change, $\Delta Pop\times H{D}_h$ represents the change in population exposure due to population change, and $\Delta HD\times \Delta Pop$ represents the change in population exposure due to interaction of climate and population change. The consistency of sigsn between these three elements and $\Delta E$ represents the contribution to the change in exposure, respectively. Consistency in sign signifies a positive effect on the change of population exposure; otherwise, it is a negative effect.

3. Results

3.1. Changes in Heatwave Events

To evaluate the ability of GCM to simulate heatwaves, the spatial distribution of HD in China is analyzed based on CN05.1 and multi-model ensemble of GCM outputs. Figure 1a,b shows the annual number of HD from 1995 to 2014 based on observational data and GCM simulations. The present number of HD is generally 3–10 days per year, and the HD simulated by the GCM is between 3 and 11 days per year. Heatwaves mainly occur in the central and eastern regions of China and the Sichuan Basin, including: the Yangtze River Basin (YRB) region, the Chongqing-Chengdu (CC) region, and the Pearl River Basin (PRB) region (blue box in Figure 1). The consistency between the historical simulation of the model and the CN05.1 observation data is high, which indicates that a multi-model ensemble of GCM can better simulate the spatial distribution of the heatwave hotspots. The simulated error of annual HD is less than 0.2 days, the regional average (areas with heatwaves) annual HD of the model simulation and CN05.1 are 4.45 and 4.63 days, respectively. As shown in Figure 1c, in places with more HD, the simulation error of HD is less, and vice versa. In general, the multi-model ensemble can predict the number of HD in China accurately, with the exception of slight overestimation in a few places.

Not surprisingly, compared with the present simulation, the HD under the 1.5 °C and 2.0 °C global warming scenarios show greater HD frequency, and the regional average annual HD increases from 4.45 days to 5.36 and 6.44 days, respectively, as shown in Figure 1d,f. The 0.5 °C of additional warming in the 2.0 °C warmer future would lead to an increase in HD by approximately 20.15 percent when compared to that under the 1.5 °C warmer climate. The annual HD at the core of the heatwave region increased by 1 and 1.8 times. Heatwave-free areas under the 1.5 °C scenario—such as Northeast China, Xinjiang, and Inner Mongolia—are expected to experience heatwaves in the 2.0 °C scenario, but the annual HD generally do not exceed 5 days in the 2.0 °C scenario. Figure 1c,g show the spatial variation of HD in the historical simulation for the 1.5 °C and 2.0 °C global warming scenarios, and more than 95 percent of the regions where heatwaves have occurred show the characteristics of increased HD, with annual HD frequency increases of 2.68 and 4.22 days, respectively. Except for newly added heatwave areas, the spatial characteristics of HD are consistent with historical periods. The HD in the Yangtze and Pearl River Basin regions showed a more obvious increasing trend under the 2.0 °C warming scenario than in the 1.5 °C warming scenario. These areas will have high population concentration in the future (as shown in Figure S1), indicating that the population may face more severe heatwave exposure risks.

3.2. Present Distribution Patterns and Future Changes in Population Exposure

Population exposure is an important parameter for revealing the risk of future heatwaves. Figure 2 shows the spatial distribution of E under the present, 1.5 °C, and 2.0 °C warming scenarios based on GCM models. During the present period, increased E to heatwaves occurred mainly in central and eastern China and the Sichuan Basin. Under the 1.5 °C and 2.0 °C warming scenarios, the annual E increased from 4.09 billion person-days in the present period to 7.98 billion and 11.22 billion person-days, respectively. The per capita annual HD frequency increased from 4.72 days to 7.0 and 9.89 days. Despite the expansion of E areas, the spatial distribution of E centers is relatively concentrated. To a certain extent, the spatial variation of E is determined by both climate change and population redistribution.

To focus on the changes of population exposure, as shown in Figure 3, the spatial variation characteristics of E are explored. The newly added heatwave regions—such as Northeast China, Xinjiang, and Inner Mongolia—show increased E, which is mainly caused by the increase in the frequency of heatwave events under climate warming. In other regions, E I ncreased or decreased differently, depending on the redistribution of population and heatwave occurrence. Compared with the 1.5 °C warming scenario, E under the 2.0 °C warming showed a steady increase in most areas of central and eastern China, and the annual E increased by 3.24 million person-days. This means that if the climate continues to warm by 0.5 °C without emission reduction measures, E related to heatwaves will increase by an additional 3.24 million person-days.

3.3. Amplified Impacts and Relative Contributions of Climate and Population

As analyzed in Section 3.2, 0.5 °C of additional warming has an absolute amplified impact of 3.24 million person-days on E. This means that the relative risk of E ($A{I}_r$) under 2.0 °C warming would increase by 40.6 percent for a 1.5 °C warming scenario. Relative to present, 1.5 °C and 2.0 °C warming increase the impact by 95.1 and 174.3 percent, respectively. Although it is known that climate and population redistribution are the main contributors to increased E by heatwaves, the magnitude of their respective contributions is unknown. Here, changes in E are decomposed into the effects of climate ($\Delta HD\times Po{p}_{his}$), population ($\Delta Pop\times H{D}_{his}$), and interactions ($\Delta HD\times \Delta Pop$), based on Equation (4).

Figure 4 shows the contribution of these three effects to the change in E to heatwaves in China under the 1.5 °C and 2.0 °C climate warming scenarios, respectively. Under the 1.5 °C climate warming scenario, the main factor causing the change in E is climate, which accounts for 71.99 percent of the change in E to heatwaves. The second factor is population, which accounts for 16.91 percent of the total changes. The contribution of interaction was the smallest, at about 11.09 percent. Under the 2.0 °C warming scenario, the contribution of climate is still larger (80.34 percent), followed by interaction (12.42 percent), and population (7.24 percent). This suggests that climate plays a more important role in changes of E with further increases in climate warming. In addition, population dynamics are also one of the key factors affecting population exposure changes. Although urbanization has led to the migration of China’s population centers to cities, the future population will begin to decline slowly after reaching a peak (1.45 billion) around 2030, as a result of many factors, including birth, death, migration, education level, and population policies under each SSP. The 1.5 °C warming scenario period coincides with the peak of China’s total population, while the 2.0 °C warming scenario period occurs during the stage after the population peaks (as shown in Figure S2). This may partially explain why the population contribution under the 2.0 °C warming scenario is smaller than in the 1.5 °C warming scenario. The decomposition of E changes from the 1.5 °C to 2.0 °C warming scenarios also illustrates this problem, with the population contributing negatively to the total change.

The relative contributions of the three elements to the overall change in population exposure to heatwaves are further analyzed at the regional scale. Figure 5 shows the spatial distribution of the relative contributions of climate, population, and interaction to changes in E under the warming scenarios. Climate and population show positive contributions in most regions, with climate contributing substantially. The proportion of positive and negative contributions for the interaction component is similar. For the 1.5 °C warming scenario, the climate elements in the newly exposed areas—the Yangtze and Pearl River Basins and the eastern Sichuan Basin—play a positive role in E. The magnitude and area of this effect are further expanded under 2.0 °C warming scenario. The negative impact area of population on E has an increasing trend. While both positive and negative contributions of the interaction were characterized by an increase, the increase in the overall positive effect was greater. In addition, we can find that population and interaction are negative in most regions, while the contribution of climate is positive in the total change in population exposure from 1.5 to 2.0 °C warming. It is important to note that at 0.25° $\times$ 0.25° resolution, the relative contribution of individual elements can exceed 100 percent because of the extreme drastic changes in population and HD at the individual grid scale. However, the sum of the relative contributions of climate, population, and interaction at each grid point equals 100 percent.

4. Discussion

This study assessed heatwave risk via E—under 1.5 and 2.0 °C warming scenarios—based on results from the latest CMIP multi-model ensemble, through changes in heatwave and regional population changes. Instead of using the Coupled Model Intercomparison Project phase 5 (CMIP5) data in most previous studies, we used the CMIP6 model output data, which shows improvements in simulation of climate indices in China, especially the daily maximum and minimum temperatures [8,25]. Results suggest that if nothing occurs to curtail continued global warming, the risk of regional heatwaves in China will increase by 2–2.8 times by mid-century. The spatial distribution of heatwave is consistent with the results of previous studies [40,42,62]. Of these, climate change was the dominant factor (>70 percent) in increasing exposure to heatwave risk.

These results are only preliminary conclusions. The main limitations include that the heatwave risk in this study mainly considers E, whereas future work should consider differences in vulnerability of different populations (including age, gender, income, education level, etc.) to quantify the assessment of heatwave exposure risk more accurately. In addition, the spatial redistribution of population is critical to the estimation of E relative to heatwave. However, urbanization is also an important factor affecting the future redistribution of China’s population. Therefore, analyzing changes in E from urban and rural perspectives is conducive to improved scientific understanding of heatwave risk changes. Recent studies have suggested that the wet bulb temperature is closer to the human apparent temperature, which can better reflect the intensity of the heatwave. Therefore, compound heatwaves should be considered more carefully in future studies.

5. Conclusions

In this study, the latest CMIP6 multi-model ensemble results were used to assess changes in frequency of HD at 1.5 °C and 2.0 °C global warming levels compared to present. On this basis, the changes in E to heatwaves were assessed based on population projection data, and a relative contribution decomposition analysis was performed on the changes in E. Results showed a significant increase in the frequency of HD at two different warming levels. The assessment highlighted that 1.5 °C and 2.0 °C warming increased E by 95.1 and 174.3 percent compared to present. An additional 0.5 °C temperature rise would significantly affect changes in E with about 40.6 percent more E. In addition, the climate element played a significant positive leading role (>70 percent) in total changes of E to heatwaves. Of the additional 0.5 °C warming, climate change positively contributes to the increase in the risk of E, with 109.69 percent of total E. Population and interaction tend to decrease the risk of E with relative contributions of –7.71 and –1.98 percent, respectively, in total change of E. Therefore, considering future heat risks, humanity benefits from a 0.5 °C reduction in warming, particularly in eastern China. This conclusion provides useful insights for advancing climate change adaptation.