This is the accepted version of the article: Chen, Jie Chen; Liu, Yujie; Pan, Tao; [et al.]. «Global socioeconomic exposure of heat extremes under climate change». Journal of cleaner production, Vol. 277 (Dec. 2020), art. 123275. DOI 10.1016/j.jclepro.2020.123275 This version is avaible at https://ddd.uab.cat/record/232214 under the terms of the license 1 Amount of words 8419 2 Global socioeconomic exposure of heat extremes 3 under climate change 4 Jie Chen1,2, Yujie Liu1,2,*, Tao Pan1, Philippe Ciais3, Ting Ma4, Yanhua Liu1, Dai 5 Yamazaki5, Quansheng Ge 1,2,& Josep Peñuelas6,7 1 6 Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and 7 Natural Resources Research, Chinese Academy of Sciences (CAS), Beijing, China 2 8 University of Chinese Academy of Sciences (UCAS), Beijing, China 9 3 Laboratoire des Sciences du Climat et de l’Environnement, IPSL-LSCE CEA CNRS UVSQ, Gif- 10 sur-Yvette, France 4 11 State Key Laboratory of Resources and Environmental Information System, Institute of 12 Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (CAS), 13 Beijing, China 5 14 Institute of Industrial Sciences, University of Tokyo, Tokyo, Japan 6 15 CSIC, Global Ecology CREAF-CSIC-UAB, Bellaterra, 08193 Barcelona, Catalonia, Spain 7 16 CREAF, Cerdanyola del Vallès, 08193 Barcelona, Catalonia, Spain 17 18 * Corresponding author at: No.11A, Datun Road, Chaoyang District, Beijing, China 19 E-mail address: liuyujie@igsnrr.ac.cn (Yujie Liu) 20 Abstract 21 Growing evidence indicates that the risk of heat extremes will increase as climate 22 change progresses and create a significant threat to public health and the economy. 23 Socioeconomic exposure is the key component for assessing the risk of such events. To 24 quantify socioeconomic exposure to heat extremes for 2016–2035 and 2046–2065, we 25 use the projections of five global climate models forced by using three representative 26 concentration pathways (RCPs) and projections of population and gross domestic 27 product (GDP), and we take into account the geographic change in the distribution in 28 shared socioeconomic pathways (SSPs). The exposure of the global population for 29 2046–2065 is the greatest under the RCP8.5-SSP3 scenario, up to 1037(±164) × 109 30 person-days, and the global GDP exposure for 2046–2065 is greatest under the RCP2.6- 31 SSP1 scenario, up to 18(±2) × 1015 dollar-days. Asia has the highest exposure among 32 all continents for both population and GDP, accounting for over half of the global 33 exposure. Africa has the largest increase in exposure, with the annual population and 34 GDP exposures increasing by over 9- and 29-fold, respectively, compared with the base 35 period (1986–2005). The effect of climate makes the dominant contribution (47%–53%) 36 globally for the change in population exposure. Changes in the geographic distribution 37 of GDP cause nearly 50% of the total change in GDP exposure for 2016–2035. 38 Mitigating emissions of greenhouse gases, either at the level of the RCP2.6 scenario or 39 at a more ambitious target, is essential for reducing socioeconomic exposure to heat 1 40 extremes. In addition, designing and implementing effective measures of adaptation are 41 urgently needed in Asia and Africa to aid socioeconomic systems suffering from heat 42 extremes due to climate change. 43 Keywords socioeconomic exposure; heat extremes; climate change; population 44 exposure; gross domestic product (GDP) exposure 45 1 Introduction 46 Climatic extremes cause extensive economic damage each year, and the risks are 47 expected to increase with continued socioeconomic development and climate change. 48 Risk is usually represented as the probability of occurrence of hazardous events or 49 trends multiplied by the impact if these events or trends occur. Risk is due to the 50 interaction of hazard, exposure, and vulnerability. Exposure usually refers to the people, 51 livelihoods, resources, infrastructure, and economic, social, or cultural assets in places 52 and settings that could be adversely affected (IPCC, 2014). The prediction of changes 53 in exposure to future climatic extremes could, therefore, contribute to the need to 54 consider effective countermeasures to reduce vulnerability and risk. Global 55 socioeconomic exposure (i.e., exposure of the population and gross domestic product, 56 GDP) under climate change has received less attention (Burke et al., 2015; Carleton 57 and Hsiang, 2016), despite recent progress in assessing the hazards of climatic extremes 58 (Hirabayashi et al., 2013; Cook et al., 2014; King et al., 2017; Huang et al., 2017; Nath 59 et al., 2017). Such estimates are urgently needed to clarify future spatiotemporal 60 variation and changes in global socioeconomic exposure and thereby avoid adverse 61 effects on public health and the economy. The importance of assessing exposure is 62 gradually being recognized, as indicated by the published reports of the 63 Intergovernmental Panel on Climate Change (Field et al., 2012; IPCC, 2013), the 64 proposal of a new set of shared socioeconomic pathways (SSPs, O’Neill et al., 2014), 65 and some recent studies of the risk of climate change that considers the effects of 66 socioeconomic factors (Ceola et al., 2015; Mora et al., 2017; Zhang et al., 2018). The 67 integration of climate change with the social economy to estimate future risk, the 68 assessment of the relative importance of different factors, and the quantification of 69 uncertainty would provide a basis for adapting to extreme climate change and reducing 70 the risk of heat extremes. 71 Because of climate change, the frequency and intensity of climatic extremes, such as 72 heat extremes, have increased in recent decades and are likely to continue to increase 73 in the coming decades (IPCC, 2013). Heat extremes have been responsible for 74 significant public health threats and economic losses over the last 100 years. Data for 75 historical damage from the International Disaster Database 76 (https://www.emdat.be/emdat_db/) indicate that heat extremes affected 1.03 × 108 77 people, which includes the deaths of 183 495 people, and caused worldwide economic 78 losses of 6.33 ×1010 USD from 1936 to 2019. Serious heat extremes have also occurred 79 more frequently in recent decades, such as those in Europe in 2003, 2018 and 2019 80 (Robine et al., 2008), Australia in 2008 (Vaneckova et al., 2008), Russia in 2010 2 81 (Trenberth and Fasullo, 2012), and China in 2013 (Sun et al., 2014). Socioeconomic 82 exposure and disaster risks will be magnified in a warmer future (Jones, and O’Neill, 83 2016; Smirnov et al., 2016), when the more frequent and intense heat extremes of this 84 century (Fischer and Knutti 2015; Kharin et al., 2013) combine with a greater 85 population and the accumulation of wealth. 86 In socioeconomic systems, population is most closely related to heat extremes 87 because of its direct impact on public health (Wang et al., 2019). Various economic 88 sectors also can be seriously affected by heat extremes, such as agriculture (water 89 shortages) and industries that rely heavily on hydropower. In addition, tourism, 90 transportation, construction, and other industries are also affected by heat extremes to 91 varying degrees (IPCC, 2014). Considering the data on historical damage of heat 92 extremes (number of people affected, total damage on the economy, etc.) and the 93 availability of simulated socioeconomic data for the future, we selected population and 94 economic activity, expressed as GDP, to determine the impact of heat extremes on the 95 socioeconomic system. Many studies that have quantified future socioeconomic 96 exposure have not considered changes in populations or GDP but assumed that these 97 variables remain constant, which is inappropriate for predicting changes in exposure 98 (Bouwer, 2013; Sun et al., 2017). Most studies have also focused on changes in 99 exposure based on a specific pathway of emission of a greenhouse gas (GHG) or a 100 target of global mean temperature (GMT) rise such as 1.5 °C or 2.0 °C (Harrington and 101 Otto, 2018; Mishra et al., 2017). These analyses of the spatiotemporal variation in 102 exposure between scenarios and time periods are thus insufficient because the 103 population and economy (usually calculated by GDP) are both essential elements in 104 socioeconomic systems and are the factors most severely affected by climatic extremes. 105 The exposures of populations and GDP are usually predicted separately, and these two 106 elements are rarely used to assess socioeconomic exposure (Bowles et al., 2014; 107 Forzieri et al., 2017), although the spatial distributions of populations and economies 108 are also consistent. Given these factors, we focus in this work on simulating the global 109 exposure to extreme heat as a function of changes in climate and population or GDP. 110 The results constitute a first step toward understanding how interactions between 111 climate change and socioeconomic systems affect exposure patterns. 112 This study systematically quantify the global spatiotemporal distribution of and 113 changes in exposure of population and GDP to heat extremes under different scenarios 114 and over different time periods. Bias-corrected projections of five global climate 115 models (GCMs, Table 1) driven by representative concentration pathways (RCPs) are 116 used to calculate the frequency of extreme heat events (see Sec. 2.1, Materials). 117 Combined with population and GDP projections in SSPs (Figs. S1 and S2), which 118 consider changes in the geographic distribution of population and GDP (see Sec. 2.1, 119 Materials), the spatiotemporal variation of global exposure to extreme heat of the 120 population and GDP is quantified in both the base period (1986–2005) and in future 121 periods (2016–2035 and 2046–2065) under various scenarios (see Sec. 2, Materials and 122 Methods). We also assess the relative importance of climatic and socioeconomic factors 123 and their uncertainties to characterize the contribution of climate change and growth in 124 population and GDP to future changes in extreme-heat exposure. The main target of 3 125 this study is quantifying the impact of heat extremes on socioeconomic system under 126 climate change to characterize variation of socioeconomic exposure among scenarios 127 and periods, distinguish high exposure regions, and identify dominant contributor for 128 exposure change, so as to support policymakers in the development of climate change 129 mitigation and adaptation strategies. 130 2 Materials and Methods 131 2.1 Materials 132 Daily climatic data were obtained from the Inter-Sectoral Impact Model 133 Intercomparison Project (ISI-MIP, Warszawski et al., 2014) for the Coupled Model 134 Intercomparison Project Phase 5 (CMIP5, Taylor et al., 2012) (Table 1), which contains 135 simulations from five GCMs based on RCPs. The RCP scenarios represent pathways 136 based on simulated impacts on land use, aerosol emissions, and GHGs (Vuuren et al., 137 2011). The four RCPs cover the period up to 2100 and have radiative forcings from the 138 open literature that vary from 2.6 to 8.5 W/m2. The scenarios, RCP2.6, RCP4.5, and 139 RCP8.5, which represent low, middle, and high GHG emissions, respectively, were 140 selected for analysis (Vuuren et al., 2011). RCP6.0, which is interpreted as either a 141 medium baseline or a high-mitigation case between RCP4.5 and RCP8.5, is not used. 142 The base period was 1986–2005, which is the commonly used reference period in 143 assessments of projected changes in extreme indices and climate impacts (Schleussner 144 et al., 2016). Future periods in the 2030s and 2050s were given lengths of 20 years 145 (2016–2035 and 2045–2065) to be consistent with the base period. The spatial 146 resolution of the output data was offset-corrected and converted to 0.5°× 0.5°latitude 147 and longitude by spatial downscaling. Statistical bias-correction methods facilitate the 148 comparison between observed and simulated data during the historical reference period 149 and for a continuous transition into the future (Hempel et al., 2013). Preservation of 150 absolute changes in monthly temperature and relative changes in monthly precipitation 151 in each grid cell implies that the global warming trend and the climate sensitivities of 152 the GCMs are preserved, and the trend and the long-term mean are well represented, 153 which ensures the credibility of the simulated data (Hempel et al., 2013; Warszawski et 154 al., 2014). 155 The United Nations, the World Bank, and other organizations proposed future 156 socioeconomic projections for population and GDP. Many previous studies also 157 combined current socioeconomic data with future climatic data for analysis, although 158 these studies neglected to consider how changes in socioeconomic factors affect 159 exposure. We used the predictions of population and GDP from the scenarios of SSPs 160 based on the selected RCP scenarios. SSPs are reference pathways describing possible 161 alternative trends in the evolution of societies and ecosystems on a timescale of 100 162 years without climate change or implementation of climate policies (Riahi et al., 2017). 163 RCP2.6, 4.5, and 8.5 generally correspond to SSP1, 2, and 3, respectively, based on the 164 correspondence between the RCPs and SSPs provided by the IPCC (O’Neill et al., 2014, 165 Table S4). We therefore selected SSP1, 2, and 3 for this study. The RCP2.6-SSP1 4 166 scenario assumes low carbon emissions, sustainable development proceeding at a 167 reasonably high pace, and fewer inequalities. The RCP4.5-SSP2 scenario assumes 168 moderate carbon emissions with medium growth in population and GDP. Finally, the 169 RCP8.5-SSP3 scenario assumes high carbon emissions with a rapidly growing 170 population and a low adaptive capacity. The projections of population and GDP were 171 obtained from the National Institute for Environmental Studies, Japan (NIES), which 172 were downscaled from the International Institute for Applied Systems Analysis 173 (IIASA). The spatial resolution was also 0.5° × 0.5° latitude and longitude. The 174 population and GDP projections were downscaled with explicitly considered spatial 175 and socioeconomic interactions between cities, and they used auxiliary variables, 176 including road network and land cover. The downscaling results were consistent with 177 the scenario assumptions and captured the difference in urban and non-urban areas in a 178 more reasonable manner, which ensures the prediction accuracy of the SSPs (Murakami 179 and Yamagata, 2016). 180 Table 1. Description of global climate models (GCMs). Atmospheric Model Institute resolution (longitude ×latitude) GFDL-ESM2M Geophysical Fluid Dynamics Laboratory 2.5°×2° National Institute of Meteorological HadGEM2-ES 1.875°× 1.25° Research/Korea Meteorological Administration IPSL-CM5A-LR Institute Pierre-Simon Laplace 3.75°× 1.875° Japan Agency for Marine-Earth Science and MIROC-ESM- Technology, Atmosphere and Ocean Research 2.8°×2.8° CHEM Institute (University of Tokyo), and National Institute for Environmental Studies NorESM1-M Norwegian Climate Centre 2.5°×1.89° 181 182 2.2 Methods 183 2.2.1 Hazards of extreme heat 184 Annual days of extreme heat, which is also the frequency of extreme heat, were 185 used to quantify the hazard, which is defined as the daily maximum temperature 186 exceeding a threshold. The threshold for extreme heat was defined as the 90th percentile 187 of daily maximum temperatures for the base period (1986–2005) and was set at 25 °C 188 when the local 90th percentile was <25 °C (Garssen et al., 2005). We chose relative 189 thresholds rather than a fixed threshold to project global spatiotemporal variation and 190 changes in exposure because no single fixed threshold suffices for the substantial 191 differences in climatic conditions around the world (Gasparrini et al., 2015). Relative 192 thresholds were, therefore, the simplest definitions of regionally relevant extreme heat 193 around the globe. The frequency of extreme heat was calculated as follows: 𝐶 = ∑365 𝑖=1 (𝑇𝐸𝑀𝑖 > 𝑇𝐻𝑅), (1) 5 ∑5𝑗=1 𝐶𝑗 𝐶̅ = , (2) 5 194 where 𝐶 is the annual number of days of extreme heat (day), i is the ith day of a year, 195 𝑇𝐸𝑀 is the daily maximum temperature (°C), THR is the local threshold (°C), 𝐶̅ is the 196 multi-model averaged value of 𝐶 (day), and j is the jth GCM. 197 2.2.2 Exposure to extreme heat 198 We measured population and GDP exposure for each grid cell as the number of 199 extreme heat days multiplied by the number of people and GDP, respectively (Jones et 200 al., 2015). Therefore, the units of population and GDP exposures are person-days and 201 purchasing power parity (PPP) dollar-days, respectively. To calculate exposure in the 202 base period and in the future periods (2016–2035 and 2046–2065), we minimized 203 interannual variations by using 20-year averages of annual extreme heat days and of the 204 projections of population and GDP. The 20-year mean exposure for each projection of 205 the five climate models was calculated for the base period and for the future periods. 206 Moreover, exposure for the grid cells was also aggregated to global and continental 207 scales for further analysis. Explicitly, we have ∑ 20 𝐶 ×𝑃 ̅̅̅ 𝐸𝑃 = 𝑚=1 𝑚 , 20 (3) 20 ̅𝐸̅̅𝐺̅ = ∑𝑚=1 𝐶𝑚 × 𝐺 , (4) 20 208 where 𝐸̅̅̅ th 𝑃 is the 20-year-averaged population exposure (person-day), m is the m year 209 of the study period, 𝐶 is the number of annual days of extreme heat (day), 𝑃 is the 210 simulated population number (person), 𝐸 ̅̅̅𝐺̅ is the 20-year-averaged GDP exposure 211 (PPP $-day), and 𝐺 is the GDP simulation (PPP $). 212 2.2.3 Analysis of relative importance of change in exposure and cumulative 213 probability 214 Using techniques from a previous study (Jones et al., 2015), we evaluated the relative 215 importance of the effects of different factors by categorizing the changes in population 216 and GDP exposures in terms of the effects of climate, population, GDP, and interactions. 217 The impact of population and GDP was calculated by holding climate constant (i.e., the 218 20 year averages of annual days of extreme heat for the base period were multiplied by 219 the populations and GDPs in the RCP-SSP scenarios). Population and GDP were 220 similarly held constant when calculating the impact of climate (i.e., the population for 221 the base period was multiplied by the 20 year averages of annual days of extreme heat 222 in the RCP scenarios). The interactive effect was also calculated to determine whether 223 the areas with continued population and GDP growth experienced more heat extremes 224 under climate change. The changes in population and GDP exposure were categorized 6 225 as follows: ∆𝐸𝑃 = 𝐶𝑏 ∆𝑃 + 𝑃𝑏 ∆𝐶 + ∆𝑃∆𝐶, (5) 226 ∆𝐸𝐺 = 𝐶𝑏 ∆𝐺 + 𝐺𝑏 ∆𝐶 + ∆𝐺∆𝐶, (6) 227 where ∆𝐸𝑃 is the total change in population exposure, ∆𝐸𝐺 is the total change in GDP 228 exposure, 𝐶𝑏 is the annual days of extreme heat for the base period, 𝑃𝑏 and 𝐺𝑏 are 229 the population and GDP for the base period, respectively, ∆𝐶 is the change in annual 230 days of extreme heat from the base period to future periods, and ∆𝑃 and ∆𝐺 are the 231 changes in population and GDP, respectively, from the base period to future periods. 232 Therefore, 𝐶𝑏 ∆𝑃 is the population effect, 𝐶𝑏 ∆𝐺 is the GDP effect, 𝑃𝑏 ∆𝐶 and 233 𝐺𝑏 ∆𝐶 are the climatic effects, and ∆𝑃∆𝐶 and ∆𝐺∆𝐶 are the interactive effects. 234 The uncertainties of changes in population and GDP exposure for future scenarios 235 were analyzed to evaluate the possible impact of climate change and growth on 236 population and GDP. The probability analysis of changes in population and GDP were 237 first separately calculated in each GCM based on the cumulative distribution function 238 (CDF). Next, the mean value and standard deviation for the five GCMs were computed. 239 The CDF of a random variable X represents the probability that X ≤ x. 240 3 Results 241 3.1 Spatial pattern of population and GDP exposures to extreme heat 242 Figure 1 shows the multi-model average exposures of the population and GDP for 243 RCP8.5-SSP3 for 2046–2065. Figure S3 shows the frequency of extreme heat, and Figs. 244 S4 and S5 show the population and GDP exposures, respectively. Tables S1 and S2 245 present the statistics for population and GDP exposures globally and continentally for 246 the RCP-SSP scenarios and the different time periods. 247 248 Fig. 1: Multi-model global projections of average exposures of (a) population and (b) 249 GDP to extreme heat for the RCP8.5-SSP3 scenario for 2046–2065. PPP is purchasing 250 power parity in USD. 251 The spatial distribution of the frequency of heat extremes indicates latitudinal 252 zonality in each time period and for each scenario (Fig. S3). The threshold in the base 253 period exceeds 25 °C, except at latitudes >50°and on the Qinghai-Tibet Plateau. Heat 254 extremes are most frequent near the equator, and their frequency gradually decreased 255 with increasing latitude in both the base period and future periods. The frequency of 7 256 extreme heat clearly increases over time. The highest frequency is 36.5 days in the base 257 period, whereas the frequency is projected to exceed 120 days in the RCP scenarios. 258 The frequency is significantly higher for 2046–2065 than 2016–2035 under each 259 scenario. The frequency is highest under the RCP8.5 scenario and lowest under RCP2.6, 260 but the difference between the scenarios is less than the difference between the time 261 periods. 262 The regions with high population and GDP exposures to extreme heat are primarily 263 concentrated in densely populated areas, such as India, China, midwestern Europe, the 264 eastern USA, and the coastal areas of South America, for both the base period and the 265 projected scenarios (Fig. 1). Exposure is also high near the equator in Africa because 266 of the frequent occurrence of extreme heat. Population exposure under the RCP 267 scenarios is highest for India and the east coast of China (>10 ×106 person-days). GDP 268 exposure is much higher for eastern China, the Indian subcontinent, western Europe, 269 and eastern North America than other locations such as northern Asia, northern North 270 America, and Middle Oceania, with annual GDP exposure >100 ×109 PPP dollar-days. 271 3.2 Global and continental change in exposure to extreme heat 272 Annual exposure of the global population for the base period is 217.80 ×109 person- 273 days, which increases to 1037.06 × 109 person-days for 2046–2065 under the RCP8.5- 274 SSP3 scenario (Table S1). The increase in population exposure is largest for the 275 RCP8.5-SSP3 scenario and smallest for RCP2.6-SSP1 (Fig. 2). The increase in global 276 annual GDP exposure is largest for 2046–2065 under the RCP2.6-SSP1 scenario (Fig. 277 3), with a 10.47-fold increase in exposure relative to the base period (Table S2). In 278 contrast, the increase is smallest for 2016–2035 under RCP8.5-SSP3, with an increase 279 of only 3.71-fold relative to the base period. The reason that the highest GDP exposure 280 appears in the RCP2.6-SSP1 scenario is that SSP1 is a “sustainable” scenario that 281 assumes economic growth is shared at the global scale, so GDP increases relatively 282 more in countries that currently have less wealth. Therefore, the exposed GDP at the 283 global scale is higher in RCP2.6-SSP1, despite the hazard of extreme heat being lower 284 in RCP2.6. 285 The population exposure is highest for Asia, followed by Africa, and lowest in 286 Oceania in both the base period and the RCP scenarios. Exposure in Asia and Africa in 287 the base period is 63% and 14%, respectively (Table S1). The percentage of population 288 exposure under the three RCP scenarios decreases for Asia and increases for Africa. 289 The percentage of population exposure for 2046–2065 under the RCP8.5-SSP3 290 scenario decreases for Asia to 53% and increases for Africa to 29% of global exposure. 291 Exposure for the continental GDP in the base period ranks as follows: Asia > North 292 America > Europe > South America > Africa > Oceania (Table S2). 293 The percent of GDP exposure increases for Asia, Africa, and South America and 294 decreases for North America, Europe, and Oceania. The increase in both continental 295 population and GDP exposures is largest for Africa, with the annual population 296 exposure >9.20-fold higher under the RCP8.5-SSP3 scenario for 2046–2065 than for 297 the base period, and the change of population exposure is 1.22-fold larger over the same 298 period and under the same scenarios for Europe than in the base period, which is the 8 299 smallest change among the continents. GDP exposure increases the most in Africa, 300 by >29.34-fold relative to the base period under the RCP2.6-SSP1 scenario for 2046– 301 2065, whereas GDP exposure in North America is 3.99-fold higher for 2046–2065 than 302 the base period, which is the lowest among continents. 303 Figure 2 presents the relative changes in population exposure for the projected 304 scenarios relative to the base period. The rate of change is highest in Africa in regions 305 near the equator. Population exposure for 2016–2035 under the RCP2.6-SSP1 scenario 306 increases more than 2-fold. The increase in the coastal regions of South America is also 307 rapid, by more than 2-fold for 2046–2065 under the RCP4.5-SSP2 scenario. The rate 308 of change, however, is <50% in areas where the rate has usually been high, such as 309 China, the USA, and Western Europe. Population exposure is much higher for 2046– 310 2065 than 2016–2035 under each scenario when the differences are compared between 311 different time periods, with time, the population increases gradually, and when that 312 increase is crossed with the more frequent heat extremes affected by climate change. 313 Population exposure is highest under the RC8.5-SSP3 scenario, where population 314 growth is rapid, followed by the RCP4.5-SSP2 and RCP2.6-SSP1 scenarios. 315 316 Fig. 2: Multi-model global projections of average relative change in population 317 exposure to extreme heat for the three RCP scenarios and two time periods relative to 318 the base period: (a) RCP2.6-SSP1, (b) RCP4.5-SSP2, (c) RCP8.5-SSP3, (1) 2016–2035, 319 (2) 2046–2065. 9 320 Figure 3 presents the spatial distributions of the average relative changes in GDP 321 exposure for the RCP scenarios projected by multiple models. GDP exposure for 2016– 322 2035 increases most rapidly in eastern China, with a more than 9-fold increase relative 323 to the base period. GDP exposure also changes for central Africa and the South Asian 324 subcontinent, with more than 6-fold increases relative to the base period. GDP exposure 325 increases for 2046–2065 relative to the base period and 2016–2035. The increase in 326 exposure is largest for Asia and is >12-fold higher than the base period in many 327 countries, such as China, India, and Mongolia. GDP exposure also changes for Africa, 328 South America, and Eastern Europe. The change in exposure is relatively small for 329 North America and Oceania, less than 6-fold relative to the base period. GDP exposure 330 is highest under the RC2.6-SSP1 scenario, where GDP growth is fastest, followed by 331 the RCP4.5-SSP2 and RCP8.5-SSP3 scenarios. 332 333 Fig. 3: Multi-model global projections of average relative change in GDP exposure to 334 extreme heat for the three RCP scenarios and two time periods relative to the base 335 period: (a) RCP2.6-SSP1, (b) RCP4.5-SSP2, (c) RCP8.5-SSP3, (1) 2016–2035, (2) 336 2046–2065. 337 3.3 Analysis of relative importance of change in exposure 338 To determine the relative importance of various factors, we categorize the changes 339 in population and GDP exposures in terms of the effects of population, GDP, climate, 10 340 and their interactions. Figures 4–6 show the changes in exposure and its components 341 for the globe and for the continents under the three RCP scenarios. The change in global 342 exposure of the population is primarily affected by the climate, whereas the change in 343 global exposure of GDP is mainly attributed to the GDP effect (Fig. 4). 344 The climate effect accounts for nearly half of the total change in population exposure 345 (47%–53%) in each scenario and for each period. The effect of population change for 346 2016–2035 is larger than the interactive effect, which reverses for 2046–2065. The GDP 347 effect is responsible for nearly 50% of the total change in GDP exposure for 2016–2035, 348 with approximately 30% and 20% attributable to the effects of climate and interactions, 349 respectively. The change in GDP exposure for 2046–2065 under the RCP4.5-SSP2 and 350 RCP8.5-SSP3 scenarios is dominated by the interactive effect, at 49% and 53%, 351 respectively, followed by the effects of GDP and climate. The increase in the influence 352 of the interactive effect highlights the importance of the interactions between GDP and 353 climate change in increasing GDP exposure under the scenarios of high GHG emissions. 354 355 Fig. 4: Categorization of projected aggregate global changes in (a) population and (b) 356 GDP exposure to extreme heat under the RCP2.6-SSP1, RCP4.5-SSP2, and RCP8.5- 357 SSP3 scenarios. Error bars are the standard deviations for the results of the five GCMs. 358 PPP is purchasing power parity in USD. 359 The relative importance of the factors of change in population exposure at the 360 continental level varies between regions (Fig. 5). As noted above, the percent of 361 population exposure worldwide is highest for Asia and Africa, but the dominant 362 contribution to the change differs between the two continents. The effect of climate is 363 the dominant contribution for Asia, Europe, North America, and South America for all 364 scenarios and periods. This result indicates that an increased frequency of heat extremes 365 amplifies population exposure, even in the absence of population increases on these 366 continents. In contrast, the contribution of the three factors (i.e., the effects of 367 population, climate, and the interaction) for Africa is nearly identical for 2016–2035 368 because of the strong population increase projected for this continent. The interactive 369 effect, however, becomes the primary contribution for 2046–2065, particularly under 370 the RCP8.5-SSP3 scenario, accounting for 53% of the total change, which is more than 371 the sum of the population and interactive effects. The contributions of the GDP and 372 interactive effects for Oceania are nearly the same for 2016–2035 under the three 373 scenarios. In contrast, the interactive effect makes the primary contribution to the 374 change in GDP exposure for 2046–2065 under the RCP4.5-SSP2 and RCP8.5-SSP3 11 375 scenarios. 376 377 Fig. 5: Categorization of projected aggregate continental change in population exposure 378 to extreme heat under the RCP2.6-SSP1, RCP4.5-SSP2, and RCP8.5-SSP3 scenarios. 379 Error bars are the standard deviations for the results of the five GCMs: (a) Africa, (b) 380 Asia, (c) Europe, (d) North America, (e) Oceania, (f) South America. 381 Between continents, the differences in relative importance to GDP exposure to the 382 effects of GDP, climate, and interactions are clear (Fig. 6), and the contribution of GDP, 383 climatic, and interactive effects under the various scenarios and periods also varies 384 among continents. The largest contribution for Asia and North America is from GDP 385 for 2016–2035 under the three scenarios and for 2046–2065 under the RCP2.6-SSP1 386 scenario, whereas the interactive effect is the dominant contribution for 2046–2065 387 under the RCP4.5-SSP2 and RCP8.5-SSP3 scenarios. The GDP effect for Africa and 388 South America is the primary contributor for 2016–2035, whereas the interactive effect 389 is the primary contributor for 2046–2065. The difference between the interactive and 390 GDP effects also increases for the pathways involving high GHG emissions in these 391 two continents. The interactive effect for the RCP8.5-SSP3 scenario accounts for 64% 392 of the total change for Africa, exceeding the contribution of 35% for the GDP effect. 393 The interactive effect accounts for 58% of the total change for South America, whereas 394 the GDP effect accounts for only 22%. The increase in GHG emissions leading to 395 climate change and more frequent heat extremes can account for the increase in the 396 difference between the interactive and GDP effects. The interaction between climate 397 and change in GDP amplifies the exposure, but the GDP effect decreases, and the 398 interactive effect increases as GDP growth slows under the pathways with high GHG 399 emissions, which leads to the large difference between the effects under the RCP8.5- 400 SSP3 scenario. The GDP effect is the primary contributor for Europe and Oceania for 12 401 all periods and scenarios, except for 2046–2065 under the RCP8.5-SSP3 scenario, 402 which implies that GDP exposure on these two continents would increase quickly 403 without the impact of climate change. For almost all continents, GDP exposure is 404 dominated by the GDP effect under the RCP2.6-SSP1 scenario because GDP grows 405 fastest under this scenario. 406 407 Fig. 6: Categorization of projected aggregate continental change in GDP exposure to 408 extreme heat under the RCP2.6-SSP1, RCP4.5-SSP2, and RCP8.5-SSP3 scenarios. 409 Error bars are the standard deviations for the results of the five GCMs. PPP is 410 purchasing power parity in USD: (a) Africa, (b) Asia, (c) Europe, (d) North America, 411 (e), Oceania, (f) South America. 412 3.4 Analysis of cumulative probability of changes in population and GDP 413 exposures 414 Figure 7 shows the cumulative distribution functions for changes in population and 415 GDP exposures under the various scenarios and time periods relative to the base period. 416 The changes are smaller for 2016–2035 than 2046–2065 for both population and GDP 417 exposure, which indicates that population and GDP exposures increase rapidly over 418 time, independent of the scenario. Population and GDP exposures, however, differ 419 between the three scenarios and the two time periods. The cumulative probability of an 420 increase in population exposure is >90% for 2016–2035 and >95% for 2046–2065. The 421 change is largest under RCP8.5-SSP3, with 90% probability of a zero- to 10-fold 422 increase for both periods, followed by RCP4.5-SSP2 and RCP2.6-SSP1, with a 90% 423 probability of a zero- to 7-fold increase relative to the base period. The future increase 424 is much faster for GDP exposure than for population exposure because of the faster 425 growth in GDP. The change in GDP exposure for 2046–2065 under the RCP2.6-SSP1 13 426 scenario is the largest among all time periods and scenarios, with 90% probability of a 427 difference of −1- to 43-fold, with a 10% probability for an increase >43-fold relative to 428 the base period. The increase for 2046–2065 is −1- to 35-fold under RCP8.5-SSP3 and 429 −1- to 31-fold under RCP4.5-SSP2 for the same probability. The changes are much 430 smaller for 2016–2035 than for 2046–2065, with >70% probability of a zero- to 7-fold 431 increase for the three scenarios. The change for 2046–2065 is largest under RCP2.6- 432 SSP1, with nearly a 15% probability of a >15-fold increase relative to the base period, 433 whereas the increases are 7- and 11-fold under the RCP4.5-SSP2 and RCP8.5-SSP3 434 scenarios, respectively, for the same probability. 435 436 Fig. 7: Cumulative probability of projected change in (a) population exposure and (b) 437 GDP exposure to extreme heat relative to the base period. Error bars are the standard 438 deviations for the results of the five GCMs. 439 4 Discussion 440 4.1 Impact of a warmer climate on socioeconomic exposure to extreme heat 441 Our study shows spatiotemporal variation and changes in exposure and its 442 components at global and continental levels. The results indicate variations among the 443 continents (Figs. S4 and S5). For example, population exposure is highest in Asia (Fig. 444 S4), accounting for >50% of the global exposure, followed by Africa, North America, 445 Europe, South America, and Oceania. The percent of population exposure for Africa 446 and South America is projected to increase over time and with an increase in GHG 447 emissions, whereas the percentage of population exposure for Asia, Europe, and North 448 America is projected to decrease over time. GDP exposure accounted for >50% of the 449 total exposure for Asia (Fig. S5), with likely increases under the future scenarios, 450 followed by North America, Europe, Africa, South America, and Oceania. The percent 451 of exposure for developing countries, such as in Africa and South America, increases 452 for the future scenarios. In contrast, decreases would be likely for developed countries, 453 such as those in North America, Europe, and Oceania. Extreme heat under climate 454 change would thus affect Asia the most, where population and GDP exposures are 455 highest. Exposure for Africa is expected to increase rapidly in the future. Therefore, 456 more attention should be directed to Asia and Africa for deeper research on extreme 457 heat exposure and risk assessment. In addition, the design and implementation of 458 effective adaptive measures are urgently needed in regions with high socioeconomic 14 459 exposure, to lessen populations suffering from heat extremes and reduce the economic 460 losses under climate change. 461 We estimate future socioeconomic exposure based on SSP projections with to various 462 RCP scenarios. Because the current data are well understood and widely accepted, 463 however, some previous research is based on current spatial distributions and 464 population and GDP exposures (Hsiang et al., 2017). We also calculate exposure to 465 future extreme heat regardless of changes in population and GDP (Table S3). The 466 results indicate that socioeconomic exposure would increase significantly under climate 467 change, with distinct differences among the RCP scenarios, even if population and GDP 468 exposures are maintained at current levels. Exposure is lowest under the RCP2.6 469 scenario and highest under the RCP8.5 scenario. Population and GDP exposures are 470 1.72- and 1.67-fold higher under RCP8.5 than under RCP2.6, respectively. Both results 471 for the SSP projections and the current fixed values highlight the importance of 472 reducing emissions. Overall, it is potential to mitigate the impact of climate change on 473 both extreme heat hazards and population exposures under the RCP2.6 scenario 474 compared to RCP4.5 and 8.5 scenarios. And the efficiency of reducing emissions 475 should be further quantified in future studies under RCP2.6 or other ambitious warming 476 target, such as 1.5°C and 2°C warming target proposed in Paris Agreement (UNFCCC, 477 2015), which will be helpful for making climate change mitigation strategies. 478 4.2 Uncertainty, limitation and further research 479 Heat extremes could be measured by using different thresholds (e.g., 90, 95, 97.5, 480 and 99th percentile of daily maximum temperatures over the base period). The spatial 481 patterns of exposure and change are generally similar, whereas the quantity of exposure 482 and the relative importance of factors differ greatly (Liu et al., 2017). Assessing 483 socioeconomic exposure under climate change has many uncertainties, except for the 484 indices that measure extreme heat. The primary sources of uncertainty include scenarios 485 of GHG emission (Maurer, 2007), GCMs (Soden et al., 2018), predictions of population 486 and GDP (Chen et al., 2018) and the calculation method of exposure (Zhang et al., 487 2018). The research of Bonan and Doney (2018) on global change in stresses on 488 terrestrial and marine ecosystems show that the uncertainty on land is mostly from 489 model uncertainty, which contributes nearly 80% of the total uncertainty in the 21st 490 century. We evaluate population and GDP exposures based on five GCM simulations 491 after downscaling and assessing the various scenarios and time periods separately. 492 Figures 4–7 show the differences between the GCMs. In addition to displaying the mean 493 value of five GCMs, the standard deviation of five GCMs is also displayed as error bars 494 in the figures. This study generates assessments by considering different sources of 495 uncertainty, so the results can be considered as reasonable with relatively high accuracy. 496 This study is also subject to some limitations. First, our study quantified 497 spatiotemporal variation of population and GDP exposures to extreme heat as an 498 important first step for estimating changes in risk, which made progress compared to 499 the previous studies ignoring the spatiotemporal variation of the socioeconomic risk 500 (Barnett et al., 2012; Gasparrini and Armstrong, 2011; Voorhees et al., 2011). However, 501 we do not estimate changes in vulnerability due to the lack of more advanced damage 15 502 data due to heat extremes. In further studies, we should also concentrate on assessing 503 vulnerability when the data are available, which is also essential for assessing risk. In 504 addition, adopting measures such as the use of air conditioning and the purchase of 505 insurance should be taken into account. These factors would likely improve resilience 506 to heat extremes and thus decrease vulnerability and limit the impact of climate change. 507 Second, the effect of urban heat islands (UHIs) in addition to global warming is not 508 considered explicitly in the climate change simulations in this study. UHIs are 509 confirmed to exacerbate the extent and intensity of heat extremes in urban areas and 510 increase the risk to urban residents in heat extremes (Dan and Bou-Zeid, 2013; Hajat 511 and Kosatky, 2010; Mishra et al., 2015). Recently, as the importance of urbanization in 512 climate change has come to be realized, several case studies simulating the impact of 513 urbanization on extreme temperature events been done (Chen and Frauenfeld, 2016; 514 Grossman-Clarke et al., 2010; Wang et al., 2012). Some research focused on intensely 515 urbanized places such as eastern China, the eastern USA, and Western Europe, which 516 is consistent with the results of our studies on high-exposure regions. However, some 517 other regions such as western Africa and northern India are also expected to suffer high 518 exposure to heat extremes in the future. More attention should be focused on these 519 regions in future studies (see Fig. 8). However, global studies are rare because of a lack 520 of precise land-use and population data. Some other studies assess the future urban 521 climate change by using GCM (Fischer et al., 2012; Oleson, 2012). However, given the 522 coarse spatial resolution, the influence of GCM on underlying urban surfaces could not 523 be fully appreciated, so the impact of future urban development on UHIs was not 524 considered in these studies. As a result, the effect of UHIs on extreme temperature 525 events may be underestimated. Therefore, in future studies, the design and nesting of 526 urban land surface models and the estimation of UHI, especially the heat island 527 estimation during heat extremes, should be taken into account, since they are critical 528 for estimating exposure and risk to heat extremes in cities due to climate change. 529 530 Fig. 8: Regions with high future exposure under climate change. E.g. Western Europe 531 (region 1), the eastern USA (region 2), eastern China (region 3), western Africa (region 532 4) and northern India (region 5), may face more risk in heat extremes coupled with the 533 effect of urban heat islands in intensely urbanized places. 16 534 5 Conclusions 535 This study generated four key findings. First, the regions with the highest population 536 and GDP exposures to extreme heat are primarily concentrated in densely populated 537 areas, such as India, China, midwestern Europe, the eastern USA, and the coastal areas 538 of South America. Second, global population exposure for the years 2046–2065 is 539 highest under the RCP8.5-SSP3 scenario, with exposure increasing 3.76-fold relative 540 to the base period 1986−2005. The RCP2.6-SSP1 scenario produces the highest global 541 GDP exposure, with exposure increasing 10.47-fold for 2046–2065. Third, exposure is 542 highest for Asia for both population and GDP, which exceeds 50% of the global 543 exposure. The increase in exposure is largest for Africa, with the annual population and 544 GDP exposures 9.20- and 29.34-fold higher, respectively, than during the base period. 545 In contrast, the relative changes in population and GDP exposures are lowest for Europe 546 and North America, respectively. Fourth, the effect of climate is the dominant 547 contribution globally to change in population exposure, accounting for nearly half of 548 the total change (47%–53%). The effect of GDP is responsible for nearly 50% of the 549 total change in GDP exposure for 2016–2035, whereas the interactive effect makes the 550 primary contribution for 2046–2065 under the RCP4.5-SSP2 and RCP8.5-SSP3 551 scenarios, accounting for 49% and 53% of the total change, respectively. In conclusion, 552 mitigating emissions of greenhouse gases, either at the level of the RCP2.6 scenario or 553 at a more ambitious target of reduction, is important for reducing socioeconomic 554 exposure to heat extremes. In addition, designing and implementing effective measures 555 of adaptation are urgently needed in Asia and Africa to aid socioeconomic systems 556 suffering from heat extremes due to climate change. 557 Acknowledgments 558 This study was supported by the National Key Research and Development Program of 559 China (Grant No. 2016YFA0602402); the Youth Innovation Promotion Association, 560 CAS (Grant No. 2016049); the Program for “Kezhen” Excellent Talents in Institute of 561 Geographic Sciences and Natural Resources Research (IGSNRR), CAS, (Grant No. 562 2017RC101); the Key Research Program of Frontier Sciences, CAS (Grant No. 563 QYZDB-SSW-DQC005); and the ERC-SyG-2013-610028 IMBALANCE-P. We also 564 thank ISI-MIP and NIES for data support. 565 References 566 Barnett, A. G., Tong, S. & Clements, A. C. A., 2012. What measure of temperature is 567 the best predictor of mortality? Environ. Res. 118, 149–151. 568 https://doi.org/10.1016/j.envres.2010.05.006 569 Bonan, G. B., & Doney, S. C., 2018. Climate, ecosystems, and planetary futures: The 570 challenge to predict life in Earth system models. Science, 359(6375), eaam8328. 571 doi:10.1126/science.aam8328 572 Bouwer, L. M., 2013. Projections of future extreme weather losses under changes in 17 573 climate and exposure. Risk Anal. 33, 915–930. https://doi.org/10.1111/j.1539- 574 6924.2012.01880.x 575 Bowles, D. C., Butler, C. D. & Friel, S., 2014. Climate change and health in Earth's 576 future. Earths Future 2, 60–67. https://doi.org/10.1002/2013EF000177 577 Burke, M., Hsiang, S. M. & Miguel, E., 2015. Global non-linear effect of temperature 578 on economic production. Nature 527, 235–239. 579 https://doi.org/10.1038/nature15725 580 Carleton, T. A. & Hsiang, S. M., 2016. Social and economic impacts of climate. Science 581 353, aad9837–aad9837. 10.1126/science.aad9837. 582 https://doi.org/10.1002/2014GL061859 583 Ceola, S., Laio, F. & Montanari, A., 2015. Satellite nighttime lights reveal increasing 584 human exposure to floods worldwide. Geophys. Res. Lett. 41, 7184–7190. 585 https://doi.org/10.1002/2014GL061859 586 Chen, J., Liu, Y., Pan, T., Liu, Y., Sun, F., & Ge, Q., 2018. Population exposure to 587 droughts in China under the 1.5°C global warming target. Earth Syst Dynam, 588 9(3): 1097-1106. https://doi.org/10.5194/esd-9-1097-2018. 589 Chen, L. & Frauenfeld, O. W., 2016. Impacts of urbanization on future climate in China. 590 Clim dynam 47, 345–357. https://doi.org/10.1007/s00382-015-2840-6 591 Cook, B. I., Smerdon, J. E., Seager, R. & Coats, S., 2014. Global warming and 21st 592 century drying. Clim. Dyn. 43, 2607–2627. https://doi.org/10.1007/s00382- 593 014-2075-y 594 Dan, L. & Bou-Zeid, E., 2013. Synergistic Interactions between Urban Heat Islands and 595 Heat Waves: the Impact in Cities is Larger than the Sum of its Parts. J Appl 596 Meteorol Clim 52, 2051–2064. https://doi.org/10.1175/JAMC-D-13-02.1 597 Field, C. B., Barros, V. & Stocker, T. F., 2012. Managing the risks of extreme events 598 and disasters to advance climate change adaptation. Special report of the 599 Intergovernmental Panel on Climate Change (IPCC). Cambridge University 600 Press. 601 Fischer, E. M. & Knutti, R., 2015. Anthropogenic contribution to global occurrence of 602 heavy-precipitation and high-temperature extremes. Nat. Clim. Change 5, 560– 603 564. https://doi.org/10.1038/nclimate2617 604 Fischer, E. M., Oleson, K. W. & Lawrence, D. M., 2012. Contrasting urban and rural 605 heat stress responses to climate change. Geophys res lett 39, L03705. 606 https://doi.org/10.1029/2011GL050576 607 Forzieri, G., Cescatti, A., Silva, F. B. E. & Feyen, L., 2017. Increasing risk over time 608 of weather-related hazards to the European population: a data-driven prognostic 609 study. Lancet Planet Health 1, e200–e208. https://doi.org/10.1016/S2542- 610 5196(17)30082-7 611 Garssen, J., Harmsen, C. & De, B. J., 2005. The effect of the summer 2003 heat wave 612 on mortality in the Netherlands. Euro. Surveill. 10, 165–168. 613 https://doi.org/10.2807/esm.10.07.00557-en 614 Gasparrini, A. & Armstrong, B., 2011. The impact of heat waves on mortality. 615 Epidemiology 22, 68–73. https://doi.org/10.1097/ede.0b013e3181fdcd99 616 Gasparrini, A. et al., 2015. Mortality risk attributable to high and low ambient 18 617 temperature: a multicountry observational study. Lancet 386, 369–375. 618 https://doi.org/10.1016/S0140-6736(14)62114-0 619 Grossman-Clarke, S., Zehnder, J. A., Loridan, T. & Grimmond, C. S. B., 2010. 620 Contribution of land use changes to near-surface air temperatures during recent 621 summer extreme heat events in the Phoenix metropolitan area. J Appl Meteoro 622 Clim 49, 1649–1664. https://doi.org/10.1175/2010jamc2362.1 623 Hajat, S. & Kosatky, T., 2010. Heat-related mortality: a review and exploration of 624 heterogeneity. J Epidemiol Commun H 64, 753–760. 625 https://doi.org/10.1136/jech.2009.087999 626 Harrington, L. J. & Otto, F. E. L., 2018. Changing population dynamics and uneven 627 temperature emergence combine to exacerbate regional exposure to heat 628 extremes under 1.5 °C and 2 °C of warming. Environ. Res. Lett. 13, 034011. 629 https://doi.org/10.1088/1748-9326/aaaa99 630 Hempel, S., Frieler, K., Warszawski, L., Schewe, J. & Piontek, F., 2013. A trend- 631 preserving bias correction - the ISI-MIP approach. Earth Syst Dynam 4, 219– 632 236. https://doi.org/10.5194/esd-4-219-2013 633 Hirabayashi, Y. et al., 2013. Global flood risk under climate change. Nat. Clim. Change 634 3, 816–821. https://doi.org/10.1038/nclimate1911 635 Hsiang, S. et al., 2017. Estimating economic damage from climate change in the United 636 States. Science 356, 1362–1369. https://doi.org/10.1126/science.aal4369 637 Huang, J. et al., 2017. Analysis of future drought characteristics in China using the 638 regional climate model CCLM. Clim. Dyn., 50, 507–525. 639 https://doi.org/10.1007/s00382-017-3623-z 640 IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of 641 Working Group I to the Fifth Assessment Report of the Intergovernmental Panel 642 on Climate Change. Cambridge, UK, New York, USA: Cambridge University 643 Press. 644 IPCC, 2014. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: 645 Global and 248 Sectoral Aspects. Contribution of Working Group II to the Fifth 646 Assessment Report of the 249 Intergovernmental Panel on Climate Change. 647 Cambridge, UK, New York, USA: Cambridge University Press. 648 Jones, B. & O’Neill, B. C., 2016. Spatially explicit global population scenarios 649 consistent with the Shared Socioeconomic Pathways. Environ. Res. Lett. 11, 650 084003 (2016). https://doi.org/10.1088/1748-9326/11/8/084003 651 Jones, B., O’Neill, B.C., Mcdaniel, L., Mcginnis, S., Mearns, L.O. & Tebaldi, C., 2015. 652 Future population exposure to US heat extremes. Nat. Clim. Change 5, 592– 653 597. https://doi.org/10.1038/nclimate2631 654 Kharin, V. V., Zwiers, F. W., Zhang, X. & Wehner, M., 2013. Changes in temperature 655 and precipitation extremes in the CMIP5 ensemble. Clim. Change 119, 345– 656 357. https://doi.org/10.1007/s10584-013-0705-8 657 King, A. D., Karoly, D. J. & Henley, B. J., 2017. Australian climate extremes at 1.5°C 658 and 2° C of global warming. Nat. Clim. Change 7, 412–416. 659 https://doi.org/10.1038/nclimate3296 660 Liu, Z., Anderson, B., Yan, K., Dong, W., Liao, H. & Shi. P., 2017. Global and regional 19 661 changes in exposure to extreme heat and the relative contributions of climate 662 and population change. Scientific Reports 7, 43909. 663 https://doi.org/10.1038/srep43909 664 Maurer, E. P., 2007. Uncertainty in hydrologic impacts of climate change in the Sierra 665 Nevada, California, under two emissions scenarios. Clim. Change 82, 309–325. 666 https://doi.org/10.1007/s10584-006-9180-9 667 Mishra, V., Ganguly, A. R., Nijssen, B. & Lettenmaier, D. P., 2015. Changes in observed 668 climate extremes in global urban areas. Environ Res Lett 10, 024005. 669 https://doi.org/10.1088/1748-9326/10/2/024005 670 Mishra, V., Mukherjee, S., Kumar, R. & Stone, D. A., 2017. Heat wave exposure in 671 India in current, 1.5 °C, and 2.0 °C worlds. Environ. Res. Lett. 12, 124012. 672 https://doi.org/10.1088/1748-9326/aa9388 673 Mora, C. et al., 2017. Global risk of deadly heat. Nat. Clim. Change 7, 501–506. 674 https://doi.org/10.1038/nclimate3322 675 Murakami, D. & Yamagata, Y., 2016. Estimation of gridded population and GDP 676 scenarios with spatially explicit statistical downscaling, available at: 677 https://arxiv.org/abs/1610.09041. 678 Nath, R. et al., 2017. CMIP5 multimodel projections of extreme weather events in the 679 humid subtropical Gangetic Plain region of India. Earths Future 5, 224–239. 680 https://doi.org/10.1002/2016EF000482 681 O’Neill, B. C. et al., 2014. A new scenario framework for climate change research: the 682 concept of shared socioeconomic pathways. Clim. Change 122, 401–414. 683 https://doi.org/10.1007/s10584-013-0905-2 684 Oleson, K., 2012. Contrasts between urban and rural climate in CCSM4 CMIP5 climate 685 change scenarios. J Clim 25, 1390–1412. https://doi.org/10.1175/JCLI-D-11- 686 00098.1 687 Riahi K et al., 2017. The shared socioeconomic pathways and their energy, land use, 688 and greenhouse gas emissions implications: an overview. Glob Environ Change 689 42:153-168. https://doi.org/10.1016/j.gloenvcha.2016.05.009 690 Robine, J. M., Cheung, S. L. K., Roy, S. L., Oyen, H.V., Griffiths, C., Michel, J. P. & 691 Herrmann, F. R., 2008. Death toll exceeded 70,000 in Europe during the summer 692 of 2003. C. R. Biol. 331, 171–178. https://doi.org/10.1016/j.crvi.2007.12.001 693 Schleussner, C. F., Lissner, T. K., Fischer, E. M., Wohland, J., Perrette, M., Golly, A., 694 Rogelj, J., CHILDERS, K., Schewe, J. & Frieler, K. 2016. Differential climate 695 impacts for policy-relevant limits to global warming: the case of 1.5 °C and 2 °C. 696 Earth Syst Dynam, 6, 2447–2505. https://doi.org/10.5194/esd-7-327-2016 697 Smirnov, O., Zhang, M., Xiao, T., Orbell, J., Lobben, A. & Gordon. J., 2016. The 698 relative importance of climate change and population growth for exposure to 699 future extreme droughts. Clim. Change 138, 1–13. 700 https://doi.org/10.1007/s10584-016-1716-z 701 Soden, B. J., Collins, W. D. & Feldman, D. R., 2018. Reducing uncertainties in climate 702 models. Science 361, 326–327. https://doi.org/10.1126/science.aau1864 703 Sun, H. et al., 2017. Exposure of population to droughts in the Haihe River Basin under 704 global warming of 1.5 and 2.0°C scenarios. Quat. Int 53, 74-84. 20 705 https://doi.org/10.1016/j.quaint.2017.05.005 706 Sun, Y. et al., 2014. Rapid increase in the risk of extreme summer heat in Eastern China. 707 Nat. Clim. Change 4, 1082–1085. https://doi.org/10.1038/nclimate2410 708 Taylor, K. E., Stouffer, R. J. & Meehl, G. A., 2012. An overview of CMIP5 and the 709 experiment design. Bull. Am. Meteorol. Soc. 93, 485–498. 710 https://doi.org/10.1175/BAMS-D-11-00094.1 711 Trenberth, K. E. & Fasullo, J. T., 2012. Climate extremes and climate change: The 712 Russian heat wave and other climate extremes of 2010. J. Geophys. Res. Atmos. 713 117, D17103. https://doi.org/10.1029/2012JD018020 714 UNFCCC Conference of the Parties (COP). 2015. Adoption of the Paris Agreement, 715 Paris, France. 716 Vaneckova, P., Beggs, P. J., de Dear, R. J. & Mccracken, K. W., 2008. Effect of 717 temperature on mortality during the six warmer months in Sydney, Australia, 718 between 1993 and 2004. Environ. Res. 108, 361–369. 719 https://doi.org/10.1016/j.envres.2008.07.015 720 Voorhees, A. S., Fann, N., Fulcher, C., Dolwick, P., Hubbell, B., Bierwagen, B. & 721 Morefield, P., 2011. Climate change-related temperature impacts on warm 722 season heat mortality: A proof-of-concept methodology using BenMAP. 723 Environ. Sci. Technol. 45, 1450–1457. https://doi.org/10.1021/es102820y 724 Vuuren, D. P. V. et al., 2011. The representative concentration pathways: an overview. 725 Clim. Change 109, 5–31. https://doi.org/10.1007/s10584-011-0148-z 726 Wang, J., Feng, J., Yan, Z., Hu, Y. & Jia, G., 2012. Nested high-resolution modeling of 727 the impact of urbanization on regional climate in three vast urban 728 agglomerations in China. J Geophys Res: Atmos 117, D21103. 729 https://doi.org/10.1029/2012JD018226 730 Wang, Y., Wang, A., Zhai, J., Tao, H., Jiang, T., Su, B., … Fischer, T., 2019. Tens of 731 thousands additional deaths annually in cities of China between 1.5 °C and 732 2.0 °C warming. Nature Communications, 10(1), 3376. 733 https://doi.org/10.1038/s41467-019-11283-w 734 Warszawski, L., Frieler, K., Huber, V., Piontek, F., Serdeczny, O. & Schewe, J., 2014. 735 The inter-sectoral impact model intercomparison project (ISI-MIP): project 736 framework. Proc. Natl. Acad. Sci. U.S.A. 111, 3228–3232. 737 https://doi.org/10.1073/pnas.1312330110 738 Zhang, W., Zhou, T., Zou, L., Zhang, L., & Chen, X., 2018. Reduced exposure to 739 extreme precipitation from 0.5 °C less warming in global land monsoon regions. 740 Nature Communications, 9(1), 3153. https://doi.org/10.1038/s41467-018- 741 05633-3 742 Zhang, Z., Li, N., Xu, H. & Chen, X., 2018. Analysis of the economic ripple effect of 743 the United States on the world due to future climate change. Earths Future 6, 744 828–840. https://doi.org/10.1029/2018EF000839 21