Contents
General observations across cities
We assessed the changes in adaptation to heat for 3820 cities worldwide from 2000 to 2100. We considered feasible combinations of the long-term ensemble mean (ENSMEAN) of each RCP and the SSPs. The projected MMT and its 2000–2100 change, \(\Delta\)MMT, were contextualised with the change in two heat exposure parameters: (1) the frequency of heat exposure (EXD) and (2) the exposure magnitude (MAG). They refer to how many days the MMT is exceeded by the daily temperature in the future and in the past annually and how large the exceedance magnitude is. We used the 1991–2000 and 2090–2099 decadal means of EXD and MAG. We evaluated the parameter changes \(\Delta\)EXD and \(\Delta\)MAG between the past and the future decades (see Supplementary Table S1 for city parameters).
MMT distributions at the end of the century varied in range and median across the RCP/SSP combinations (Fig. 1a). The distributions’ medians increase with an rising forcing level, culminating in SSP5 combinations. Combinations with SSP1 and SSP4 display second and third largest median values. Distributions associated with SSP5 and SSP1 stretch towards high MMT values. More cities reach higher adaptation in 2100 than in other combinations. The contrary applies for SSP3. SSP5 combinations, especially with RCP8.5, cover an enormous \(\Delta\)MMT range, while in SSP1 it is smaller (Fig. 1b). Although most distributions show low EXD, the tail towards higher EXD enlarges with increasing forcing, as obvious in RCP8.5/SSP5, and in RCP6.0 combinations (Fig. 1c). In RCP2.6 few cities show a positive \(\Delta\)EXD towards additional EXD in 2090–2099 but many display a negative \(\Delta\)EXD, a reduction in EXD (Fig. 1d). A higher forcing relates to additional EXD in more cities. RCP8.5/SSP5 exhibits a tail into larger \(\Delta\)EXD. Simultaneously, the number of cities experiencing reductions in EXD remains high across RCPs, notably in SSP5 and SSP1 combinations. A pattern similar to EXD is obvious in MAG (Fig. 1e). Distributions involving SSP3 have largest tails towards high MAG. The median is lowest in SSP1 and SSP5 combinations. Reductions in \(\Delta\)MAG are dominant in SSP1 and SSP5 combinations (Fig. 1f). Contrary, SSP3 displays highest \(\Delta\)MAG. The distribution of \(\Delta\)MAG RCP8.5/SSP5 can be distinguished by prominent tails into both \(\Delta\)MAG directions and by the highest median. This investigation implies, changes in adaptation and exposure across the city sample are divers, explicitly under RCP8.5/SSP5 conditions.
Figure 1
Distributions of studied adaptation and exposure parameters for the RCP/SSP combinations. Absolute MMT for 2100 (a), 2090–2099 exceedance days EXD (c), 2090–2099 exceedance magnitude MAG (e). Parameter changes are \(\Delta\)MMT 2000–2100 (b), and exposure from 1991–2000 to 2090–2099: \(\Delta\)EXD (d), \(\Delta\)MAG (f). Vertical lines indicate the distributions’ medians. Distributions (c–f) were trimmed.
Full size image
How RCPs and SSP influence adaptation and exposure
An investigation of mean adaptation and exposure across the cities supports the previous findings. Figure 2 summarises the effects on MMT and EXD and their changes in the cities across all feasible RCP/SSP combinations according to the Scenario Matrix26,27. For the city sample, the highest change in mean adaptation across all scenario combinations is a mean \(\Delta\)MMT of 7.9 °C, which is reached by RCP8.5/SSP5, the highest forcing level and the most rapid unsustainable economic growth. This also yields the highest absolute mean MMT for the city sample, 34.4 °C. This combination shows a small \(\Delta\)EXD of − 7.6 until the decade 2090–2099 and thus deviates from SSP5 combinations paired with lower forcing (Fig. 2). The lowest mean \(\Delta\)MMT (3.7 °C) and the lowest absolute mean MMT (30.2 °C) are displayed for RCPs 4.5 with SSP3. SSP3 is characterised by a low future socio-economic level. Combinations with RCPs 4.5 and 6.0 yield the smallest mean reductions in EXD (\(\Delta\)EXDs of − 5.2 and − 1.4) across the city sample until the future. A small mean \(\Delta\)MMT of 4.3 (and a moderate future mean MMT of 30.8 °C) across the cities is produced by RCP2.6/SSP1. This scenario combination relying on sustainable growth exhibits the second highest increment in socio-economic development until 2100. It yields a relatively large mean reduction of − 16.6 \(\Delta\)EXD until the future decade and equals \(\Delta\)EXD in RCP6.0/SSP5.
We observe that with an increasing forcing, except for SSP5 combinations, the mean \(\Delta\)MMT, but also the mean \(\Delta\)EXD become larger across the city sample. The mean \(\Delta\)MAG behaves likewise. It ranges between − 0.2 \(^{\circ }\)C in RCP2.6/SSP1 and +0.2 \(^{\circ }\)C in RCP8.5/SSP5 (Supplementary Fig. S1). Even though increasing forcing levels achieve higher adaptation and larger adaptation rates until 2100, they amplify heat exposure frequency and magnitude.
SSP5 combinations, excluding that with RCP8.5, show high socio-economic levels across the cities and perform well concerning exposure reductions and high future MMTs. Hence, a large projected GDP/capita is an advantageous precondition for high adaptation. For our city sample, SSP5 generates the highest future mean country-based GDP/capita (Int$ 101 205) compared to the beginning of the century (Int$ 9 849). The 2100 GDP/capita related to SSP1 is less, but still comparatively high (Int$ 63 346). This suggests a high GDP/capita does not automatically lead to a better bearable situation regarding future heat and its related mortality for urban populations. In the following, we contrast two scenario combinations that generate the most optimistic socio-economies and thus enable highest adaptation gains (SSP5 and SSP1) but that are contrary in future exposure (RCP8.5/SSP5 and RCP2.6/SSP1).
Figure 2
Systematic overview of the change in adaptation and exposure for the city sample according to each RCP/SSP combination and the future socio-economic level per SSP. Lower panel: The 1991–2000 to 2090–2099 \(\Delta\)EXD (orange boxes) in context of \(\Delta\)MMT (in-graph text annotations) for all possible RCP/SSP combinations. RCP2.6/SSP5 seems implausible26. Parameters of all scenario combinations are presented in Supplementary Table S2. Upper panel: Unique country-based GDP/capita per SSP in 2100 (mean from IIASA and OECD data). Green line in upper panel denotes the country-based GDP/capita as of 2000 [in 2011 int.$].
Full size image
Drivers of changes in adaptation 2000–2100
An analysis of adaptation change for the scenario combinations RCP2.6/SSP1 and RCP8.5/SSP5 unveils regionally distinct change pattern of \(\Delta\)MMT (Figs. 3a, 4a). Cities in the northern hemisphere, especially in western Europe, show a larger \(\Delta\)MMT than subtropical and tropical cities or cities in the southern hemisphere, except Oceania. Evidence from the top and the bottom of the \(\Delta\)MMT distributions confirm these findings: Under RCP2.6/SSP1, highest \(\Delta\)MMTs are 7.2 \(^{\circ }\)C in Ostrava and Brno (Czech Republic), 7.1 \(^{\circ }\)C in Olomouc (Czech Republic), 7.0 \(^{\circ }\)C in Prague and Plzen (Czech Republic) and 6.9 \(^{\circ }\)C in Râmnicu Vâlcea (Romania). Lowest \(\Delta\)MMTs range between 1 \(^{\circ }\)C (Nouakchott, Mauritania) and 1.5 \(^{\circ }\)C (Umm Durman, Sudan) covering further cities in Sudan and Chad. In a RCP8.5/SSP5 future, the highest adaptation gain is a \(\Delta\)MMT of 13 \(^{\circ }\)C in Râmnicu Vâlcea (Romania), followed by Ostrava (Czech Republic, 12.9 \(^{\circ }\)C ), Piatra Neamt and Olomouc (Romania and Czech Republic, 12.8 \(^{\circ }\)C ), and Kislovodsk and Uhta (Russia, 12.7 \(^{\circ }\)C ). The smallest gains in MMT range between 1.8 \(^{\circ }\)C (Nouakchott, Mauritania) and 2.5 \(^{\circ }\)C (Umm Durman, Sudan) and cover further cities in Sudan and Chad.
To shed light on the principal drivers behind \(\Delta\)MMT across cities around the globe, we calculated the changes in each variable (climate variables and GDP/capita) until 2100 and their respective isolated effects on \(\Delta\)MMT. The greatest weighted variable change was identified for each city and compared to the aggregated weight of the remaining variables’ changes. The resulting primary contributors to \(\Delta\)MMT, either a single variable’s influence or the sum of two, are illustrated in Fig. 3b for RCP2.6/SSP1 and Fig. 4b for RCP8.5/SSP5. In the former case, large changes in adaptation are solely driven by high gains in GDP/capita cities in Western Europe, North America, East Asia, Oceania and coastal South America (Fig. 3b). The change in climate as a primary driver leads to moderate increments in \(\Delta\)MMT in Eastern European cities. Depending on the increment in variable change, this can also result in low \(\Delta\)MMT, as in cities in Northern Africa, the Middle East and coastal Nigeria. This concerns cities in the Sahel, the Arabian Peninsula, and Pakistan and India, where climate-driven \(\Delta\)MMTs range between 0 and 2 \(^{\circ }\)C . Thus, the adaptation gain is lowest and slowest until 2100.
RCP8.5/SSP5 portrays higher increments in \(\Delta\)MMT than RCP2.6/SSP1, while the regional distribution of change pattern roughly remains similar (Figs. 3a, 4a). However, completely different variable effects dominate \(\Delta\)MMT in RCP8.5/SSP5 (Fig. 4b). The extensive effect of GDP/capita on high \(\Delta\)MMTs is either masked by an even larger climate influence or conjoined by the climate effect. Few cities in northwestern Europe, in China and in the southernmost latitudes remain whose large \(\Delta\)MMT is uniquely defined by socio-economic gains until 2100. In RCP8.5/SSP5 the changes in climate variables until 2100 act as primary contributors to \(\Delta\)MMT. Highest \(\Delta\)MMTs across Europe are driven by the 30-year mean temperature in the east and additionally by the 30-year mean amplitude in southern Europe. In RCP8.5/SSP5, the 30-year mean temperature dominates the \(\Delta\)MMT in a larger share of cities. Across African cities, Tmean30 is associated with rather low \(\Delta\)MMTs.
Especially in a fossil-fuel-based future with rapid socio-economic growth as in RCP8.5/SSP5, the adaptation to heat will largely be driven by a strong physiological acclimatisation until 2100 and outweigh the already strong effect of economic growth. In a sustainable prosperous future, the gain in wealth until 2100 is the primary contributor to achieve heat adaptation.
Figure 3
Changes in adaptation until 2100 and their primary contributors to \(\Delta\)MMT in a RCP2.6/SSP1 future. \(\Delta\)MMT 2000–2100 for RCP2.6/SSP1 for major world cities (a). Contributions of the single variables’ 2000–2100 changes to \(\Delta\)MMT 2000–2100 for RCP2.6/SSP1 (b). (Map created in R (version 3.6.2)28 using the tmap package (version 2.3-2)29).
Full size image
Figure 4
Changes in adaptation until 2100 and their primary contributors to \(\Delta\)MMT in a RCP8.5/SSP5 future. \(\Delta\)MMT 2000–2100 for RCP8.5/SSP5 for major world cities (a). Contributions of the single variables’ 2000–2100 changes to \(\Delta\)MMT 2000–2100 for RCP8.5/SSP5 (b). (Map created in R (version 3.6.2)28 using the tmap package (version 2.3-2)29).
Full size image
Changes in exposure
Comparing the two highlighted scenario combinations regarding changes in exposure parameters until 2090–2099 across the cities reveals distinguished characteristics for \(\Delta\)EXD and \(\Delta\)MAG. In RCP2.6/SSP1, EXD reductions until 2090–2099 are largest with − 140 EXD in the Chilean city Antofagasta, in Mossoró (Brazil, \(\Delta\)EXD − 128), Coquimbo and La Serena (Chile) with \(\Delta\)EXD of − 98 EXD. The maximum increase in EXD is projected for Jhang Maghiana (Pakistan, \(\Delta\)EXD + 40). Further Pakistani cities follow (Gojra \(\Delta\)EXD + 38, Faisalabad and Jaranwala, Sahiwal, and Okara \(\Delta\)EXD + 3 7). RCP8.5/SSP5 yields more extreme EXD changes than the sustainable scenario. Still, its mean \(\Delta\)EXD across the cities remain twice as high (Figs. 1d, 2). The largest \(\Delta\)EXDs reductions are − 185 EXD in Antofagasta (Chile), − 149 EXD in Coquimbo and La Serena (Chile). Mossoró (Brasil), Copiapó (Chile), Downey (USA) follow with − 116, − 109 and − 104 EXD. The maximum EXD increase for this scenario is a \(\Delta\)EXD of + 92 EXD in Ciudad Obregón (Mexico), which exceeds the maximum increase in RCP2.6/SSP1 by factor 2.3. Further Pakistani cities rank high in \(\Delta\)EXD: Faisalabad and Jaranwala (\(\Delta\)EXD + 86), Sahiwal, Okara and Bahawalnagar (\(\Delta\)EXD + 82).
The global perspective on \(\Delta\)EXD shows prominent differences between the two highlighted scenario combinations in southern European, North American, Subsaharan, Indian, and East Asian cities, (Fig. 5). Here, in RCP2.6/SSP1, the negative \(\Delta\)EXDs cause reduced future EXD (− 50 to 0 EXD) (Fig. 5a). In RCP8.5/SSP5, the positive \(\Delta\)EXDs will yield additional EXDs (0–50 EXD) in those cities until 2090–2099 (Fig. 5b). Large scenario disparities are obvious in cities in northwestern Mexico, Brazil, Pakistan, and India. In a fossil-fuel dependent future, these cities will have to cope with 50 to 92 additional EXD until 2090–2099. In RCP2.6/SSP1 \(\Delta\)EXD will be less in and even bring EXD reductions in some of these cities (− 50 to 0 EXD).
In terms of \(\Delta\)MAG in the sustainable scenario, the largest MAG decreases concern Copiapó (Chile) with a \(\Delta\)MAG of − 3.5 \(^{\circ }\)C , Dunhuang (China, \(\Delta\)MAG − 3.2 \(^{\circ }\)C ), Coquimbo and La Serena (Chile, \(\Delta\)MAG − 3 \(^{\circ }\)C ), Geermu (China, \(\Delta\)MAG − 2.9 \(^{\circ }\)C ). The maximum increments in \(\Delta\)MAG are expected in Pakistan (Fig. 6a): Chishtian Mandi (\(\Delta\)MAG + 1.9 \(^{\circ }\)C ), Sahiwal, Okara, Bahawalnagar and Kamalia (\(\Delta\)MAG + 1.8 \(^{\circ }\)C ), and Gojra (\(\Delta\)MAG + 1.7 \(^{\circ }\)C ). Some of these cities overlap with highest ranks in \(\Delta\)EXD. Generally, for RCP2.6/SSP1 we record a \(\Delta\)MAG between − 2 and 0 \(^{\circ }\)C for most cities. Some cities in the subtropics and tropics display a positive \(\Delta\)MAG between 0 and 2 \(^{\circ }\)C until the future decade (Fig. 6a). MAG reductions in RCP8.5/SSP5 are more extreme compared to the sustainable future. The mean of the RCP8.5/SSP5 \(\Delta\)MAG across the cities is still higher and positive (Supplementary Fig. S1). Maximum MAG reductions are projected in Copiapó and Antofagasta (Chile, \(\Delta\)MAG − 4.4 \(^{\circ }\)C and − 3.9 \(^{\circ }\)C , Geermu (China, \(\Delta\)MAG − 3.8 \(^{\circ }\)C ), Coquimbo and La Serena (Chile, \(\Delta\)MAG − 3.5 \(^{\circ }\)C ) and Dunhuang (China, \(\Delta\)MAG − 3.2 \(^{\circ }\)C ). Maximum \(\Delta\)MAG increments are observed in Bechar (Algeria, \(\Delta\)MAG + 5.4 \(^{\circ }\)C ), Karbala, al-Fallujah (Iraq) and Zambol (Iran) (\(\Delta\)MAG +5 \(^{\circ }\)C ), ar-Ramadi and as-Samawah (Iraq, \(\Delta\)MAG + 4.8 \(^{\circ }\)C ). RCP8.5/SSP5 conveys an intensified situation concerning the extremes, while \(\Delta\)MAG is still small in many cities (Fig. 6b). Mainly cities in the Sahel, the Middle East into Pakistan and India, in northern and southern Africa, in the Southwestern USA and in northern Mexico show high increments in \(\Delta\)MAG (+ 2 \(^{\circ }\)C to +6 \(^{\circ }\)C ) until 2090–2099. Some severely exposed cities would profit from MAG reductions in RCP2.6/SSP1 instead (Fig. 6a,b).
Figure 5
Changes in future heat exposure frequency in major cities worldwide until 2100. \(\Delta\)EXD 1991–2000 to 2090–2099 in case of RCP2.6/SSP1 (a) and in case of RCP8.5/SSP5 (b). (Map created in R (version 3.6.2)28 using the tmap package (version 2.3-2)29).
Full size image
Figure 6
Changes in future heat exposure magnitude in major cities worldwide until 2100. \(\Delta\)MAG 1991–2000 to 2090–2099 in case of RCP2.6/SSP1 (a) and in case of RCP8.5/SSP5 (b). (Map created in R (version 3.6.2)28 using the tmap package (version 2.3-2)29).
Full size image
Further evidence sharpens the disparity in future exposure concerning the selected scenario combinations (Table 1). In a RCP8.5/SSP5 future, a total of 2285 (60%) cities in our sample will profit from an EXD reduction and 1968 (51%) from a MAG reduction. Still, 1150 (30%) cities will experience additional EXD and 1459 (38%) a larger MAG. No EXD or MAG changes will concern 385 (10%) and 393 (10%) cities cities. In our sample, 705 (18%) cities will face more than one fourth of the year being EXD and 185 (5%) cities half the year being EXD. For five cities we project almost the entire year to be EXD (> 350 EXD) considering such future. RCP2.6/SSP1 in contrast is associated with less cities experiencing aggravated exposure changes until 2090–2099. Only 224 (6%) cities will face additional EXD and 338 (9%) a higher MAG. The majority of cities in 2090–2099 will profit from less EXD (3207 cities, 84%) and from a lower MAG (3074 cities, 80%). No changes in EXD or MAG will affect 389 (10%) and 408 (11%) cities. These findings imply that a lower forcing yields a larger reducing effect on the heat exposure parameters for more cities due to a milder climate change. This suggests, in an ideal future without climate change but high wealth-driven adaptation, exposure measures would be minimised and a massive reduction in mortality could be expected. Against this ideal future, RCP2.6/SSP1 constitutes a slight impairment leading to a higher mortality. RCP8.5/SSP5 signifies a substantial worsening of prospects because positive wealth effects on adaptation are likely annihilated.
Table 1 Number and share of cities affected by changes in exposure from 1991–2000 to 2090–2099. Exposure outcomes by indicator across the 3820 world cities comparing RCP2.6/SSP1 and RCP8.5/SSP5.
Full size table
Adaptation and heat exposure in context
It is indispensable to view adaptation and exposure jointly. We aggregated the adaptation and exposure parameters across our sample of 3 820 cities and provide their outcomes (P05, P95, Mean, Min and Max) in Supplementary Table S2. By 2100, a higher adaptation can be achieved by RCP8.5/SSP5 compared to RCP2.6/SSP1 because the \(\Delta\)MMT increment is greater. This also requires a faster adaptation rate until 2100. However, a future according to RCP2.6/SSP1 is able to minimise heat exposure for our city sample in contrast to RCP8.5/SSP5. In both scenarios \(\Delta\)EXD and \(\Delta\)MAG values mostly correlate as in cities in the Middle East, Pakistan and in parts of the USA and northern Mexico. High increases in heat exposure also coincide with small \(\Delta\)MMTs in these regions (Fig. 7). Such circumstances are obvious in Asian cities at the tip of the distribution in RCP2.6/SSP1, and in Asian and African cities in RCP8.5/SSP5 (Fig. 7). This suggests these cities might be prone to increases in mortality until 2100.
Figure 7
Regional disparities in \(\Delta\)MMT 2100–2000 and \(\Delta\)EXD 1991–2000 to 2090–2099 for the selected scenario combinations RCP2.6 paired with SSP1 and RCP8.5 paired with SSP5. Midbar indicates the mean of the \(\Delta\)EXD distribution across the sample cities.
Full size image