Figure 1 displays the projected changes in excess mortality by comparing the 1.5- vs 2-°C increase scenarios in each country, geographic region, and climate zone (actual numbers in Tables S3–S6). Overall, an increase from 1.5 to 2 °C in GMT could generate a substantial rise in heat-related mortality in most countries included in the analysis. Specifically, assuming no changes in population and vulnerability, heat-mortality impacts could increase between + 0.11 and + 2.13%, with most countries in South Europe and South-East Asia showing increments above + 1%. In contrast, cold-related mortality could decrease in all countries between − 0.27 and − 0.98%. These decrements are of a lower magnitude compared to the corresponding heat-related risks, producing a net increase in excess mortality in about half of the countries but with large uncertainty in most of them. For instance, only Italy and Spain, the two countries in the South of Europe included in the study, could face a significant net increase in total excess temperature-related mortality of about + 0.66 and + 0.77%, respectively. Similar or larger but not statistically significant increases, reaching + 1.15%, could be observed in other warm countries in South-East Asia, such as the Philippines, Thailand, and Vietnam, despite having different climate conditions (Fig. 1, Fig. S1, Table 1). In contrast, total excess mortality could decrease in cooler areas such as North Europe and East Asia. For example, Ireland and Japan could register a small decrease between − 0.51 and − 0.27%. Other large countries (Canada, USA, and Chile), and some countries located in central areas such as Czech Republic and Moldova, might not face substantial changes in total excess mortality. However, many of the net change estimates are not statistically significant, reflecting the uncertainty in the estimates.
Figure 2 illustrates the trends in differences in excess mortality by geographic region and climatic zone projected under other warming scenarios, including those beyond the Paris Agreement targets, corresponding to increases in GMT of 2, 3, and 4 °C, relative to 1.5 °C (figures in Tables S3–S6). As expected, we observed a common pattern of rising mortality associated with heat and a moderate attenuation in cold-related impacts. Under more extreme scenarios, most regions could experience considerably larger heat-mortality risks that would not be balanced by the projected decreases in cold-related excess mortality. Specifically, central and southern regions of America, Europe, and East-Asia are projected to experience increases in heat-related mortality impacts ranging between + 3.53 and + 8.86% in the most extreme 4-°C scenario, while cooler regions in Europe and Asia could face smaller increases below + 2%. In the latter regions, the contribution of the reduction in cold-related mortality impacts would be similar to the increase observed for heat, ranging between − 1.88 and − 2.23%, translating into small or even null and non-significant net reductions in total excess mortality. The most affected areas in terms of large increases in net temperature-mortality impacts would be again the warmest regions in the south of Europe and South-East-Asia (+ 4.41 and + 7.07%), and in a smaller magnitude in South America and Central Europe (+ 2.60 and + 2.63%). However, estimates on net changes are highly uncertain for most of the regions and warming levels, showing statistically significant results only for the central and southern European regions. A different pattern is observed for Australia, with negative net changes in excess mortality across the warming scenarios. This would be due to the milder temperature projections obtained for this region, compared to other areas of similar climate, as shown in Fig. S3 and in a previous work (Gasparrini et al. 2017).
The pattern of results suggests an association between projected total temperature impacts and current climate conditions. This is particularly clear in Europe and Asia, where the impact of a change in temperature applied to the current population shows moderate decreases in excess mortality in the colder areas of the north, nearly null changes or small increases in the temperate central areas, and larger increases in warmer southern regions. This pattern is better illustrated in Fig. 3 showing the geographical distribution of the location-specific estimates for the 2- vs 1.5-°C difference. We can also observe a within-country positive gradient from temperate to warmer or equatorial areas in Brazil, although this was not clear in other countries in Central America and USA; the results obtained by different climate zones confirm this gradient (Figs. 1 and 2, bottom panel). Although imprecise, our projections indicate that locations in equatorial climates could experience a + 0.66% increase in total excess mortality from 1.5 to 2 °C warming, followed by a smaller + 0.16% increase in arid zones, while cooler regions (warm temperate and snow climates) could register a decrease or close-to-null change in total excess temperature-related mortality (Fig. 1, Table S6) based on the current population. Similarly, equatorial locations would be the most affected areas by far under more extreme scenarios, experiencing an increment in total mortality impacts above + 6% if GMT increases by 4 °C.
Figure S4 shows the results from a sensitivity analysis in which we compare the GCM-specific and ensemble excess mortality estimated for each geographic area for 1.5-°C warming scenario. We observe little variation across models, thus confirming that impact calculation does not depend on the timing in which each GCM reach a specific warming level, and therefore on the choice of specific climate models or emission pathways.