Estimating future heat-related and cold-related mortality under climate change, demographic and adaptation scenarios in 854 European cities

Abstract

Previous health impact assessments of temperature-related mortality in Europe indicated that the mortality burden attributable to cold is much larger than for heat. Questions remain as to whether climate change can result in a net decrease in temperature-related mortality. In this study, we estimated how climate change could affect future heat-related and cold-related mortality in 854 European urban areas, under several climate, demographic and adaptation scenarios. We showed that, with no adaptation to heat, the increase in heat-related deaths consistently exceeds any decrease in cold-related deaths across all considered scenarios in Europe. Under the lowest mitigation and adaptation scenario (SSP3-7.0), we estimate a net death burden due to climate change increasing by 49.9% and cumulating 2,345,410 (95% confidence interval = 327,603 to 4,775,853) climate change-related deaths between 2015 and 2099. This net effect would remain positive even under high adaptation scenarios, whereby a risk attenuation of 50% is still insufficient to reverse the trend under SSP3-7.0. Regional differences suggest a slight net decrease of death rates in Northern European countries but high vulnerability of the Mediterranean region and Eastern Europe areas. Unless strong mitigation and adaptation measures are implemented, most European cities should experience an increase of their temperature-related mortality burden.

Main

Heat and cold are established health risk factors with a notable impact on mortality across Europe1,2. Estimates generally report that there are roughly ten cold-related deaths for each heat-related death3,4,5,6; some studies suggested that temperature-related mortality in Europe could overall decrease with climate change7,8. However, the balance between heat-related and cold-related mortality varies substantially across regions and over time with climate change. The latter has been associated with an important increase in heat-related deaths in the twenty-first century2,9. Increases in temperature are coupled with growth of urban areas and populations, which enhance exposure to high temperatures10. Given the current balance between heat-related and cold-related mortality burdens, the question of whether a decrease in cold exposure would offset the adverse increase in high heat exposure under climate change remains.

The balance between increased heat-related and decreased cold-related mortality, hereby referred to as the net effect of climate change, can be influenced by many factors. Previous studies provided inconsistent estimations of the net effect, depending on the location and considered scenarios11,12,13,14,15,16,17,18. Indeed, both extreme heat and cold change at different rates with climate change, resulting in narrowing or broadening of the temperature distribution depending on the region19. In addition, the exposure-response function (ERF) for temperature and mortality is complex, being usually U-shaped or J-shaped; it is often steeper on the heat side, although it varies widely between locations4. Previous evidence has been either too limited in scope or at a resolution too coarse to provide a representative impact estimate at the European level, while neglecting large portions of the continent, such as the Nordic and Baltic countries, and the Balkans11,12,13,18.

An additional complexity of projecting the net effect of climate change lies in the adaptive capacity of European populations. Several studies estimated a substantial attenuation of the heat risk on mortality over the last decades, generally linked to the increase in mean temperature or air conditioning penetration20,21,22; however, trends in cold-related mortality risks are less clear22. Despite some attempts at integrating heat adaptation into impact projections, through shifts in the minimum mortality temperature (MMT) or risk attenuation13,15, methodologies differ widely and with little empirical evidence to guide the modeling of adaptation. Adaptation to heat is additionally interlinked with underlying demographic and socioeconomic trends that necessitates integration within the shared socioeconomic pathways (SSP) framework23,24. Population aging results in increased vulnerability to both heat and cold1,25,26, while a general improvement in socioeconomic conditions and health systems under some SSP scenarios could, on the other hand, reduce the overall impacts that heat and cold have on mortality27. Given the complexities exposed above, projecting heat-related and cold-related mortality–and related net effect–under future conditions is inherently difficult because it depends on temperature projections from climate models and complex ERFs derived from epidemiological analysis, in addition to varying pathways in socioeconomic, demographic and vulnerability changes. An appropriate assessment of future temperature-related mortality must isolate the specific impact of climate change in a wide range of societal scenarios, while accurately propagating uncertainty from climate and epidemiological models.

In this study, we aimed to provide a comprehensive assessment of the net effect of climate change on temperature-related mortality across 854 cities spanning the whole European continent for the period 2015–2099 and for several levels of warming above preindustrial levels. We sought to provide insights on the expected evolution of the net effect in Europe, and under which conditions an increase of this net effect can be avoided. We evaluated a range of future demographic, mitigation and adaptation scenarios represented by a matrix of three SSP scenarios and four different heat adaptation scenarios.

Results

Study design

We considered three SSP scenarios based on European downscaling of the global scenarios and their effect on temperature-related adaptation28: (1) a more equitable Europe committed to sustainability and low-consumption lifestyles resulting in substantial action toward both mitigation and adaptation (SSP1-2.6); (2) a Europe maintaining current inequalities with increased privatization and slow progresses toward mitigation and adaptation (SSP2-4.5); and (3) a Europe with growing instability, regional conflicts and inequalities resulting in little to no effort toward mitigation and adaptation (SSP3-7.0). In each SSP scenario, we initially considered a baseline ‘no adaptation’ scenario in which the vulnerability to heat only depended on the local age distribution to provide a picture of the mortality burden of inaction toward adaptation to heat. We then evaluated a range of adaptation scenarios to heat by attenuating the heat-related mortality risk across ages by 10%, 50% and 90%. Attenuating the risk was done by shrinking the local age-specific ERF for temperatures above the MMT toward no association, according to the prespecified level.

This work builds on a published assessment of historical temperature-related mortality in 854 European urban areas with a population above 50,000, spanning a total of around 40% of 30 European countries29. We used the published city-specific ERFs derived for five age groups1,30, integrated them with projected temperature series and age-specific population and death rates for each SSP scenario, and performed comprehensive health impact projections31. In this assessment, we isolated the part specifically attributed to climate change by quantifying the burden as the difference in temperature-related deaths between two subscenarios: (1) ‘full’, in which both temperature and demographic projections are considered; and (2) ‘demographic change only’, in which only the demography changes while the temperature distribution from the period 2000–2014 is kept constant across the century. This allowed us to control for population aging and changes in mortality rates to isolate heat, cold and net effects of climate change directly attributable to the evolution of the temperature distribution and the population adaptation to heat. For each scenario described above, we accounted for climate uncertainty by considering bias-adjusted temperature outputs from 19 general circulation models (GCMs) extracted from the NASA Earth Exchange Global Daily Downscaled Projections database, based on the output from phase 6 of the Coupled Model Intercomparison Project (CMIP6)32. We additionally propagated the uncertainty from the epidemiological analysis by performing projections for 500 Monte Carlo simulations of the ERFs30. The methodology and the assumptions related to the several scenarios are fully detailed in the online methods and illustrated in the extended data.

European-level results

For the three considered SSP scenarios, the no adaptation scenario resulted in an increase in net temperature-related excess death rates, related to climate change only, across the whole 2015–2099 period (Fig. 1). In all cases, the increase in heat-related deaths outweighed the reduction in cold-related deaths, although the magnitude differed across SSP scenarios. For the SSP1-2.6 scenario, the net increase in temperature-related deaths peaked at 7.6 (95% confidence interval (CI) = −14.5 to 25.8) deaths per 100,000 person years in 2060, and decreased slightly afterward. In the SSP2-4.5 scenario, climate change-related death rates plateaued between eight and ten deaths per 100,000 person years from 2070 to the end of the century. In contrast, under the SSP3-7.0 scenario, the net effect substantially increased over the century to reach 45.4 (95% CI = 0.7 to 106.0) deaths per 100,000 person years (Table 1). This represents a 49.9% increase compared to the historical levels of 91 deaths per 100,000 person years1,30. Additionally, while temperature-related deaths almost disappeared for the youngest age groups under the SSP1-2.6 and SSP2-4.5 scenarios, rates consistently increased across all ages under SSP3-7.0 (Extended Data Fig. 1).