Has Global Warming Made Heat Waves Deadlier in Sweden?

Has global warming made heat waves more lethal in Sweden? That’s the conclusion of a study by Swedish scientist Daniel Oudin Åström and colleagues, published last October in Nature Climate Change (NCC). The researchers examined the association of mortality and extreme temperatures in Stockholm. They found that the number of “heat extremes” increased from 220 in 1900-1920 to 381 in 1980-2009. After adjusting for urban heat-island effects, they conclude that climate change was responsible for 288 out of 689 heat-related deaths in the latter period.

Why mention this now? Last week, NCC published a rebuttal by Chip Knappenberger, Patrick Michaels, and Anthony Watts. The authors also posted commentaries on their respective blogs.

One thing that puzzled me right off the bat is Åström et al.’s definition of “heat extreme”: any two-day period when the temperature exceeds 67.2ºF. To some of us who hail from the Sun Belt, 67°F is still sweater weather.

Knappenberger et al. find two major flaws in the Åström study. First, the Swedish scientists mistakenly assume that all warming not due to urban heat islands must be due to anthropogenic climate change. But Stockholm’s climate is also affected by a natural mode of climate variability called the Atlantic Multidecadal Oscillation. The AMO was primarily in its negative (cold) phase during 1900-1929 and primarily in its positive (warm) phase during 1990-2009. The difference between the two phases “is likely to be responsible for some portion of the increase in extreme-heat events identified by Åström et al. and inappropriately attributed to global climate change,” Watts writes.

Second, and more importantly, the Åström team ignores a relevant finding from another Åström et al. study on extreme temperatures and mortality in Stockholm. The key concept here is “relative risk” — an estimate of how much likelier an individual is to die from exposure to a particular risk factor relative to individuals who are not exposed.

In that study, Åström and colleagues found that the relative risk of dying from extreme heat in Stockholm was about 20% in the beginning of the 20th century. But in the NCC study, they estimate that the relative risk of dying from extreme heat in 1980-2009 was 4.6%. In other words, people in Stockholm today are only about one-fourth as likely to die during heat waves than was the case in the early 20th century.

This should not be surprising. People are not dumb. When climate changes, they adapt. Adaptations include physical acclimatization; behavioral adjustments; improvements in medical care, public health programs, buildings, and infrastructure; and increased use of air conditioning.

In a study of 28 U.S. cities, Knappenberger, Michaels, Robert Davis, and Wendy Novicoff (Davis et al.) found that as urban air temperatures increased, average heat-related mortality declined. On average, annual excess heat-related deaths per million declined from 41.0 in the 1960s and 1970s, to 17.3 in the 1980s, to 10.5 in the 1990s. Moreover, cities with the most frequent hot weather, such as Tampa, Florida and Phoenix, Arizona, had practically zero heat-related mortality, and increases in “apparent temperature” (heat intensified by humidity) did not increase their mortality rates.

On their blog, Michaels and Knappenberger excerpt from a new paper in Environmental Health Perspectives, which finds that heat-mortality risk in U.S. cities has continued to decline since the 1990s. The researchers, Bobb et al. (2014), examined summer temperature data and all-cause mortality in 105 U.S. cities during 1987-2005. They found that the heat-mortality risk of elderly people has declined to levels about the same as people in middle age:

While heat-related mortality risk for the ≥75 age group was greater than for the <65 group at the beginning of the study period, by 2005 they had converged to similar levels.

Bobb et al. point out that their study confirms and updates the research of Davis et al. and others who have found declining sensitivity to heat extremes in U.S. and European populations:

This study provides strong evidence that acute (e.g., same-day) heat-related mortality risk has declined over time in the US, even in more recent years. This evidence complements findings from US studies using earlier data from the 1960s through mid-1990s on community-specific mortality rates (Davis et al. 2003a; Davis et al. 2003b), as well as European studies that found temporal declines in heat-related mortality risk (Carson et al. 2006; Donaldson et al. 2003; Kysely and Plavcova 2011; Schifano et al. 2012), and supports the hypothesis that the population is continually adapting to heat.

In contrast, Åström et al., noting that few Swedes have air conditioning, worry that their fellow citizens won’t or can’t adapt:

There is very low awareness of the negative health effects of heat in Sweden. Future changes in the frequency and intensity of heat waves might be of a magnitude large enough to overwhelm the ability of individuals and communities to adapt. The expected increase in the number of elderly and other potentially vulnerable groups, in absolute numbers and as a proportion of the population, could make the impact of temperature extremes on human health even more severe.

Is it reasonable to predict that, even if Swedes start dropping like flies in the summer heat, they still won’t install air conditioning?

No. Adaptation via the marketplace, even if not through central planning, is obviously working in Sweden, which is why the relative risk of heat mortality today is one-quarter of what it was in the early 20th century.

Based on the Åström team’s data and relative risk estimates, Knappenberger et al. calculate that if adaptation had not occurred, the heat-associated death toll in 1980-2009 would be 2,994 deaths instead of 689. In other words, accepting Åström et al.’s data and relative risk estimates, adaptation reduced heat-related mortality in Stockholm by 2,305 deaths during the 1980-2009 period. The mortality reduction from adaptation is 8 times greater than the mortality increase from the extra heat waves.

Impressive! But I was still troubled by one thing. Even if adaptation makes people better off than they would otherwise be, Stockholm had more extreme heat events in 1980-2009 than in 1900-1920. The Åström study seems to suggest that more people died from extreme heat in the later period than in the earlier period. So has global warming made heat waves more deadly?

Mr. Knappenberger kindly answered my query as follows. The population of Sweden today is larger than it was in 1900-1920. Therefore, more people die from all causes under the baseline or normal daily mortality rate. So the relevant question is: How many people would have died from heat mortality in 1900-1929, given the climatic conditions of the time, had Stockholm had the same population and baseline mortality rate as in 1980-2009?

For the 1980-2009 period, Åström et al. assume a baseline death rate in Stockholm of 40 people per day. For 1900-1929, they estimate a relative risk of dying from heat extremes of 20%.

The arithmetic works out like this. In 1900-1920, if the population and baseline death rate were the same as in 1980-2009, each extreme event would have killed 8 people (40 x 0.20 = 8). There were 220 extreme heat events in 1900-1920. Eight extra deaths times 220 extreme events equals 1,760 heat-related deaths. 1,760 heat-related deaths is 2.5 times larger than the 689 heat-related deaths that occurred in 1980-2009.

Knappenberger sums it up:

So in absolute numbers, given the same size population, more than 1,000 fewer died in heat events during 1980-2009 because the population adapted — even despite an increase in the frequency of heat events. Game. Set. Match.