Warming added 1,700 heat deaths in Zürich. Adaptation took at least 700 back

Heat attribution studies normally produce a single headline number: this many deaths would not have happened without human-caused warming. A study published this month in Climatic Change takes a Swiss city apart over half a century and finds that the number is really two numbers moving in opposite directions [s1].

What the study did

The authors assessed heat mortality attributable to human-induced climate change in Zürich, Switzerland, across 50 years — 1969 to 2018 — with a closer case study of the summer of 2018 [s1]. Their stated aim was methodological: to address limitations in the standard epidemiological approach and apply refined versions of it [s1].

Three refinements matter. The first changes how the counterfactual climate scenario — the temperatures that would have occurred without human influence — is derived [s1]. The second allows the analysis to account for changing vulnerability over time, rather than holding a population's heat sensitivity fixed across five decades [s1]. The third separates impacts occurring during heatwaves from those occurring outside them [s1].

That last point is not a technicality. The paper's framing is that heat-related deaths occur throughout the summer months and peak during heatwaves, rather than being confined to them [s1] — which means an attribution method anchored to named heatwave events will systematically undercount.

The two numbers

Across 1969 to 2018, the study attributes nearly 1,700 heat-related deaths in Zürich to human-induced climate change [s1].

Over the same period, declining vulnerability to heat avoided at least 700 heat-related deaths [s1].

Those two figures are not netted against each other in any simple way — they are separate accountings of a warming signal and an adaptation signal. But their relative size is the finding. In one European city over 50 years, the reduction in how likely people were to die at a given temperature offset a substantial share of what the added temperature itself caused.

The authors do not attribute that declining vulnerability to any single cause, and the abstract does not break it down. Air conditioning, housing stock, health system capacity, heat-health warning systems and the age structure of the population all plausibly contribute, and disentangling them is a different study.

Why this cuts both ways

The optimistic reading is that adaptation works and is measurable. The cautious reading is that a vulnerability decline is not guaranteed to continue, and an attribution method that assumes it will — or that ignores it entirely — will produce projections that are wrong in one direction or the other.

There is also a scope limit. This is one city, with good long-run mortality and meteorological records, in a high-income country with substantial adaptive capacity. The authors present the approach as transferable — they write that it could be applied elsewhere to quantify the effect of climate change on other health outcomes [s1] — but the specific ratio of warming harm to adaptation benefit in Zürich should not be read as a European average, let alone a global one.

A contrasting picture from İstanbul

A separate analysis published this month, of temperature and mortality in İstanbul, illustrates how different the same question looks in a different city [s2]. Using a distributed lag non-linear model with quasi-Poisson regression across lags of up to 21 days, the study examined respiratory, cardiovascular and total non-accidental mortality between 2007 and 2012 [s2].

Its central result is that extreme cold temperatures carried the highest relative risk for cardiovascular mortality, while extreme heat carried the highest relative risks for respiratory and total mortality [s2]. Quantified, extreme hot days — above the 97.5th percentile — accounted for 1.9 excess deaths per 1,000 cardiovascular deaths (95% CI 0 to 7.5), while extreme cold days below the 2.5th percentile accounted for 9.0 (95% CI 3.1 to 21.0) [s2].

The confidence interval on the heat estimate includes zero, so that figure is not statistically distinguishable from no effect in this dataset. The cold estimate is more precise and considerably larger. In a Mediterranean city that the authors describe as facing rising temperatures and rising urban heat stress [s2], cold still dominated cardiovascular mortality across the study window.

What to take from the pair

Neither study argues that heat is unimportant. What they jointly show is that "heat deaths" is a compound quantity — temperature, vulnerability, cause of death, and which end of the temperature distribution you look at all move the answer, and the İstanbul window (2007–2012) predates most of the warming the Zürich analysis tracks [s1][s2].

For anyone reading heat-mortality headlines, the practical implication is that the number depends heavily on method. Two competent teams analysing two European cities in the same month produced results that are not directly comparable, and said so.

Sources

• Refining methods for attributing health impacts to climate change: a heat-mortality case study in Zürich — Climatic Change, 2025-09-10
• The impact of air temperature on mortality in İstanbul — International Journal of Biometeorology, 2025-09-18