Unraveling the debate: Does heat or cold cause more deaths?
it's more complicated than you think
I have gotten several requests to write about quantifying heat- and cold-related deaths. And with recent research reporting that 61,000 people died last summer during a European heat wave, the time seems right. This post is part one, which goes over previous work on this topic. In part two, I talk about some new research that my former graduate student Jangho Lee and I have done on this topic. In part three, I talk about adaptation as a response and why it’s going to be harder than you think it will be.
If you peruse Twitter during a heat wave you’ll inevitably encounter arguments about whether extreme cold or heat causes more deaths. The answer isn't as straightforward as one might think.
A lot of research has been done on this question and probably the most well known is by Gaspirrini et al. entitled “Mortality Risk Attributable to High and Low Ambient Temperatures.” It’s definitely worth reading.
In this and many similar papers, temperature-related mortality is calculated statistically by looking at how total (non-accidental) deaths vary with temperature. This approach generates plots of relative risk (RR) versus temperature:
Relative risk is the fraction of excess deaths at a particular temperature relative to the number of deaths at the Minimum Mortality Temperature (MMT), the temperature where the death rate is the lowest, typically around 20C. For example, if there are 100 deaths per day in London at the MMT (19C), then an RR of 1.2 at 5C means that there are 20 additional deaths, with that increase assumed to be due to temperature.
The conclusion that “cold kills more” emerges from assigning any temperature-related deaths occurring below the MMT to be “cold-related deaths”. This would include, for example, those at 17C (63F), a temperature that almost no one would consider to be extreme, or even “cold”.
The plot above also shows a histogram of the deaths of London over the last few decades as a function of temperature. You can take this as a proxy for temperature and, as you can see, the temperature rarely reaches above the MMT of 19C. Thus, for London, whose temperature is nearly always below the MMT, the majority of temperature-related mortality is “cold-related”.
If you add up all cities around the world, this result holds up and we find that, in today’s climate, worldwide cold-related mortality exceeds heat-related mortality.
The future
The Earth is warming. Humans are to blame. And this global warming will push the temperature distributions of all cities to hotter temperatures. This will have two effects: fewer cold-related deaths and more heat-related deaths. The arguments over temperature-related deaths due to climate change therefore reduce to determining which of these terms is larger.
In some cases, the answer is “it depends”. For London, small increases in temperature will lead to reductions in overall temperature-related deaths because typical London temperatures are so far below the MMT.
But, with enough warming, even London will experience a net increase in temperature-related deaths. To see this, let’s hypothetically shift London’s temperature distribution by 5C, crudely simulating a 3C global average warming above a baseline of the last few decades, corresponding to ~3.5-4C warming above pre-industrial:
With this warming, there will clearly be fewer cold-related deaths. But you can see that a lot more days are on the “hot side” of the MMT. And since the hot side of the RR curve rises steeply with temperature, this will drive an enormous increase in heat-related deaths that will eventually outstrip any lives saved from the reduction of cold-related deaths.
Is climate change going to cause more or fewer deaths?
The upshot of all of this is that, for many cities in the mid-latitudes, where the MMT is above typical temperatures, small amounts of warming will reduce temperature-related mortality while a lot of warming will increase it.
But there are some cities, typically in the tropics, where any warming at all will increase temperature-related mortality. For example, here is the curve for Bangkok:
This city’s temperature distribution is centered on the MMT. Any global warming will increase the frequency of temperatures on the hot side and decrease it on the cold side about equally. But since the hot side curve rises more rapidly than the cold side curve, this will lead to a net increase in mortality.
In fact, even in places where we might not expect it, we’re already seeing bursts of heat-related deaths during extreme heatwaves. For example, a paper in Nature Medicine that was just published estimates that 61,000 people died last summer in Europe due to the high temperatures there. Here is the RR curve from that paper:
Temperature were about 3C warmer than the summertime baseline, which is at least partially due to climate change, and this pushed summertime temperatures that would normally be near the MMT well into the hot side of the steeply rising RR curve (see Fig. 2 of the paper), leading to the huge number of deaths.
I’ll talk more about impacts and adaptation in posts 2 and 3 of this series, but it’s worth noting that most of the people who die in events like this are poor, vulnerable, or old.
NOAA’s estimate
NOAA states that extreme heat is the biggest weather killer today, a result at odds with the results described above. The reason for this is that NOAA only counts deaths where the cause of death is determined to be due to heat/cold. In contrast, the Gaspirrini method is a statistical approach that regresses “excess deaths” against temperature. Hence, these are very different methodologies.
The debate around whether cold or heat causes more deaths thus requires nuanced understanding and is dependent on several factors including the methodologies used for calculating these deaths, current climate conditions, and how the climate will change in the future. I will pick up on these in posts 2 and 3 of this series.
Related posts:
My interview with Jeff Goodell about his book, The Heat Will Kill You First
Thanks for this post. It's very illuminating and well presented.
It's interesting that the MMT varies so much by city. This makes sense - Bangkok is a hot city in the tropics; a normal day in Bangkok would be far out of normal for London or Boston.
But this suggests two things:
1. People can adapt to, and flourish in, a wide range of temperatures.
2. We should start paying attention to adaptation. Some cities and countries that have historically used little air conditioning should perhaps plan to use more. Further temperature increase is inevitable, but a little prudent adaptation would keep it from being fatal.
"More people die of cold than heat" - the climate denier mantra.
While I'm grateful to have an expert answer, the question is silly. First, the correct answer to (say)100k dying from cold is money for heating, not global warming. (Is the answer to the higher frequency of deaths on pedestrian crossings to remove the crosswalks?)
Business as usual - millions eventually billions will die when they can no longer grow food where they live.