Here’s a tale of how to generate headlines from circular reasoning built on brave assumptions. All it requires are some unskeptical science journal editors and gullible journalists. Et Voila!

Congratulations to Chip Knappenberger, Pat Michaels, and Anthony Watts, whose response to Åström et al was published Wednesday.

In October 2013 Åström et al claimed that global warming had killed lots of people in Stockholm, hundreds. But the first thing you need to know is that they don’t appear to start with actual mortality data in the early 1900’s. Surprised? Me too. Anthony Watts found it hard to believe . The other thing worth knowing is that extreme heat was defined as the top 2% of hot days, and in Stockholm that mean everything above a terrifying 2-day-moving-average mean temperature of 19.6C (67 F).

From the methods:

We collected daily mortality during the period 1980-2009 and daily temperature data for the period 1900-2009 for Stockholm County, Sweden.

Åström 2013:  Figure 2 j Temperature distribution of 2-day moving average of mean temperatures during summer months. Grey distribution, 1900–1929; black distribution, 1980–2009.

It appears the authors compared calculated death rates (using a model) from 1900-1929 with rates from 1980-2009 and concluded that mortality from heat was twice as high as it would have been which appears to be a product of their assumptions. They also find there was “no evidence”  that humans adapted to extreme temperatures. It’s a notion that seems self-evidently silly. We don’t need statistical tests to figure out if humans adapt to heat and cold, we see hot people turn on air-conditioners  — and any statistical test is confounded by improvements that are difficult to quantify, like medical knowledge, diet, exercise and changes in population diversity. Our species thrived through 4 million years of rolling ice-ages, and spread out to inhabit nearly every weather-zone from the Arctic to the Sahara. Why would homo sapiens find 0.7 of a degree warming to be difficult to adapt too? How many Swedes die of heatstroke when they holiday on the Mediterranean? Assume the paper is correct and Scandinavians couldn’t adapt to heat — wouldn’t there be shorter lifespans among those who emigrated to Australia, California and Florida?

Here’s a brave assumption:

“…we assumed constant exposure–response relationships.”

Perhaps I’m missing something. If we assume that exposure to a certain temperature causes a certain death rate (is that what they are saying?), I would have thought it follows automatically that adaption is not possible. If people adapted to hotter days the exposure–response rate would not stay the same?

Cold is deadly

Past studies established that winter kills more people than summer, even in hot countries like Australia. Cold times in China mean more death, war, rebellion and general other bad stuff that we don’t want (Lee 2010 and Zhang 2010). Evidence suggests that global warming is good for our health as I said in 2011:

The statistics on cardiovascular disease make it clear that cold weather is deadly. In Russia, ischemic stroke is 32% more likely on colder days; in Norway, cardiovascular deaths are 15% higher in winter months; in Israel, cardiovascular deaths were 50% higher in winter, even though Israeli winters are not exactly cold. Likewise in California heart disease mortality in 220,000 deaths was 33% higher in winter. A study in Brazil found that deaths were 2.6% more likely for every degree the temperature fell below 20°C. Need I go on?

Plus there is the other assumption in Åström et al, that the cause of the warming was human emissions of CO2, which for lots of reasons we know is a spurious exaggeration. There are plenty of factors that could make Stockholm warmer that have nothing to do with coal fired power plants. For starters, it is one of Europe’s fastest growing cities. The North Atlantic Oscillation had different effects in the latter 30 years than the earlier period.

Congratulations to Chip Knappenberger, Pat Michaels, and Anthony Watts, who show that warmer temperatures  led to a reduction in the rate of heat-related mortality in Stockholm which is pretty much what we would expect, and consistent with most other literature.

Anthony Watts notes that Nature Climate Change took months to publish their comment, used a long review process, and only published with a reply from Åström running along-side.  All of which is fine, if only they applied the same rules to both sides of the equation, they might not publish so many weak studies in the first place.

This study was all so circular.

We quantified the number of deaths that could be attributed to climate change
through a change in the frequency of extreme heat and cold events. If the numbers
of extreme events were roughly the same during the reference period as during
1980-2009, it would not have been possible to attribute any change in the extremes
or their associated mortality to climate change.

This is the third time in a week I’ve hit a paper with the marvel factor — I marvel that anyone thought it should be done in the first place, let alone submitted. And what were the editors thinking?

(I guess this is not a serious science journal, only Nature, right?)

Did a single person die in 1917 because the daily 2-day running average temperature hit 20C in Stockholm? Or were deaths on those days occurring because their health was so precarious they would likely have died anyway in the next few weeks or months?

Dare I suggest that people have to die sometime, sooner or later. Even if they lived under continuous ideal temperature conditions we would still see deaths on days of ideal weather. It’s true there are curves, and more people  die at the hotter and colder extremes, but there is no perfect temperature, and deaths at the warm end are often people who were likely to die. Mortality rates rise during heatwaves but its a well known phenomenon that they often fall below the expected afterwards. After mild heatwaves, the deficit mortality as it is known, “was close to 1.0” meaning most of those particular deaths were likely to have occurred soon anyway.  (See Saha et al.) This does not seem to be the case for severe heatwaves like the one in France in 2003. (Though here, confoundingly, the mortality deficit was inexplicably higher than expected, but did not fit the spacial distribution pattern. The authors could only speculate as to why thousands less people died over the following year than were expected too). But in Stockholm, remember, we are not talking about extreme heatwaves so much, but 2 day pairs of mean temperatures above 19.6C degrees (about 70 deg F).

These results are completely contrary to others, like Keatinge in BMJ which looked at mortality rates and temperatures across Europe and concluded that “Populations in Europe have adjusted successfully to mean summer temperatures ranging from 13.5°C to 24.1°C, and can be expected to adjust to global warming predicted for the next half century with little sustained increase in heat related mortality.” (H/t to Ferdinand Engelbeen)

Read this one carefully:

All temperature-related mortality is potentially preventable, making mortality during extreme temperature events a public  health concern.

What exactly is a “temperature related death? ” There is not a single definitive temperature death curve for homo sapiens. What is a “heat extreme” and “cold extreme” changes with latitude. Feeling a bit wicked, can I suggest that if “all  temperature related mortality is potentially preventable” what we really need are days with no temperature, so no one ever dies. (Ban thermometers: cut the celcius, and kill the kelvin.)

But dutiful gullible science journalists soaked in the statistics and generated these headlines:

Keep reading  →