Lomborg-errors: "Cool it!"

Heat- and cold-related deaths
 
    Home                                                                                                                                   Cool it


 "Cool it!", chapter 2: It´s getting hotter: The short story
 
On deaths related to heat and cold, pages 13-24 and 44-48.


 HEAT- AND COLD-RELATED DEATHS - THE SHORT VERSION

Lomborg postulates that rising temperatures will cause fewer people to die. He postulates that in Europe and North America today, many more people die due to excess cold compared to those that die due to excess heat. And with global warming, the decline in numbers of cold-related mortalities will be much larger than the rise in heat-related mortalities.
    But is it true that so many die from cold? Only if you make a trick by including the normal seasonal variation in death rate. It has always been so, and will remain so, that more people die during the winter months than during the summer months. Old and seriously sick people have less vitality in the dark season. It is too bold to say that the excess deaths during the dark part of the year are `deaths due to excess cold´. There is no evidence that a warmer climate will alter the seasonal variation. These people would soon die in any case, even if winters became warmer. Indeed, cold and warm climates, like Finland and Greece, have aprroximately the same seasonal variation in mortality. Thus, the figures in the book for cold-related deaths lack sound foundation.
    Lomborg also writes about the effect of a temperature rise not just for people in the rich western world, but for the whole world. However, his sources use the equations that are based on the inclusion of European seasonal variation in mortality, which means once again that figures for the projected change in deaths due to cardiovascular disease are flawed. The postulated declines in cold-related deaths worldwide are very unrealistic. Several sources of heat-related mortality have been omitted: studies dealing with hot climates are more or less neglected; deaths due to heat waves are not included; and with deaths due to cardiovascluar disease, only increases in urban populations are counted, whereas increases in rural populations are excluded.

DETAILED REMARKS ON HEAT- AND COLD-RELATED DEATHS

Lomborg claims p. 18 and p. 46 that the drop in death rates due to fewer cold-related deaths will be larger than the rise in death rates due to more heat-related deaths. This is opposite of what is claimed in the IPCC reports.
How does this discrepancy between IPCC and Lomborg come about?

Seasonal variation
To understand this, we will first look at the seasonal variation in death rate in temperate climates.
An example is Washington State, USA, where the daily number of deaths out of hospitals are 0.54 per 100,000 in the summer season, but 0.64 per 100,000 in the winter season (Cagle, cited by Lomborg). So there is an excess mortality of 0.10 per 100,000 in the winter months. The main cause of this is that old people who are about to die, die relatively often during the cold and dark season.
    One may choose to plot the daily death rate not as a function of season, but as a function of the temperature of the preceding day. That is done by many of the scientists that Lomborg refers to. The figure shown here (below) is a schematic representation of such a plot of the death rate versus temperature, such as it is presented for instance in Keatinge et al. (2000). We see that death rates are lowest in the small temperature sector designated C. The exact position of sector C differs between countries. It is centered around 15° C in Finland, 21° C in London and 24° C in Athens. In these places, the depicted curve has approximately the same shape, but the warmer the climate, the more the curve is shifted to the right. That is, irrespectively of the actual temperatures, the seasonal variation is the same. On average for the seven localities studied by Keatinge et al. (2000), there are each year 40 days in the favourable sector C, 38 days in the warmer sector D, and 287 in the colder sectors A and B (The data for London are extreme in this respect; in London, only 5 days per year are warmer than the optimal sector C. No wonder that projections for England show that reduction in cold has more effect than increase in heat).
    So, not only is the daily death rate lower in summer than in winter; it is especially low during rather few days in summer - on average 40 days - where temperatures are most pleasant, neither cool nor very hot. 

  

Relation between daily temperature and daily death rate


    What is done in the papers utilised by Lomborg, is that the death rate in sector C is taken as the basic death rate. Whenever the daily death rate is higher, the surplus is designated as `excess deaths´. It is implicitly assumed that if we had pleasant summer temperatures all year round, then mortality would stay at the low level all year round, and fewer people would die altogether. In Washington State, for instance (referred to above), death rates would stay at 0.54/100,000 all year round. But this is unrealistic. If climate were constant, there would be an average death rate (say 0.59/100,000) every day. With seasonal variation, deaths rates can be lower in summer only because many susceptible persons had died already during the preceding winter. To say that all deaths in excess of 0.54/100,000 are deaths due to `cold´, is a misuse of statistics. This misuse becomes even worse when we are said to have the basic death rate only during the 40 days per year, whereas we are said to have excess deaths due to cold not just during winter, but during 287 days per year, that is during 79 % of the year. We get an awful lot of `excess deaths´ in that way.  Most of these are deaths of elderly people that would have deceased anyway within about half a year.
    For each location and each temperature, the average daily mortality is taken from the death statistics. It is then believed - falsely, it seems - that even if climate warms, the death rate at each temperature will remain the same. So, if there will be fewer days with relatively low temperatures then the total mortality will be lower. This belief is not founded on any concrete knowledge. It would mean that the seasonal variation in death rate would become less marked, but what we see today is that the seasonal variation is equally marked in cold climates (Finland) and warm climates (Greece); the warmer the climate, the more is the whole curve shifted to the right, without a change in shape. It would therefore be a natural expectation that if the climate warms slightly in future, then the same will happen: the whole curve will shift to the right at each locality, without any change in the total number of deaths. But the scientists on whose papers Lomborg supports his text, do not project that. They project that the total number of deaths will go down. However, such a reduction cannot last for very long, for all people have to die sooner or later.
    What is relevant is to look at excess deaths in the true meaning of the word. We may be talking of a man of 50 years of age, with a weak heart, who may either survive during an extreme weather event and live for another 20 years; or he may not survive the weather event, in which case we are seeing an excess death which might have been avoided. We should not be talking of a man 80 years of age, who is about to die, and who will most likely die already in December rather than later, in June.
    To avoid misleading statistics due to seasonal variation, one may `de-season´ the death rates by subtracting from each day´s mortality count the average mortality for the month in which the deaths occurred. If that is done so that we look only at the excess death rate above what is normal for the season, then we get a death rate curve that is  horizontal during the whole range of sector B and C. What remains will be the death anomalies at extremely cold temperatures (sector A) and extremeley hot temperatures (sector D), that is at temperatures that are unsual for the season. This approach is used for instance in the papers by Cagle & Hubbard and by Davis et al. cited by Lomborg.

Extreme weather events
    With the de-seasoned data, we can focus on excess deaths during extreme weather events.
    The question we try to answer is: What will happen if climate warms? Will death rates change?
     A plausible theory is that what causes excessive death rates are not extreme temperatures, but unusual temperatures, that is temperatures that people did not expect and were not prepared for. There is no excess winter mortality in Yakutsk, the coldest larger city in the world. But if London were suddenly to have winter temperatures like in Yakutsk, mortality in London would soar. In conclusion, when we are talking about temperate climates, the temperature as such is irrelevant. What is relevant is the variability in temperature. And the projections are that this variability and unpredictability will increase. This in itself will lead to higher mortality. The question is then if society will be able to cope by improving warning systems.
    Lomborg says (p. 18) that we will adapt to higher temperatures, for instance by having more widespread air conditioning, and by improved warning systems. Therefore heat waves will not be as damaging as before. He is partially right. But even when we include the mitigating effects of adaptation, there will probably be an increase in heat-related deaths. See the text on adaptation in the list of flaws below.
     The sources used in the book assume that increased mortality from cardiovascular disease during hot weather will occur only in cities. But there is recent evidence that on the contrary, rural and suburban areas will be most affected. This kind of increased heat-related mortality in rural districts is omitted in the figures presented in the book.
   
Lomborg also deals with the situation for the world in total, that is he includes projected mortalities for people living in third world countries. Here, Lomborg relies on the report by Bosello et al. (2006). This in turn relies on other papers by Richard Tol, which in the end are based on a paper by Martens (1998) (cited by Lomborg and by IPCC (2001)). In the introduction part of Martens´ paper, we read that most of the recent research has been related to the impacts of heat waves, and references are given to a few papers, especially papers by Kalkstein´s research group, some of which Lomborg has read. In contrast, Martens describes the theme of his own paper as follows: "However, this paper . . . will not focus on periods of extreme heat or cold, but will consider the potential changes in numbers of deaths associated with moderate warmth and cold . . It should be noted that that studies on mortality effects of short-term periods of excessive heat address another question than studies which focus on mortality changes of moderate warmth and cold". So, this study on which Lomborg relies, excludes the effect of heat waves. The same is true for the papers by Keatinge and Donaldson, and this is especially true for their data on London which do not include data from any very hot days at all. All the papers on which Lomborg supports his calculations deal only with mortality due to moderate temperature changes, and they find for temperate climates - not surprisingly - that few people die when summer temperatures increase moderately.
    In Lomborg´s text on heat waves on page 18, he refers directly to Martens (1998), even though Martens does explicitly not deal with heat waves.

The original paper is manipulated
    The paper by Martens (1998) is in itself manipulating. It is a metastudy, compiling data from many original studies. In the middle of the paper, the overall combined trend from all these studies is stated as follows by Martens: "For total mortality, the weighted effect estimate for an increase of 1°C in the "cold" range is -1.0 % . . . In the "warm" temperature range, the weighted mean estimate of an increase of 1°C for total mortality is 1.4%." Thus, what the compiled data show is that total mortality will increase more by one extra °C in the warm range, than it will decrease by one less °C in the cold range. The net effect of a temperature increase will be an increase in mortality. But, towards the end of his paper, Martens manipulates the data. He focuses on mortality due to one cause, cardiovascular disease, where there is no clear overall increase of mortality with a temperature rise, and assumes that people will adapt so that the end effect will be reduced mortality due to cardiovascular disease. And next he says that the effect of a temperature increase on other diseases, and on total mortality, is doubtful. In the end the conclusion is approximately the opposite of what the data analysis actually shows - the conclusion is turned around to be that total mortality will decrease, when in fact the data indicate that total mortality will increase.
   All the studies that Lomborg refer to are based on Martens, and therefore the flaw is passed on to them all.

The situation in The Third World
    The above studies are not very relevant for poor countries with hot climates. Here a temperature rise may in itself increase the number of days where temperatures surpass human physiological limits. The increased mortality in tropical countries due to increased heat may have been underestimated by the sources used by Lomborg. There are other studies indicating greater sensitive to temperature increases. There exists for instance a study by Kalkstein & Moyer who conclude that a warmer climate will have worse effects in Egypt and China than in North America. Also, a study cited by Lomborg on his page 126 (Campbell-Lendrum et al. 2003) indicates that an extra °C will cause a net increase in mortality due to cardiovasular diseases in New Delhi as large as that in The Netherlands. ( But there are other places (Sao Paulo) where extra heat may cause a net decline in cardiovasular mortality).

What the IPCC reports say
    Lomborg´s conclusions are at variance with the conclusions in the IPCC reports. In the third IPCC report from 2001, the most relevant section is 9.4.2. It contains those references that up to 2001 were relevant for the balance of heat deaths relative to cold deaths. Additional studies after 2001 are presented in the fourth IPCC report from 2007 in its table 8.1, and related text is found in sections 8.2.1.3 and 8.4.1.3 in that report. Altogether, the reports cite rather many studies demonstrating the risk of an increase in heat-related deaths, e.g. in central and south Europe (according to IPCC, many regions, including south Europe, will experience more severe heat waves in the future).The overall pattern is summed up in a figure (Fig. 8.3) in the fourth IPCC report, and  according to this figure the reduction in cold-related deaths is considerably smaller than the increase in deaths due to heat and extreme weather episodes.
    



Flaws on particular pages in Lomborgs text:


(COMMENT)
Page 16: ". . the total death toll ran to more than 35,000."
Comment:
According to newer information, the excess number of deaths ran to more than 70,000.


FLAW
Page 16 bottom: ". . . equal or more unusual warm anomalies have occurred regularly since 1979."
Flaw:
This is misleading. The cited paper refers to the whole northern hemisphere, not just to western Europe. But according to the 2007 IPCC report, part I, page 311, the heat wave in western Europe in the summer of 2003 was indeed warmer than any other episode from 1780 onwards.

(COMMENT)
Page 17 top: ". . the BBC recently ran a very quiet story telling us that deaths caused by cold weather . . "
Comment:
The excess deaths in winter are a normal phenomenon in most countries in temperate climates. Nearly 90 % of the excess deaths are of elderly people. See the introductory text on this  web page.

FLAW
Page 18 top: "It is reasonable to estimate that each year about 1.5 million people die from excess cold in Europe."
Flaw:
No, it is not reasonable. When one talks of dying from "excess cold", one would usually think of deaths due to unusually cold weather, deaths that would otherwise not have occurred. But most of Lomborg´s 1.5 million deaths are simply deaths of old people that would die soon in any case; they die most often during the relatively cold months. Lomborg mixes up normal seasonal variations in mortality (which will always occur, irrespectively of climate change) with excess deaths due to extreme weather events (which may be affected by climate change). Thereby he obfuscates the whole issue.

FLAW
Page 18 top: "This is more than seven times the number of heat deaths."
Flaw:
No it is not, because Lomborg includes only deaths due to moderate warmth and excludes deaths during unusual heat waves.

FLAW
Page 18 top: "Europe has lost more than 10 million people to the cold, . ."
Flaw:
This is more or less nonsense. We are talking about the dying of old people at temperatures mostly in the interval 0° C to 20° C, people who would have died in any case. They are not lost to the cold. They died because they would so anyhow.

(COMMENT)
Page 18 bottom: "For Britain, it is estimated that. ."
Comment:
Here Lomborg refers to Keatinge and Donaldson (2004) which has the main point that over time, people adjust to temperature changes, e.g. by installing air conditioning, but where it is also said that "Populations accustomed to heat or cold adjust to them, but record high temperatures for a locality during global warming have caused high mortality rates." So again, the study cited by Lomborg refers to moderate warmth, and does explicitly not include what happens at record high temperatures.


(COMMENT)
Page 18 bottom: "global warming may cause a decrease in mortality rates. ."
Comment:
This quote leaves out the preceding sentences, which say: ". . this study has tried to answer the question; `What is the annual balance between a change in moderate cold and warmth-related deaths due to global climate change . .?´ Although the overall balance remains difficult to quantify and would depend on adaptive responses and existing health levels, global warming  may cause a decrease in mortality rates, especially of cardiovascular diseases." It is important to note that this conclusion deals only with moderate cold and warmth, not with heat waves, and that the word may should be understood in the sense that the conclusion is uncertain.

FLAW
Page 18 bottom: "Several recent studies have looked at adaptation . ."
Flaw:
The cited studies have considered only adaptation to heat, not adaptation to cold. They conclude that deaths due to heat waves in USA have gone down because of adaptation to the heat. There is evidence from elsewhere that there has also been an adaptation to cold (IPCC, 2007 report, WG II, section 8.2.1.3). Also, the paper by Keatinge & Donaldson (2004), cited by Lomborg, has the following text:
"In recent years, temperature and mortality data from several countries shows that cold-related deaths in each age group are falling in most countries. Much of that was due to rising climatic temperature and better home heating. The reduced frequency and severity of new epidemics of winter influenza has contributed, but since 1976, campaigns for warmer housing and advice on clothing and exercise to keep warm outdoors can take most of the credit. A surprising finding is that the heat-related mortality rate has stabilized or fallen, despite rising temperatures. Air conditioning has been a major factor in the United States. . . An extension of air conditioning was accompanied by the virtual disappearance of heat-related death in North Carolina, despite summers becoming hotter. . .  Britain and the rest of northern Europe still have little air conditioning, and the heat-related mortality rate in London has not fallen. Nor has it risen, however, despite a 3.6° F rise in summer temperature since 1971. Such factors as more relaxed lifestyle, more informal clothing, and purchase of electric fans as prosperity increased have apparently countered heat stress there."
    So, Lomborg´s sources speak of adaptation to cold as well as to heat. But Lomborg chooses to talk only about the latter. Before the 1980s American cities had more excess deaths due to unusual heat than due to unusual cold. After adaptation to both extremes, the ratio between the two causes of death may possibly have remained unchanged. As Lomborg himself writes in a note to page 17: ". . both heat and cold deaths will tend to revert to their previous values, because of acclimatization. . " On page 18, however, he writes as if this were not so.
    Lomborg does only vaguely mention possible limits to adaptation (page 15). Air conditioning in the South Atlantic region of the United States extended from 56 to 72% of homes from 1976 to 1997 (Keatinge & Donaldson (2004), cited by Lomborg), and many cities now have more than 90 % saturation with air conditioning. In addition, deaths during heat waves have become much reduced due to better public warning systems. This means that there remain no obvious possibilites for futher adaptation. Davis et al.  2003 (cited by Lomborg) write: "The impact of heat waves on mortality there may provide some case studies of how future populations might respond to heat stress under full air-conditioning saturation conditions." In other words - we do not yet know what will happen there when temperatures increase further.
    In more northerly cities there is still a scope for adaptation, such as more widespread air conditioning. However, a recent study by Kalkstein (link here) finds that a warmer climate will cause more heat-related deaths here, even when adaptation is included in the models.

FLAW
Page 18: "However, more people still die because of colder weather."
Flaw:
As explained above, Lomborg obfuscates the issue by mixing up seasonal variation and excess deaths due to unusual weather. The reader may understand the above sentence in the way that there has been no adaptation to cold. As explained above, this is not true.

FLAW
Page 21: " . . the average minimum temperature over the past 40 years for Tucson has increased more than 6° C."
Flaw:
In the reference (Comrie 2000), it is stated that there has been approximately a 6° C increase from the early part of the 20th century. This agrees with Lomborg´s figure 6. You can only claim that the rise has occurred over 40 years, if you count from the atypical year 1963. A statistician should not do that. Out of the 6° rise, the reference states that probably 46 % is due to the urban heat island effect. Notice also that according to the reference, urban-rural temperature contrasts are more noticeable in daily temperature minima (in contrast to maxima).

(COMMENT)
Page 23 top: " Although deaths have in general been declining . ."
Comment:
According to data from the US National Weather Service (link), deaths that can be directly attributed to heat have not declined in recent decades. The average number of heat fatalities per million people was 0.47 in the period 1986-1996, but 0.61 for the period 1997-2006.

FLAW
Page 46 top: ". . published in 2006. What it shows us very clearly is that climate change will not cause . . "
Flaw:
As explained in the introduction to this page, the study referred to is based on calculations that deal with seasonal variation and moderate warmth, but not with heat waves. Therefore, what Lomborg concludes on the basis of this study does not refer to heat waves.