According to various studies and numerous web postings climate change is already causing hundreds of thousands of deaths and will cause millions more in the future, dominantly in poor countries (see inset). In this post we take a brief look at how these estimates were arrived at and whether they have any firm observational or statistical backup. The conclusions are, well, interesting.
I looked at three major reports that estimated how many deaths climate change is causing and is going to cause in the future. In order of appearance these were:
Only the second two reports had enough banner headline material to catch the attention of the media. According to the Guardian the Global Humanitarian Forum report showed that “Climate change is already responsible for 300,000 deaths a year and is affecting 300m people …. It projects that increasingly severe heatwaves, floods, storms and forest fires will be responsible for as many as 500,000 deaths a year by 2030”. According to Reuters the DARA report showed that “More than 100 million people will die and global economic growth will be cut by 3.2 percent of gross domestic product (GDP) by 2030 if the world fails to tackle climate change”. Both of these reports, however, contain over a hundred pages (the second over 300) of text, graphics and photographs, and as far as I can see without going through them in detail, which I don’t have the time to do, no hard backup data. They are more political manifestos than scientific assessments.
The first report (the “2007 report”) did not catch the attention of the media because it gave little in the way of banner headline material. One has in fact to go all the way down to table A2 on page 58 to find the report’s estimate of climate change deaths – 166,000 in 2000. However, it did adopt a somewhat more scientific approach and even contained some backup data, so this is the report we will take a closer look at here.
The methodology the 2007 report uses to estimate climate change deaths is too complex to be described in detail, but the basics are summarized in Section 2.5:
This approach is reasonable, but the 2007 report gives only one example where disease rates can be quantitatively related to a climate variable – child diarrhea versus temperature in Lima, Peru, which reportedly shows an 8% change in incidence for each 1 degree C change in temperature:
Figure 1: Child deaths from diarrhoea versus temperature, Lima, Peru
If these are all the hard data the 2007 report can provide then it clearly does not have enough data to estimate worldwide climate deaths from other sources (malnutrition, malaria, cardiac arrest, extreme weather events etc.) The report also uses only data from Peru, Australia and Fiji, and one can question whether these are applicable to the rest of the world. It does, however, provide a long list of references, and we will assume that the data necessary to make meaningful quantitative global, regional and source-by-source mortality assessments are to be found somewhere in them.
Now to the 2007 report’s results. These are summarized in Table A2 in the Appendix, which presents the report’s estimates of climate change deaths in 2000 by cause and subregion:
Subregion locations are shown in Figure X below.
Figure 2: World Health Organization subregions (note that no map of these subregions is provided in the 2007 report)
A defect of the 2007 report is that while it estimates deaths from climate variables it never gets round to comparing them with a climate variable. If it did, and if the report’s conclusions were correct, we would expect to see a positive relationship between deaths and changes in the climate variable. I rectified this omission by constructing XY plots of deaths against temperature change, which is the variable the 2007 report makes most use of. I used several hundred of the unadjusted GHCN v2 temperature records contained in “Roger Andrews climate data” (also accessible in the sidebar) to estimate surface air temperature changes between the 1951-80 “baseline” temperatures and 2000-2010 mean temperatures in the different subregions, thereby matching the period shown in Figure 1b of the 2007 report. It was not possible to do this for all the subregions, but by reconfiguring boundaries and population-weighting the mortality and temperature change estimates to take the boundary changes into account I came up with the results shown on the following table:
I then used these results to construct an XY plot of climate-caused deaths against 1951-80 to 2000-2010 surface air temperature change:
Figure 3: Deaths from climate change versus change in surface air temperature between 1951-1980 and 2000-2010 means, degrees C
The data points show considerable scatter and the trend line shows only weak levels of correlation (R squared is close to zero). Yet the overall trend is negative, not positive, which is the opposite of what we would expect to see if rising temperatures indeed caused increased mortality. But unfortunately the data are too scattered to allow us to conclude with any confidence that climate change is actually good for us.
Figure 3, however, is a global plot. It takes no account of the fact that warming in recent decades has been stronger in the Northern Hemisphere than in the Southern. When we segregate the subregions into those entirely or dominantly in the Northern Hemisphere and those dominantly in the Southern we see a more consistent picture:
Figure 4: Figure 3 data segregated by hemisphere
The correlations are now much stronger (R squared = 0.71 in the Northern Hemisphere and 0.83 in the Southern). Taken at face value these results can be interpreted to mean that rising temperature save lives in the Northern Hemisphere but kill people in the Southern. A more realistic interpretation, however, is that with the arguable exception of deaths caused by winter cold versus deaths caused by summer heat, we don’t have enough data to say whether climate change has any impact on human mortality, if indeed it has any impact at all.