Revisiting Svensmark & Friis-Christensen (1997)
Posted by Ari Jokimäki on February 15, 2010
Lately I have been putting together stuff with intention to write a thorough review of the situation with the role of the cosmic rays in the climate. For that, I read for the first time the original Svensmark & Friis-Christensen (1997, “SF97” from hereafter). To my surprise, I found out that their whole thing seems to be based on false cloud trends. Some of you might have read my previous writing of ISCCP problems. It turns out that SF97 use ISCCP data to show a correlation between clouds and cosmic rays.
At one point, they say this about cloud cover:
“In Fig. 2 it is seen that a pronounced variation (corresponding to 3%-4%) takes place during this period with a maximum around 1986-1987, close to the minimum in solar activity.”
Their Figure 2 then shows that situation with the changes in cosmic rays and there indeed seems to be good correlation between the two in the short interval they are presenting. Both increase roughly from 1984 to 1987 and then decrease from 1987 to 1990.
Among others, Evan et al. (2007) have studied the problems in ISCCP cloud data. They noted that addition of satellites to the measurement network causes decrease in the measured cloud cover due to the changes in the satellite viewing angle. They listed some points in time when there had been remarkable changes in the satellite network. One of the changes is the launch of satellite GOES 7. It happened in February 1987 and fits well to the strong decrease in the ISCCP cloud data at that time. Evan et al. also showed how the trends change if one uses only the regions where viewing angle problem does not affect the data much. They presented it in their Figure 3. Figure 1 here shows the Fig. 2 of SF97 and Fig. 3 of Evan et al. Note that the cloud cover changes are little different in these two because SF97 only used data taken over oceans.
Figure 1. Figure 2 from SF97 (top panel) and Figure 3 from Evan et al. (bottom panel). The data from regions affected less by the viewing angle problem has been highlighted with red between the time period with ISCCP data (about 1984-1990) used in SF97.
When we look at the data highlighted with red that presents the time period where SF97 showed the correlation (shown in top panel of Figure 1 above), we notice that the cloud trends between that time have disappeared almost completely. What follows is that the correlation presented in SF97 is most likely not real but a result of ISCCP viewing angle problem.
After this, SF97 show the situation with other cloud data in their Figure 4, but that image is very unclear. It is difficult to estimate if the other cloud data supports the correlation or not. I first thought that this was just due to scanning the original paper to electronic format, but Laut (2003) has also noted the blurred appearance of the figure, so the problem was already in the original publication. However, it can be seen from Laut’s discussion that the other cloud data don’t support the correlation (putting the blurred appearance of SF97 figure and its presentation as if it would support the correlation in not so pleasing light), so the correlation presented in SF97 seems to rest on the ISCCP data alone.
SF97 also show that the correlation is less near the equator and they think it supports their view. If the apparent correlation would be due to ISCCP viewing angle problem, then the geometry of the problem would also cause the correlation to appear less near the equator because at the equator there is generally less affected areas than elsewhere.
So, there you have it. It seems in a strange way quite amusing to think that the whole of this cosmic ray issue might have been resting on the ISCCP viewing angle problem in the original paper and the later works have just been stamp collecting the apparent correlations here and there. In this light it is not surprising that we keep finding lot of problems with the cosmic ray hypothesis.
Laut (2003), “Solar activity and terrestrial climate: an analysis of some purported correlations”, Journal of Atmospheric and Solar-Terrestrial Physics, Volume 65, Issue 7, May 2003, Pages 801-812, [abstract, full text]
Svensmark & Friis-Christensen (1997), “Variation of cosmic ray flux and global cloud coverage—a missing link in solar-climate relationships”, Journal of Atmospheric and Solar-Terrestrial Physics, Volume 59, Issue 11, July 1997, Pages 1225-1232, [abstract, full text]