Gothmog
Dread Enforcer
- Joined
- Sep 19, 2002
- Messages
- 3,352
OK, here’s my promised thread on cosmic ray climate connections…
I currently am involved in studying the influence of cosmic rays on lightning initiation. Thus, I do have some stake in finding a correlation between cosmic rays and various climate parameters. However, as a scientist I also must try and take data at face value. That is what I am going to do here.
Now the first thing to address is that the total solar irradiance (TSI) and the galactic cosmic ray flux (GCR) are roughly inversely proportional.
This is because the TSI is proportional to the solar wind, which interacts with the earth's magnetic field in such a way as to make periods of high solar wind periods of low GCR.
Here I've plotted historical data from NOAA for sunspots and from Climax, the gold standard Cosmic Ray Observatory. Both are available by ftp on the web.
These are our main proxy measurements for TSI and GCR. Sunspot observations go back a few hundred years and the Climax record can be extended with measurements of 10Be or other cosmogenic isotope.
For TSI we also have direct measurements from satellites, though absolute calibration and inter-calibration are difficult and topics of some dispute.
http://www.pmodwrc.ch/pmod.php?topic=tsi/composite/SolarConstant
Here’s a composite:
There are additional technical details I am happy to discuss in this context.
We easily see that there is no systematic trend in TSI, Sunspots, or GCR on these timescales. This alone makes it hard to believe that solar activity can explain the recent trend in global temperatures.
The main hypotheses in this area are:
1) Change in TSI leads directly to change in earth climate
2) Change in TSI 11 year cycle length leads to change in earth climate
3) Change in GCR leads to change in total cloud coverage
4) Change in GCR leads to change in low cloud coverage.
Hypothesis 1 involves a direct forcing from changes in the TSI of the Sun. The mechanism here is clear, more energy input to the earth system. This hypothesis actually has quite a bit of success helping to explain the Medieval Warm Period and the Little Ice Age (along with other factors, again lots of technical details); and there are trends in TSI and GCR on those timescales. However, as we see in the figure above there is no trend in TSI or GCR on timescales relevant to modern climate.
Hypothesis 2 involves the length of the 11 year cycle. There is no physical mechanism that I know of here, though it is thought that the length of the 11 year cycle is related to the intensity of the 11 year cycle (but then we could just investigate the correlation of climate with the intensity of the 11 year cycle – there is no there there). This is the famous Lassen et. al. 1991 paper (also 1995 and 2000). This hypothesis has been seriously rebuked, both for misrepresentation of the data in the original paper (e.g. P Laut, Journal of Atmospheric and Solar-Terrestrial Physics 65 (2003) 801-812) and for a failure to continue the correlation since 1991 (or maybe 95’
. That’s why we don’t hear much about that hypothesis any more.
Hypothesis 3 involves a correlation between GCR flux and total cloud cover as measured from satellites. The mechanism involves changes in GCR flux changing the rate of electroscavenging in the atmosphere, and so changing the size distribution and number density of liquid aerosols in the atmosphere (e.g. clouds, condensation nuclei, and cloud condensation nuclei). This is actually a very old hypothesis and has popped up a number of times, though in many of its original forms it was solar ionization and not GCR that was thought to be important but empirical evidence was lacking. There are some reasonable theoretical reasons to think some correlation might exist. The seminal empirical paper was published in 1997 by Svensmark and Friis-Christensen. But once again this paper has been panned for misrepresentation of data (e.g. Kristjansson and Kristiansen Journal of Geophysical Research 105 (D9), 11851-11863), which (to his credit) the principle author owned up to. The basic issue was that the 97’ paper combined two different data sets, which he said represented ‘total cloud cover’. One of the data sets showed an increasing trend and the other a decreasing trend, and so combined they showed a periodic trend like the GCR. This topic has again dropped off the radar due to lack of empirical corroboration.
Hypothesis 4 involves a correlation between GCR flux and ‘low cloud cover’. The mechanism here is that GCR dominate atmospheric ionization at low altitudes because only high energy rays can penetrate that far into the atmosphere. At higher altitudes UV radiation, and other solar components (such as muon flux), compete with GCR in terms of ion production. There is a mechanism by which ions can become condensation nuclei (ion mediated condensation). If condensation nuclei originating in ions then become cloud condensation nuclei (CCN), and if this source of CCN dominates other sources of CCN (such as formation from condensation nuclei originating in SO2 oxidation, or heterogeneous nucleation on sea salt, dust, or soot), then we expect GCR flux to be important to cloud formation and character.
The seminal empirical paper in this area is by Marsh and Svensmark (Physical Review Letters 85 (23) 5004-5007, 2000). Yes, that’s the same Svensmark as in Hypothesis 3, which he has now apparently discarded. Here he uses a derived quantity called ‘low cloud cover’. He finds a good correlation with GCR, but he chooses to use GCR flux from a Peruvian station known as Huancayo. On its face this is fine, Huancayo is a reputable station and the data should be fine. Trends between GCR stations are very similar, but there are some differences due to GCR interactions with the earths magnetic field and the change of that interaction with latitude. So because Svensmark used Climax data in the 97’ paper, it is a little strange to start using Huancayo data now. Then in 2002 a paper comes out (by Kristjansson again, he works on the satellite cloud data archive in use here - ISCCP), showing much poorer agreement and making the point that there is somewhat of a phase shift (when higher frequency data is looked at), which contradicts known CCN from ion mediated nucleation theories. The paper also brings to light other issues that have been discussed by scientists more informally: 1) Marsh (2000) uses only IR data to arrive at the correlation, not the more common IR/VIS data; 2) low clouds are tough to detect from satellites because they are often blocked by higher clouds; 3) the analysis by Marsh only uses data out to 1995, the Kristjansson (2002) paper shows the correlation deteriorates in the 1995-2001 data analyzed by them.
Then in 2002, Marsh and Svensmark incorporates a calibration change to the satellite data in 1994, that no one else uses. While it is possible that such a change is warranted, and he does present a justification, the ISCCP team does not use such a change. Future papers on this calibration issue also do not suggest that such a change be incorporated.
Other work is still being done in this area. Other low cloud data sets (including ground based observations) are set to the task, none find a significant correlation (e.g. Sun and Bradley Journal of Geophysical Research, 107 (D14): article 4211, 2002). Kristjansson publishes another paper in 2002 suggesting that it is the TSI that correlates better with low cloud cover (mechanism involving stratospheric temperatures influencing gravity waves, which influence tropospheric dynamics). A number of other papers have suggested that there has been an increase in total cloud cover in the last century, or few centuries, and that solar proxies show an increase over that time. This again suggests TSI, and not GCR, as the causative mechanism (remember inverse relation). This is important because it helps informs us of which physical mechanisms should be investigated, and which ruled out.
Another thing to keep in mind is that even if there is a connection between GCR or TSI and low cloud cover, it would still have to be shown that this change was consistent with empirical measures of earth’s climate.
People are now looking at GCR proxies that extend farther into the past and their correlation with climate proxies. Some find correlations, others don’t. Many scientists think that it is the variation in UV radiation that is important, but since there is an inverse relation with GCR flux (mediated through the solar wind/earth’s magnetic field issue mentioned above), it is nearly impossible to tease these issues apart in the absence of a complete mechanism and plenty of modeling studies.
So part of the problem is we lack a well defined mechanism, and an associated model created to test that mechanism. Thus, we will continue to see publications with new mechanisms, and new models, that claim to reproduce all data. These will need to be put through the grinder of the scientific endeavor and should not be taken as consensus.
I have loads of other references, so feel free to ask. Here is a good overall review by the Hadley center, freely available, and with an extensive bibliography.
http://meteo.lcd.lu/globalwarming/Gray/Influence_of_Solar_Changes_HCTN_62.pdf
I currently am involved in studying the influence of cosmic rays on lightning initiation. Thus, I do have some stake in finding a correlation between cosmic rays and various climate parameters. However, as a scientist I also must try and take data at face value. That is what I am going to do here.
Now the first thing to address is that the total solar irradiance (TSI) and the galactic cosmic ray flux (GCR) are roughly inversely proportional.
This is because the TSI is proportional to the solar wind, which interacts with the earth's magnetic field in such a way as to make periods of high solar wind periods of low GCR.
Here I've plotted historical data from NOAA for sunspots and from Climax, the gold standard Cosmic Ray Observatory. Both are available by ftp on the web.
These are our main proxy measurements for TSI and GCR. Sunspot observations go back a few hundred years and the Climax record can be extended with measurements of 10Be or other cosmogenic isotope.
For TSI we also have direct measurements from satellites, though absolute calibration and inter-calibration are difficult and topics of some dispute.
http://www.pmodwrc.ch/pmod.php?topic=tsi/composite/SolarConstant
Here’s a composite:
There are additional technical details I am happy to discuss in this context.
We easily see that there is no systematic trend in TSI, Sunspots, or GCR on these timescales. This alone makes it hard to believe that solar activity can explain the recent trend in global temperatures.
The main hypotheses in this area are:
1) Change in TSI leads directly to change in earth climate
2) Change in TSI 11 year cycle length leads to change in earth climate
3) Change in GCR leads to change in total cloud coverage
4) Change in GCR leads to change in low cloud coverage.
Hypothesis 1 involves a direct forcing from changes in the TSI of the Sun. The mechanism here is clear, more energy input to the earth system. This hypothesis actually has quite a bit of success helping to explain the Medieval Warm Period and the Little Ice Age (along with other factors, again lots of technical details); and there are trends in TSI and GCR on those timescales. However, as we see in the figure above there is no trend in TSI or GCR on timescales relevant to modern climate.
Hypothesis 2 involves the length of the 11 year cycle. There is no physical mechanism that I know of here, though it is thought that the length of the 11 year cycle is related to the intensity of the 11 year cycle (but then we could just investigate the correlation of climate with the intensity of the 11 year cycle – there is no there there). This is the famous Lassen et. al. 1991 paper (also 1995 and 2000). This hypothesis has been seriously rebuked, both for misrepresentation of the data in the original paper (e.g. P Laut, Journal of Atmospheric and Solar-Terrestrial Physics 65 (2003) 801-812) and for a failure to continue the correlation since 1991 (or maybe 95’
. That’s why we don’t hear much about that hypothesis any more.Hypothesis 3 involves a correlation between GCR flux and total cloud cover as measured from satellites. The mechanism involves changes in GCR flux changing the rate of electroscavenging in the atmosphere, and so changing the size distribution and number density of liquid aerosols in the atmosphere (e.g. clouds, condensation nuclei, and cloud condensation nuclei). This is actually a very old hypothesis and has popped up a number of times, though in many of its original forms it was solar ionization and not GCR that was thought to be important but empirical evidence was lacking. There are some reasonable theoretical reasons to think some correlation might exist. The seminal empirical paper was published in 1997 by Svensmark and Friis-Christensen. But once again this paper has been panned for misrepresentation of data (e.g. Kristjansson and Kristiansen Journal of Geophysical Research 105 (D9), 11851-11863), which (to his credit) the principle author owned up to. The basic issue was that the 97’ paper combined two different data sets, which he said represented ‘total cloud cover’. One of the data sets showed an increasing trend and the other a decreasing trend, and so combined they showed a periodic trend like the GCR. This topic has again dropped off the radar due to lack of empirical corroboration.
Hypothesis 4 involves a correlation between GCR flux and ‘low cloud cover’. The mechanism here is that GCR dominate atmospheric ionization at low altitudes because only high energy rays can penetrate that far into the atmosphere. At higher altitudes UV radiation, and other solar components (such as muon flux), compete with GCR in terms of ion production. There is a mechanism by which ions can become condensation nuclei (ion mediated condensation). If condensation nuclei originating in ions then become cloud condensation nuclei (CCN), and if this source of CCN dominates other sources of CCN (such as formation from condensation nuclei originating in SO2 oxidation, or heterogeneous nucleation on sea salt, dust, or soot), then we expect GCR flux to be important to cloud formation and character.
The seminal empirical paper in this area is by Marsh and Svensmark (Physical Review Letters 85 (23) 5004-5007, 2000). Yes, that’s the same Svensmark as in Hypothesis 3, which he has now apparently discarded. Here he uses a derived quantity called ‘low cloud cover’. He finds a good correlation with GCR, but he chooses to use GCR flux from a Peruvian station known as Huancayo. On its face this is fine, Huancayo is a reputable station and the data should be fine. Trends between GCR stations are very similar, but there are some differences due to GCR interactions with the earths magnetic field and the change of that interaction with latitude. So because Svensmark used Climax data in the 97’ paper, it is a little strange to start using Huancayo data now. Then in 2002 a paper comes out (by Kristjansson again, he works on the satellite cloud data archive in use here - ISCCP), showing much poorer agreement and making the point that there is somewhat of a phase shift (when higher frequency data is looked at), which contradicts known CCN from ion mediated nucleation theories. The paper also brings to light other issues that have been discussed by scientists more informally: 1) Marsh (2000) uses only IR data to arrive at the correlation, not the more common IR/VIS data; 2) low clouds are tough to detect from satellites because they are often blocked by higher clouds; 3) the analysis by Marsh only uses data out to 1995, the Kristjansson (2002) paper shows the correlation deteriorates in the 1995-2001 data analyzed by them.
Then in 2002, Marsh and Svensmark incorporates a calibration change to the satellite data in 1994, that no one else uses. While it is possible that such a change is warranted, and he does present a justification, the ISCCP team does not use such a change. Future papers on this calibration issue also do not suggest that such a change be incorporated.
Other work is still being done in this area. Other low cloud data sets (including ground based observations) are set to the task, none find a significant correlation (e.g. Sun and Bradley Journal of Geophysical Research, 107 (D14): article 4211, 2002). Kristjansson publishes another paper in 2002 suggesting that it is the TSI that correlates better with low cloud cover (mechanism involving stratospheric temperatures influencing gravity waves, which influence tropospheric dynamics). A number of other papers have suggested that there has been an increase in total cloud cover in the last century, or few centuries, and that solar proxies show an increase over that time. This again suggests TSI, and not GCR, as the causative mechanism (remember inverse relation). This is important because it helps informs us of which physical mechanisms should be investigated, and which ruled out.
Another thing to keep in mind is that even if there is a connection between GCR or TSI and low cloud cover, it would still have to be shown that this change was consistent with empirical measures of earth’s climate.
People are now looking at GCR proxies that extend farther into the past and their correlation with climate proxies. Some find correlations, others don’t. Many scientists think that it is the variation in UV radiation that is important, but since there is an inverse relation with GCR flux (mediated through the solar wind/earth’s magnetic field issue mentioned above), it is nearly impossible to tease these issues apart in the absence of a complete mechanism and plenty of modeling studies.
So part of the problem is we lack a well defined mechanism, and an associated model created to test that mechanism. Thus, we will continue to see publications with new mechanisms, and new models, that claim to reproduce all data. These will need to be put through the grinder of the scientific endeavor and should not be taken as consensus.
I have loads of other references, so feel free to ask. Here is a good overall review by the Hadley center, freely available, and with an extensive bibliography.
http://meteo.lcd.lu/globalwarming/Gray/Influence_of_Solar_Changes_HCTN_62.pdf
