The IPCC, 2013 report page 16:
“Equilibrium climate sensitivity is likely in the range 1.5°C to 4.5°C (high confidence), extremely unlikely less than 1°C (high confidence), and very unlikely greater than 6°C (medium confidence).”
“It is extremely likely that more than half of the observed increase in global average surface temperature from 1951 to 2010 was caused by the anthropogenic increase in greenhouse gas concentrations and other anthropogenic forcings together.”
Soon, Connolly and Connolly (“SCC15”) make a very good case that the ECS (equilibrium climate sensitivity) to a doubling of CO2 is less than 0.44°C. In addition, their estimate of the climate sensitivity to variations in total solar irradiance (TSI) is higher than that estimated by the IPCC. Thus, using their estimates, anthropogenic greenhouse gases are not the dominant driver of climate. Supplementary information from Soon, Connolly and Connolly, including the data they used, supplementary figures and their Python scripts can be downloaded here.
It is clear from all satellite data that the Sun’s output varies with sunspot activity. The sunspot cycle averages 11 years, but varies from 8 to 14 years. As the number of sunspots goes up, total solar output goes up and the reverse is also true. Satellite measurements agree that peak to trough the variation is about 2 Watts/m2. The satellites disagree on the amount of total solar irradiance at 1 AU (the average distance from the Earth to the Sun) by 14 Watts/m2, and the reason for this disagreement is unclear, but each satellite shows the same trend over a sunspot cycle (see SCC15 Figure 2 below).
Prior to 1979 we are limited to ground based estimates of TSI and these are all indirect measurements or “proxies.” These include solar observations, especially the character, size, shape and number of sunspots, the solar cycle length, cosmogenic isotope records from ice cores, and tree ring C14 estimates among others. The literature contains many TSI reconstructions from proxies, some are shown below from SCC15 Figure 8:
Those that show higher solar activity are shown on the left, these were not used by the IPCC for their computation of man’s influence on climate. By choosing the low TSI variability records on the right, they were able to say the sun has little influence and the recent warming was mostly due to man. The IPCC AR5 report, in figure SPM.5 (shown below), suggests that the total anthropogenic radiative forcing (relative to 1750) is 2.29 Watts/ m2 and the total due to the Sun is 0.05 Watts/ m2.
Thus, the IPCC believe that the radiative forcing from the Sun is relatively constant since 1750. This is consistent with the low solar variability reconstructions on the right half of SCC15 figure 8, but not the reconstructions on the left half.
The authors of IPCC AR5, “The Physical Science Basis” may genuinely believe that total solar irradiance (TSI) variability is low since 1750. But, this does not excuse them from considering other well supported, peer reviewed TSI reconstructions that show much more variability. In particular they should have considered the Hoyt and Schatten, 1993 reconstruction. This reconstruction, as modified by Scafetta and Willson, 2014 (summary here), has stood the test of time quite well.
The main dataset used to study surface temperatures worldwide is the Global Historical Climatology Network (GHCN) monthly dataset. It is maintained by the National Oceanic and Atmospheric Administration (NOAA) National Climatic Data Center (NCDC). The data can currently be accessed here. There are many problems with the surface air temperature measurements over long periods of time. Rural stations may become urban, equipment or enclosures may be moved or changed, etc.
“…a major survey of the degree to which these [weather station] instrument housings were correctly placed and maintained in the United States was made by a group of 600-odd followers of the blog Climate Audit; … “[In] the best-sited stations, the diurnal temperature range has no century-scale trend” … the relatively small numbers of well-sited stations showed less long-term warming than the average of all US stations. …the gridded mean of all stations in the two top categories had almost no long term trend (0.032°C/decade during the 20th century), Fall, et al., 2011). “
The GHCN data is available from the NCDC in raw form and “homogenized.” The NCDC believes that the homogenized dataset has been corrected for station bias, including the urban heat island effect, using statistical methods. Two of the authors of SCC15, Dr. Ronan Connolly and Dr. Michael Connolly, have studied the NOAA/NCDC US and global temperature records. They have computed a maximum possible temperature effect, due to urbanization, in the NOAA dataset, adjusted for time-of-observation bias, of 0.5°C/century (fully urban – fully rural stations). So their analysis demonstrates that NOAA’s adjustments to the records still leave an urban bias, relative to completely rural weather stations. The US dataset has good rural coverage with 272 of the 1218 stations (23.2%) being fully rural. So, using the US dataset, the bias can be approximated.
The global dataset is more problematic. In the global dataset there are 173 stations with data for 95 of the past 100 years, but only eight of these are fully rural and only one of these is from the southern hemisphere. Combine this with problems due to changing instruments, personnel, siting bias, instrument moves, changing enclosures and the global surface temperature record accuracy is in question. When we consider that the IPCC AR5 estimate of global warming from 1880 to 2012 is 0.85°C +-0.2°C it is easy to see why there are doubts about how much warming has actually occurred.
Further, while the GHCN surface temperatures and satellite measured lower troposphere temperatures more or less agree in absolute value, they have different trends. This is particularly true of the recent Karl, et al., 2015 “Pause Buster” dataset adopted by NOAA this year. The NOAA NCEI dataset from January 2001 trends up by 0.09° C/decade and the satellite lower troposphere datasets (both the RSS and the UAH datasets) trend downward by 0.02° to 0.04° C/decade. Both trends are below the margin of error and are, statistically speaking, zero trends. But, do they trend differently because of the numerous “corrections” as described in SCC15 and Connolly and Connolly, 2014? The trends are so small it is impossible to say for sure, but the large and numerous corrections made by NOAA are very suspicious. Personally, I trust the satellite measurements much more than the surface measurements, but they are shorter, only going back to 1979.
The IPCC computation of man’s effect on climate
Bindoff, et al., 2013 built numerous climate models with four components, two natural and two anthropogenic. The two natural components were volcanic cooling and solar variability. The two anthropogenic components were greenhouse warming due mainly to man-made CO2 and methane, and man-made aerosols which cool the atmosphere. They used the models to hindcast global temperatures from 1860-2012. They found that they could get a strong match with all four components, but when the two anthropogenic components were left out the CMIP5 multi-model mean hindcast only worked from 1860 to 1950. On the basis of this comparison the IPCC’s 5th assessment report concluded:
“More than half of the observed increase in global mean surface temperature (GMST) from 1951 to 2010 is very likely due to the observed anthropogenic increase in greenhouse gas (GHG) concentrations.”
The evidence that the IPCC used to draw this conclusion is illustrated in their Figure 10.1, shown, in part, above. The top graph (a) shows the GHCN temperature record in black and the model ensemble mean from CMIP5 in red. This run includes anthropogenic and natural “forcings.” The lower graph (b) uses only natural “forcings.” It does not match very well from 1961 or so until today. If we assume that their models include all or nearly all of the effects on climate, natural and man-made, then their conclusion is reasonable.
While the IPCC’s simple four component model ensemble may have matched the full GHCN record (the red line in the graphs above) well using all stations, urban and rural, it does not do so well versus only the rural stations:
Again, the CMIP5 model is shown in red. Graph (a) is the full model with natural and anthropogenic “forcings,” (b) is natural only and (c) are the greenhouse forcings only. None of these model runs matches the rural stations, which are least likely to be affected by urban heat or urbanization. The reader will recall that that Bindoff, et al. chose to use the low solar variability TSI reconstructions shown in the right half of SCC15 Figure 8. For a more complete critique of Bindoff, et al. see here (especially section 3).
The Soon, et al. TSI versus the mostly rural temperature reconstruction
So what if one of the high variability TSI reconstructions, specifically the Hoyt and Schatten, 1993
This is Figure 27 from SCC15. In it all of the rural records (China, US, Ireland and the Northern Hemisphere composite) are compared to TSI as computed by Scafetta and Willson. The comparison for all of them is very good for the twentieth century. The rural temperature records should be the best records to use, so if TSI matches them well; the influence of anthropogenic CO2 and methane would necessarily be small. The Arctic warming after 2000 seems to be amplified a bit, this may be the phenomenon referred to as “Polar Amplification.”
Discussion of the new model and the computation of ECS
A least squares correlation between the TSI in Figure 27 and the rural temperature record suggests that a change of 1 Watt/m2 should cause the Northern Hemisphere air temperature to change 0.211°C (the slope of the line). Perhaps, not so coincidentally, we reach a value of 0.209°C assuming that the Sun is the dominant source of heat. That is, if the average temperature of the Earth’s atmosphere is 288K and without the Sun it would be 4K, the difference due to the Sun is 284K. Combining this with an average TSI of 1361 Watts/m2 means that 1/1361 is 0.0735% and 0.0735% of 284 is 0.209°C per Watt/m2. Pretty cool, but this does not prove that TSI dominates climate. It does, however, suggest that Bindoff et al. 2013 might have selected the wrong TSI reconstruction and, perhaps, the wrong temperature record. To me, the TSI reconstruction used in SCC15 and the rural temperature records they used are just as valid as those used by Bindoff et al. 2013. This means that the IPCC and Bindoff’s assertion that anthropogenic greenhouse gases caused more than half of the warming from 1951 to 2010 is questionable. The sound alternative, proposed in SCC15, is just as plausible.
The SCC15 model seems to work, given the data available, so we should be able to use it to compute an estimate of the Anthropogenic Greenhouse Warming (AGW) component. If we subtract the rural temperature reconstruction described above from the model and evaluate the residuals (assuming they are the anthropogenic contribution to warming) we arrive at a maximum anthropogenic impact (ECS or equilibrium climate sensitivity) of 0.44°C for a doubling of CO2. This is substantially less than the 1.5°C to 4.5°C predicted by the IPCC. Bindoff, et al., 2013 also states that it is extremely unlikely (emphasis theirs) less than 1°C. I think, at minimum, this paper demonstrates that the “extremely unlikely” part of that statement is problematic. SCC15’s estimate of 0.44°C is similar to the 0.4°C estimate derived by Idso, 1998. There are numerous recent papers that compute ECS values at the very low end of the IPCC range and even lower. Fourteen of these papers are listed here. They include the recent landmark paper by Lewis and Curry, and the classic Lindzen and Choi, 2011.
Is CO2 dominant or TSI?
SCC15 then does an interesting thought experiment. What if CO2 is the dominant driver of warming? Let’s assume that and compute the ECS. When they do this, they extract an ECS of 2.52°C, which is in the lower half of the range given by the IPCC of 1.5°C to 4.5°C. However, using this methodology results in model-data residuals that still contain a lot of “structure” or information. In other words, this model does not explain the data. Compare the two residual plots below:
The top plot shows the residuals from the model that assumes anthropogenic CO2 is the dominant factor in temperature change. The bottom plot are the residuals from comparing TSI (and nothing else) to temperature change. A considerable amount of “structure” or information is left in the top plot suggesting that the model has explained very little of the variability. The second plot has a little structure left and some of this may be due to CO2, but the effect is very small. This is compelling qualitative evidence that TSI is the dominant influence on temperature and CO2 has a small influence.
The CAGW (catastrophic anthropogenic global warming) advocates are quite adept at shifting the burden of proof to the skeptical community. The hypothesis that man is causing most of the warming from 1951 to 2010 is the supposition that needs to be proven. The traditional and established assumption is that climate change is natural. These hypotheses are plotted below.
This is Figure 31 from SCC15. The top plot (a) shows the northern hemisphere temperature reconstruction (in blue) from SCC15 compared to the atmospheric CO2 concentration (in red). This fit is very poor. The second (b) fits the CO2concentration to the temperature record and then the residuals to TSI, the fit here is also poor. The third plot (c) fits the temperatures to TSI only and the fit is much better. Finally the fourth plot (d) fits the TSI to the temperature record and the residuals to CO2 and the fit is the best.
Following is the discussion of the plots from SCC15:
“This suggests that, since at least 1881, Northern Hemisphere temperature trends have been primarily influenced by changes in Total Solar Irradiance, as opposed to atmospheric carbon dioxide concentrations. Note, however, that this result does not rule out a secondary contribution from atmospheric carbon dioxide. Indeed, the correlation coefficient for Model 4[d] is slightly better than Model 3[c] (i.e., ~0.50 vs. ~0.48). However, as we discussed above, this model (Model 4[d]) suggests that changes in atmospheric carbon dioxide are responsible for a warming of at most ~0.12°C [out of a total of 0.85°C] over the 1880-2012 period, i.e., it has so far only had at most a modest influence on Northern Hemisphere temperature trends.”
This is the last paragraph of SCC15:
“When we compared our new [surface temperature] composite to one of the high solar variability reconstructions of Total Solar Irradiance which was not considered by the CMIP5 hindcasts (i.e., the Hoyt & Schatten reconstruction), we found a remarkably close fit. If the Hoyt & Schatten reconstruction and our new Northern Hemisphere temperature trend estimates are accurate, then it seems that most of the temperature trends since at least 1881 can be explained in terms of solar variability, with atmospheric greenhouse gas concentrations providing at most a minor contribution. This contradicts the claim by the latest Intergovernmental Panel on Climate Change (IPCC) reports that most of the temperature trends since the 1950s are due to changes in atmospheric greenhouse gas concentrations (Bindoff et al., 2013).”
So, SCC15 suggests a maximum ECS for a doubling of CO2 is 0.44°C. They also suggest that of the 0.85°C warming since the late-19th Century only 0.12°C is due to anthropogenic effects, at least in the Northern Hemisphere where we have the best data. This is also a maximum anthropogenic effect since we are ignoring many other factors such as varying albedo (cloud cover, ice cover, etc.) and ocean heat transport cycles.
While the correlation between SCC15’s new temperature reconstruction and the Hoyt and Schatten TSI reconstruction is very good, the exact mechanism of how TSI variations affect the Earth’s climate is not known. SCC15 discusses two options, one is ocean circulation of heat and the other is the transport of heat between the Troposphere and the Stratosphere. Probably both of these mechanisms play some role in our changing climate.
The Hoyt and Schatten TSI reconstruction was developed over 20 years ago and it still seems to work, which is something that cannot be said of any IPCC climate model.
Constructing a surface temperature record is very difficult because this is where the atmosphere, land and oceans meet. It is a space that usually has the highest temperature gradients in the whole system, for example the ocean/atmosphere “skin” effect. Do you measure the “surface” temperature on the ground? One meter above the ground? One inch above the ocean water, at the ocean surface or one inch below the ocean surface in the warm layer? All of these temperatures will always be significantly different at the scales we are talking about, a few tenths of a degree C. The surface of the Earth is never in temperature equilibrium.
“While… [global mean surface temperature] is nothing more than an average over temperatures, it is regarded as the temperature, as if an average over temperatures is actually a temperature itself, and as if the out-of-equilibrium climate system has only one temperature. But an average of temperature data sampled from a non-equilibrium field is not a temperature. Moreover, it hardly needs stating that the Earth does not have just one temperature. It is not in global thermodynamic equilibrium — neither within itself nor with its surroundings.”
“One fundamental flaw in the use of this number is the assumption that small changes in surface air temperature must represent the accumulation or loss of heat by the planet because of the presence of greenhouse gases in the atmosphere and, with some reservations, this is a reasonable assumption on land. But at sea, and so over >70% of the Earth’s surface, change in the temperature of the air a few metres above the surface may reflect nothing more than changing vertical motion in the ocean in response to changing wind stress on the surface; consequently, changes in sea surface temperature (and in air a few metres above) do not necessarily represent significant changes in global heat content although this is the assumption customarily made.”
However, satellite records only go back to 1979 and global surface temperature measurements go back to 1880, or even earlier. The period from 1979 to today is too short to draw any meaningful conclusions given the length of both the solar cycles and the ocean heat distribution cycles. Even the period from 1880 to today is pretty short. We do not have the data needed to draw any firm conclusions. The science is definitely not settled.