This post was updated August 25, 2015 to clarify the text on sea surface temperatures, provide a correlation coefficient between the two warming periods and to discuss the Planck parameter.

Many writers, including Professor Richard Lindzen and Ed Caryl have noticed the remarkable similarity in global warming observed from around 1910 to 1944 and 1975 to 2009. The similarity in slopes exists in all global surface temperature datasets. Figure 1 shows the HadCRUT version 4 dataset and the NASA GISS land (GHCN v3) and ocean (ERSST v4) temperature dataset. We’ve identified the two periods of interest on the figure. All datasets also show some cooling between 1945 and 1975.

Figure 1

Figure 2 shows the two periods overlain with data from the HadCRUT version 4 dataset. This display is scaled to actual average temperature. Unlike Figure one this figure uses monthly smoothed data. In that way, we can see some of the variation within each year.

Figure 2

The left side of Figure 2 represents 1910 for the blue line and 1975 for the orange line. On average the earlier blue line is 0.36°C cooler than the later line. The later line also has a steeper slope, the earlier represents 0.144°C of warming per decade and the later line shows 0.192°C warming per decade. Figure 3 shows the yearly HadCRUT v4 anomalies from the mean with the two means forced to be the same.


Overlay of the two 20th century HadCRUT warming periods

Figure 3

Now we can easily see the similarity in the two warming periods. The linear correlation coefficient (also called the Pearson product moment correlation coefficient) between the two periods is 0.81, so they are highly correlated.  Yet the IPCC in their AR5 Summary for Policy Makers states on page 17:

“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.”

On page 14 of the Summary for Policy Makers they provide a description of the anthropogenic “radiative forcing” from man’s emissions and other actions. This is shown in Figure 4.

Figure 4

At the bottom of the figure the total IPCC estimated anthropogenic climate radiative forcing is given for three years 1950, 1980 and 2011. The IPCC man-made radiative forcing for 2011 is 4 times the forcing for 1950. According to the IPCC, CO2 and methane (CH4) are the primary influences on climate. Land use change and variations in solar irradiance are very minor in their estimation. Soon, Connolly and Connolly (SCC15) and others have criticized this view and think that solar variability could play a larger role in climate change. One problem is the long term variation in total solar irradiance and in the amount of that radiation that reaches the earth is unknown at this time. Many different estimates have been published. Unfortunately the IPCC only chose four low variability estimates (as identified in SCC15 in their Figure 8) and ignored the others. Further they assume that the only natural influence on climate for the whole period of the IPCC study (roughly 150 years) is the variation in solar radiation, if we ignore episodic volcanism. This assumption has been criticized by Professor Judith Curry and others. Variations in the strength of the sun’s magnetic field, the Earth’s orbit and inclination may be important. Very long term cycles in ocean currents might also be affecting this relatively short 150 year period.

The change in slope between the earlier HadCRUT line and the later line (see Figure 2) is about 0.05°C/decade. The later rate of 0.192°C/decade represents an increase of about 33% in the warming rate. So we are comparing a quadrupling of man’s influence to a 33% change in the rate of warming assuming that the natural forces were the same in both warming periods. It is understandable if this doesn’t make sense to you. Below we discuss this conundrum at more length.

The warming in the early 20th century has always been a bit of a mystery. Attempts to model this warming event have mostly failed. An excellent overview of the peer-reviewed literature on this warming period by Ari Jokimaki can be seen here. Generally it is considered to be natural and roughly equivalent to the warming since 1950, at least in the northern hemisphere and particularly north of 60°N. We have some indications that warming in the United States was more severe in the late 1930’s than today. In particular 1936 has the most US all-time records for daily maximum temperature and 1930 is second.

Measuring the global average surface temperature accurately is problematic. Land based measurements are affected by weather station siting problems and the changing environment around long term weather stations as people have become more urbanized. Attempts at “homogenizing” the temperatures can induce a warming trend because urban areas are warmer than rural areas and many previously rural weather stations have had urban areas surround them over time. In Connolly and Connolly (2014) they point out that the unadjusted US climate network data (their Figure 5) shows that the 1930’s were at least as warm as today. However, once the data are homogenized by the National Centers for Environmental Information, the 1930’s are suddenly cooler (Connolly and Connolly 2014, Figure 20) than today.  Further, most weather stations in the world between 1850 and the present day are in urban areas and Connolly and Connolly argue that most of the urbanization bias remains in the data.  For example, only 24.7% of the GHCN network is fully rural.

Only 30% of the surface of the Earth is on land. Oceans cover the largest area and have a correspondingly larger effect on the average temperature. Here the problem is the ocean skin effect. The temperature difference between the air just above the water, the temperature at the surface of the water and the temperature just below the surface is often large. On “average” the temperature of the mixed layer (roughly the upper 50 meters of the ocean) is very similar and slightly higher than the temperature of the air above the ocean. But, the ocean mixed-layer temperature varies much more slowly due to a higher heat capacity. The mixed layer heat capacity is almost 23 times the heat capacity of the entire atmosphere.

Marine air temperatures are measured, but they are not used in the calculation of global average temperatures due to difficulties in interpreting them.  Instead the global average surface temperature datasets use temperatures measured with sea surface temperature buoys and at the water intakes of ships.  The depth of the water intake ports varies making these measurements problematic.  Older measurements, especially before World War II, include bucket samples.  Bucket samples are taken over the side of a ship.  A thermometer is placed in the retrieved bucket to obtain a water temperature.  All of these methods are perfectly adequate for ballpark estimates of the ocean surface temperature +-2°C or so.  But, we are interested in very small changes in temperature of only +-0.2°C.  None of these methods, with the exception of the buoys, is that accurate.  To make it worse, the highly accurate buoy data has been adjusted to the ship measurements, not the other way around.  This is also explained in Karl, et al 2015.  As a result, as more buoys are deployed the “average” ocean surface temperature goes up artificially because 0.12°C is added to the measurements.  Two hypothetical temperature profiles of the upper ocean are presented in Figure 5.  These are from Dr. Peter Minnet, Rosenstiel School of Marine and Atmospheric Science at the University of Miami.  As you can see the upper layer of the ocean is almost always 0.1K to 0.5K cooler than the immediate subsurface water because of evaporation and the fact that the ocean is normally warmer than the atmosphere.  The temperature difference varies a lot depending upon weather, time of day and cloud cover.

Figure 5

Besides an informative discussion of surface temperatures, SCC15 also provides a new land only northern hemisphere surface temperature dataset based mostly upon long term rural temperature stations. A comparison of the early 20th century and the later 20th century using their dataset is shown in Figure 6.

Figure 6

The difference in the two lines is reduced from the HadCRUT value of 0.36°C to 0.02°C. Probably, this is mostly due to using rural data and minimizing the processing and homogenization. This dataset is also northern hemisphere land only and not directly comparable to the HadCRUT or NASA datasets. Although the means have moved closer together, the difference in the slopes is similar. The HadCRUT increase is 33% and the mostly rural increase is 29%.

Both the HadCRUT v4 and the SCC15 records agree that the rate has increased.

Figure 7

In Figure 7, the NASA GISS data also shows an increase in the slope from the first period to the second. Here it increases 0.0046°C/year or 0.046°C/decade. This is very similar to the increase of 0.048°C/decade for the HadCRUT v4 dataset and not too different from the northern hemisphere, rural, land only difference of 0.07°C/decade observed with the SCC15 dataset. Like the HadCRUT dataset, this one shows a large offset (0.44°C) between the periods.

The ultimate, presumably natural, cause of the early 20th Century warming is unknown. But, Wyatt and Curry have observed and documented a series of cyclical patterns in numerous climatic records that they collectively call a “Stadium Wave.” This wave is illustrated in Figure 8. They believe that these cycles act in concert, like a stadium wave, to form our current natural climate cycle. The reverse could also be true, a single factor may be causing all of these observed effects, but with different time delays.

Figure 8

The climatic records they used include the Atlantic Multidecadal Oscillation (AMO), the Pacific Decadal Oscillation (PDO), and various sea ice records. The curves in Figure 8 are normalized climate indices created from the records. They are presented so that up (positive) is warmer and down (negative) is cooler. The various indices are derived from records of atmospheric, oceanic and sea ice data gathered since 1900. The two most important components turned out to be the Atlantic Multidecadal Oscillation and the sea ice extent in the western Eurasian arctic. Since the depths of the Little Ice Age in the early 19th century, we have been in a period of long term natural warming. The stadium wave periodically enhances or dampens that trend. As the figure shows, from 1910 to 1940 it was enhancing the warming trend. From 1940 to 1970 the trend was dampened, warming resumed in the 1970’s. This corresponds well with the temperature records. For an explanation of the “segments” I, II, III, and IV I refer you to the paper. Figure 9 shows how Wyatt and Curry interpret the various records in terms of climatic effect:

Figure 9

They place the start of early 20th Century warming at about 1918 and the start of the most recent warming at 1976. These dates are not very far from what we picked off of the actual global temperature records. This is a statistical study and it has extracted a cyclic pattern from observations. It does not offer a cause for the pattern.

We can speculate that the natural forces causing the warming trend in the early 20th century are about the same as those acting on us from 1975 to roughly 2009. If this is true, then the increase in warming rate (roughly 30% or 28%-33%) might be due to man’s influence. The extra radiative forcing estimated by the IPCC (bottom of Figure 4, 1950 to 2011) is about 1.72 Watts/m2. They have also estimated that more than half of the warming since 1951 was due to man. No warming occurred between 1945 and 1975, so we are really talking about 1975 to 2009. The increase in the rate of warming from the HadCRUT record is 35 years x 0.0048°C or 0.168°C. The NASA GISS dataset gives us a virtually identical 0.0046°C increase in slope. We assume that the natural influences from 1910 to 1945 were the same as those from 1975 to 2009. We further assume that difference in the two slopes is due to man’s influence. The actual temperature increase from 1975 to 2009, from the best fit line to the HadCRUT record, is 0.672°C. So using our estimate of man’s contribution of 0.168°C, we can estimate that man’s contribution is 25%, much less than half.

SCC15 provides another record based mostly on rural northern hemisphere (land only) weather stations. Temperature swings on land are higher than over the oceans, the high ocean heat capacity has a dampening effect.  Here the difference in the two slopes is 0.0074°C/year. So for 35 years the difference is 0.259°C, a little more than the HadCRUT difference. The total temperature change, from the best fit line, is 1.165°C from 1975 to 2009. SCC15 then suggests that man’s contribution is 22%. Very similar to the estimate using the HadCRUT record.


The temperature records, except for SCC15, and Wyatt and Curry’s stadium wave are presented here as global. But, in reality all of these records are based mostly on northern hemisphere data. We simply have very little climate data for the southern hemisphere prior to 1979 when satellite microwave sounding units were first put in orbit. We have made our estimates of man’s influence on climate by comparing two 35 year periods of time out of a total record of 136 years. Our sole reason for choosing the two periods is that they looked similar and the earlier one was before man could have had much influence on climate. Choosing one short period as our example of a “natural” warming cycle is very speculative. Then comparing it to a later period and assuming that the entire difference is due to man is even more speculative. All we can say is this scenario is plausible given the data we have today. We would need much longer and better records of our climate and the solar climate to reach a firmer conclusion.

But, the same uncertainties exist for the IPCC’s estimate that man is causing more than 50% of current warming and their estimate that man’s radiative forcing is 4 times what it was in 1950. They picked only one natural radiative forcing, variations in solar irradiance (excluding episodic volcanism) and they picked only low variability total solar irradiance (TSI) records. They ignored equally well supported high variability TSI records. In one respect the estimate presented in this paper is superior to the IPCC estimate. In our estimate we used actual data for the calculation. The IPCC estimate of more than 50% is based only on unvalidated computer models. They are unvalidated because they have not successfully predicted the Earth’s climate to date. Therefore their results should not be used in calculations. A detailed description of their calculation can be found in IPCC Report Chapter 10, page 879. A more compact description is half way down “Facts and Theories.” You can see in the IPCC figures 10.1a and 10.1b (see Figure 10) how poorly their model reproduces the warming from 1910-1945. Yet they still ascribe nearly all of the warming from 1950 to 2014 to man. This is illogical.


Figure 10

Given the 20th century temperature record, the IPCC summary is internally inconsistent when it claims that man has increased his radiative forcing on the climate 1.72 Watts/m2 from 1950 to 2011 and has caused more than 50% of the warming since 1951. It is very difficult for both of these statements to be true. A rise of 1.72 Watts/m2 represents a global average temperature increase of 2°C using the conversion (1.18°C per Watt/m2) from section 5.1 of SCC15. But, temperatures have only risen 0.57°C in that period using SCC15’s record and 0.55°C using the HadCRUT record. If we cherry-pick the maximum warming in the period (1955 to 2006) we get a maximum warming of 1.1°C from the SCC15 record. The HadCRUT cherry-picked maximum warming is 0.62°C. So, we can get man’s influence to be over 50%, barely, by assuming no natural warming and using the cherry-picked warming from the SCC15 record. But, this is not reasonable.

The actual warming from 1951 to 2011 is likely under 0.6°C. If we assume the radiative forcing values from IPCC AR5 Chapter 10 and we use the Planck’s parameter from IPCC AR5 Working Group 1, Chapter 9, page 818; we get 1.72 x 0.312 = 0.54°C of warming. Page 818 is a table, the value we want is the inverse of the column labeled “Planck Feedback” which is in (Watts/m2)/°C. The Planck parameter is in °C /(Watts/m2). This parameter is the direct change in temperature for a radiative forcing without any net feedback, positive or negative. Yet, it is barely less than the HadCRUT measured temperature change of 0.55°C. Is the net feedback zero or close to it? If so, why do we need the IPCC at all? Too many uncertainties.
The early 20th century warming is very similar to the warming from 1975 to 2009 and no warming occurred at all from 1945 to 1975. Wyatt and Curry have shown that (statistically) a similar long term climate pattern existed in the two periods. Dr. Roy Spencer also suspects that most of the warming seen in the 20th century is natural and explains this very well in congressional testimony.

The early 20th century warming is very similar to the warming from 1975 to 2009 and no warming occurred at all from 1945 to 1975. Wyatt and Curry have shown that (statistically) a similar long term climate pattern existed in the two periods.

It is very hard to claim that mostly natural forces caused the warming from 1910 to 1945 and mostly man-made forces caused the similar warming from 1975 to 2009. The simplest explanation, given the data before us, is that the natural forces were the same in the two periods. That being said and accepting that man does have some influence on climate today with his CO2 and methane emissions, it seems more likely that our influence is, at most, in the 22% to 25% range. “More than half” is not credible to this observer.