Human emissions of carbon dioxide (CO2) are the primary factor contributing to the warming of the Earth’s surface over the past half-century.

However, for the past few years global temperatures have been stagnant or slightly decreasing even as atmospheric CO2 concentrations have been increasing faster than ever. This situation has led some voices in the media and blogging world to challenge the relationship between the CO2 concentrations and warming. These critiques are flawed, however, as short-term changes in global temperature are driven by numerous factors going beyond CO2, and the recent disconnect between the two is not particularly unusual in light of their past relationship.

There is little debate over the basic idea that, all things being equal, higher levels of atmospheric carbon dioxide would warm the planet’s surface. This is a basic result of the radiative properties of CO2 as a greenhouse gas: it absorbs a certain range of longwave radiation wavelengths and reradiates them down to Earth. Solar energy hits the Earth as shortwave radiation like light, and is absorbed and re-emitted as longwave radiation (e.g. heat). CO2 and other atmospheric greenhouse gases absorb this longwave radiation coming up from the Earth’s surface and re-emit a portion of it downwards, heating the Earth. This process, the Greenhouse Effect, is well understood and extensively backed up by empirical laboratory and satellite measurements.

The radiative forcing associated with greenhouse gases generally follows a logarithmic function: The amount of extra heating of the Earth’s surface resulting from an additional amount of CO2 in the atmosphere decreases as there is more CO2 in the atmosphere. This result occurs because parts of the CO2 absorption spectrum become saturated, leading to less additional forcing for additional CO2.

This saturation is never complete, however, as even at astronomically high CO2 concentrations like 20,000 parts per million (ppm) additional CO2 has a slight positive forcing. An easy way to grasp the logarithmic properties of CO2 is to think about it as a fixed temperature increase every time atmospheric concentrations double. So increasing atmospheric CO2 from 250 ppm to 500 ppm would result in a certain amount of warming, and increasing concentrations from 500 to 1000 would result in an additional warming of the same magnitude, all things being equal. In reality, global temperatures respond to various feedbacks and also to the direct forcing from CO2, and these feedbacks may not increase linearly with CO2 forcing.

By itself, doubling atmospheric CO2 would increase global temperatures by about 1.2 degrees C. Even most of the scientists skeptical of the severity of climate change agree on this basic point. The big question is the climate sensitivity of the Earth: how much that 1.2°C worth of forcing from CO2 increases the Earth’s temperature when various positive and negative feedbacks are factored in. These feedbacks include water vapor, clouds, the lapse rate, albedo (the reflectivity of the Earth’s surface), and other factors.

Each of these feedbacks has an estimated range of uncertainty (as shown in the figure below), but in combination they magnify the 1.2°C warming resulting from doubling CO2 into 2 to 4.5°C warming. This is a wide range of possible warming, as 2°C per doubling would not be that terrible, while 4.5°C per doubling would be potentially catastrophic. There is a healthy and impassioned debate in the scientific literature over what the actual climate sensitivity is likely to be.

View larger image
Range of estimated magnitudes of major climate feedbacks from the most recent IPCC report and Colman 2003. Figure taken from Soden and Held 2006 via Realclimate

In addition to feedbacks and greenhouse gas forcings, the Earth’s climate is affected by a number of natural factors such as solar cycles, El Niño Southern Oscillation (ENSO), and large volcanic eruptions that result in stratospheric aerosols. When combined together, these natural factors can have a strong effect on the climate for a few years.

For example, the current strong solar minimum combined with a modest La Niña (the cool phase of ENSO) likely contributed to colder temperatures in 2008. However, these natural factors are largely cyclical: solar cycles switch between minima and maxima, ENSO switches between hot El Niños and cold La Niñas.

Volcanoes are more sporadic and unpredictable, but have only a short-term cooling effect as a result of the limited lifespan of aerosol particles. While these natural factors can combine to create stagnant temperatures over a few years, over the long term their cyclical effects cancel out, leaving us with the underlying warming trend.

Over the past few years, global temperatures have slightly declined. This doesn’t mean that the last few years have been particularly cold; on the contrary, 11 of the past 13 years have been the warmest on record over the past 130 years. These declining temperatures have occurred during a time when atmospheric CO2 was increasing faster than ever, leading to those assertions challenging the connection between CO2 and temperature. The figure below shows atmospheric CO2 and global surface temperature over the past 28 years, with the divergence post-2004 clearly visible.

Global monthly surface temperature data is from HadCRUt and atmospheric CO2 data is from the Mauna Loa Observatory. Temperature data is smoothed using a 12-month running mean.

This figure shows that CO2 and temperature do not necessarily increase in lockstep. CO2 increases every year (with a strong annual cycle resulting largely from changes in hemispheric vegetative photosynthesis and respiration rates caused by seasonal effects), while temperature fluctuates as a result primarily of natural variability. If the divergence between temperature and CO2 over the past few years were unusual, one might maintain that the relationship between the two is not as strong as anticipated, or, more likely, that the recent years have seen natural forcings of a magnitude stronger than that generally modeled by climate scientists.

However, examining the present divergence in light of the past variability between the two factors suggests that the past few years’ divergence is not particularly unusual.

Simple linear regression of atmospheric CO2 concentrations on HadCRUt temperatures from 1980 to 2002 in red, as well as the line of best fit for the data projected out through 2009. 2003 to 2009 observations are shown in blue so they can be evaluated in light of the prior trend without influencing it.

The figure above shows a linear regression of temperature on CO2 concentration, and also the projection one would expect of 2003 to 2009 CO2 concentrations given the relationship between the two factors from 1980 to 2002. This projection allows testing of the post-2003 values consistent with the prior relationship between the two factors, without allowing the post-2003 observations to influence the trend. There is a noticeable dip at the end (and a spike a bit earlier representing the 1998 El Niño event), but it is difficult to see if it is significant, or if it is well within the bounds of the “noisy” relationship between CO2 and temperature.

To actually determine the significance of the recent divergence, one can remove the trend from the data and plot just the difference between the observed value and the expected value.

Residuals of the prior regression of atmospheric CO2 and temperature with the two standard deviation confidence intervals based on 1980 to 2002 observations shown as black lines. A 12-month running mean of the data is shown in red.

Based on the figure above, there are three months in the past five years (January 2008, in particular) that fall below the 95% confidence interval associated with the 1980 to 2002 trend in the relationship between CO2 and temperature. Given that there are 60 months in the past five years, one would expect that 5 percent of them, or exactly three months, would fall outside the 95 percent confidence interval.

So while the past five years have been decidedly below the long-term trend, one cannot say that it is particularly unusual given the past variability in the data. Further illustrating this point, the figure has a 12-month running average of the residuals shown in red (which we would expect to exceed the 95% confidence interval only once every 20 years or so), and this 12-month running mean is still within two standard deviations of the trend.

The declining temperatures and rising atmospheric CO2 concentrations over the past two years is best understood as the result of the natural variability of temperature, without casting any doubt on the underlying long-term relationship between CO2 and temperature.

Considerable uncertainties remain regarding the actual climate sensitivity (whether it is on the low or high side of the uncertainty range), and also the magnitude of short-term cyclical natural variability. Further studies will help diminish remaining uncertainties over the next few years, but for now the evidence of a fundamental long-term positive relationship between atmospheric greenhouse gases and the Earth’s temperature remains strong.

Zeke Hausfather

Zeke is an energy systems analyst and environmental economist with a strong interest in conservation and efficiency. He was previously the chief scientist at C3, an energy management and efficiency company,...