Atmospheric CO2: Principal Control Knob Governing Earth’s Temperature

See allHide authors and affiliations

Science  15 Oct 2010:
Vol. 330, Issue 6002, pp. 356-359
DOI: 10.1126/science.1190653

This article has a correction. Please see:


Ample physical evidence shows that carbon dioxide (CO2) is the single most important climate-relevant greenhouse gas in Earth’s atmosphere. This is because CO2, like ozone, N2O, CH4, and chlorofluorocarbons, does not condense and precipitate from the atmosphere at current climate temperatures, whereas water vapor can and does. Noncondensing greenhouse gases, which account for 25% of the total terrestrial greenhouse effect, thus serve to provide the stable temperature structure that sustains the current levels of atmospheric water vapor and clouds via feedback processes that account for the remaining 75% of the greenhouse effect. Without the radiative forcing supplied by CO2 and the other noncondensing greenhouse gases, the terrestrial greenhouse would collapse, plunging the global climate into an icebound Earth state.

It often is stated that water vapor is the chief greenhouse gas (GHG) in the atmosphere. For example, it has been asserted that “about 98% of the natural greenhouse effect is due to water vapour and stratiform clouds with CO2 contributing less than 2%” (1). If true, this would imply that changes in atmospheric CO2 are not important influences on the natural greenhouse capacity of Earth, and that the continuing increase in CO2 due to human activity is therefore not relevant to climate change. This misunderstanding is resolved through simple examination of the terrestrial greenhouse.

The difference between the nominal global mean surface temperature (TS = 288 K) and the global mean effective temperature (TE = 255 K) is a common measure of the terrestrial greenhouse effect (GT = TSTE = 33 K). Assuming global energy balance, TE is also the Planck radiation equivalent of the 240 W/m2 of global mean solar radiation absorbed by Earth.

The Sun is the source of energy that heats Earth. Besides direct solar heating of the ground, there is also indirect longwave (LW) warming arising from the thermal radiation that is emitted by the ground, then absorbed locally within the atmosphere, from which it is re-emitted in both upward and downward directions, further heating the ground and maintaining the temperature gradient in the atmosphere. This radiative interaction is the greenhouse effect, which was first discovered by Joseph Fourier in 1824 (2), experimentally verified by John Tyndall in 1863 (3), and quantified by Svante Arrhenius in 1896 (4). These studies established long ago that water vapor and CO2 are indeed the principal terrestrial GHGs. Now, further consideration shows that CO2 is the one that controls climate change.

CO2 is a well-mixed gas that does not condense or precipitate from the atmosphere. Water vapor and clouds, on the other hand, are highly active components of the climate system that respond rapidly to changes in temperature and air pressure by evaporating, condensing, and precipitating. This identifies water vapor and clouds as the fast feedback processes in the climate system.

Radiative forcing experiments assuming doubled CO2 and a 2% increase in solar irradiance (5) show that water vapor provides the strongest climate feedback of any of the atmospheric GHGs, but that it is not the cause (forcing) of global climate change. The response of the climate system to an applied forcing is determined to be the sum of the direct (no-feedback) response to the applied forcing and the induced radiative response that is attributable to the feedback process contributions. The ratio of the total climate response to the no-feedback response is commonly known as the feedback factor, which incorporates all the complexities of the climate system feedback interactions. For the doubled CO2 and the 2% solar irradiance forcings, for which the direct no-feedback responses of the global surface temperature are 1.2° and 1.3°C, respectively, the ~4°C surface warming implies respective feedback factors of 3.3 and 3.0 (5).

Because the solar-thermal energy balance of Earth [at the top of the atmosphere (TOA)] is maintained by radiative processes only, and because all the global net advective energy transports must equal zero, it follows that the global average surface temperature must be determined in full by the radiative fluxes arising from the patterns of temperature and absorption of radiation. This then is the basic underlying physics that explains the close coupling that exists between TOA radiative fluxes, the greenhouse effect, and the global mean surface temperature.

An improved understanding of the relative importance of the different contributors to the greenhouse effect comes from radiative flux experiments that we performed using Goddard Institute for Space Studies (GISS) ModelE (6). Figure 1 depicts the essence of these calculations, including the separation of the greenhouse contributors into feedback and forcing categories.

Fig. 1

Attribution of the contributions of individual atmospheric components to the total terrestrial greenhouse effect, separated into feedback and forcing categories. Horizontal dotted and dashed lines depict the fractional response for single-addition and single-subtraction of individual gases to an empty or full-component reference atmosphere, respectively. Horizontal solid black lines are the scaled averages of the dashed- and dotted-line fractional response results. The sum of the fractional responses adds up to the total greenhouse effect. The reference atmosphere is for conditions in 1980.

In round numbers, water vapor accounts for about 50% of Earth’s greenhouse effect, with clouds contributing 25%, CO2 20%, and the minor GHGs and aerosols accounting for the remaining 5%. Because CO2, O3, N2O, CH4, and chlorofluorocarbons (CFCs) do not condense and precipitate, noncondensing GHGs constitute the key 25% of the radiative forcing that supports and sustains the entire terrestrial greenhouse effect, the remaining 75% coming as fast feedback contributions from water vapor and clouds.

We used the GISS 4° × 5° ModelE to calculate changes in instantaneous LW TOA flux (annual global averages) in experiments where atmospheric constituents (including water vapor, clouds, CO2, O3, N2O, CH4, CFCs, and aerosols) were added to or subtracted from an equilibrium atmosphere with a given global temperature structure, one constituent at a time for a 1-year period. Decreases in outgoing TOA flux for each constituent relative to the empty or the full-component atmosphere define the bounds for the relative impact on the total greenhouse effect. Had the overlapping absorption been negligible, the sum of the flux differences would have been equal to the LW flux equivalent of the total greenhouse effect (GF = σT4S – σT4E = 150 W/m2), where σ is the Stefan-Boltzmann constant. We found the single-addition flux differences to be overestimated by a factor of 1.36, whereas in the single-subtraction cases, the sum of the TOA flux differences was underestimated by a factor of 0.734. By normalizing these fractional contributions to match the full-atmosphere value of GF, we obtained the fractional response contributions shown in Fig. 1.

Because of overlapping absorption, the fractional attribution of the greenhouse effect is to some extent qualitative (as shown by the dashed and dotted extremum lines in Fig. 1), even though the spectral integral is a full and accurate determination of the atmospheric greenhouse strength for the specified global temperature structure. Still, the fractional attribution is sufficiently precise to clearly differentiate the radiative flux contributions due to the noncondensable GHGs from those arising from the fast feedback processes. This allows an empirical determination of the climate feedback factor as the ratio of the total global flux change to the flux change that is attributable to the radiative forcing due to the noncondensing GHGs. This empirical determination leads then to a climate feedback factor of 4, based on the noncondensing GHG forcing accounting for 25% of the outgoing flux reduction at the TOA for the full-constituent atmosphere. This implies that Earth’s climate system operates with strong positive feedback that arises from the forcing-induced changes in the condensable species.

A direct consequence of this combination of feedback by the condensable and forcing by the noncondensable constituents of the atmospheric greenhouse is that the terrestrial greenhouse effect would collapse were it not for the presence of these noncondensing GHGs. If the global atmospheric temperatures were to fall to as low as TS = TE, the Clausius-Clapeyron relation would imply that the sustainable amount of atmospheric water vapor would become less than 10% of the current atmospheric value. This would result in (radiative) forcing reduced by ~30 W/m2, causing much of the remaining water vapor to precipitate, thus enhancing the snow/ice albedo to further diminish the absorbed solar radiation. Such a condition would inevitably lead to runaway glaciation, producing an ice ball Earth.

Claims that removing all CO2 from the atmosphere “would lead to a 1°C decrease in global warming” (7), or “by 3.53°C when 40% cloud cover is assumed” (8) are still being heard. A clear demonstration is needed to show that water vapor and clouds do indeed behave as fast feedback processes and that their atmospheric distributions are regulated by the sustained radiative forcing due to the noncondensing GHGs. To this end, we performed a simple climate experiment with the GISS 2° × 2.5° AR5 version of ModelE, using the Q-flux ocean with a mixed-layer depth of 250 m, zeroing out all the noncondensing GHGs and aerosols.

The results, summarized in Fig. 2, show unequivocally that the radiative forcing by noncondensing GHGs is essential to sustain the atmospheric temperatures that are needed for significant levels of water vapor and cloud feedback. Without this noncondensable GHG forcing, the physics of this model send the climate of Earth plunging rapidly and irrevocably to an icebound state, though perhaps not to total ocean freezeover.

Fig. 2

Time evolution of global surface temperature, TOA net flux, column water vapor, planetary albedo, sea ice cover, and cloud cover, after the zeroing out of the noncondensing GHGs. The model used in the experiment is the GISS 2°× 2.5° AR5 version of ModelE, with the Q-flux ocean and a mixed-layer depth of 250 m. Model initial conditions are for a preindustrial atmosphere. Surface temperature and TOA net flux use the lefthand scale.

The scope of the climate impact becomes apparent in just 10 years. During the first year alone, global mean surface temperature falls by 4.6°C. After 50 years, the global temperature stands at –21°C, a decrease of 34.8°C. Atmospheric water vapor is at ~10% of the control climate value (22.6 to 2.2 mm). Global cloud cover increases from its 58% control value to more than 75%, and the global sea ice fraction goes from 4.6% to 46.7%, causing the planetary albedo of Earth to also increase from ~29% to 41.8%. This has the effect of reducing the absorbed solar energy to further exacerbate the global cooling.

After 50 years, a third of the ocean surface still remains ice-free, even though the global surface temperature is colder than –21°C. At tropical latitudes, incident solar radiation is sufficient to keep the ocean from freezing. Although this thermal oasis within an otherwise icebound Earth appears to be stable, further calculations with an interactive ocean would be needed to verify the potential for long-term stability. The surface temperatures in Fig. 3 are only marginally warmer than 1°C within the remaining low-latitude heat island.

Fig. 3

Zonally averaged annual mean surface temperature change after the zeroing out of noncondensing GHGs.

From the foregoing, it is clear that CO2 is the key atmospheric gas that exerts principal control over the strength of the terrestrial greenhouse effect. Water vapor and clouds are fast-acting feedback effects, and as such are controlled by the radiative forcings supplied by the noncondensing GHGs. There is telling evidence that atmospheric CO2 also governs the temperature of Earth on geological time scales, suggesting the related question of what the geological processes that control atmospheric CO2 are. The geological evidence of glaciation at tropical latitudes from 650 to 750 million years ago supports the snowball Earth hypothesis (9), and by inference, that escape from the snowball Earth condition is also achievable.

On million-year time scales, volcanoes are the principal source of atmospheric CO2, and rock weathering is the principal sink, with the biosphere acting as both source and sink (10). Because the CO2 sources and sinks operate independently, the atmospheric level of CO2 can fluctuate. If the atmospheric CO2 level were to fall below its critical value, snowball Earth conditions can result.

Antarctic and Greenland ice core data show atmospheric CO2 fluctuations between 180 to 300 parts per million (ppm) over the glacial-interglacial cycles during the past 650,000 years (11). The relevant physical processes that turn the CO2 control knob on thousand-year time scales between glacial and interglacial extremes are not fully understood, but appear to involve both the biosphere and the ocean chemistry, including a significant role for Milankovitch variations of the Earth-orbital parameters.

Besides CO2, methane is another potent greenhouse control knob, being implicated in the Paleocene-Eocene thermal maximummass extinction 55 million years ago, when global warming by up to 5°C (12) occurred because of a massive release of methane from the disintegration of seafloor clathrates (13, 14). Methane is the second most important noncondensing GHG after CO2. Of the 2.9 W/m2 of GHG radiative forcing from 1750 to 2000, CO2 contributed 1.5 W/m2, methane 0.55 W/m2, and CFCs 0.3 W/m2, with the rest coming from N2O and ozone (15). All of these increases in noncondensing GHG forcing are attributable to human activity (16).

Climate control knobs on the solar side of the energy balance ledger include the steady growth in luminosity since the beginning of the Solar System (from about 70% of present luminosity, depending on the postulated early solar mass loss), as hydrogen is consumed in nuclear reactions in the solar interior (17, 18). Milankovitch variations of the Earth-orbital parameters, which alter the relative seasonal distribution as well as the intensity of incident solar radiation within the polar regions, are another important solar energy control knob that is intimately associated with glacial-interglacial cycles of climate change. For solar irradiance changes over the past several centuries, an increase by about 0.1 W/m2 is inferred since the time of the Maunder minimum, based on trends in sunspot activity and other proxies (19).

Of the climate control knobs relevant to current climate, those on the solar side of the energy balance ledger show only negligible impact. Several decades of solar irradiance monitoring have not detected any long-term trends in solar irradiance beyond the 11-year oscillation associated with the solar sunspot cycle. Large volcanic eruptions can happen at any time, but no substantial eruptions have occurred since the eruption of Mt. Pinatubo in the Philippines in 1991.

In a broader perspective, CO2 greenhouses also operate on Mars and Venus, because both planets possess atmospheres with substantial amounts of CO2. The atmospheric greenhouse effect requires that a substantial fraction of the incident solar radiation must be absorbed at the ground in order to make the indirect greenhouse heating of the ground surface possible. Greenhouse parameters and relative surface pressure (PS) for Mars, Earth, and Venus are summarized in Table 1.

Table 1

Planetary greenhouse parameters.

View this table:

Earth is unique among terrestrial planets in having a greenhouse effect in which water vapor provides strong amplification of the heat-trapping action of the CO2 greenhouse. Also, N2 and O2, although possessing no substantial absorption bands of their own, are actually important contributors to the total greenhouse effect because of pressure-broadening of CO2 absorption lines, as well as by providing the physical structure within which the absorbing gases can interact with the radiation field.

The anthropogenic radiative forcings that fuel the growing terrestrial greenhouse effect continue unabated. The continuing high rate of atmospheric CO2 increase is particularly worrisome, because the present CO2 level of 390 ppm is far in excess of the 280 ppm that is more typical for the interglacial maximum, and still the atmospheric CO2 control knob is now being turned faster than at any time in the geological record (20). The concern is that we are well past even the 300- to 350-ppm target level for atmospheric CO2, beyond which dangerous anthropogenic interference in the climate system would exceed the 25% risk tolerance for impending degradation of land and ocean ecosystems, sea-level rise, and inevitable disruption of socioeconomic and food-producing infrastructure (21, 22). Furthermore, the atmospheric residence time of CO2 is exceedingly long, being measured in thousands of years (23). This makes the reduction and control of atmospheric CO2 a serious and pressing issue, worthy of real-time attention.

Supporting Online Material

SOM Text

Figs. S1 and S2

Table S1


References and Notes

  1. We thank B. Carlson, A. Del Genio, J. Hansen, G. Russell, R. Stothers, and L. Travis for comments and the NASA Earth Science Research Division managed by J. Kaye and D. Considine for support.

Stay Connected to Science

Navigate This Article