Technical Comments

Comment on “Long-term climate forcing by atmospheric oxygen concentrations”

See allHide authors and affiliations

Science  08 Jul 2016:
Vol. 353, Issue 6295, pp. 132
DOI: 10.1126/science.aad6976

Abstract

Poulsen et al. (Reports, 12 June 2015, p. 1238) argued that lower atmospheric oxygen levels during the Phanerozoic would have given a warmer climate. However, radiative and atmospheric structure changes under lower pressure both cause cooling, making their result unusual in that a hierarchy of models gives opposing results. Scrutiny of how radiative and cloud processes were represented, and a mechanistic explanation of the results, are required.

Poulsen et al. (1) present results from the Global Environmental and Ecological Simulation of Interactive Systems (GENESIS) global climate model (GCM) that show a warmer climate under lower atmospheric pressure. This contrasts to established radiative-convective model (RCM) results in which lower pressure causes cooling (2). Changes to the clear-sky radiation under less pressure have a net cooling effect (in their paper, Poulsen et al. described only the warming effects). Likewise, atmospheric structure changes weaken the greenhouse effect, causing cooling. Yet their GCM model output shows a warming via less cloud reflection. This raises fundamental questions: Are the clear-sky processes correctly represented in GENESIS? If they are, what is the physical mechanism by which a negative clear-sky forcing induces a positive cloud forcing strong enough to reverse the sign of the overall change?

Less atmosphere leads to less Rayleigh (molecular) scattering. This can be treated with a simple scaling analysis, given that the amount of scattering is proportional to the number of molecules in the column and that the scattering cross sections for N2 and O2 are similar. (The scaling with surface density used by Poulsen et al. was erroneous.) To estimate the change in absorbed sunlight, consider a single-layer, scattering only, model atmosphere over a black surface. Let the direct solar beam incident at the top of the atmosphere (TOA) be Idir(t) = I0. Using Beer’s law, the direct beam at the surface is found Idir(s) = I0e−τ, where optical depth (τ) is directly proportional to the number of molecules in the atmosphere. The Rayleigh scattering phase function is symmetrical, so equal amounts of light are scattered up and down and Iscat(s) = Iscat(t) = ½I0(1−e−τ). τ = 0.1 gives a good fit to Earth’s atmosphere [Iscat(t) = 16 W m−2; observations (3) and models (4) show that Rayleigh scattering is 15 W m−2 today]. Decreasing the number of molecules by 11% would correspond to decreasing τ by 11%, reducing the amount of reflected sunlight at the TOA by 1.7 W m−2. This would tend to warm the planet slightly.

In their scaling analysis, Poulsen et al. (1) conflate Mie scattering by clouds with Rayleigh scattering by molecules. Quite obviously, there will be less Rayleigh scattering when there are fewer oxygen molecules. However, there is no a priori justification for the number of cloud droplets scaling with the number of air molecules. The decrease of reflected sunlight of 16 W m−2 suggested by the Poulsen et al. (1) scaling cannot be supported.

The other consequence of fewer molecules is less pressure broadening of the absorption lines of radiatively active species like CO2, so that these absorb less radiation. The natural width of absorption lines is very narrow, but molecular collisions widen them so that more radiation overall is absorbed. Lower palaeopressure would weaken the greenhouse effect and tend to cool the planet. Changes to pressure broadening dominate over Rayleigh scattering, so the net clear-sky radiative effect is negative forcing and cooling. (2)

Mean atmospheric structure is well approximated by a moist adiabat. Given that saturation vapor pressure depends only on temperature, the amount of water vapor does not change with pressure, but less dry air molecules make the saturation mixing ratio of water larger. Thus, lapse rate is lower (less steep), there is less temperature difference between the surface and the atmosphere, and the greenhouse effect is further weakened. (2)

A straightforward way to evaluate the combination of the above is to calculate the radiative forcing, with fixed surface temperature and CO2 inventory but varying O2 abundance (Fig. 1). The radiative transfer here was performed at line-by-line resolution using the Spectral Mapping Atmospheric Radiative Transfer (SMART) model, with absorption coefficients calculated from HITRAN2012 (High-Resolution Transmission Molecular Absorption Database) (5) specifically for these atmospheric profiles, so these results are of reference accuracy. For a decrease in O2 to 10% of the atmosphere, the radiative forcing at the tropopause is −3.0 W m−2, consisting of thermal and solar contributions of −4.4 and +1.4 W m−2 (note the dominance of pressure broadening over Rayleigh scattering). To first approximation, surface temperature change is proportional to radiative forcing. With a typical climate sensitivity of 0.8 K/W m−2), this implies that reducing O2 to 10% would lead to a planetary cooling of 2.4 K.

Fig. 1 Forcing from a change in number of moles of oxygen in the atmosphere relative to standard conditions (nstd).

Colors are red for thermal, green for solar, and black for net. Forcings are positive downward, so a positive forcing warms everything below that level. Atmospheric profiles are constructed as follows. Troposphere (surface to pressure, p = 104 Pa): a moist adiabat is followed from a 289 K surface to 240 K, and from p (240 K), temperature varies linearly with pressure to 206 K at p = 104 Pa, resulting in a profile that approximates a global annual mean (GAM) profile well when surface pressure is 105 Pa. Stratosphere (above p = 104 Pa): GAM profile. The CO2 mixing ratio is adjusted to conserve moles of CO2 between profiles. Water vapor varies linearly with pressure from the surface to the tropopause [similar to Manabe and Wetherald (12)].

Given that a warming is seen in the GCM, one must first ask whether the driving radiative processes are represented correctly. Away from modern conditions, errors in GCM radiative transfer codes are, unfortunately, rather common (6, 7). They have to be heavily parametrized for speed to run a single profile in a fraction of a second (compared to ~20 min for SMART). In palaeoclimate, we take these codes well away from the modern conditions for which they were tuned. Accuracy cannot be presumed, and testing the code in the conditions of interest is a due-diligence step (8). Poulsen et al. (1) did not do this verification for the legacy radiation code, National Center for Atmospheric Research Community Climate Model 3 (CCM3), used in GENESIS.

The standard method for testing GCM radiation codes is off-line comparison to a line-by-line code, comparing radiative fluxes calculated over fixed atmospheric profiles, so that the accuracy of the radiation code may be separated from other model components [see Collins et al. (6) for a thorough discussion of this methodology]. In correspondence with Poulsen et al., before this comment, I compared results of SMART (Fig. 1) and CCM3 (kindly run by the authors), but the authors did not want their results included here.

Next, one asks whether the model cloud response accurately represents the real world. Generally, the problem of modeling change to marine stratus is fraught (9, 10). Circumstantial evidence of a potential problem comes from an intercomparison project that addressed clouds for a different palaeoclimate topic: GENESIS gave cloud radiative forcing 15 to 20 W m−2 higher than any other model, which was linked to how cloud radiative properties were represented (11). These should be independent of climate state, so evidence of anomalous warming from clouds raises substantial concerns.

Climate modeling has developed through a hierarchy of models: For the canonical problem of temperature response to CO2 doubling, all model classes from the original RCMs (12) to modern GCMs give an answer of the same sign and similar magnitude. Departure from this precedent is an exceptional result, which requires exceptional evidence. GCMs have the great utility of resolving the interaction of radiative transfer, dynamics, and the hydrological cycle. However, the results are only meaningful if all individual components are accurate for the conditions in which the model is used. For Poulsen et al.’s response to this Comment, I would lay down two challenges. First, to demonstrate that clear-sky and cloud radiative processes are verified as accurate for the conditions at hand. Second, in the context of other climate forcings indicating cooling, to provide a direct, physically based, mechanistic explanation for why lower pressure should lead to less low cloud.

References and Notes

Acknowledgments: This work was supported by a Natural Sciences and Engineering Research Council of Canada Discovery grant to C.G.
View Abstract

Navigate This Article