by Judith Curry
Three new papers highlight how atmospheric radiative transfer, particularly how it is treated in climate models, is not ‘settled science.’
The greatest uncertainties in simulating climate change from increasing CO2 is generally regarded to be associated with cloud feedbacks and ocean circulations (there are many more, but these stand out). Atmospheric radiative transfer is regarded to be among the most certain aspect of simulating climate change. For some background on this issue, see this previous post: Confidence in radiative transfer models
Three new papers highlight how atmospheric radiative transfer, particularly how it is treated in climate models, is not ‘settled science.’
An assessment of methods for computing radiative forcing in climate models
Eui-Seok Chung and Brian J Soden
Abstract. Because the radiative forcing is rarely computed separately when performing climate model simulations, several alternative methods have been developed to estimate both the instantaneous (or direct) forcing and the adjusted forcing. The adjusted forcing accounts for the radiative impact arising from the adjustment of climate variables to the instantaneous forcing, independent of any surface warming. Using climate model experiments performed for CMIP5, we find the adjusted forcing for 4 × CO2 ranges from roughly 5.5–9 W m−2 in current models. This range is shown to be consistent between different methods of estimating the adjusted forcing. Decomposition using radiative kernels and offline double-call radiative transfer calculations indicates that the spread receives a substantial contribution (roughly 50%) from intermodel differences in the instantaneous component of the radiative forcing. Moreover, nearly all of the spread in adjusted forcing can be accounted for by differences in the instantaneous forcing and stratospheric adjustment, implying that tropospheric adjustments to CO2 play only a secondary role. This suggests that differences in modeling radiative transfer are responsible for substantial differences in the projected climate response and underscores the need to archive double-call radiative transfer calculations of the instantaneous forcing as a routine diagnostic.
Published in Environmental Research Letters [link]
I find this to be pretty astonishing: The adjusted forcing for 4 × CO2 ranges from roughly 5.5–9 W m−2 in current models.
Some interesting insights from this paragraph:
Our assessment of the intermodel spread in the instantaneous forcing from CO2 is similar to that obtained by Collins et al (2006) for both the shortwave and longwave components. Collins et al (2006) documented that at the top of model the range of instantaneous forcing for a doubling of CO2 is ∼1.2Wm−2 for the longwave part of the electromagnetic spectrum and ∼0.5Wm−2 for the shortwave part. These ranges, respectively, correspond to ∼2.4Wm−2 and ∼1.0Wm−2 for a quadrupling of CO2 if the curve of growth of forcing withCO2 holds. This agreement further supports the validity of the kernel methodology. The spread is significantly larger than that obtained by Collins et al using line-by-line calculations, indicating that the spread in forcing calculations does not reflect uncertainties in radiative transfer theory, but in the fidelity of its implementation in climate models.
These discrepancies are surprisingly large when considered as a fraction of the total forcing.
Back when I was deeply involved in radiative transfer research (1990’s), it was clear that many climate models were using substandard and erroneous radiative transfer codes, largely because radiation codes are very computationally intensive. I imagine the situation has improved, but apparently there are still substantial issues. For an evaluation of climate model radiation codes against observations, see this paper The continual inter comparison of radiation codes: Assessing anew the quality of GCM radiation algorithms.
———–
The next paper, while not a new one (published in 2006) gives an example of something missing from radiative transfer models (someone tweeted this paper last week, bringing it to my attention).
Parameterization of the Absorption of the H2O Continuum, CO2, O2, and Other Trace Gases in the Fu-Liou Solar Radiation Program
ZHANG Feng, ZENG Qingcun, Y. GU, and K. N. LIOU
Abstract. The absorption properties of the water vapor continuum and a number of weak bands for H2O, O2, CO2, CO, N2O, CH4, and O3 in the solar spectrum are incorporated into the Fu-Liou radiation parameterization program by using the correlated k-distribution method (CKD) for the sorting of absorption lines. The overlap absorption of the H2O lines and the H2O continuum (2500–14500 cm−1) are treated by taking the two gases as a single-mixture gas in transmittance calculations. Furthermore, in order to optimize the computation efforts, CO2 and CH4 in the spectral region 2850–5250 cm−1 are taken as a new singlemixture gas as well. For overlap involving other absorption lines in the Fu-Liou spectral bands, the authors adopt the multiplication rule for transmittance computations under which the absorption spectra for two gases are assumed to be uncorrelated. Compared to the line-by-line (LBL) computation, it is shown that the errors in fluxes introduced by these two approaches within the context of the CKD method are small and less than 0.48% for the H2O line and continuum in the 2500–14500 cm−1 solar spectral region, 1% for H2O (line)+H2O (continuum)+CO2+CH4 in the spectral region 2850–5250 cm−1, and 1.5% for H2O (line)+H2O (continuum)+O2 in the 7700–14500 cm−1 spectral region. Analysis also demonstrates that the multiplication rule over a spectral interval as wide as 6800 cm−1 can produce acceptable errors with a maximum percentage value of about 2% in reference to the LBL calculation. Addition of the preceding gases increases the absorption of solar radiation under all sky conditions. For clear sky, the increase in instantaneous solar absorption is about 9%–13% (12Wm−2) among which the H2O continuum produces the largest increase, while the contributions from O2 and CO2 rank second and third, respectively. In cloudy sky, the addition of absorption amounts to about 6–9 W m−2. The new, improved program with the incorporation of the preceding gases produces a smaller solar absorption in clouds due to the reduced solar flux reaching the cloud top.
Published in Advances in Atmospheric Science [link] …
The Introduction provides some context:
The absorption of a number of gases in the earth’s atmosphere makes an important contribution to the radiation budget of the Earth-atmosphere system. In the discussion of solar absorption, Liou (2002) presented numerous solar absorption bands of gases that have not been properly accounted for in radiation parameterizations. These include absorption lines associated with H2O, CO2, O3, O2, N2O, CH4, CO, and NO2. It is noted that only the major absorbers (H2O near-infrared bands, O2, CO2 near-infrared bands, and O3 UV and visible bands) have been considered in the radiative transfer parameterizations in the majority of current general circulation models (GCMs). A common feature in most GCMs to date has shown that the simulated net solar fluxes at the top of the atmosphere (TOA) are smaller than the observed values, indicating a cold bias in the GCMs (Gu et al., 2003). Introducing the neglected absorbers in the radiation model can correct this cold bias and at the same time improve the performance of the GCMs.
From the conclusions:
Under all sky conditions, the new version of the Fu- Liou radiation parameterization has produced larger solar absorption than the original one. Contribution from the absorption of the H2O continuum is most important, followed by O2, CO2, the H2O visible band, and CH4. The contributions of N2O, the O3 3.3 μm band and CO on solar absorption are quite small and can be neglected for most practical applications. In cloudy sky, the new version has generated a smaller solar absorption in the cloud due to less solar flux reaching the cloud top. Finally, it is our intent to integrate the new version of the Fu-Liou radiation program into the IAP AGCM II to determine the contributions of the preceding absorption bands to the heating of the Earth-atmosphere system for climate study.
The Fu-Liou radiation code is a state-of-the-art research quality radiation code. It is very computationally intensive, and hence not used in climate models for production runs for CMIP/IPCC. Although I did spot this presentation where the Fu-Liou code was incorporated into the UCLA GCM, although it doesn’t directly address the issues raised in the 2006 paper. The presentation is worth reading since it highlights additional uncertainties in radiative transfer modeling.
——-
The final paper is a Ph.D. thesis from Germany, the most interesting thing I’ve read in a long time, and it takes me back to my own thesis (on radiative transfer in the Arctic [link])
Antarctic specific features of the greenhouse effect
Holger Schmithusen
Abstract. CO2 is the strongest anthropogenic forcing agent for climate change since pre-industrial times. Like other greenhouse gases, CO2 absorbs terrestrial surface radiation and causes emission from the atmosphere to space. As the surface is generally warmer than the atmosphere, the total long-wave emission to space is commonly less than the surface emission. However, this does not hold true for the high elevated areas of central Antarctica. For this region, it is shown that the greenhouse effect of CO2 is around zero or even negative. Moreover, for central Antarctica an increase in CO2 concentration leads to an increased long-wave energy loss to space, which cools the earth-atmosphere system. These unique findings for central Antarctica are in contrast to the well known general warming effect of increasing CO2. The work contributes to explain the non-warming of central Antarctica since 1957.
PhD thesis from the Alfred Wegener Institut, Universitat Bremen [link].
Current explanations for the cooling in Antarctica include ozone hole and the Southern Annular Mode. This thesis argues for a negative greenhouse effect (GHE), whereby there is more radiation emitted from the top of the atmosphere over Antarctica than by the surface. The reason this happens is because of a combination of temperature inversions (the temperature increases with height in the lower atmosphere over Antarctica) and the high elevation. The negative GHE is most prominent in austral autumn, because the stratosphere is still warm while the surface is cold. In spring the stratosphere is warmed up rapidly by the absorption of ozone, while the surface has just started to recover from its winter temperature, causing a strong negative GHE in October.
Schmithusen explains the effect:
The term negative GHE might seem to sound odd, as we think of GHGs to act like a blanket for the planet, shielding terrestrial radiation from being emitted to space. “Anti-shielding” does not make sense. The following thought experiment demonstrates that GHGs can actually help the planet to lose energy, that would not be emitted without them:
Say, there were no GHGs in the Earth’s atmosphere. Clouds shall be neglected as well, to make things easier. The planet gains energy over the tropics (positive budget) and loses this extra energy over the poles (negative budget). The energy transport in-between is carried out by the atmosphere. The ocean, of course, also contributes to this meridional transport of energy, but this is not of importance here.
The energy gained over the tropics, which is then transported to the poles, must enter the ground in the polar regions before it can be emitted to space. This is because no GHGs and no clouds, also no aerosol, shall be contained in this hypothetical atmosphere. The atmosphere cannot emit energy directly to space, as it lacks long-wave emitters. Consequently, any “imported” energy that shall leave the Earth-atmosphere system in the polar regions, must be transported via sensible heat flux into the ground. From there it can then be emitted to space.
Now, GHGs shall be introduced. Sure, they have a “shielding” effect over the tropics by causing long-wave downwelling radiation to heat the surface. The same happens, to some smaller extent though, in the polar regions. In addition to that, GHGs give the atmosphere the ability to emit energy directly into space, without the need to transport it through the surface first. This increases the ability of the planet to get rid of energy at the poles, which has been collected over the tropics. In essence, this helps the atmosphere to perform its “task” of meridional energy transport; GHGs help to balance the radiative imbalance between the tropics and the poles.
The conditions in central Antarctica, being a high-altitude plateau and having a continental climate, are such, that the “shielding” effect of GHGs is excelled by the “helping in losing energy” effect. This, one can name negative greenhouse effect.
From the concluding section:
A better linkage between the reported phenomena and the widely discussed surface temperature can be provided from analyses of GCM results. For this, it is crucial that the surface temperatures on the Antarctic plateau are modelled correctly. The CMIP5 comparison shown here demonstrates that this is not the case for many state-of-the-art climate models: most models evaluated here overestimate the surface temperature. Consequently, many models do not reproduce the observed negative GHE over central Antarctica. Furthermore, GCM analyses shall ensure that the surface temperature inversion is correctly reproduced. Both the strength and the height of the inversion influence the changes in LWD caused by increasing GHGs. If the surface inversion is too weak in a model, the increase of LWD caused by increasing GHGs will be overestimated.
Further observational proof of the phenomena reported here could be gained from long-term analysis of TOA thermal infrared emission spectra. Satellite records of such measurements date back to the launch of the Nimbus 4 satellite in 1970. Given the comparability of the different sensors, that have been in space since then, and given sufficient data coverage, a correlation of GHE of CO2 over central Antarctica with the atmospheric CO2 concentration should be feasible. This kind of analysis is expected to resemble the results of RF of CO2 presented here, essentially showing no or slightly negative correlation.
JC comment: this is a very readable and informative thesis, it is well worth reading for anyone interested in radiative transfer. A nice thing about Ph.D. theses is that they really explain things. It is also a potentially important thesis, although I may be biased since this makes me nostalgic about my own Ph.D. thesis completed over 30 yrs ago.
————
JC reflections
Back in the 1990’s, when radiative transfer was a central research topic for me, we didn’t worry too much about the details of radiative transfer codes in climate models, because any errors were swamped by the errors in the modeled distribution of clouds, which had a much greater impact on the planetary energy balance than did any errors in the radiative transfer code.
But the issue is that all of the errors/uncertainties highlighted in the above 3 papers are systematic errors in a given model, directly giving rise to errors in sensitivity to CO2, although it is difficult to infer anything quantitatively re climate sensitivity from these papers.
While the state of understanding of atmospheric radiative transfer is pretty high, there remain some significant uncertainties and unaddressed problems. The bigger issue is the slow translation of this understanding into the radiation codes used in climate models. I recall ECMWF was using a neural network approach based on a sophisticated radiative transfer model, this seems a promising approach.
I spent the 1990’s working on issues related to radiative transfer in the Arctic. A key issue in polar regions is the ‘dirty window’ in the far infrared around 20-30 microns (fig 2.20 in the Antarctic thesis illustrates this). The dirty window is also an issue in the upper troposphere. Getting this wrong in your climate model will cause all sorts of problems, including too much heating in the polar regions. I know a few climate models that treat the dirty window in a reasonable way, but I suspect that most don’t.
It is an important but fully tractable challenge to bring climate model radiation codes up to the level of our understanding that is reflected by state-of-the-art radiation codes such as Fu-Liou.
Moderation note: This is NOT the thread to discuss theories of the greenhouse effect or its nonexistence. Keep such discussion on the Week in Review thread.
Filed under: climate models, Greenhouse effect