I had been intending to write a post about aerosols, given what’s been said about them recently. In particular, Nic Lewis at the Parliamentary Select Committee said a couple of things that I wanted to check. Essentially he was arguing that the aerosol estimates had changed from AR4 to AR5 and that climate models were using an aerosol forcing that was too high (negative) and hence over-estimating climate sensitivity.

Fortunately, someone else has done the job for me. Will Morgan has written what seems to be a very informative and clear post about what did the IPCC say about aerosols. I encourage you to read it, but I’ll my summarise my understanding below. This isn’t so that you can learn from me (read Will Morgan’s post), it’s so that those more informed than me can correct my understanding, so that I might learn something.

  • Anthropogenic aerosols can influence the radiation budget in a number of ways : they can scatter incoming sunlight and produce cooling, they can seed clouds also producing cooling, and dark aerosols can absorb incoming radiation and produce warming.

  • Overall it is thought to be that aerosols produce a net cooling. The exact amount (i.e., the radiative forcing) is, however, very uncertain.

  • Determining the net radiative forcing due to aerosols is a crucial part of estimating climate sensitivity using recent observations (energy budget constraints). The large uncertainty, however, means that such estimates are likely to be very uncertain.

  • There are two ways in which the aerosol forcing is estimated. Satellite based measurements (giving a value of -0.85 Wm-2 with a range from -0.93 Wm-2 to -0.45 Wm-2) and using a subset of climate models (giving -1.38 Wm-2 with a range from -1.68 Wm-2 to -0.81 Wm-2).

  • The value presented in AR5 was based on expert judgement – using the two methods described above – and was -0.9 Wm-2 with a range from -1.90 Wm-2 to -0.1 Wm-2. The aerosol forcing in AR5 was reduced slightly compared to AR4 (-1.2 Wm-2 with a range from -2.30 Wm-2 to -0.2 Wm-2).

  • Here’s the kicker (I think) – the claim, by Nic Lewis and others, that climate models are using aerosol forcings that are higher (more negative) than the IPCC best estimate appears to be based only on the climate models used to estimate the aerosol forcing. Therefore, I don’t believe that the claim that climate models – overall – have too high an aerosol forcing is necessarily correct.

The table below gives the various estimates and the change from AR4 to AR5

Aerosol radiative forcing estimates (credit : Will Morgan)

Aerosol radiative forcing estimates (credit : Will Morgan)

So, my basic summary is that the anthropogenic aerosol radiative forcing is uncertain and hence means that energy budget estimates of the climate sensitivity will also be very uncertain (possibly to the extent that they’re not particularly useful). The range has changed from AR4 to AR5, but not by much. The claim that climate models are using aerosols forcings that are too high is confusing the subset of models used to estimate the aerosol forcing with all climate models, and hence isn’t – strictly speaking – correct. Corrections and clarifications through the comments are welcome though.

This entry was posted in Climate change, Climate sensitivity, IPCC, Science and tagged , , , , , , , . Bookmark the permalink.

20 Responses to Aerosols

  1. OPatrick says:

    Is there a good understanding of local variation in aerosols? How much does this affect the measurements and modelling results and are they affected by it in different ways? (Sorry to be adding more questions rather than any answers.)

    I see the IPCC observe:

    Cloud and aerosol properties vary at scales significantly smaller than those resolved in climate models, and cloud-scale processes respond to aerosol in nuanced ways at these scales.

  2. OPatrick,
    I don’t know the answer, unfortunately, but I vaguely remember someone criticising a paper because they were using very outdated regional aerosol distributions.

    You’ve reminded me that there was something I meant to add to the post, but forgot. My understanding is that aerosols tend to precipitate out fairly quickly. Therefore, even if we don’t reduce aerosol emissions, I don’t think the aerosol forcing can increase (negatively) at the same rate as the GHG forcings (assuming we follow a BAU scenario). Therefore, the cooling influence should reduce with time. Furthermore, I think that we expect aerosol emissions to reduce as we become more energy efficient, so we should expect the cooling influence to reduce and the net anthropogenic forcing to increase as a result.

  3. johnrussell40 says:

    I thought before I read your or Will’s posts, I’d gen up on atmospheric aerosols so I know what you’re both talking about. I found this quite informative: It certainly gives a good feel for why volcanoes disrupt the planet’s weather so dramatically when they erupt.

  4. BBD says:

    I notice that Will Morgan sounded a cautionary note over the various estimates of aerosol negative forcing:

    The satellite based central value of -0.85 W/m^2 is less negative than the central value from the climate models, which means the models indicate more cooling than the satellite based estimate. Compared to the subset of climate models that the IPCC used for their radiative forcing judgement, there is little overlap between their ranges also.

    This lack of agreement is a big driver for the large uncertainty range. It’s important to stress that there isn’t a strong reason to “trust” one set of results over another here as both satellite observations of aerosol properties and their representation in climate models are prone to many biases and it isn’t currently clear how these will impact the results.

    Both methods are very sensitive to the assumed pre-industrial conditions assumed by the studies. Both methods have difficulties dealing with clouds but for different reasons.

    So if I understand Will correctly, he warns against making strong claims using a very low estimate for aerosol forcing as, say, Nic Lewis does.

  5. BBD,
    Yes, I think that’s probably right – although we shouldn’t assume it’s high either 🙂 . I also think, as I pointed out in the post, that Nic Lewis is confusing the aerosol forcing from models (which, as Will’s post points out is as valid as the satellite estimates) with what climates models typically use.

  6. BBD says:


    Yes, of course the same applies to the higher estimate, although as you know, this is one of the reasons why I am leery of estimates of S derived from “observational” data and give more weight to those derived from paleoclimate behaviour.

    I did notice the point you made about “the models” vs “the models used to estimate aerosol forcing”, which, if correct, is an important distinction that may shed further light on issues with L13.

  7. BBD,
    Actually, I think the “models used to estimate aerosol forcing” vs “the models” issue makes his argument that model estimates of climate sensitivity must be too high, questionable. I think there are other – but related – issues with L13 🙂

  8. johnrussell, have just had a chance to start looking at the link in your comment. Haven’t gone through it all, but quite a resource it seems. Thanks.

  9. Paul S says:

    I’ve said before but all is not quite as it seems with the satellite-based estimates. The -0.85W/m2 figure depends on a non-observational adjustment of +0.2W/m2 to account for cloud adjustments causing longwave forcing – most of the estimates based on satellite data only look at the shortwave.

    Also, the treatment of estimates given in a couple of the contributing papers looks wrong:
    – The given estimate in AR5 for Lohmann and Lesins (2002) is -0.85, which appears to have been taken directly from the published paper, presumably because that study includes longwave forcing (?). However, this result is generally understood as an estimate of indirect-only forcing, requiring a direct forcing contribution to get the total.
    – Sekiguchi et al. (2003) estimated total indirect aerosol forcing to be between -0.6 and -1.2W/m2. According to how other papers are treated in theory they should have taken the centre of the range, slapped on a direct forcing estimate and added the longwave adjustment to arrive at a total of about -1.15W/m2. However, their given value is -0.93W/m2. I’m not entirely sure how they got this to be honest, but one possibility is that they found a longwave forcing estimate in the paper and decided to use that instead of the generic +0.2W/m2. The estimate given is much larger (+0.87W.m2) but the authors also clearly state that the observations they have don’t provide support for it – it’s essentially a hypothetical calculation of the maximum possible forcing through the mechanism they are studying:

    According to the tendency shown in Figure 1f, the cloud-top temperature did not significantly change. In this case there is no radiative forcing caused by a change in the cloud-top temperature. This may be the most possible scenario of the temperature change. However, the maximum possible forcing can be given by the forcing estimated by supposing that the cloud-top temperature has changed according to the correlation similar to that of T14. In this case, it is found that the cloud-top temperature change is thought to cause a positive forcing of about +0.87 ± 0.37 Wm2

    If you take out the dubious longwave adjustment and treat the two estimates above in a manner which I believe is more correct the median changes from -0.85 to -1.1W/m2 for studies using satellite observations.

  10. PaulS, thanks, I was hoping yourself and/or Karsten would pop along and clarify some of the issues 🙂

  11. BBD says:

    So there is a reasonable basis to argue that very low estimates of aerosol forcing may be incorrect. There’s a parallel with low estimates of S from the LGH/Holocene transition that are influenced by a possible warm bias in the LGM SST estimates (eg MARGO). The wide-screen picture still seems to be ~3C/2xCO2.

  12. Paul S says:


    Generally observational aerosol forcing estimates use 2-dimensional global aerosol optical depth (AOD, or sometimes AOT) satellite retrievals to determine where aerosols are.

    For direct forcing the typical process is to first relate the AOD field to a global Direct Radiative Effect (DRE), which is the TOA radiative flux impact of all aerosols (no distinction between natural and anthropogenic at this point). Then the aim is to determine the anthropogenic fraction of this AOD field. This is sometimes done by reference to MODIS AOD fine-mode retrievals, because apparently anthropogenic aerosols tend to be smaller than natural ones. Some studies instead use aerosol models to understand where the anthropogenic particles are. Often some combination of the two.

    For indirect effects my understanding is that the pure satellite estimates are obtained by tracking how cloud behaviours, and ultimately radiative fluxes, alter in relation to changes in AOD (no differentiation between anthropogenic and natural). They then simply apply this derived scaling factor to an assumed change in total aerosol burden from the preindustrial. A common amount to use is an increase by 30%. I can’t think how they’d get around it so presumably there is an implicit assumption that the spatial distribution of aerosols was not significantly different in the preindustrial.

    Other observationally-informed estimates use models to produce both preindustrial and present day conditions, then relate regional modelled aerosol-cloud behaviours to the equivalents found using satellite retrievals, and scale the modelled TOA flux change to fit.

  13. IPCC: “Cloud and aerosol properties vary at scales significantly smaller than those resolved in climate models, and cloud-scale processes respond to aerosol in nuanced ways at these scales.”
    Cloud variability is very important for radiative transfer through the atmosphere. I have tried to explain this with some pictures in an old post.

    Clouds change the chemical properties of aerosols and rain wash them out. I guess it was a special case, but I ones saw a nice picture of an aerosol field from the top with a beautiful round hole in it, which was caused by a cloud that had evaporated by them.

    Next to the direct effect of aerosols on the radiative balance of the Earth, they have many effects via their influence on clouds. They can make clouds whiter, the can make the life time of clouds longer (both by providing more cloud condensing nuclei, which causes droplets to be smaller. Smaller droplet with the same water content have a larger area and thus reflect light better (whiter). Smaller droplets also make the production of precipitation more difficult, more droplets have to be combined to produce one rain drop). Those are just two important effects, there are many and they are hard to quantify..

    Yes, aerosols have a quite short life time, typically up to days (Volcanic aerosols that make it to the stratosphere, however, can have life times up to years). As a consequence they do not build up like CO2. Many aerosols come from the burning of fossil fuels, like most greenhouse gases. Thus while they initially counter the greenhouse effect from fossil fuels, this soon saturates and the effects starts to be dominated by CO2 and Methane.

  14. chris says:

    I wonder if this might shed some light on the subject of aerosols…

  15. Thanks ATTP for highlighting the aerosol issue and thanks Paul S for providing additional insight. I might add a few points.

    But first, let me start with a question for Paul S (or whoever feels knowledgeable enough) re AR5: Do you know where the -0.27 W/m2 in Fig SPM.5 for the direct forcing of aerosols and precursors comes from? Cloud adjustments (-0.55 W/m2) are ERFaci (-0.45 W/m2) plus adjustment of RFari (-0.1 W/m2). Re direct forcing, in the caption of Fig TS.7 (Techn. Summary) it says: “Note that the total RF due to aerosol-radiation interaction (-0.35 W/m2) is slightly different from the sum of the RF of the individual components (-0.33 W/m2). The total RF due to aerosol-radiation interaction is the basis for Figure SPM.5.” What is it after all? I just can’t bring the -0.27 W/m2 (SPM.5) and -0.35 W/m2 (TS.7) together.

    Regarding the satellite estimates, the -0.85 W/m2 figure given for Bellouin et al. 2013 (see Table 7.4 in AR5) seems indeed to be calculated “manually” on the basis of the underlying AR5 assumptions (I got confirmation from Piers Forster). Basically, their (i.e. Bellouin et al. 2013) 1860-2011 RFari (-0.4 W/m2) becomes a ERFari of -0.55 W/m2 after scaling for 1750 and accounting for rapid adjustments (-0.1 W/m2). The 1860-2011 RFaci (all-sky IRF in the paper) of -0.6 W/m2 is “treated” with an pre-industrial-correction factor (the same which turns their inital RFari of -0.7 W/m2 into -0.4 W/m2), the 1750 scaling and the dubious LW adjustment (+0.2 W/m2), which makes it an ERFaci of approximately -0.3 W/m2. Total ERF hence -0.85 W/m2. As Paul S pointed out, some of these assumptions should be considered rather hypothetical. The indirect forcing in the paper is the change in cloud albedo exerted by a change in cloud droplet number concentration (CDNC). The susceptibility of liquid cloud albedo depends on cloud fraction and AOD. Quite a few assumptions, but the methodology has been tested several times so that people got some handle on it, despite the fact that the uncertainty margins remain fairly large. Cloud amount and determination of anthropogenic fraction as explained by Paul S.

    After all, the interesting bit is how strong an aerosol forcing has been “observed” in the past century. In my opinion, the best attempt to pin down the forcing past 1900 has been made by Wilcox et al. 2013. Ellie Highwoods Blog entry is well worth a read in that respect. They provide a direct estimate for the associated temperature response based on very plausible physical assumptions such as the inter-hemispheric temperature difference or changes in precipitation.

    Sorry for the technical details, but I’m sure Paul S is getting to grips with it 😉

  16. Karsten, thanks. I’m going to have to read your comment a few more times 🙂

    Ellie Highwoods blog post is very interesting. Thanks for highlighting that.

  17. Paul S says:

    Hi Karsten,

    I think the table is intended to represent radiative forcing, as opposed to the ERF discussed in Chapter 7. As such the total net direct aerosol forcing is -0.35W/m2, as per Table 8.6 and Figure 8.15 in Chapter 8. However, they’re also including the effect of black carbon on snow since it’s about sources rather than mechanisms, which is estimated at +0.04W/m2 (Table 8.6, Figure 8.15 again). On top of that the caption for Figure 8.16, which appears to be related to Figure SPM.5, states that SOA are not included. That accounts for the remaining 0.04W/m2 discrepancy.

    They appear to have added the direct ERF component (-0.1W/m2) to the indirect ERF to get the -0.55W/m2 cloud-adjustment figure.

    One thing I didn’t notice before looking at these figures was the huge stated uncertainties for direct RF and ERF in this final published version*: -0.85 to +0.15 and -0.95 to +0.05 respectively. That compares to -0.9 to -0.1 in AR4. Despite the larger uncertainty the stated confidence level increased from Medium in AR4 to High in AR5. I guess you could argue having a wider range should provide greater confidence that the true value lies in that range, but that’s not typically how these things are defined as far as I can see.

    * As late as the second order draft the ranges were -0.7 to -0.1 and -0.9 to -0.1.

  18. Pingback: Another Week of Anthropocene Antics, February 16, 2014 – A Few Things Ill Considered

  19. Hi Paul S,
    thanks for doing the detective work for me (and others) ;). I knew you would know. Although I skimmed through chapter 8, I hadn’t spotted the precise explanation.
    Regarding uncertainty, it might be the same issue as with attribution. The sum of the uncertainty range is larger than the actual uncertainty as nicely explained by Gavin. Not sure though … haven’t really thought it through yet.

  20. BBD says:

    Great thread! Thanks Karsten and Paul S. Will Morgan and Ellie Highwood blogs bookmarked too. I’d almost pay for this blog at times, ATTP 😉

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.