_{2}doubling, pointing out that Lewis, while not explicitly disagreeing with SOD's data, used his own substantially lower estimate for heat uptake.

On the other hand, Lewis does explain at some length why he prefers a lower number to the IPCC's median estimate of -0.9 W m

^{-2 }for (non-volcanic, mainly sulphate) aerosol forcing - what the IPCC calls

*Adjusted Forcing from Combined Aerosol-Radiation and Aerosol-Cloud Interactions*, or AF

_{ari+aci}for short. He points out that that estimate is "not what the observations indicate: it is a composite of observational, GCM-simulation/aerosol model derived, and inverse estimates." He wants to use observational data only, so he takes the figure of -0.73 W m

^{-2}from section 7.5.3 of SOD, and corrects SOD's estimate of total forcing for the difference. Paul S, commenting here, says that that value wrongly treats estimates in some source papers of the

*indirect effect*only as the whole forcing - the principal aerosol forcing mechanisms are the direct interaction with radiation, and the

*first indirect effect*, which is caused by a change in cloud albedo resulting from droplet nucleation.

I've looked up the five papers listed on page 7-109 as sources for the estimate. I present a crude summary of their results:

Paper | Direct | Indirect | Total | Note |
---|---|---|---|---|

Bellouin | -0.5 | -0.4 | -0.9 | |

Lohmann | -0.85 | [powerpoint] | ||

Quaas 2006 | -0.53 | Using LMDZ General Circulation Model | ||

Quaas 2006 | -0.29 | Using ECHAM4 General Circulation Model | ||

Quaas 2009 | -0.4 | -0.7 | -1.2 | See paper for why it doesn't add up |

Sekiguchi | -0.4 | -0.9 | -1.3 | Not including cloud fraction change |

To get more confidence in the numbers, I've had a look at the difference between

*Adjusted Forcing*and

*Radiative Forcing*. SOD estimates a centre value for RF

_{ari}of -0.4 W m

^{-2}, and explains (page 7-45)

AFSo its estimate of the direct effect is consistent with the numbers in my table._{ari}adds the radiative effects from rapid adjustments onto RF_{ari}...Rapid adjustments are principally caused by cloud changes...Overall a best estimate for the rapid adjustment is taken to be –0.1 W m^{-2}...[giving] an assessment for AF_{ari}of –0.5 W m^{-2}

SOD's preferred method for evaluating AF

_{aci}is to back it out from AF

_{ari+aci}:

We produce a best estimate ... for AF_{ari+aci}in the following way. The global CMIP5 models and inverse estimates are grouped together and a bootstrapping method is used to estimate a mean ... of –0.98 W m^{-2}. Processing the satellite-based estimates in the same way leads to a mean ... of –0.73 W m^{-2}... We combine these two estimates into a best estimate ... for RF_{ari+aci}of –0.9 W m^{-2}...

The AF[I think the "RF_{aci}is estimated as the residual between AF_{ari+aci}and AF_{ari}. We further assume that AF_{ari}and AF_{aci}are additive ..., which yields to our assessment of AF_{aci}of –0.4 W m^{-2}. Models indicate that RF_{aci}is less than AF_{aci}, which implies an estimate of –0.3 W m^{-2}...

_{ari+aci}" there is a typo for "AF

_{ari+aci}"]

Comparing the numbers in my table with the scatter plot in SOD Figure 7.10 on page 7-130, which shows seven points with the largest at about -1.0, it's apparent that the

*SAT*points plotted are for indirect forcing only. The two values in Quaas 2006 seem to have been plotted separately. I can't see where the seventh point comes from. Averaging the six points I've got gives -0.6, and adding a direct effect of -0.5 onto that gives a total adjusted forcing of -1.1 W m

^{-2}. Whence the big difference from SOD's combined estimate of –0.73 W m

^{-2}? It's hard to be sure, but it looks very much as if SOD has used the satellite estimates for indirect forcing only as if they're the whole forcing effect. Three of the papers explicitly disagree with that, because they talk about additionally about the direct effect and a total effect. The other two papers are clear that they're discussing indirect effects only. If the combined effect is really –0.73 W m

^{-2}and the estimate quoted above of -0.5 W m

^{-2}for the direct effect is good, that implies an indirect effect of only -0.23 W m

^{-2}, lower than the estimates in any of the studies referenced. This seems obviously wrong.

I've also looked up the seven papers listed as sources for the estimate of RF

_{ari+aci }(two of them the same as above)

Paper | Direct | Indirect | Total | Note |
---|---|---|---|---|

Bellouin | -0.5 | -0.4 | -0.9 | |

Dufresne | -0.5 | -0.22 | -0.72 | Change in sulphate aerosols 1860 to 1995 |

Lebsock | -0.42 | - | ||

Quaas and Boucher | -0.35 | - | Average of results for MODIS and POLDER satellite data | |

Quaas 2008 | -0.9 | -0.2 | -1.1 | |

Quaas 2009 | -0.4 | -0.7 | -1.2 | See paper for why it doesn't add up |

Storelvmo | -0.94 | Mode of four values |

Here simply summing the average direct (-0.58) and indirect (-0.46) forcings gives a total forcing of -1.04 W m

^{-2}. (I think the Dufresne result should be corrected for this purpose by adding in the 1860 forcing, but I've not done that.)

So the sources SOD references suggest an RF

_{ari+aci}of about -1.0 W m

^{-2}and an AF

_{ari+aci}of about -1.1 W m

^{-2}.

In conclusion, SOD mentions the value of -0.73 W m

^{-2}only in passing, and does not offer it as an estimate. And the source papers it refers to seem not to justify using it. Lewis might reasonably have take SOD's explicit estimate for AF

_{ari+aci}of -0.9 W m

^{-2}, or he could have gone to the sources using his preferred method and come up with a consensus value of -1.1 W m

^{-2}. I think he is not justified in using -0.73 W m

^{-2}.

Using a value of -0.9 together with my previous heat uptake estimate increases the median sensitivity to 1.95. Using -1.1 increases median sensitivity to 2.29 .

__

It's very possible that this post could be improved by expert interpretation of the source materials. I welcome suggestions and corrections, and may undertake extensive revision in their light.

[I've omitted throughout estimates and discussion of uncertainties. Which is not to say that they're unimportant, but I'm concentrating here on the headline numbers.]

Having (finally) read this I'd say it looks reasonable, though I've not checked back against the SOD. As to possible internal errors/inconsistencies in the SOD, that fits with what I've heard behind the scenes; I'm hopeful of hearing more "soon".

ReplyDeleteHi, just found that you'd written this. Thanks for following up.

ReplyDeleteSOD mentions the value of -0.73 W m-2 only in passing, and does not offer it as an estimate.This was my initial thought on skim-reading the SOD. However, on further reading of the chapter it's pretty clear the erroneous -0.73 value is a strong influence on their final estimate of -0.9. Specifically they mention four factors that contribute towards this estimate: 1) CMIP5 model median (~-1.0) 2) hi-res cloud-resolving model result (-1.1) 3) inverse estimates (median ~-1.0) 4) satellite-obs based (average given as -0.73). Given that three of these values are more negative than -0.9 (and this estimate was apparently arrived at

after1850->1750 adjustment so their basic average was probably closer to -0.8), the satellite study average they produce must have been given a strong weighting. That may be justified of course, but obviously only if they produce an accurate and representative synthesis of results.I've been attempting a brief reconciliation of the obs-based studies, satellite and inverse, to IPCC-compatible values both in terms of direct/indirect discrepancies and also the effective timescale of the estimates - most use 1850 or later as pre-industrial rather than 1750.

Assuming directRF=-0.4, directAF=-0.1 and indirectAF=-0.1 to fill in any gaps and a 20% amplification for 1850->1750 conversion where applicable (based on the SOD forcing time series), I make it that satellite obs-based studies span a range of -1 to -1.5, mainly clustering around -1.2W/m2.

Taking the inverse estimates at face value as representing total aerosol AF (though in reality "aerosol forcing" implicitly means different things across these studies), and adding the 20% 1850->1750 conversion where applicable, I make it that they span a range of -0.5 to -1.9, with median about -1.2W/m2.

Using the SOD's given justifications for arriving at a best estimate of -0.9 and applying a more representative synthesis of each factor I make it that the best estimate should be -1.2W/m2. It'll be interesting to see what we get in the final report next month.