by Frank Bosse
A few days ago a paper (Sato et al) dealing with some aspects of the “Aerosol Cloud Interactions”, (ACI, also called “aerosol indirect effects”) was released. It bolsters the conclusions of earlier papers: the effective radiative forcing from ACI (ERFaci) is smaller than thought, perhaps near zero .
A discussion of Malavelle et al. (2017) and Stevens (2017, an accompanying comment on the Malavelle paper) was released in July 2017.
The Sato et al. paper refers to a new model which makes it possible to study the globally impacts of aerosols on warm clouds (T>0° C) thanks to a much higher resolution than former approaches. Some details were discussed in this blogpost. Nic Lewis made a comment in which he states:
” By contrast, in all but a few models that forcing is significantly negative, and is one of the main reasons why current climate models match observed historical warming despite their generally high (transient) sensitivity.”
I was interested in the quantitative order of the mean overestimation of the total effective aerosol forcing (ERFaero) of the CMIP5 models. It is well known that the ERFaero , the sum of direct aerosol forcing ( ERFari) and ERFaci is by far the greatest source of uncertainty when it comes to observationally based estimates about the transient sensitivity (TCR) and the expected warming in this century.
Data
I used the IPCC AR 5 forcing data, revised (Myhre 2017 , Etminan (2016), updated, and the GMST from Cowtan & Way both for the time span 1950…2015 (not 2016 due to very strong ENSO influence). I excluded the volcano forcing and the years with ERFvolcano >1W/m² to avoid some bias due to the timing of these events and the known lower impact of volcanism on the GMST than other forcings.
Estimation of ERFaero in the CMIP5 mean vs. Sato et al.
All calculations (i.e. here or here ) using the regression method- observed GMST vs. the total forcings- come to TCR estimates which are well below the mean of the CMIP5 models of 1.8 K/doubling CO2. Despite this, the CMIP5 mean historical warming reasonably matches the observed warming. The explanation for this should be the use of different ERFaero values because all other forcing data are much better constrained than ERFaero.
I used the regression method the other way around — the TCR of 1.8 °C/2*CO2 for the CMIP5 mean is given, implying a target sensitivity of 0.48 °C/W/m² with 3.8W/m² per doubling CO2. The required correction to total forcing in order for the regression of GMST on total forcing to produce that sensitivity should represent the additional ERFaero included in the CMIP5 models on average.
With an empirical method I found that a linear additional trend from 1950 to 2015 of 0.085 W/m² /decade to ERFaero made the correlation coefficient match the CMIP5 mean sensitivity. The regression given this is shown on Fig.1.
Fig.1: The 1950-2015 regression of the GMSTA (Global Mean Surface Temperature Anomaly) on the total forcings that achieves the wanted TCR of 1.8 °C/doubling CO2 for the CMIP5 mean.
To estimate a possible bias from the exclusion of volcano years in the regression I used the changes in forcing and GMST from the 1950-1962 mean to the 1996-15 mean, the longest early and late volcanic free periods. This gives a TCR (based on F2x= 3.80 W/m2) of 1.76, close enough to 1.8.
In Sato et al. there is no quantitative forcing estimation included about the described finding that one of the two major components of the models, the cloud lifetime effect, has in most regions the opposite sign in the real world. The authors give some hint when they write:” This suggests that estimates of the net negative radiative forcing due to the total ACI can also be significantly reduced and its uncertainty range could even include positive values.”.
As a first guess, I reduced the ERFaci by about 50%, this gives for the ERFaero a scaling of 0.75 vs. the AR5 ERFaero.
With this input I regressed the GMSTA vs. the “Sato-forcings” — the total AR5 forcing estimate with ERFaero scaled by 0.75:
Fig. 2: The 1950…2015 regression of the GMSTA on the total forcings with reduced ERFaero following Sato et al., assumed to imply that ERFaero is 0.75 of the level estimated in AR5. The TCR regression estimate gives 1.3 with 3.8W/2*CO2.
Closing this section, I compare the ERFaero estimates for the time span 1950…2015:
Fig.3: The ERFaero for the CMIP5 mean (red), for the IPCC AR5 estimate (black) and the Sato et al. estimated findings (blue), all relative to 1750. The described additional linear trend as a rough approach to estimate the ERFaero of the CMIP5 mean is also justified by the agreement between the determined -1.25 W/m² for the years around 2000 and the given values in this paper.
The difference in the ERFaero between the CMIP5 mean and that implied by the latest research results increased to more than 100% for the years after 2000.
Conclusions
The sensitivity of the earth’s climate vs. GHG depends on how much GHG warming has been offset by cooling from aerosols. The forcing due to greenhouse gases is well known: an average of 3 W/m² in the last decade. The “damping” of the greenhouse gas warming in the past due to aerosols is poorly known. Many CMIP5 models (influencing strongly the model mean) have an inherent strong aerosol forcing. Their high TCR is partly compensated by this. The latest research in this field points to much lower ERFaero values, roughly half as much. This contrast will lead to greater model divergences from the real climate system. The next generation of models will appear and hopefully close this gap. The latest progress makes hope.
