By Willis Eschenbach – Re-Blogged From http://www.WattsUpWithThat.com
In the comments to Christopher Monckton’s latest post, Nick Stokes drew attention to Soden and Held’s analysis of feedback in the climate models. I reproduce their Table 1 below:
Figure 1. Soden and Held’s Table 1, showing all of the feedback parameters calculated from the models.
I found several amazing things in this table. The first is the huge range of values for the various parameters. While all of the Planck parameters are within a few percent of each other, the lapse rate feedback varies by more than three to one from smallest to largest; the surface albedo feedback varies by nearly five to one; and the cloud feedback varies by an amazing factor of more than eight to one from smallest to largest.
Despite these huge variations, all of them can (relatively) successfully emulate the historical record when they are each fed their own special brand of forcings … which should tell us something about the models. But I digress.
What I want to look at today is the cloud feedback. This is measured as something called the net cloud radiative effect (CRE), which is the sum of the solar and longwave radiation from the clouds. There is general agreement that as a global average the CRE is negative with a value of about -21 W/m2, meaning that in general the clouds cool the earth.
However, there is little agreement about the size or even the sign of the cloud feedback. Cloud feedback is the change in the net CRE that we can expect from a 1°C change in temperature. The models say that cloud feedback is a 0.69 ± 0.10 W/m2 INCREASE in downwelling radiation for each additional degree of temperature. In other words, if there is a small warming, the models say the clouds amplify it to make a large warming. This implies a positive correlation between temperature and the net CRE.
Fortunately, the CERES data can give us actual observational data regarding this question. Figure 2 shows the correlation between temperature and the net CRE.
Clearly, this is a hugely complex system, where in some parts of the world the correlation is strongly negative, and in some parts it is positive. Note that the inter-tropical convergence zone (ITCZ), which I have long held is a crucial part of understanding the climate, is negatively correlated. And so is the land area north of about 50°N or so.
Overall, we can calculate the global correlation by looking at the global area-weighted average net CRE versus area-weighted global average temperature. Figure 3 shows that result:
This is very bad news for the models … they all claim that there is a positive correlation between CRE and temperature, which makes the model-projected warming much larger … but in fact the global average correlation is negative.
And from this same data, of course, we can calculate the global average cloud feedback parameter, viz:
Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -0.0006277 0.0392260 -0.016 0.987250 (Temperature) -1.0121541 0.2695399 -3.755 0.000234 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 0.5263 on 178 degrees of freedom Multiple R-squared: 0.0734, Adjusted R-squared: 0.0682 F-statistic: 14.1 on 1 and 178 DF, p-value: 0.0002345
The CERES observational data says that for every additional degree of warming, the net cloud radiative effect becomes 1 W/m2 more negative, meaning a cooling effect. So despite the fact that the models claim that the cloud feedback parameter is positive (0.69 ± 0.10 W/m2 per degree C), ugly reality disagrees. According to the CERES data, the real number is not only negative, it is strongly negative ( – 1.0 ± 0.27 W/m2 per °C) and is strongly significant.
Not much more I can say about that, except that it totally confirms my long-held belief that given the amazing stability of the climate system, and my hypothesis about how clouds and thunderstorms regulate the temperature, cloud feedbacks perforce must be net negative, not net positive as the models claim.