In #1 we took a brief look at Natural Variation - climate varies from decade to decade, century to century. In #2 we took a brief look at attribution from “simple” models and from climate models (GCMs).
Here’s an example of the problem of “what do we make of climate models?”
I wrote about it on the original blog - Opinions and Perspectives – 6 – Climate Models, Consensus Myths and Fudge Factors. I noticed the paper I used in that article came up in Hourdin et al 2017, which in turn was referenced from the most recent IPCC report, AR6.
So this is the idea from the paper by Golaz and co-authors in 2013.
They ran a climate model over the 20th century - this is a standard thing to do to test a climate model on lots of different metrics. How well does the model reproduce our observations of trends?
In this case it was temperature change from 1900 to present.
In one version of the model they used a parameter value (related to aerosols and clouds) that is traditional but wrong, in another version they used the best value based on recent studies, and they added another alternate value.
What happens?
The black and gray lines are observations of temperature change over 100+ years. The green line is the climate model simulation using the traditional but probably wrong value, the blue line is using the “best value” and the red line is the “worst value”:
We see that while temperatures have risen about 0.8°C from the late 1800s to 2000, the model with the correct value (blue line) of this parameter produces only about 0.2°C from 1860 to 2000. The red line, with the worst value, is very close.
CM3w predicts the most realistic 20th century warming. However, this is achieved with a small and less desirable threshold radius of 6.0 um for the onset of precipitation. Conversely, CM3c uses a more desirable value of 10.6 um but produces a very unrealistic 20th century temperature evolution. This might indicate the presence of compensating model errors.
The paper notes that the present day climate produced in the different versions of these models is very similar.
References
Cloud tuning in a coupled climate model: Impact on 20th century warming, Jean-Christophe Golaz, Larry W. Horowitz & Hiram Levy II, GRL 2013
And finally, AR6 WG1 recognized the failure of AOGCMs to properly handle clouds (which you exemplified above) and noted that about half produce estimates of current warming that are inconsistent with observations. They decided evidence form models was no longer suitable for quantitatively estimating climate sensitivity. Section 7.5.6 and Figure 7.19. (:)) However, everywhere else in the report other authors rely heavily on the output from climate models.
If AOGCMs are ruled out, estimates of climate sensitivity from other lines of evidence might be worthy of examination, sun as Figure 7.12. Of course, all six studies cited rely on models to some extent too.
True Frank. There are quite convincing reasons why climate models can't be even close to predicting patterns of change. Their resolution is way too course, probably at least 50 times too course. As pointed out in this post, their subgrid models appear inadequate as well.