Well, every time I go to look at the climate models, I come away more confused. And today is no exception. I decided to take a look at the relationship between the change in forcing (downwelling radiation) and the change in temperature.
Forcing datasets are somewhat hard to come by, but the Computer Model Intercomparison Project 5 (CMIP5) forcings for the GISS-E2 model are here. The sum of all the forcings is shown in Figure 1, along with the CEEMD smooth of the data.
.
.
Figure 1. Forcings used in the CMIP5 runs by the GISS E2 model.
The big dips in the forcings are the theoretical changes in forcing due to volcanic eruptions.
Next, here is the average surface temperature output of six runs of the GISS E2 model using those forcings, available from the marvelous KNMI website here.
.
.
Figure 2. Average of surface temperature (“tas”) output from six CMIP5 GISS model runs.
.
Already, although the model does a decent job hindcasting the past global surface temperature, we have problems. The difficulty is, back here in the real world, the volcanoes have not had that large an effect on the surface temperature.
.
.
Figure 3. Average surface temperature of six CMIP5 GISS model runs compared to Berkeley Earth surface temperature.
.
Why the exaggeration of the volcano effects? Me, I say it is because as the eruptions cool the surface, the climate responds by the daily tropical cumulus field forming later in the day, and the tropical thunderstorms also forming later or not at all. This warms the surface, counteracting the effects of the eruptions. However, YMMV …
But none of that is what I started out to look at. I wanted to see if the CMIP5 modeled temperatures slavishly follow the forcings. Thirteen years ago, I showed that the temperature output of the CCSM3 climate model could be very closely emulated by a simple one-line formula, viz:
T(n+1) = T(n) + λ ∆F(n+1) * (1- exp( -1 / τ )) + ΔT(n) * exp( -1 / τ )
See the linked post for the description of what the formula means.
So I used that formula, to see how well I could emulate the temperature output using nothing but the forcing applied to the model. Here’s the result:
.
.
Figure 3. “Black box” emulation of the GISS-E2 model temperatures, calculated from the forcing alone.
.
So the situation is unchanged. An R^2 of 0.97 says the emulation is doing a most excellent job. In passing, it’s interesting that the volcanic action in the model averages is a bit larger than in the calculations from the forcing, just as happened with the model average compared to the real world.
In any case, to recap the bidding: The GISS-E2 climate model has 440,000+ lines of code. It has over two million gridcells representing the world, and it takes a whole day to do just one model run on a parallel-processing computer with 88 processors.
And after all that, it merely spits out a lagged and resized version of the input forcing.
Hmmm …
.
Now, in the formula used to emulate the model output, the variable “lambda” (λ) is the change in temperature resulting (in modelworld) from a 1 watt per square meter (W/m2) increase in forcing. This is a measure of the “transient climate response” (TCR), how the temperature responds in the short term to a change in forcing. It’s usually expressed as the amount of change from a doubling of CO2 (2xCO2).
And the doubling of CO2 is said by the IPCC to increase forcing by 3.7 W/m2. This would make the TCR have a value of 0.41 °C/Wm2 * 3.7 W/m2 per 2xCO2 = 1.5°C / 2xCO2.
.
Now, I wanted to calculate the equilibrium climate sensitivity (ECS) from that TCR value. However, when I went to research that … let me say I was shocked. There’s a good discussion of the issues in an article with the long title “Emergent constraints on transient climate response (TCR) and equili...“.
.
What was shocking? Well, several things. First, according to the article, every different model uses a different value for the increase in forcing from a doubling of CO2. Remember above where I noted that the IPCC says forcing increases by 3.7 W/m2 from a doubling of CO2 (2xCO2)? Here’s what the models say:
.
.
Figure 4. Change in forcing from a doubling of CO2 as used by 31 different climate models.
.
Zowie! The models are all over the place, with values for 2xCO2 forcing ranging from 2.6 to almost 4 W/m2. And the range of the uncertainty on the median value (width of the notch in the sides of the box) doesn’t even include the canonical IPCC value of 3.7 W/m2 … how these jokers have the nerve to call climate science “settled” is a mystery.
Not only do we not have agreement on the values of the ECS and the TCR, but we don’t even agree on how much the forcing changes when CO2 doubles! Who knew?
I remember Steve McIntyre asking for an engineering-level derivation of the 3.7 W/m2 figure for 2xCO2, but I had no idea that the models disagreed so much on this central figure.
Then there was another surprise.
It turns out that the ECS is not some fixed multiple of the TCR. Instead, the ECS of small values of TCR is a smaller multiple of the TCR than for large values of the TCR. Figure 5 shows that relationship.
.
.
Figure 5. Scatterplot, ECS values versus TCR values for 31 different climate models.
.
This relationship would mean that the GISS E2 model would have an ECS on the order of 2°C / 2xCO2 … but of course, that’s only if the 2xCO2 forcing is 3.7 W/m2, when in fact we have values from 2.6 to almost 4 W/m2.
To compound the problem, as you can see, the different models give widely different values for equilibrium climate sensitivity (ECS). They range from 1.8°C to 5.7°C for a doubling of CO2, more than a three-to-one variation.
Here’s the final impossibility. Despite the different models having wildly different equilibrium climate sensitivities and different transient climate responses and different forcing changes from doubling CO2 … somehow they all do a pretty good job of hindcasting the actual temperature record.
And if the models were actually “physics-based” as every modeler claims, this would not be possible. I call this “Dr. Kiehl’s Paradox”, since he noted it first, and I discuss this impossible result here.
Gotta love that settled science …
Look, the ECS is a central, vitally important number in mainstream climate science. The forcing change from doubling CO2 is another central, vitally important number in mainstream climate science. They’ve been studying the subject for a half-century, and both numbers still have an enormous uncertainty range.
And worse than that, the uncertainty of the ECS has gotten wider with more study, not narrower …
.
Figure 6. ECS estimates over time, from 172 different sources.
.
In any other science, a half-century spent studying such a central number would result in reduced uncertainty … but in climate science, it’s going the other way.
.
To me, this is evidence that the basic paradigm of climate science is wrong. This basic paradigm is the claim that the temperature change is a linear function of the forcing change.
I think that is not true. I think that is simplistic nonsense. I think that the climate actively responds to changing conditions, and that there are a host of emergent climate phenomena that oppose any warming from any source, including changes in forcing.
Anyhow, that’s the story of my latest wandering through the models … I ended up knowing less than when I started.
You need to be a member of Citizens' Task Force on Wind Power - Maine to add comments!
Join Citizens' Task Force on Wind Power - Maine