Guest contribution by Willis Eschenbach
There is a new open access paper in Nature Magazine titled “A Stricter Limitation on the Sensitivity of the Earth System Based on Long-Term Observations of Temperature and the Carbon Cycle” by Wong et al., Hereinafter Wong2021. Gavin Schmidt, GISS programmer for the stars, praises it on Twitter. The summary says:
The long-term temperature response to a particular change in CO2 forcing, or Earth System Sensitivity (ESS), is a key parameter that quantifies our understanding of the relationship between changes in Earth’s radiative forcing and the resulting long-term Earth system response. Current ESS estimates are subject to considerable uncertainty. Long-term models of the carbon cycle can be a useful way of constraining the ESS, but previous efforts have either used more informal statistical approaches or focused on discrete paleoevents. Here we are improving previous ESS estimates by using a Bayesian approach to merge in-depth CO2 and temperature data from the last 420 myrs with a long-term carbon cycle model. Our mean ESS estimate of 3.4 ° C (2.6-4.7 ° C; 5-95% range) shows a narrower range than previous ratings. We show that weaker chemical weathering compared to the a priori model configuration over reduced weatherable land area gives better agreement with the temperature records during the Cretaceous Period. Research to improve our understanding of these weathering mechanisms therefore offers potentially effective ways to further limit this fundamental property of the Earth system.
So I had to think about her paper. The first thing that made my urban legend detector ring was a statement in the summary above that you may have walked right by, which is:
We show that weaker chemical weathering compared to the a priori model configuration over reduced weatherable land area gives better agreement with the temperature records during the Cretaceous Period.
Translated from Scientese into English, one possible meaning of this is:
We adjusted the adjustable parameters of the climate model to make the output more in line with our theory that CO2 controls the climate.
Not a promising start …
All of this is based on a computer model called GEOCARBSULF, a long-term (millions of years) carbon and sulfur cycle model that is used to estimate past levels of CO2. So I asked myself … how many adjustable parameters are there in the GEOCARBSULF model?
But before I discuss the number of adjustable GEOCARBSULF parameters, why is the number of adjustable parameters important? There is a famous story about Freeman Dyson and Enrico Fermi that explains this problem well. Here it is, in Dyson’s own words:
We started by calculating meson proton scattering using a theory of strong forces known as pseudoscalar meson theory. By the spring of 1953, after heroic efforts, we had drawn theoretical graphs of meson-proton scattering. We happily observed that our calculated numbers agreed pretty well with Fermi’s measured numbers. So I made an appointment with Fermi and showed him our results. I proudly took the Greyhound bus from Ithaca to Chicago with a package of our theoretical graphics that I wanted to show Fermi.
When I got to Fermi’s office I gave the graphics to Fermi, but he barely looked at them. He invited me to sit down and asked me kindly about the health of my wife and our newborn baby boy, who is now fifty years old.
Then he announced his judgment in a low, even voice. “In theoretical physics there are two ways to do calculations,” he said. “One way, and this is the way I prefer, is to have a clear physical picture of the process you are calculating. The other way is a precise and self-consistent mathematical formalism. You do not have. ”
I was slightly stunned, but dared to ask him why he didn’t see pseudoscalar meson theory as a self-consistent mathematical formalism. He replied, “Quantum electrodynamics is a good theory because the forces are weak and when the formalism is ambiguous we have a clear physical picture that can guide us. With pseudoscalar meson theory there is no physical picture and the forces are so strong that nothing converges. To get your calculated results, you had to introduce arbitrary boundary methods that are neither based on solid-state physics nor solid-state mathematics. “
Desperate, I asked Fermi if he was not impressed with the correspondence between our calculated numbers and his measured numbers. He replied, “How many arbitrary parameters did you use for your calculations?” I thought about our shutdown procedures for a moment and said, “Four.” He said, “I remember my friend Johnny von Neumann saying that with four parameters I can fit an elephant, and with five I can make it wiggle its trunk.” That ended the conversation. I thanked Fermi for his time and effort and unfortunately took the next bus back to Ithaca to bring the bad news to the students.
So … how many adjustable parameters does the GEOCARBSULF model have? From the Wong2021 paper …
There are 68 GEOCARB model parameters, 56 of which are constants and 12 of which are time series parameters. The constant parameters have well-defined earlier distributions from previous work, and the time series parameters have central estimates and independent uncertainties that are defined for each point in time15.
Hmmm, me … 68 parameters … not a good sign.
To see if “the constant parameters have well-defined previous distributions from previous work” as claimed above, I looked at reference 15 listed in the above quote. It is called “ERROR ANALYSIS OF CO2 AND O2 ESTIMATES FROM THE LONG-TERM GEOCHEMICAL MODEL GEOCARBSULF”. There the abstract closes with the words:
The late Mesozoic model-proxy mismatch can be removed with a change in the GYM within its plausible range, but no change within the plausible ranges can correct the early Cenozoic mismatch. Either the true value for one or more of the input parameters during this interval is outside of our sampling range, or the model is missing one or more key processes.
Hmmm, sez I … doesn’t sound like that which confirms Wong2021’s claim that “the constant parameters have well-defined prior distributions from previous work and the time series parameters have central estimates and independent uncertainties defined for each point in time15.”
Aside from the fact that the model has enough adjustable parameters to get an elephant to put on a tutu and play the Swan Lake ballet, I’ve looked at their results. First, here is her graphical representation of her results.
Figure 1. This is Figure 4 in Wong2021. ORIGINAL SUBTITLE: “Model hindcast using CO2 and temperature data for pre-calibration and a% outbound threshold of 30% (shaded areas). The gray shaded areas show the data compilations for CO2 (ref. 26) and temperature12. The lightest shaded areas denote the 95% probability range of the pre-calibrated ensemble, the middle shading denotes the 90% likelihood range, the darkest shading denotes the 50% likelihood range, and the solid lines show the ensemble medians. To represent the limit of each data set, the dashed lines show the 95% probability range of the pre-calibrated ensemble when only temperature data is used (a) and when only CO2 data is used (b). “
(A brief digression. In Figure 1 I have taken into account the fact that dinosaurs lived on the planet from about 245 million years ago to 66 million years ago. Mammals first appeared on the planet 178 million years ago. During this time, temperatures were measured according to Figure 1 between 6 ° C and 12 ° C warmer than now. And people are hyperventilating about another half a degree Celsius as an “emergency” that will ruin our lives and drive extinction through the roof? … but I digress.)
Their claim is now that their results more severely limited the sensitivity of the planet’s temperature (lower panel) to atmospheric CO2 levels (upper panel). I blinked at this graphic and said “Hmmm …”. Didn’t look too likely.
So I did what I usually do when the writers aren’t scrupulous enough to archive their findings. I digitized and graphed the Wong 2021 temperature and CO2 data. Figure 2 shows this result.
Figure 2. Scatter plot, paleotemperature versus logarithmic (base 2) paleo CO2 values from Wong2021
If the CO2 levels were indeed the control knob that regulates global temperature, we would see all the points fall on a nice straight line … but we are not far from it. There is no statistically significant relationship between temperature and CO2 levels reported by Wong et al.
So I have to say that the data reported in the Wong2021 paper are far from justifying the information given in their abstract. Even after carefully tweaking the tunable parameters of the GEOCARBSULF model in their favor, their results support the null hypothesis that CO2 is not the global temperature regulator.
I wish everyone, dinosaurs and mammals, all the best.
w.
PLEASE: If you comment, quote the exact words you are discussing. I can and am happy to defend what I have written. But I cannot defend your interpretation of what I wrote.
Like this:
Loading…
Comments are closed.