What Exxon Knew in 1982

If you've participated in discussion about climate change, you've probably come across the slogan "Exxon Knew." This came out as private, internal communications from within Exxon (and other fossil fuel companies) revealed that their scientists had been investigating potential catastrophic consequences of fossil fuel emissions, and they made predictions that turned out to be pretty accurate. To what extent Exxon is culpable for this is difficult to determine. However, what is clear (to me) is that Exxon and other fossil fuel companies ought to be held liable for knowingly promoting misinformation when their own scientists were telling them that there were valid concerns that continued fossil fuel emissions would have harmful consequences for humanity.

Papers were published in 2021[1] and 2023[2] that evaluated the accuracy of the scientific work that Exxon knew by evaluating the internal studies that have now been made public. The above graph has been popularly shared as evidence that their predictions were startlingly accurate. A summary of their analysis of Exxon's scientific work is below. The model above is represented by the third column below (1982/84 Glaser Shaw). The implied TCR from this, they say, was just over 3°C for 2xCO2.

I thought it would be fun to evaluate this to see for myself. I've seen several marked up versions of the above graph arguing that the model was accurate, but a TCR or 3°C would be a bit warm. As a general rule of thumb ECS ≈ 1.5*TCR, so if ECS is 3°C, then we'd expect TCR to be closer to 2°C. The first step for me was to duplicate the model results, which I did below.

After getting this correct, it was just a matter of plotting the data with CO2 on the x-axis and GMST on the y-axis. The slope of the resulting graph should be the model's estimate for TCR.

The slope of the above graph is 0.826°C/W/m^2, which is consistent with a TCR of 0.826*3.71 = 3.06°C. This precisely agrees with Supran's paper. I then plotted CO2 forcings on the x-axis and HadCRUT5 on the y-axis. The slope of that correlation is 0.739°C/W/m^2 ± 0.053°C/W/m^2 (2σ) from 1958-2024. I then plotted Exxon's model with a slope of 0.826°C/W/m^2. Exxon's model is outside the confidence envelope for HadCRUT5, though HadCRUT5 is responding to all forcings (GHGs, aerosols, etc) not just CO2. Nevertheless, you can see the performance of Exxon's model was slightly warmer than than the correlation between CO2 and HadCRUT5. The graph below shows HadCRUT5 and Exxon's model set to a 1961-1980 baseline.

If I plot this same data with the year in the x-axis.

But from what I see here, if the Exxon model's TCR is 3°C, that would imply an ECS for their model of ~4.5°C. This would put Exxon's model in the ballpark of Hansen 1988, which calculated ECS to be 4.2°C, and Hansen's model has performed quite well. Hausfather's analysis showed that when Hansen's model is compared to actual forcings since 1988, the climate response from the model performed quite well.

So it seems to be a fair conclusion that 1982/84 Glaser Shaw runs a bit hot due to a TCR that may be as much as 50% too warm, but it's in the ballpark with what Hansen's model was producing. Supran et al 2023 appears to be a fair assessment of Exxon's knowledge, at least with respect to this 1982 model. The science that Exxon was producing in the 1980s was probably comparable to Hansen, and if anything it would have led them to conclude that continued fossil fuel emissions would cause more warming than we've actually experienced.

References:

[1] Supran, G., & Oreskes, N. (2021). Rhetoric and frame analysis of ExxonMobil’s climate change communications. One Earth, 4(5), 696–719. doi:10.1016/j.oneear.2021.04.

[2] G. Supran et al. ,Assessing ExxonMobil’s global warming projections.Science379,eabk0063(2023).DOI:10.1126/science.abk0063