How Well Has Hansen's 1988 Climate Model Performed?

We hear a lot that climate models predict too much warming of global mean surface temperature (GMST), but just about every graph intended to back this up contains flaws or misrepresentations, either by sloppy comparisons of models vs observations or using a subset of models that are heavily influenced by those with high sensitivities. I think it's valuable to show how well models form, since the sensitivities of those that perform the best are likely to be good indicators of the sensitivity of the real climate system. So I thought I'd toss my hat in the ring by comparing Hansen's 1988 climate model with observations. This has been done before, most notably by Zeke Hausfather in 2019, but it's been 6 years since then, and I thought it might be good to update it. To do this, I downloaded Hansen's data for GHG forcings and temperatures for his Scenarios A, B and C (available here and here), and I plotted these scenarios with the major GMST datasets against their 1958-1987 means. This sets the mean 30-year span to be the historical period of his model. All data from 1988 to 2019 is predictive. Here's what I came up with.

Observations fall squarely in the middle between Scenarios A and C. They follow Scenario B much more closely than the other two, but observational warming trends are still a little slower than Scenario B. So to evaluate Hansen's model, I decided to compare the forcings for each of Hansen's Scenarios to the forcings that actually occurred. To do this, I used the projected GHG concentrations for each model and I calculated the associated forcings for each using formulas that NOAA has published to calculate forcings they put into a data table of forcings from 1979 to 2023. Since I'm using the same calculations NOAA used, I can make apples-to-apples comparisons between the forcings associated with Hansen's Scenarios and observational forcings. Hansen published temperature predictions through 2019, model scenarios went through 2050, which is why I can extend them beyond 2023. This is what I ended up with.

So clearly observed forcings track with Scenario B up through 1988, but then diverge. Hansen's Scenario B shows larger forcings through 2023 than observations.

2023 Forcings	Scenario B	Observed	% Change
CO2	2.212	2.286	1.0335
CH4	0.719	0.565	0.7861
N2O	0.244	0.223	0.9153
CFCs	0.650	0.301	0.4634
Total	3.824	3.375	0.8826

As you can see, Hansen's Scenario B overpredicted concentrations of CH4 and CFCs, so observed forcings were about 88% of Scenario B (Hansen also did not include forcings for HCFCs and HFCs, which I included in the chart above). So since forcings are directly proportional to temperature changes (ΔT = S*ΔF) I figured we could scale the the warming of Scenario B by 88% (ΔFo/ΔFb) to approximate what Hansen's model would produce if it was making predictions on observed forcings. Here's what I came up with.

I wish I had Hansen's data for aerosols; that would let me make this even better. The H88 Scaled plot agrees even more closely to observations, which much of the differences occurring between he 2005 and 2015. I then calculated the trends for observations and my H88 Scaled reconstruction for the years since 1988 to see how well my H88 Scaled predicts observations. Here's are the trends I came up with.

    H88 Scaled    0.249°C/decade

    HadCRUT5    0.220°C/decade

    BEST              0.222°C/decade

    NASA             0.223°C/decade

    NOAA            0.220°C/decade

    JRA-3Q          0.221°C/decade

    ERA5             0.235°C/decade

The average of observations is 0.221°C/decade or about 89% of the warming trend in my H88 Scaled. But since Hansen's model predicted ECS to be 4.2°C, I'm glad for that. I think it's fair to say that Hansen's model performed quite well. It's certainly skillful (the uncertainties of the model predictions include observations, and it certainly supports the IPCC's estimate that ECS is ~3°C (between 2.5 and 4°C).

References:

[1] Hausfather, Z., Drake, H. F., Abbott, T., & Schmidt, G. A. (2020). Evaluating the performance of past climate model projections. Geophysical Research Letters, 47, e2019GL085378. https://doi.org/10.1029/2019GL085378