Phanerozoic Climate Reconstructions and Sensitivity

I watched a lecture recently with Dr. Jessica Tierney, who is a professor of geosciences at the University of Arizona. Her lecture was fascinating to me because she explains how reanalyses (what she calls Data Assimilation or DA) can be used to reconstruct past climates. The lecture also gives us a preview to an upcoming paper with Tierney as a co-author that reconstructs global temperatures over most of the Phanerozoic (almost 500 million years). The method they use is notable to me for it's use in Osman's recent reconstruction for which Tierney is a co-author. I won't go into a lot of detail here, since she explains it in the above video, but essentially DA starts with  an ensemble of model simulations of past climates as a "prior." The model ensemble is spatially complete, but it isn't real - it's a simulation of what the past may have been like. DA then corrects the model results with proxy data. The proxy data has a huge advantage over models in that it's real - it's empirical data that clues us into actual climates in the past, at least locally or regionally. However, it's not spatially complete. We often do not have enough proxy evidence to measure the climate changes globally. So DA combines model simulations and proxy data to produce a reconstruction that is both spatially complete and controlled by the proxy evidence. It's essentially the kind of work done in weather forecasts, though with much lower resolution.

Illustration of Data Assimilation

As you can see in the illustration above, DA corrects model simulations with proxy evidence by pushing model results towards the proxy data. Notice it does not go exactly through the proxy data points, since the data points also have a margin of error. But the more data you add, the more corrections you can apply to the model results. The following reconstruction covers the last 1000 years, and as you can see, it agrees pretty well with the large network of reconstructions using proxy data without DA.

PaleoDA Analysis for the last 1000 Years[1]

In this lecture, Tierney covers some of her work reconstructing climate for the Cenozoic (last 66 million years), but I've covered a related study already published very recently. What I'd like to share here is her work covering most of the Phanerozoic. What I find interesting is that their model prior was an ensemble that was too insensitive to temperature. Temperature variability reconstructed in their model ensemble was much smaller than what was likely, so it was the empirical data that pushed temperature variability to those found in the final reconstruction. In other words, it wasn't that the model prior was making sensitivity high; it was the proxy evidence. The reconstruction below is a screenshot from the lecture, which is why Tierney shows up in top right corner.

GMST for the Last 480 Million Years

The graph helpfully shows both confidence intervals and the latitudes for which we have evidence of large scale glaciation. Mass extinctions are indicated on the graph by upside-down fish bones.  There are a few observations from this that I think are helpful:

• There's pretty good agreement between the reconstruction and geologic evidence of glaciation.
• Mass extinctions can be associated with rapid cooling (the end Ordovician extinction) or with rapid warming (end Permian extinction), but more often than not, these mass extinctions happened during warming events, even if temperatures alone weren't the main cause of the extinction. The coldest temperatures (and times of greatest glaciation) are not associated with mass extinctions. It's sudden changes in climate that drive extinction, not just warm or cold temperatures. In another lecture, she points out that the major mass extinctions of the last 500 million years are associated with large, rapid changes in climate.

• The long-term (on time scales of millions of years) data for CO2 and GMST correlate well, and a regression of CO2 and GMST has a slope of ~8°C for 2xCO2.

Her plot of the correlation between CO2 and GMST for the last 480 million years has a slope of 7.7°C for 2xCO2 (r = 0.72), with most of the scatter in the data coming from the Mesozoic, a point she acknowledges but doesn't explain.

Phanerozoic Correlation of CO2 to GMST

For the Paleozoic and Cenozoic, the correlation is stronger and the slope is larger, averaging 8.0 ± 0.4°C for 2xCO2 (r = 0.83). The correlation across the Cenozoic is even stronger at 8.2 ± 0.4°C for 2xCO2 (r = 0.97). Now this value is NOT the same as ECS, which is an estimate of how much GMST changes for 2xCO2 after climate reaches equilibrium with rapid feedbacks (on time scales of decades or so). Tierney refers to this as "apparent climate sensitivity" (ACS). Interestingly she does not use the term Earth System Sensitivity (ESS), which is typically thought to be ~2xECS. I'm not sure why she does this, but I think it has to do with the fact that this is just a simple regression analysis of CO2 and temperature. It's possible that that there were forcings that were correlated with CO2 but not dependent on warming from CO2. For instance, if methane and carbon dioxide typically increase in concentration together with methane increasing independently from CO2, then the methane forcings would lower the slope of this regression line.

What's surprising to me about this, though, is that the figure of ~8°C for 2xCO2 appears to be a robust feature of the climate system, despite the fact that the Sun was weaker in the geologic past. And warm climates and cold climates still seem to have similar sensitivities, so the ice-albedo feedback, though real, perhaps gets replaced by other feedbacks when the globe is too warm for ice sheets. It would be interesting, though, to know if the age of the data point explains some of the scatter in the data during the Paleozoic and Mesozoic - that is, if older datapoints tend to show up below the regression line. It may be helpful to see her results as an upper-bound estimate for ESS, consistent with another paper that arrived at a value of 5-8°C for ESS in the Cenozoic.

Incidentally, Tierney's reconstruction constitutes a major improvement over this sometimes misused graph from the Smithsonian. Tierney points out that the Smithsonian graph below isn't really scientific before 100 million years ago, and her work was partly done to improve on this curve.

This Smithsonian Reconstruction is "Not Scientific" Before 100 million years ago

It looks like Jessica Tierney, Emily Judd and their colleagues have done a lot of work collecting proxy data from the literature to make this study possible.[3] It sounds like they are making the proxy database available to others so that they don't have to do the same work. More proxy data can be added, and improvements can be made to the reconstruction as the reanalysis effort continues to be improved.

References:

[1] Tardif, R., Hakim, G. J., Perkins, W. A., Horlick, K. A., Erb, M. P., Emile-Geay, J., Anderson, D. M., Steig, E. J., and Noone, D.: Last Millennium Reanalysis with an expanded proxy database and seasonal proxy modeling, Clim. Past, 15, 1251–1273, https://doi.org/10.5194/cp-15-1251-2019, 2019.

[2] The Cenozoic CO2 Proxy Integration Project (CenCO2PIP) Consortium, Toward a Cenozoic history of atmospheric CO2. Science 382,eadi5177(2023). DOI:10.1126/science.adi5177. Accepted version online at: https://oro.open.ac.uk/94676/1/Accepted_manuscript_combinepdf.pdf

[3] Judd, E.J., Tierney, J.E., Huber, B.T. et al. The PhanSST global database of Phanerozoic sea surface temperature proxy data. Sci Data 9, 753 (2022). https://doi.org/10.1038/s41597-022-01826-0.