If the model doesn't seem too pretty at first glance

No, not that kind of model...

As I discussed in an earlier post, climate modelling studies have generally been unable to simulate the magnitude of greening indicated by proxy evidence. Up until now, it has served my narrative nicely to treat the palaeodata as if they were gospel. By ignoring the possibility of any uncertainty in observations, it follows logically that any model-data disagreement is due to error in the model. If your experiments results aren't pretty then they must be wrong, right?

It should be pretty obvious that this isn't true.

The cool tropics paradox is a a textbook example of the dangers of having too much faith in palaeodata. During the greenhouse of the late Cretaceous (~66 Ma), conditions at the poles were positively balmy, warm enough for crocodiles to thrive. However, the sea-surface temperature signal recorded in benthic foraminifera fossils (δ¹⁸O) appeared to indicated that tropics were cooler than they are today. As illustrated in the plot below, climate models were unable to reproduce this relatively flat latitudinal temperature gradient.

Model temperatures indicated by dotted line, palaeodata temperatures by solid line. Reproduced from Poulsen et al. (1999).

In fact, for GCMs to simulate the conditions indicated by the proxy temperature signal would have required an overhaul of the understanding of the physics governing poleward heat transport. Thankfully, it turned out that much of the model-proxy disagreement was due to error in the palaeodata — poor preservation of the foram fossils altered the δ¹⁸O signal to produce unrealistically low temperatures at low latitudes. This later inspired the development of more robust proxy archives such as TEX₈₆ and Mg/Ca, which provided further support of models.

I'm discussing the cool tropics paradox to illustrate why, when comparing models and data, it is important to validate each with the other. While it is expected for modellers to be critical of their experiment results, it is key to remember that observations, too, cannot always be trusted. I will be exploring an example that is more relevant Green Sahara example in my next post.