Public Climate Data Explorations: Pangea Proxima

This is part of an ongoing series of mine looking at publicly available data on climate modelling to produce climate maps. To reiterate from previous posts, I didn't run any of these models myself, unlike my separate series on ExoPlaSim climate explorations, I'm just relying on what data researchers have made available from their own work.

Today we'll be looking at the results from a single study, "Climate extremes likely to drive land mammal extinction during next supercontinent assembly", Farnsworth et al. 2023. In this series' first exploration we looked at potential climates for the proposed future continents of Amasia (where the Arctic Ocean closes and most continents converge on the north pole) and Aurica (where Eurasia rifts apart, both the Atlantic and Pacific close, and the continents converge near the equator) modelled with ROCKE-3D (see here as well for their appearance in my bioclimate system and here for explanation of that system). This study looks at the Pangea Proxima model, where just the Atlantic closes and an equatorial supercontinent forms, but with a large interior sea as a remnant of the Indian Ocean (for those keeping score, that just leaves Novopangea , where the Pacific closes, as the only prominent model for the next supercontinent that I haven't seen modeled).

This is also the first study we've looked at using the HadCM3 climate model, developed in the UK and used extensively by researchers at the University of Bristol, primarily for studying climate change or paleoclimates but with a few oddball experiments like this. The outputs from many of these studies are available online (here's the repository for this specific study), though to be frank from a somewhat clunky and arcane interface; I ended up figuring out how to use python and Wget to query the servers directly for my purposes. But the data is quite detailed and available on a monthly basis, so can be used for any of our usual climate classification systems. The models here were all run with a resolution of 73 by 96 cells and with dynamic ocean currents, and conveniently enough they include a control model of modern Earth with a pre-industrial CO₂ level of 280 ppm:

A good overall match to the real climate, but of course it has a few quirks, including a certain semiarid bias that seems to be a common issue for lower-resolution models, making many deserts too wet but also many rainforests and wet temperate areas too dry, creating an excess of savanna, steppe, and Mediterranean zones (as per usual I did tune the threshold for Mediterranean zones in the bioclimate system, setting it to 0.6 GrS to try to balance between having too much in regions it shouldn't be and too little in regions it shouldn't). Much of this may be just because the model happened to come out a bit cold here, with a global average around 12 °C. Also note that the paper doesn't specify if they used Earth's real orbital eccentricity here, which may factor into these results.

HadCM3 also often runs with the BIOME4 vegetation model, a distant descendent of the Prentice et al. 1993 biome model that partially inspired my own climate classification system. This latest version uses a more complex model of photosynthesis and the relative growth rates of different plant types, and then determines biome distribution based on the resulting densities of different types.

BIOME4 predictions based on real climate data. PMIP 2

If it seems like this system is oddly granular in how it classifies tundra, that's because the latest iteration was developed as part of a study on arctic vegetation. Anyway, it was easy enough to extract this data from the model outputs and map it out as well:

Now, as to the main results, the study uses a somewhat simplistic model of Pangea Proxima which essentially just retains the modern topography of each continent but smoothed a bit between them (the Himalayas and Rockies seem to be notably reduced, but I'm not too sure why); no new mountains are added to represent orogenies along collision zones or new subduction zones. So this is not terribly accurate to exactly how we'd expect Pangea Proxima's terrain to look, but constructing such a map was perhaps considered out of scope for the climate study.

They model Pangea Proxima with CO₂ levels varying from 0 ppm to 2240 ppm, 8 times industrial levels, both with modern levels of sunlight and 2.5% increased light, appropriate to roughly 250 million years in the future. I'll just use the latter set of results with increased light here, and I'll skip the 0 ppm model, which just promptly forms a global snowball with no habitable land.

The results (with nonzero CO₂) are all warmer than the control, ranging from 14 °C with 70 ppm to 33 °C with 2240 ppm. The paper doesn't actually mention the 2240 ppm model, perhaps indicating that the authors either don't think such a scenario is realistic or don't expect the model to be reliable at such high temperatures, and their analysis mostly focuses on the 280, 560, and 1120 ppm models, with some broad predictions at the end of the paper suggesting that the 560 ppm case at 23 °C may be the most likely average climate.

As you might expect for a supercontinent, vast deserts stretch across the interior, but a temperate refuge survives along the northern coast even in the hottest cases, and there's a strip of wetter climates along the equator. There is some notable disagreement between these climate classification schemes on what exactly that equatorial strip looks like, though. The Pasta classification suggests generally drier conditions for the hotter climates, and while the Koppen-Geiger and BIOME4 maps broadly agree, they also suggest patches of rainforest appearing in some of the hotter cases. The BIOME4 model involves the most detailed representation of vegetation growth patterns and restrictions, and so perhaps should be trusted most here, and it may also be reflecting how higher CO₂ levels in the hotter cases may make it easier for plants to survive drier conditions (because plants lose much of their water in the process of absorbing CO₂), but I'm not sure how well the model is tuned for very high CO₂ levels or summer temperatures over 50 °C. It's also possible that in gathering data for the Pasta classification, I've somehow misinterpreted how it reports potential or actual evapotranspiration. Evapotranspiration isn't reported as a single value by the HadCM3 model, but separately as forest canopy evaporation, soil evaporation, plant transpiration, and ice sublimation, which I summed together, and this seemed to give good agreement with the other classifications for the control case, but perhaps there's some subtleties I'm missing that become more important in hotter climates.

The paper makes the case that, at some point after Pangea Proxima has assembled, it's fairly likely that there would be a volcanic event that pushes the climate towards one of the higher-CO2 cases here, and in that case the hot, dry conditions would exceed known thermal tolerances for mammals across almost all land area, leading to their mass extinction. Personally I'm a bit more skeptical that we can so confidently extrapolate the tolerances of modern mammals to their descendants 250 million years in the future, or that the conditions described in these models would necessary preclude survival of even modern mammals from at least the cooler polar regions. But, to be fair, globally successful animal groups have been driven to extinction in the past not because their survival became strictly impossible but because their reduced diversity after a major extinction event made it more difficult for them to sustain a stable population in the short term or avoid replacement by more diverse competitors in the long term. The paper also doesn't explicitly claim that mammals would go completely extinct, so the softer claim that the conditions of Pangea Proxima might substantially reduce their range and dominance over niches for large terrestrial animals might be easier to accept. Another broader point the study makes is that you can't judge the habitability of a planet by global statistics like average temperature or ocean area alone; the particular arrangement of continents and climate patterns can substantially influence how much of the planet's land surface would actually be hospitable to humans or similar life.

That'll do for this little look at one study, we may see more from HadCM3 in the future but our next public data exploration should return to looking at more exotic climate states better covered by other models.

Buy me a cup of tea (on Patreon—and join the Discord!)