Public Climate Data Explorations: Future Supercontinents

 Ever since I found that dataset of Phanerozoic climate models, I’ve been keeping a lookout for other published datasets with climate model data I can work into climate maps, and there actually is a fair amount of this sort of data out there, more than I could comfortably fit in one post. So this will be the first of a new little series of posts, each going over one or a few of these datasets roughly organized into reasonable categories. For this first post I’ll start small and look at some models for the climate of potential future supercontinents.

To make it clear (because this is sometimes misunderstood about that last post): I DID NOT RUN THESE MODELS MYSELF, and that will be true for all the posts in this series. I’m just taking the output of models other people have already run and interpreting the data into climate maps. Most of these models weren’t even run in ExoPlaSim, but a range of other, typically more complex climate models.

I’ve looked specifically for models that might be interesting to see as climate zone maps and are directly associated with published research that gives a bit of context on how these models were run and you can read for a deeper analysis of the implications of these climates. So I’ve excluded models run as “aquaplanets”, with a surface covered in uniform ocean, or “land planets” with flat global land, or other studies similarly concerned with over-idealized cases or non-earth-like climates that wouldn’t look like much on a climate map. I also won’t be digging into all the individual parameters as much as I do in my own climate explorations; the scientists who ran these models and wrote the associated papers know far better than I how to interpret the results, and in general I want to keep these posts are bit shorter than my norm. Most of these datasets are in the netCDF format and all are publicly available, so you can generally use Panoply to look through them yourself if you’re particularly interested.

For the first several posts of this series, I’ll be going through this data repository of ROCKE-3D model outputs produced as part of the research of the planetary atmosphere modelling group at NASA’s Goddard Institute for Space Studies. You can think of ROCKE-3D as sort of the industrial-grade counterpart to ExoPlaSim, with a similar focus on flexibility for different types of exoplanets, but with rather more complexity (in particular including a dynamic ocean circulation model) and correspondingly far greater runtime. I know of at least one person who’s got it running at home on a custom rig, but I don’t intend to try it myself any time soon.

Now, most of the model outputs included in this dataset include only averaged data across many simulation years, without any of the data on monthly (or at least seasonal) variation necessary for determining Koppen-Geiger climate zones (save for a couple cases with zero obliquity and no other seasonal forcings, where we can assume the climate varies little from the average). However, this is enough data for determining Holdridge life zones, which depend only on average data.

But there are a couple wrinkles. For one, Holdridge zones depend on average “biotemperature”, which isn’t the same as average temperature because all temperatures below 0 °C are counted as 0 °C and all above 30 °C are counted as 30 °C; so a region with a long winter below freezing will have a higher average biotemperature than actual average temperature. To account for this, I looked at some temperature data from Earth (monthly data, still not perfect for determining biotemperature but it’ll do for our purposes) and came up with a scheme for estimating average biotemperature from average temperature that about matches up with the real data without any odd discontinuities:

Tavg = Average temperature in °C
Tbio = Average biotemperature in °C
Ta = Tavg – 15
Tb = Tavg - 25

Where Tavg < -20,        Tbio = 0
Where -20 < Tavg < 15,   Tbio = Tavg + 0.013(Ta²) – 0.12(Ta)
Where 15 < Tavg < 25,    Tbio = Tavg
Where 25 < Tavg < 35,    Tbio = Tavg – 0.046(Tb²) – 0.04(Tb)
Where 35 < Tavg,         Tbio = 30

Both this conversion and the construction of the Holdridge system itself assume climates with about the same pattern of seasonal temperature variation of Earth, which will be roughly true for a lot of the models we’ll look at but certainly not all of them (though for the cases with no seasonal variation we can just assume average temperature and biotemperature are equivalent).

Also, the ROCKE-3D outputs includes data on potential evaporation, so I’ve switched my script from indexing Holdridge by biotemperature and total precipitation to doing it by biotemperature and ratio of potential evaporation to precipitation, which I figure is probably a more reliable indicator of actual water availability (I’m specifically using the “Penman Potential Evaporation” rather than “Potential Evaporation” parameter in the files, as the former seems to be closer to what the Holdridge zones were built around). I wasn’t brave enough to try constructing a system to reference all 3 parameters as intended for Holdridge zones, and frankly I’m not sure how one’s even supposed to approach that.

And finally for the oceans, I’ve excluded the tropical/temperate distinction but marked seasonal and permanent ice cover based on 20% and 80% thresholds of average ice cover.

This is far from an ideal approach but at least we can get a general idea of where major deserts, forests, tropical climates, and ice caps might be, even if we won’t be getting a lot of the nuances of seasonal climate types like monsoon and mediterranean regions.

As a reminder, here’s the color scheme I’m following for Holdridge zones:

 

 

And here’s for Koppen-Geiger zones:

 

“The Climates of Earth’s Next Supercontinent: Effects of Tectonics, Rotation Rate, and Insolation”

Way et al. 2021

Paper

Dataset

This paper models the climate for two scenarios of a speculative future supercontinents, Amasia and Aurica. For each scenario, it tests 3 types of topography: A roughly flat but slightly randomized case with all land below 200 meters, a randomized case with terrain in each cell varying between 1 and 4000 meters, and a more tailored case with mostly flat terrain below 200 meters but mountain ranges drawn in roughly appropriate areas for the tectonic reconstruction at heights of 2000 to 7500 meters. There’s also a slight difference in coastlines between models due to how their downsampling worked.

In addition to annual averages, for a couple of the cases (the randomized terrain, and also a couple control cases with modern Earth topography) the dataset also includes 3-month averages for June, July, and August and for December, January, and February. This, along with the average data, is about enough information to start estimating Koppen-Geiger zones, if I’m willing to make some assumptions. By calibrating my script on the control model of modern Earth also included in the dataset, I worked out this general scheme for estimating the necessary data (where “summer” and “winter” averages refer to whichever of the two 3-month averages was warmer and colder in a particular spot):

• On land, the max temperature is taken as 2 °C higher than the summer average, and min temperatures as 2 °C lower than the winter average (sea surface temperature is taken as about the same as the average).

• If the summer average temperature is below 15 °C, temperatures over 10 C are presumed to last under 4 months (for purposes of the Xxb/Xxc distinction).

• Average summer precipitation (for the whole warm half of the year, rather than just 3 months) is taken as half-way between the annual average and the summer average.

• For each season, the max precipitation is taken as 50% higher than the reported average, and the min precipitation as 30% lower, and the lower of the 2 is the yearly min precipitation (this was based mostly on the distribution of tropical and Mediterranean climates; I’m still getting too much of the latter but I couldn’t figure out how to get less without removing it entirely).

Here’s how this all looks for the control case (for whatever reason these ROCKE-3D models of Earth always have the Mediterranean, Baltic, Caspian, and Hudson bay filled in as land while filling in various islands as sea):

 

Even taking into account this model came out a tad cooler than modern Earth (13.5 °C average), it’s not a perfect result, but again, it’ll do.

And for reference, here’s the Holdridge zones for the control case as well, based just on annual averages:

 

In addition to the control, the dataset also includes annual average data for modern Earth topography with the day lengthened to 24.5 hours as is expected to happen in ~200 million years, coming out much the same but slightly cooler:

 

And another case with the longer day and insolation increased by 2.6%, appropriate for 250 million years in the future, which raises the average temperature to 17.7 °C:

 

All the supercontinent models are also run with longer days and higher insolation, but an atmosphere equivalent to preindustrial Earth for simplicity.

First, let’s look at Amasia, a supercontinent scenario assuming that the Pacific ocean continues to close, bringing the Americas, Eurasia, and Australia together, and then North America and Eurasia push closer together over the north pole, closing the North Atlantic, leaving a supercontinent centered over the north pole in about 200 million years, but with Antarctica remaining more-or-less static over the south pole.

 

Black lines are subduction zones, white lines are mid-ocean ridges, red lines are structures in the mantle that can influence plate motion but you don't need to worry about here. Davies et al. 2018

The result is a planet with some hefty ice caps over much of the continents:

 

Flat Topography:

 

Randomized Topography:

Mountain Ranges: 

 

Despite appearances, all these cases have higher average temperatures than Earth, it’s just that the landmasses are mostly concentrated in the colder areas, though not without some substantial tropical and temperate regions, some of why see some fairly intense monsoon rains as you might expect of subtropical continents next to an almost uninterrupted equatorial ocean. Why exactly the last case has warmer poles, I’m not sure, perhaps it’s a quirk of circulation patterns.

Next, the Aurica scenario also assumes closure of the Pacific, pulling in Antarctica as well this time, but also closure of the Atlantic, with Eurasia splitting int two along an enlarged Baikal rift, ultimately ending up with an equatorial supercontinent in about 250 million years (In these maps there’s still a strip of land at the south pole, which is required for ROCKE-3D due to a quirk in how it handles ocean circulation).

 

Davies et al. 2018

This gives us a rather different result:

 

Flat Topography:

 

Randomized Topography:

 

Mountain Ranges :

 
Despite being not much warmer on average than the Amasia models, the different arrangement of landmass gives us a starkly different outcome, with no permanent ice caps but an impressive interior desert, closer to the climate of Pangea (though this more compact arrangement is less likely to have extreme monsoon patterns).

You can perhaps see the appeal of choosing these two supercontinent scenarios: aside from predicting what future earth might look like, it shows two extremes for the ways in which land distribution affects climate and conditions for life on the surface, even if all other circumstances are roughly the same.

That’ll do for today; this isn’t the only study to model future supercontinent climates, but so far it’s the only one I’ve found so far with publicly accessible climate data outputs. If you find any others (on supercontinents or other climates), feel free to send them my way.

Buy me a cup of tea (on Patreon)