Public Climate Data Re-Explorations with the Pasta Bioclimate System

Now that I've worked out my new Hersfeldt/Pasta bioclimate classification system, I figured it'd be worth quickly looking over some of the past climate data explorations and seeing how they look in this scheme.

 

First off, I didn't originally classify my look at the Phanerozoic climate dataset as a public data exploration but it was effectively the first in the series. The dataset provides monthly data on temperature and precipitation, which as I've mentioned before is sufficient to estimate all the necessary classification parameters for the Pasta scheme. Based on the control model of modern Earth, this seems to work fairly well (with the threshold for Mediterranean zones tuned to 0.9 GrS), bearing in mind that we've already seen that the model seems a bit biased towards wet climates:

I've passed the whole dataset through the Pasta algorithm and added the maps to the repository I made for the original post, but here's a sample of the notable points I showed in that post:

I won't reiterate all that I said there, but you might have noticed that this predicts large stretches of hot or extraseasonal climates, defined by having summers hot enough to be hazardous to modern plant life and substantially inhibit photosynthesis, throughout Earth's major greenhouse episodes:

• Hot summers in the Cambrian and Ordovician wouldn't have mattered much for lack of much complex life on land, and at no point in this dataset do sea surface temperatures exceed 40 °C. Even in the Silurian and Devonian, plant life was probably largely restricted to wetter lowland climates, though our fossil record is too patchy to get a clear view of the exact distribution.
• Hot summers largely disappear in the Carboniferous but then return at the end of the Permian and remain widespread in Pangea throughout the Triassic; heat stress on vegetation from extreme temperatures has been suggested as a contributor to the End-Permian Mass Extinction, but it's a bit hard to sort out from all the other hazards and ecological stresses at that point. The extinction is also followed by a fairly slow recovery, particular in terms of vegetation cover, which is partially attributed to a series of hyperthermal events when volcanic activity caused sharp temperature rises.
• Hot-summer zones recede fairly dramatically at the start of the Jurassic but then expand again in the late Cretaceous. I have found a few papers pointing out that the high average temperatures predicted for this period could imply significant heat stress for plants and inhibited photosynthesis. There is some evidence for sparser vegetation at particularly hot points, but it's not clear if this was due to direct heat stress or just greater aridity, and it's perhaps not as dramatic a difference as we might expect if daytime temperatures were too high for productive photosynthesis for much of the year. It seems like an open question whether this implies that our estimates for Cretaceous temperatures are too high or if we're underestimating the capacity for plant life to adapt to these temperature extremes, and perhaps adaptation to heat stress may even have played a role in the diversification of flowering plants at this time.

One other note is that I've had a few questions about how boreotropical biomes might fit into my scheme, referring to regions of fast-growing, warm-climate vegetation that appear in the paleontological record at high latitudes during warm periods like the Paleocene-Eocene thermal maximum. I actually added quasitropical climates partially with these areas in mind, imagining that some areas could be warm enough for growth in winter but lack the light for it, but in practice this model predicts that these areas are, at best, temperate or subtropical; substantially warmer and lusher than today, but still with cool winters and potential frosts, and so not strictly tropical in the sense I'm using in my scheme (though this dataset doesn't quite capture the PETM specifically, I may see if I can find one that does). More broadly speaking, when paleobotanists refer to "tropical" vegetation, they're not necessarily using the term in exactly the same way as modern climatologists. Just to be safe, I did check for any potential light limits on growth by estimating the insolation at each point based on latitude and time of year (assuming Earth's modern obliquity as in the model, but I didn't bother with eccentricity) and presuming 1/4 is lost in the atmosphere, but it didn't make much difference.

 

The other public data explorations used data from ROCKE-3D, as will several future explorations. Most available ROCKE-3D datasets contain only annual, not monthly, data, which is why I mostly only tried using Holdridge life zones with them at first. However, on closer inspection I think I can pull out enough useful information from these datasets to reconstruct at least vaguely reasonable estimates for most of the parameters I need for classifying Pasta bioclimate zones. There is some variance in the model behavior depending on how it's configured, but for the studies we're interested in this seems to be a decent approach:

• Average near-surface air temperature, diurnal highs and lows, and diurnal range are provided, as well as the diurnal high of the coldest day; why this specific parameter is included, I don't know, but if we subtract the average diurnal range this gives us a rough estimate of MinT, with thresholds for cool, cold, and frigid winters tuned to 12, -6, and -25 °C.
• Minimum sea temperature can then be estimated from MinT by subtracting the difference between average air and sea surface temperatures and adding half the diurnal range to represent negligible daily change.
• Growing season length is provided as a number of days; based on some quick correspondence with some of the ROCKE-3D maintainers, this is measuring the number of days without frost, though it's counted in a slightly peculiar way such that it has a non-zero baseline, counts down in the southern hemisphere, and is essentially only counting for half the year and so has to be doubled for the full growing season. This is all easy enough to account for, and taking the resulting length of above-freezing temperatures and multiplying it by MaxT/20 °C (to account for peak temperatures during the growing season) seems to give a decent proxy for GDD, at least around the relevant biome boundaries; a resulting value of 100 works for the CD/CE boundary, 6 works for the CE/CF boundary, and no growing season at all will do for the CF/CG boundary.
• "Heating degree-days" are included, which is similar to GDD but measured as degrees Fahrenheit below 65 °F (about 18.3 °C), and seems to be reported as a monthly average; this is generally used to help estimate the energy demand for heating in cold climates, and is presumably inherited from the predecessor to ROCKE-3D, ModelE, which was built more for study of climate change and its impacts on modern Earth; but it's conveniently similar to my measure of GInt; so we can estimate the CT/CD boundary as 400 average °F-days.
• For light-limited GDD (and associated GInt), I can again estimate monthly average insolation based on the latitude, obliquity, eccentricity, and time of year, and then tune the atmospheric losses in each cell such that the average of these monthly rates matches the annual average value reported in the dataset; even though I can't exactly match up light and warm periods, I can at least take these results as a limit to the potential growing period (with the same boundary thresholds I use in ExoPlaSim).
• If we presume that the temperature extremes are roughly symmetric around the average temperature, we can very roughly estimate MaxT from MinT. Based on that, it doesn't appear that any of the cases we're looking at today have hot or extraseasonal zones, so we don't have to worry about them just yet, but if we do in the future, we can perhaps approximate the whole yearly temperature cycle as a sinusoidal wave, and then use that to estimate GInt and GDD losses in the hot periods.
• Average precipitation and evapotranspiration are directly provided, which gives us Evr.
• Potential evapotranspiration is also provided, but the values appear greatly out of scale with PET values I've usually seen reported and the model's own precipitation and evapotranspiration, about an order of magnitude too high; I haven't been able to track down an exact reason for this, but dividing the value by 7 seems to give a good match to real Ar values.
• Land and sea ice are provided as annual average cover, though in some models this seems to be limited to a maximum of 95%; I'll take an average of 94% and to represent permanent ice and 6% for seasonal ice.
  

The one major missing element I haven't been able to infer is any indication of precipitation variation across the year, so we can't map Mediterranean zones and we'll have to define semiarid zones by 0.5 Ar rather than GAr. That issue aside, the result is a pretty good match considering the limitations, though ROCKE-3D seems to tend towards a very noisy precipitation distribution:

The issue with PET does make my previous results a bit suspect, but oh well, let's have a look back through them now:

 

To start off, we first looked at models of future supercontinents, from Way et al. 2021, "The Climates of Earth’s Next Supercontinent: Effects of Tectonics, Rotation Rate, and Insolation". As mentioned in that post, in a few cases these included some additional 3-month average data for the peak summer and winter months, so for those cases I can estimate Mediterranean distribution by comparing precipitation in the hottest part of these averages for each cell to the annual average evaporation as a rough measure of GrS, with a threshold of 0.8 working well enough (I tried to do something similar to get GAr but it didn't work as well, so I left it as equal to Ar). I'll mark each of the cases where I could do this with asterixes (*).

 

Here's how that looks for the control models of modern Earth, though with simplified coastlines as is often used in ROCKE-3D:

Then for the supercontinents, each modeled 3 times with essentially flat terrain, randomized terrain, and predicted terrain with mountain ranges, with 3-month averages provided for the randomized terrain. First, Amasia, the polar supercontinent:

And next, Aurica, the equatorial supercontinent:

No major surprises compared to what we saw in original post, though this perhaps emphasizes how much of a difference the choice in topography makes (and that ROCKE-3D seems to be a bit better than ExoPlaSim at modelling the orthographic effects that produce pluvial zones along mountainsides).

 

Next, we looked at a couple studies modelling a habitable early Mars with a Hydrogen-rich atmosphere. First, Guzewich et al. 2021, "3D Simulations of the Early Martian Hydrological Cycle Mediated by a H₂-CO₂ Greenhouse", which looked at Mars 3.8 billion years ago with a thick CO₂-H₂-CH₄ atmosphere. Originally I showed only a few models because most turned out largely the same, but with PET tuned down there does seem to be more variation here (the depths noted here indicate that this depth of water was placed over the planet's surface and then allowed to gather in lakes—or mostly freeze onto Tharsis, in practice—and the first case had only wet initial soil for moisture):

Rather than Arrakis we're getting a bit more interesting hints here of a world that could support small patches of forest or broad stretches of desert annuals despite lacking any major bodies of water (presuming such vegetation is present on the planet, of course; we can only speculate on whether Mars could have developed any life in its past in reality, but I'm thinking more generally about the sorts of habitable dry planetary climates these models suggest could appear elsewhere). And then in that last case a single small sea provides enough moisture to sprout forests over much of the sourthern hemisphere. This is a point I'm sure I'll return to in later posts, but I think people sometimes imagine too much of a hard binary between dry Dune-like planets and wet Earth-like worlds, where there's a lot of potential for semiarid in-betweens.

But for those who do like wetter worlds, the most interesting case is the model with a large northern ocean:

One thing to note is that, with 32% the solar flux of Earth, light levels are a much more stringent restriction here, hence the widespread proliferation of quasitropical and boreal zones (though it also helps make the planet less arid, as less light reduces evaporation even at a set temperature). Just in case anyone thinks hypothetical Martian life might be better adapted for low light or is just interested in the effects of temperature alone, I've also worked up maps with no added restrictions on growth periods from light levels; for the dry models, it mostly makes little difference:

But for the model with oceans, there's a starker difference between the tropical northern lowlands and boreal southern highlands:

And of course the study also included a number of models with a hypothetical early Martian topography with a shifted polar axis and lower Tharsis plateau, here with light-restricted growth:

And here without:

A number of more interesting cases of partially habitable worlds here, though as mentioned in the first post, bear in mind that the "lakes" in some of these later cases are really more of broad seas. If we count all areas with over 94% lake area in the model as seas, we get some bizarre coastlines due to the irregularity of the topography but a better idea of how much open water there is:

Most of the models with modern Mars topography don't gather such large lakes, but a few crater lakes do show up in the case with oceans:

Moving on, the second paper we looked at in that post,  Schmidt et al. 2022, "Circumpolar Ocean Stability on Mars 3 Gy ago", modeled Mars 3 billion years ago with a somewhat thinner CO₂-H₂ atmosphere, investigating a potential "cold and wet" climate state. First, here's the models with 20% H₂, with varying tilts and with and without restrictions on GDD from light:

Similarly warm to the global oceans case in the last paper, but a tad drier, with a vast equatorial desert dividing lusher polar regions (depending on obliquity). This model also used dynamic lake hydrology and there are a few in craters in the south like the last model, but none particularly large so I won't bother showing them. For the colder cases with 10% H, the dataset did include monthly data, so we can approach these a bit more normally with the full range of bioclimates:

This helps to give us a bit more nuanced of a look at the cold interior: large stretches are below freezing but too dry to accumulate ice and glaciers, and between these barrens and the hotter lowland deserts are some potentially hospitable patches of tundra or steppe. The high-obliquity case even gives us an interesting suggestion of a vast circumpolar steppe, sustained by short, hot summers.

That's all the public datasets we've looked at so far. There's plenty more to look through using this new approach, including some of habitable Venus that I'll probably look at next time we do a public climate data exploration. See you then.

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

Also, for anyone not aware, I'm participating along with a number of other channels in a worldbuilding festival this weekend, September 5-7. I'll particularly be involved in the opening event on the 5th, host a "speed worldbuilding improv" event on the 6th, and then join Conlangery for a tutorial on building naming languages later that day. If you'd like to participate a little, you can suggest instructions for me to follow for the speed worldbuilding event (instructions on this sheet; don't worry if you can't attend at the time, I'll leave the video up on my channel afterwards). Check out the other events on the playlist, there's a lot of good stuff in there.