Climate Mini-Exploration: Checking Randall Munroe's Work

Way back in 2012, one of the early entries in the xkcd What if? series pondered the potential climate of Earth if its geography were rotated by 90 ° relative to its current axis of rotation; specifically, it treated the Cassini map projection, a transverse aspect of the equirectangular projection, as if it were the normal aspect, with the equator running along what was once the prime meridian. It's a neat little case study in how global air circulation patterns determine local climate, and was recently reworked into a video, so now may be a good time to mention that I've run a proper climate model of this scenario:

 

For anyone not familiar with my earlier work with ExoPlaSim, the above is a map of this world's climate classified according to the Köppen-Geiger climate classification system, a popular standard for classifying Earth's climate, as simulated by the ExoPlaSim climate model. ExoPlaSim is a GCM, which stands for either "Global Climate Model" or "General Circulation Model" depending on whom you ask and is the standard type of model used for simulating the climate of both Earth and exoplanets. These are generally fairly complex and specialized packages of software run on supercomputers, but ExoPlaSim is designed as an "intermediate" model, with the same general construction and similar level of detail as top-end GCMs, but with various simplifications to the internal calculations that make it much faster. The original idea was that this would allow researchers to run dozens or hundreds of models of slightly different scenarios for exoplanets on supercomputers rather than having to pick one—and then perhaps they could pick the most interesting of those results and run them again in one of the top-end models—but this coincidentally also makes it possible to run the model on a regular consumer-grade laptop; though it's not particularly easy and a typical model run can still take days to weeks to complete (I ran this one a while ago and was going to post it with a grab-bag of other one-off models but I figured I'd show it now that it's a bit topical).

So, the blocky appearance of the map above shows the way the model represents a planet's surface, as a 128 by 64 grid of cells, each with a given surface type (land or sea), elevation, and properties like temperature, vegetation, snow cover, etc.; and then above each cell, the model simulates 10 atmospheric layers. The model starts out in a simple starting state and then moves forward in short timesteps (usually 30 minutes in my runs), predicting the shift in air masses, heat, and moisture over that time and simulating daily and seasonal sunlight patterns, precipitation, ice formation and retreat, etc. It runs through simulated days and years in that way until it reaches a state of radiative equilibrium, where the total energy radiated from the planet into space as heat through the year is about equal to the total energy absorbed from sunlight, which is generally a good indication that the climate is stable and unlikely to change much more year-to-year; in this case this took 70 model years and about a week of real time. The model than writes out data on various climate parameters for each grid cell throughout the year, such as average temperature and precipitation, and I run that data through a python script I made that compares it against the standard definitions for Köppen-Geiger zones and draws out a map of the results.

One trick I sometimes like to pull is to interpolate the data to higher resolution (essentially just smoothing the data into continuous curves between the blocks) and then adjust the temperature at each point based on elevation differences too small to appear at the model resolution. This isn't a perfect representation of how air masses will move around topographic features, and I also can't predict how they'll influence precipitation patterns, so this isn't really an accurate picture of how these climates would appear at this level of detail, but it at least looks a bit nicer:

If you're having trouble reading these colors, though, I also sometimes like to turn these maps into a "true color" approximation; I compared a map of Earth's real Köppen-Geiger zones to a satellite map, found the average color of land in earth zone, and then copied that to the corresponding zones on this map:

Before looking at these results in more detail, a few notes on some of the weaknesses of this model:

First, ExoPlaSim doesn't properly model ocean currents. Various processes on the surface of oceans are modelled like ice formation and heat exchange with the atmosphere, which moderates seasonal temperature variation along coastlines, but the oceans aren't really treated as fluids, and aren't modeled at all below about 50 meters' depth (which is one of those simplifications that substantially reduces the model runtime). The influence on ocean currents on global climate is often somewhat overstated—much of their purported impact of the North Atlantic Current on Europe, for example, has as much to do with seasonal wind patterns as ocean heat transport—but there's still a general tendency for this model to make west-facing coastlines at mid-to-high latitudes a bit too cold, east-facing coastlines a bit too warm, and equatorial climates overall a bit too warm compared to the polar regions.

Second, you may have noticed the lack of any large ice caps over the poles here: ExoPlaSim has a fairly simplistic model of glaciers that doesn't represent how they flow across the surface, so it tends to underestimate their extent, though even if it did have a better model it would probably have to be run for thousands of model years to fully equilibriate. But even aside from the quirks of this model, there are some tricky feedbacks around ice formation: ice reflects away large amounts of sunlight, thus cooling the planet and so preventing itself from thawing to some extent, which means that a planet that already has some ice to start with will tend to equilibriate with a cooler climate than a planet without any initial ice. This means that often there's not really a single definitive climate for a given set of parameters for a planet, but multiple possible stable climate states. Climate researchers will often talk of "hot start" and "cold start" models (though there's not necessarily always just two possible states); Earth today is essentially a cold start case, having warmed up from a colder glacial period, whereas ExoPlaSim models tend to be more-or-less hot starts, with low initial temperatures but no initial ice cover so it warms up pretty fast; there are ways to run it as more of a cold start, but this tends to take much longer to reach equilibrium. Given the above issue with ocean heat transport, there's also an extent to which these two issues somewhat cancel out to give the closest match to Earth's mid-latitude climates.

Past that, a few more local issues: the map that I used for Earth's elevation was measured from the top of the ice sheets on Greenland and Antarctica, so they're effectively treated here as high plateaus. There are some models for the likely topography of these landmasses were all ice removed and the crust allowed to rebound to a new equilibrium, but these still have some substantial highlands and don't necessarily represent how these landmasses would develop in this world, with different patterns of erosion and deposition, so in this case I didn't spend the time to try to splice them into my elevation map (not to say that it wasn't also out of laziness); suffice it to say that with more realistic elevation, these landmasses would probably still have colder highlands, but perhaps not as widespread as shown here.

And finally, the coarse resolution of the model also means it can't always resolve small-scale climate dynamics along coastlines or along the edges of mountain ranges, so it tends to underestimate rain a bit in these areas. ExoPlaSim also seems to have a bit of a bias towards Mediterranean-like dry-summer, wet-winter climates, and I'm not totally sure why; perhaps as a side-effect of the colder summers at high latitudes.

You can look at my previous analysis of how well this model replicates Earth's real climates for more detail on the effects of these errors, but the upshot of all that is that while this model gives us a more reliable and detailed look at the potential climate of other worlds like this, it's still best approached in broader strokes, looking at the spread of different climates across continents, rather than getting too caught up in the exact boundary lines or local oddities.

Now, the original What if? article is actually a pretty good template for how to sketch out an Earthlike planet's climate if you're not interested in dealing with fancy climate models or finding specific measures of temperature and precipitation. It does gloss over a few details, though, such as how Randall goes from the main circulation cells to his map of prevailing wind, which includes details like the tendency for winds to form anticlockwise loops in the mid-latitude oceans:

The predicted climate in the article, showing blue cold climates, green wet climates, and tan dry climates on land, prevailing winds, and hurricane regions highlighted in red.

I couldnt say if this comes from a deeper theoretical understanding he isn't mentioning or just staring at maps of Earth's prevailing winds long enough to develop an intuitive sense for how winds move around landmasses, which can work better than you might expect for a case like this where the global parameters of the climate aren't too different.

At any rate, the modelled wind patterns are generally pretty close to what the article predicts over the oceans, but the continents are a bit trickier. But an important element of Earth's climate is seasonal variation in wind patterns; as the continents in either hemisphere alternately warm and cool, the convergence zone between the Hadley cells shifts towards the hottest areas, and the other atmospheric circulation cells shift in turn, influenced as well by temperatures at their latitudes.

Average prevailing winds in the model on top with arrows indicating direction, and monthly winds at different points in the seasonal cycle below.

The bands of high rainfall at the convergence zones follow along, which creates heavily monsoonal rain patterns on many of the subtropical landmasses: This is partially why parts of South America and Australia are perhaps a tad wetter than the article predicts. But western Asia and North America are a bit drier than expected, despite bordering the tropical Arctic. This is because the vast landmasses of Eurasia and North America form a particularly strong monsoon pattern, and as tropical monsoon winds cross the equator they tend to curve towards the east; thus, for most of the year the moisture from the Arctic is carried either northeast into coastal Europe or southeast into Greenland, with relatively little travelling west like tropical winds usually do. The article actually somewhat anticipates this possibility, mentioning a similarly monsoon pattern around the real Indian Ocean that keeps Somalia dry.

Grey areas show cells with no reported preipitation in these months.

Africa also has heavy monsoon rains but they're a bit more broadly distributed, so rather than many particularly wet or dry regions there's more a range of semiarid climates with bursts of summer rain and long winter droughts—somewhat like Africa's real climate, but rearranged and with the northwest desert being not quite as dry.

I won't comment too much on the predicted storm patterns; ExoPlaSim is theoretically capable of predicting storm activity, but this is only really reliable at higher resolutions, which run much slower and are more prone to crashing. I'm a bit skeptical on the predicted tornado activity in Africa—I would've thought central Eurasia here would be the better bet—but I don't have any data to back that up. With regards to the predictions for local climates, though, a few notes beyond what I've already said:

• In addition to monsoon rains, South America gets consistent onshore winds from the Atlantic to its west and gets fairly heavy snowfall against the edges of the Andes in winter, so is overall a fairly wet continent, though the total discharge of the Amazon basin is still probably a good deal lower.
• Even accounting for the overestimated highlands here, there's still a decent chance that some part of the northwest coast of Antarctica would lie under a rainshadow created by the interior highlands, much like the equatorial west coast of real South America.
• Japan and Korea here might have similar temperatures to their real counterparts, if not a bit warmer, but are rather drier, with a Mediterranean-like pattern of dry summers and wet winters.
• Cairo here is in one of the driest parts of the African desert, with intensely hot summers, low winds, and little moisture coming off the Mediterranean.
• Los Angeles is in a cool submediterranean climate here; the heavy rains the article is expecting come further to the south, in Baja California.

These are nitpicks, really; overall the predictions are a pretty good effort which hold up fairly well, and this little article is deservedly well-regarded in certain internet cultural niches.

I will take the opportunity to address one incredibly minor and esoteric gripe in relation to this article, though: as an example of the complexities of global climate, it mentions a study claiming that a large amount of the nutrient-rich dust falling on the Amazon comes from one valley in Africa. This is a fair enough point to be made in this article, but I've often seen this claim misinterpreted elsewhere to mean that the Amazon rainforest depends on this single nutrient source and would be absent without it. For one, the particular claim of that study hasn't been borne out, but at any rate this vastly overestimates how sensitive global biogeography is to factors like this. A hot, wet climate like that of the Amazon basin fosters thick rainforests in plenty other regions around the globe without analagous sources of dust, and seems to have generally done so in the past; the Amazon forest itself has persisted for at least 66 million years despite substantial changes in geography and global climate over that period. There is a degree to which rainforests will encourage their own rainfall, but that only enhances a preexisting pattern. Extra nutrient sources like wind-blown dust can further help to create a broader, lusher forest, but again that's only an addition to rainforest distribution rather than a necessary prerequisite.

There's perhaps a broader point to be made here about the tendency to assume that because Earths climate system is so complex and interrellated, any small change must have dramatic and obvious consequences (though to be clear, increasing atmospheric CO₂ levels by over 50% doesn't constitute a "small" change), but that's a matter for another time. If you liked this little look at the climate of an alternate Earth, I've done a number of explorations of substantially more exotic climate scenarios you can check out. I've also recently been working on constructing my own climate classification system to better describe these worlds, so this gives me as good an opportunity as any to give it a little test run; which also highlights how the different distribution of continents seems to create greater seasonal extremes of temperature than any we see on Earth, peaking at daytime maximums of over 50 °C over much of the subtropical arid and semiarid regions (as indicated by the H and E climates here, see that previous post for all the specific definitions).

 That'll do for today, see you next time.

Buy me a cup of tea (on Patreon)