[WIP] Building a world from tectonics onward

[MrBragg]

Another piece finally in an alpha state!

rhur_scaled_colored_v1.png

[Peter Toth]

As always, congrats on achieving such realism with your topography! I'm sometimes wonder if our methods have any similarities or common starting points. But if there's one thing I know for sure, your method is FAR superior over mine.

A question I've been wondering: When you determine your climate zones (I'm assuming near the end of your project), are you planning to reduce some of the erosion on some of your mountains where there is desert, or is the river flow / erosion already accounted for, assuming that you've already determined your basic climatology in advance? (In my case, I worked out the basic climate first, before detailing the topography, just to gauge where the river channels and erosion should be. When I'm finished the fine-tuning, however, I'm planning on doing a final run with ExoPlaSim using the T63 model, to finalize the climatology.

At any rate, keep at it, for I'm sure we're all looking forward to that final project!

Peter

[MrBragg]

Thanks for the kind words, Peter. I think there's definitely some similarity between our approaches, though it sounds like you have way more patience for Wilbur then I do Just about all of my work goes into getting the heightmap the way I want it in Gimp and Wilbur is just the final step to add the highest frequency details / rivers. In any case I quite like your results and always enjoy when you post something new.

Your question about climate is a good one since there's definitely feedback between topography and climate. Far back in this thread--after a few pretty disastrous initial attempts--I got a rough sketch of climates that seemed plausible given a rough idea of what the topography would be. Using that as a guide, I use partial opacity masks to tailor the amount of erosion applied in Wilbur when I do the detailed topography; e.g. wet areas use 100% opacity while putative deserts use something like ~25% opacity. It's an undeniably imperfect approach, and if I had infinite time and patience I could in theory iterate between topography and climate generation, though I'm hoping it captures things well enough to not be immediately offensive I'm also planning to dig into ExoPlaSim once my topography is sufficiently done, so I'll be very interested to see your results since you'll invariably get there before I do!

[MrBragg]

Finally closing in on finishing the first round of topography here. At this point I'll probably start playing around with ExoPlaSim a bit to try and figure out the extent of the ice cap on the remaining southern continent, then do a first pass there and then maybe take a bit of a break

map6_topography_smallish.png
Downsized from the orig 24k x 12k

[Peter Toth]

As usual, the topography looks absolutely stunning. Is there a particular reason why you didn't include polar continents? I always admire the way some mappers can seamlessly meld two different projections to account for curvature near the poles. That is evident on your southern continent (on the bottom of your conworld.)

I'm looking forward to seeing the ExoPlaSim results. Congratulations on making it this far, for your are very much an inspiration!

Peter

[MrBragg]

Thanks Peter. I actually did initially start with a south polar continent, but in order to get a decent tectonic model that needed to move a bit north and off the pole. To account for the projection of continents with severe distortion (anything poleward of ~45 N/S), the gist is that I typically make the “canonical” map of a continent in oblique equirectangular and then reproject that finalized map back to normal equirectangular.

Now, onto my first foray into ExoPlaSim. Before diving into ExoPlaSim, I need to give a shout out to Nikolai over at Worldbuildingpasta for his excellent introduction and overview of the program as well as for making his conversion scripts available. Without that kind of support, this type of program would be totally inaccessible, so a big thanks!

Part I: Resolution

I first wanted to get a sense for how the resolution of the input map affects the climate output. I used input resolutions of T21 (32 x 64), T42 (64 x 128 ), and T85 (128 x 256) and kept all parameters the same between runs except for the addition of the physics filter for T85; all parameters were kept to Earth values except for the days / year, which I set to 360 to make math easy. Using 16 cores @ 2.3 GHz, running these simulations from start to balance took roughly 6 hours, 24 hours, and 94 hours, respectively (~100 ± 10 model years), which scaled just about linearly with the increased number of pixels. To generate the Köppen maps, I averaged 20 model years together (all after balance was achieved) and shifted the coldest month in the northern hemisphere to ~January; from there I output the climate map without interpolation or other fiddling with the defaults.

scaled_T21_16proc.png

At T21 resolution it’s hard to make out much beyond vague climate bands: There’s tropical, desert, temperate, and... not much else. Things are a fair bit warmer than I would initially expect—21.4 C global average—which is probably at least partially due to the lack of polar ice caps and lack of any land at the north pole (ClimaSim gave similarly warm northern temperatures). There’s also likely a systematic temperature error: Even when working with earth, Nikolai observed an average temperature of 16.5 C with a 300 ppm CO2 model, which is several degrees warmer than the 20th century average of 13.9 C. Anyway, BWh also seems a bit more widespread than expected, but at this resolution a lot of nuance is just missing and so it’s hard to say too much more.

scaled_T42_8mo_scaled.png

At T42, it’s immediately apparent how helpful quadrupling of the number of pixels is. Areas that used to be flat Cfb or BWh are revealed to have a lot more texture, so that’s nice and fun. A lot of the systematic errors that Nikolai has previously identified are apparent here—ice caps being too small, northwests of continents being too cold, and perhaps somewhat more expansive arid climates—and things are still too warm (20.9 C) to my totally untrained eye (though notably the “seasonal sea ice” has expanded quite a bit). There are also still some other oddities; ExoPlaSim really seems to like hot deserts and goes so far as to put them at the same latitude as tundra just a few pixels away. Generally though, this seems to capture the trends we’d expect for a fairly warm planet.

scaled_T85.png

Finally, as expected, T85 resolution is even more detailed and a lot less blocky. Unlike the change from T21 to T42 though, the overall picture here doesn’t change all that much; there’s a lot more detail, but there aren’t really any regions of major qualitative shifts to the climate. The temperature here is 22.1 C, making it appreciably higher than for either T21 or T42 and signalling that there may not be a consistent trend between resolution and computed temperature.

Overall, there seems to be a significant qualitative shift in going from T21 to T42 and a more subtle refinement at T85. This is kind of good news for my planned approach, since it seems to indicate that already at T42 the major effects are captured and so I can refine parameters at T42 and wait until things are set before doing a final pass at T85. I do like the improved detail at T85, so at the end I can probably justify the ~4 days of computational time that it takes.

Part IIish: CO2

scaled_T42_230ppm.png

As a bit of a side experiment, at T42 resolution I also lowered the CO2 a bit to try and tone down the average global temperature. Starting with the output from the 300 ppm model and then re-converging to balance at 230 ppm definitely results in significant cooling; the average temp drops from 20.9 C to 18.9 C, still quite warm but better than before. The poles are affected much more strongly than the tropics (a southern ice cap starts to develop), which is consistent with our own current global experiment with CO2 levels. Precipitations patterns also change, though there it’s a bit more of a mixed bag despite an overall drop in global precipitation.

230_minus_290ppm_tas.png 230_minus_290ppm_precip.png

[Caenwyr]

The poles are affected much more strongly than the tropics (a southern ice cap starts to develop), which is consistent with our own current global experiment with CO2 levels.

You made me snort coffee out of my nose with that offhand remark. You monster!

[Naima]

Thanks Peter. I actually did initially start with a south polar continent, but in order to get a decent tectonic model that needed to move a bit north and off the pole. To account for the projection of continents with severe distortion (anything poleward of ~45 N/S), the gist is that I typically make the “canonical” map of a continent in oblique equirectangular and then reproject that finalized map back to normal equirectangular.

Now, onto my first foray into ExoPlaSim. Before diving into ExoPlaSim, I need to give a shout out to Nikolai over at Worldbuildingpasta for his excellent introduction and overview of the program as well as for making his conversion scripts available. Without that kind of support, this type of program would be totally inaccessible, so a big thanks!

Part I: Resolution

I first wanted to get a sense for how the resolution of the input map affects the climate output. I used input resolutions of T21 (32 x 64), T42 (64 x 128 ), and T85 (128 x 256) and kept all parameters the same between runs except for the addition of the physics filter for T85; all parameters were kept to Earth values except for the days / year, which I set to 360 to make math easy. Using 16 cores @ 2.3 GHz, running these simulations from start to balance took roughly 6 hours, 24 hours, and 94 hours, respectively (~100 ± 10 model years), which scaled just about linearly with the increased number of pixels. To generate the Köppen maps, I averaged 20 model years together (all after balance was achieved) and shifted the coldest month in the northern hemisphere to ~January; from there I output the climate map without interpolation or other fiddling with the defaults.

scaled_T21_16proc.png

At T21 resolution it’s hard to make out much beyond vague climate bands: There’s tropical, desert, temperate, and... not much else. Things are a fair bit warmer than I would initially expect—21.4 C global average—which is probably at least partially due to the lack of polar ice caps and lack of any land at the north pole (ClimaSim gave similarly warm northern temperatures). There’s also likely a systematic temperature error: Even when working with earth, Nikolai observed an average temperature of 16.5 C with a 300 ppm CO2 model, which is several degrees warmer than the 20th century average of 13.9 C. Anyway, BWh also seems a bit more widespread than expected, but at this resolution a lot of nuance is just missing and so it’s hard to say too much more.

scaled_T42_8mo_scaled.png

At T42, it’s immediately apparent how helpful quadrupling of the number of pixels is. Areas that used to be flat Cfb or BWh are revealed to have a lot more texture, so that’s nice and fun. A lot of the systematic errors that Nikolai has previously identified are apparent here—ice caps being too small, northwests of continents being too cold, and perhaps somewhat more expansive arid climates—and things are still too warm (20.9 C) to my totally untrained eye (though notably the “seasonal sea ice” has expanded quite a bit). There are also still some other oddities; ExoPlaSim really seems to like hot deserts and goes so far as to put them at the same latitude as tundra just a few pixels away. Generally though, this seems to capture the trends we’d expect for a fairly warm planet.

scaled_T85.png

Finally, as expected, T85 resolution is even more detailed and a lot less blocky. Unlike the change from T21 to T42 though, the overall picture here doesn’t change all that much; there’s a lot more detail, but there aren’t really any regions of major qualitative shifts to the climate. The temperature here is 22.1 C, making it appreciably higher than for either T21 or T42 and signalling that there may not be a consistent trend between resolution and computed temperature.

Overall, there seems to be a significant qualitative shift in going from T21 to T42 and a more subtle refinement at T85. This is kind of good news for my planned approach, since it seems to indicate that already at T42 the major effects are captured and so I can refine parameters at T42 and wait until things are set before doing a final pass at T85. I do like the improved detail at T85, so at the end I can probably justify the ~4 days of computational time that it takes.

Part IIish: CO2

scaled_T42_230ppm.png

As a bit of a side experiment, at T42 resolution I also lowered the CO2 a bit to try and tone down the average global temperature. Starting with the output from the 300 ppm model and then re-converging to balance at 230 ppm definitely results in significant cooling; the average temp drops from 20.9 C to 18.9 C, still quite warm but better than before. The poles are affected much more strongly than the tropics (a southern ice cap starts to develop), which is consistent with our own current global experiment with CO2 levels. Precipitations patterns also change, though there it’s a bit more of a mixed bag despite an overall drop in global precipitation.

230_minus_290ppm_tas.png 230_minus_290ppm_precip.png

~4 days ??? Wow that's dedication surely. I could barely wait 1 hour.
What pc do you have ?

[MrBragg]

~4 days ??? Wow that's dedication surely. I could barely wait 1 hour.
What pc do you have ?

Haha, I actually do a fair amount of computational chemistry, so my computer has 20 physical cores and I'm used to waiting days (or weeks) for results

[Tiluchi]

This looks amazing so far! Will be interested to see what happens when you compare the Azelor tutorial version with the output from ExoPlaSim. Still impressed you were able to get it all running- I took one look at Nikolai's code and turned tail and ran.

Seeing these outputs makes me wonder if part of what's going on is that ExoPlaSim (and ClimaSim apparently) really like continental climates, and overstate the high summer temperatures that come with low pressure systems over land, or maybe understate the effect of high pressure systems in the wintertime. I don't know enough about how the program works to know if it's overestimating the atmospheric pressure itself, or just the temperature effects of that. Either way, maybe the trick is to figure out what attenuates atmospheric pressures systems (or their temperature effects) on Earth and determine if that also applies to our conworlds. But this is all conjecture, and I'm not really knowledgeable enough about climatology to say one way or the other.