Home Creators Posts Import Register Favorites Logout

Content

A couple experiments with ExoPlaSim at t85 resolution (128 by 256 cells). The first is one for Earth, done by a friend at cartographer's guild, which shows a somewhat interesting mix of improvements and issues compared to T42 which I usually run at. The second is a T85 run for Teacup Ae which came out a tad warmer than I wanted but should help as input into the terrain generation process.

Files

Comments

Stephen Rider

Looking at the previous Teacup Ae map and this one I have to wonder if what we're seeing is primarily the tightening of costal influence. I see some small regional changes and the biggest change is on the Eastern side of the ">" shaped continent.

Nikolai Lofving Hersfeldt

Bear in mind that this run is both hotter and higher-obliquity than the one I put up for the exoplasim tutorial, though it does look like the better-resolved coastal geology in several areas is having an effect on precipitation patterns.

Stephen Rider

that makes 1:1 comparison a little tougher. I really need to figure out why I keep getting an error with HDF5 when I try to get ExoPlaSim to initialize. And also finish my tectonics.

Nikolai Lofving Hersfeldt

Is it happening at the end of the first year? Might be an issue with how you've set the postprocessor.

MorelockTheWarlock

I take it this was done on a personal machine and not on AWS as previously discussed? Also, I don't suppose T63 resolutions are working yet? - Erik M.

Nikolai Lofving Hersfeldt

Both were on personal, and T63 seems to still be broken. I'm still mostly going to be doing T42 runs, I might just do another T85 of teacup once I have the topography hammered out.

MorelockTheWarlock

I see. Do you have an idea of the machine specs and runtime?

Nikolai Lofving Hersfeldt

The teacup run was on a laptop with an 8-core i7-4710HQ and 8 gb RAM, the earth run all I really know was it was 16-core, I didn't track the exact runtime, but it seems that it scales roughly linearly with the number of cells; all else being equal (including number of layers), a T85 run should take 4 times as long as a T42 run. The hitch is that T85 is usually less stable and so may require more restarts or shorter timesteps, which will add to a greater effective runtime.

MorelockTheWarlock

That's a little less imposing than I was expecting and probably further reinforces the idea that my own issues arose from my world and not the hardware. That and I was trying to run it without restarts. What timestep and how many layers did you settle on?

Stephen Rider

I pushed a little more on the HDF5 error and it turns out that was solved by installing netcdf-devel, which got me to the next problem, the exoplasim.pyfft. I tried deleting the files and reinstalling and that doesn't fix the problem. Any chance reinstalling from the OS up would fix it? Because that's on the table.

Nikolai Lofving Hersfeldt

The pyfft problem I think relates to exoplasim needing a slightly older version of numpy (before 1.22 I think?), best fix seems to be to remove exoplasim and numpy and then let pip install exoplasim deal with the numpy install

Nikolai Lofving Hersfeldt

I used 10 layers for this, which I think is fine but one of these days I should test how much of a difference 20 layers would make for either t42 or t85. I have a sort of management script set up to restart runs when they crash and reduce the timestep if they do it a lot; it starts at 30 minutes and goes down to 10, I didn't check where it ended up in this case.

MorelockTheWarlock

The management script sounds useful! When it’s convenient, if you would share it with any necessary instruction, I would be appreciative.

MorelockTheWarlock

Additionally, do you know if ExoPlasim considers smaller straits and waterways that are otherwise not found on low resolution maps? Something like the Bosporus or Gibralter won’t make it onto the final product for a T42, but I was curious whether they still factored into its calcalculaitons. I realize their effect on a global scale might be limited, but again, just curious.

Nikolai Lofving Hersfeldt

It does not. The ocean model is fairly limited (in essence, it's modeled not so much as a fluid as just a box in each cell that can hold heat and potentially freeze over) and the only topography input is a single elevation value in each cell

MorelockTheWarlock

Interesting. I'm not sure how simple the oceanic current modeling is, but the lack of Panama in your T42 explorations could be making Europe colder than it should be.