How to achieve thickness variation of the subducting slab along-strike

[WenjingZhu56]

Hello everyone,

I am using the World Builder to define a 3D subducting plate with along-strike variations in slab thickness, but encountered a "slab gap".

I tried to define this subducting slab with thicknesses variation (e.g., 66 km at Y=0 km→ 37 km at Y=400 km → 66 km at Y =800 km) by using "sections":

{"model":"subducting plate", "name":"Slab",
     "coordinates":[[1000e3,0],[1000e3,400e3],[1000e3,800e3]],
     "dip point":[1500e3,800e3],
     "segments":[{"length":400e3, "thickness":[66e3], "angle":[10,60]}],
     "sections":[
	{"coordinate":1,
	 "segments":[{"length":400e3, "thickness":[37e3], "angle":[10,60]}]}
      ],
...

But the World Builder generates a discontinuity in the slab near the section boundary. This appears in both composition and temperature. The figure below shows the discontinuity of the oceanic mantle lithosphere (red) around approximately 360–400 km.

I am looking forward to any guidance. Thanks to all of you!

[danieldouglas92]

Hi @WenjingZhu56,

Can you please upload the whole worldbuilder file and the grid file that you used to generate this discontinuity in the slab?

Cheers,
Daniel

[WenjingZhu56]

Hi @danieldouglas92,

Thanks for your request.

I am not sure about the grid properties in the grid file. Therefore, I have uploaded the worldbuilder file and parameter file I used to test the initial geometry in Docker. The version of ASPECT is 3.1.0-pre, and the version of Geodynamic World Builder is 1.0.0. May be you could directly generate a solution.pvd through them.

Thicknessvariation.zip

Thank you!

[danieldouglas92]

Hi @WenjingZhu56

I was able to reproduce your issue (thanks for sending all the files over!). I was able to fix the gap in your slab and the result looks correct to me (see the screenshots below). I did this using the changes that are being proposed in this Pull Request #799. Can you please try running your model with this branch and see if it works for you as well?

Here is the view of the slab from the top comparing the current main branch with the branch introduced in PR #799

and here is the view of the slab from below, you can see the variable thickness when using PR #799.

[WenjingZhu56]

Dear @danieldouglas92,

Thanks a lot for testing my files and showing the fixed results which look great！

Another question is how to use the branch introduced in PR #799 on the supercomputer.

ASPECT is already installed on the supercomputer I use, and I normally run my models by simply loading the ASPECT module. If I want to use the changes from PR #799, does this mean I need to compile and install a new version of ASPECT myself (including this branch) in my own directory? Or is there another way to use the updated World Builder with the system-installed ASPECT module?

Thanks again for your help!

[danieldouglas92]

Hey @WenjingZhu56

Happy to help! Hopefully it isn't too much of an issue to do this on the supercomputer you're using, but you can tell ASPECT which version of World Builder to use when you compile ASPECT. By default, ASPECT will use the version of World Builder stored in the contrib folder in the ASPECT directory. To explicitly specify a World Builder version, you need to add the flag -D WORLD_BUILDER_SOURCE_DIR when running cmake. So when compiling ASPECT (assuming your build directory is within the ASPECT_SOURCE_DIRECTORY), the command would look like this:

cmake -DDEAL_II_DIR=/path/to/dealii -D WORLD_BUILDER_SOURCE_DIR=/path/to/WorldBuilder ..

and as long as you point to the WorldBuilder branch in #799 this should work with ASPECT.

Let me know how this goes and if you have any other questions!
Cheers,
Daniel

[danieldouglas92]

@WenjingZhu56

Alternatively, something that will work for you right now (and will be much easier) is if you just make your trench points not exactly colinear. The changes introduced in #799 handles the edge case where trench points lie exactly on the same line, so if you shift your middle trench point by just 1 m to [[1000e3,0],[1000.001e3,400e3],[1000e3,800e3]] instead of [[1000e3,0],[1000e3,400e3],[1000e3,800e3]] this also removes the slab gap when I tested it!

[WenjingZhu56]

Thank you so much @danieldouglas92 !

Indeed! Just slightly shifting the middle point generates a continuous slab! That is a much easier way to handle my problem.