search

geodynamics / aspect

A parallel, extensible finite element code to simulate convection in both 2D and 3D models.
https://aspect.geodynamics.org/
Other
228 stars 237 forks source link

Chunk geometry model gives different results when not centered around zero meridian #3969

Open anne-glerum opened 3 years ago

anne-glerum commented 3 years ago

While working on #978, I noticed that for both the Chunk and the TwoMergedChunks the velocity solution becomes several OOM bigger when the chunk is no longer centered around the (0,0) coordinate. E.g. my reference model runs from -20 to +20 for both latitude and longitude, and the other models change the longitude from [-80,-40], [-40,0], [0,40], [40,80], [160,200]. The velocity becomes higher (1.5 m/yr) on the bottom boundary, which has tangential boundary conditions. This velocity is directed diagonally, see figure below. The setup has prescribed nonzero velocities on the top 290 km of the lateral boundaries, with zero velocities below. Velocities for the centered model are on the order of the prescribed velocities (e.g. RMS, max velocity: 0.00772 m/year, 0.0284 m/year for a prescribed velocity of 0.01 m/yr at low resolution).

The figure shows inner boundary of the centered Chunk model and one from 40 to 80 degrees longitude. Only the Minimum longitude and Maximum longitude are changed. Velocity_anomalies_with_varying_longitude_correct_vel

Centered around 180 degrees longitude, the velocities are of the correct OOM, but not the same as for the model centered around 0. Gravity is in the right radial direction. Inflow should equal outflow (set Function expression = if(sqrt(z*z+y*y+x*x)>6081e3,-sin(atan(y/x))*vel_mag,0); if(sqrt(z*z+y*y+x*x)>6081e3,cos(atan(y/x))*vel_mag, 0); 0).

mibillen commented 3 years ago

I don't know if this is the issue, but when moving the chunk model off the equator, do the east and west sides of the box follow meridian lines? If they do, then a 1 deg x 1deg box at the north end of the box is a smaller than a 1 deg x 1deg box at the south end of the model. Therefore the force from a given density anomaly in an element at the south end will be bigger than the same density anomaly at the north end.

anne-glerum commented 3 years ago

Hi @mibillen , the east and west boundaries do indeed follow meridian lines. I'm only moving the chunk along the equator though, it's still centered around zero latitude with the northern part being the mirror image of the southern part. So I wouldn't think the location along the equator matters.

bangerth commented 3 years ago

@mibillen made an interesting observation about the size of the cells. If you plot the volume of each cell (Visit can do this), what do you see for cells at the same depth?