Refining the mesh at the surface
In this tutorial we run the same short mantle cooling case twice: once on the uniform mesh and once with a 1 km top cell. We then compare the top cell, the surface heat flux and the surface temperature T_magma. Mesh refinement explains how the refined mesh is built.
Assumptions
- Aragog is installed (see Installation).
ARAGOG_TEST_EOS_DIRpoints at the SPIDER test tables (step 1 of the First-run tutorial shows how to fetch them).
1. Run the two meshes
The script loads the bundled abe_solid.toml (1000 km mantle, 100 nodes, mass-coordinate mesh), sets the CMB cell to 0 (uniform at that end; the default is 3000 m) and the top cell, shortens the run to 100 years and starts from a uniform entropy of 2900 J/kg/K. Save it as refined.py:
import os
from pathlib import Path
import numpy as np
import aragog
from aragog.solver import EntropySolver
cfg = Path(aragog.__file__).parent / 'cfg' / 'abe_solid.toml'
eos_dir = os.environ['ARAGOG_TEST_EOS_DIR']
for top in (0.0, 1000.0):
solver = EntropySolver.from_file(filename=str(cfg), eos_dir=eos_dir)
solver.parameters.mesh.cmb_cell_thickness = 0.0 # m; 0 = not set
solver.parameters.mesh.surface_cell_thickness = top # m; 0 = not set
solver.parameters.solver.end_time = 100.0 # yr
solver.initialize()
solver.set_initial_entropy(2900.0)
solver.solve()
out = solver.get_state()
r = np.ravel(solver.evaluator.mesh.basic.radii)
print(
f'top cell {r[-1] - r[-2]:8.1f} m: status {out.status}, '
f'surface flux {np.ravel(out.heat_flux)[-1]:.4e} W/m^2, T_magma {out.T_magma:.1f} K'
)
Run it:
python refined.py
The two runs take about 100 s together on one core. The script prints
top cell 10866.6 m: status 0, surface flux 1.4383e+05 W/m^2, T_magma 1262.7 K
top cell 1000.0 m: status 0, surface flux 1.4530e+05 W/m^2, T_magma 1265.9 K
2. Read the result
The first line is the uniform mesh: in mass-coordinate mode its top cell is 10.9 km, thicker than the 10.1 km radial average. The second line has the 1 km top cell that the config sets. Both runs end with status 0.
After 100 years the surface heat flux on the refined mesh is 1.0 % higher than on the uniform mesh, and the surface temperature T_magma is 3.2 K higher. The top cell changes the result, so a run with a thin top cell is not the same model as the uniform run. Which top cell is thin enough is the question of the convergence study on the Validation page.
3. Set the cell in a config file
For a config file, set the field in the [mesh] table instead of in Python:
[mesh]
cmb_cell_thickness = 0.0 # m; without this line the CMB cell is 3000 m
surface_cell_thickness = 1000.0 # m
Refining the mesh towards the surface and the CMB lists the allowed values and the errors.