Jump to content
GMS, SMS, and WMS User Forum

Convergance and strange boundary condition efftecs with GMS:FEMWATER


woodward
 Share

Recommended Posts

Hi

I am applying FEMWATER at a small field research site (80m x 40m, 20m deep), and having problems with flow convergance and boundary conditions. The domain is roughly rectangular (slightly converging) and slopes downhill. I have set constant head bcs at the top and bottom boundaries (according to measured well levels) and no-flow on the side boundaries (assumed to follow streamlines). There is constant recharge on top, and no flow on the bottom. At the moment I am looking for a steady-state flow solution. I've attached a plot of a moisture content solution (which didn't converge properly).

Firstly, convergance problems seem to arise near the water table due to the sharpness of the van Genuchten release curves for the soil (ignumbrite, which I have approximated as a loamy sand). The linear iteration converges fast enough when I use the PGCM (Choleski preconditioner) option but the non-linear (Picard) iteration is painfully slow. Could this be due to my spatial scale being too coarse? I was wondering what grid resolution I am likely to need to get a fast, stable, correct result. Currently my (unstructured mesh) elements are about 4.0 m horizontally and 0.2 m vertically. Are there any general guidelines for grid resolution that I should be aware of? I have read a little about numerical dispersion re Peclet and Courant numbers, but what about for unsaturated flow solutions?

Secondly, I am seeing some strange artifacts near the boundaries. The unsaturated nodes directly under the specified head boundaries (top and bottom) are much drier than their neighbours, giving a large horizontal gradient in the unsat zone near the boundaries. It doesn't look right. It's almost like no recharge is being applied at the boundary. Has anyone seen this kind of thing before and know what to do about it?

Thanks in advance for your suggestions.

post-958-124942331909_thumb.jpg

  • Like 1
Link to comment
Share on other sites

Further to this...

I have learned that Dirichlet boundaries cannot properly be imposed in unsaturated conditions, because when there is recharge, the total head in the unsaturated zone will not be constant. An appropriate approach is to apply the Dirichlet boundary condition only below the water table, and use a no flow boundary condition above the water table. Therefore, I created a Matlab script that removes the DB1 cards of nodes above the water table from the 3bc file.

Edited by woodward
Link to comment
Share on other sites

  • 3 weeks later...

Convergance remains painfully slow.

Has anyone had experience in getting variably saturated models to converge? Any tips?

Link to comment
Share on other sites

Convergance remains painfully slow.

Has anyone had experience in getting variably saturated models to converge? Any tips?

Link to comment
Share on other sites

  • 2 weeks later...

Convergance remains painfully slow.

Has anyone had experience in getting variably saturated models to converge? Any tips?

The model convergence is now much faster. The slow convergence seems to be caused because the van Genuchten curves generated by GMS are not very smooth. Increasing the number of points in the curve (from 60 to 500) gives much faster convergence. I had to recompile FEMWATER to do this because the GMS version of FEMWATER allows a maximum of 100 points. Tightening the convergence tolerance and/or turning the spline interpolation on/off seem to make little difference to the convergence rate.

I am still having problems with one node being drier than the surrounding nodes in the solution.

Link to comment
Share on other sites

The model convergence is now much faster. The slow convergence seems to be caused because the van Genuchten curves generated by GMS are not very smooth. Increasing the number of points in the curve (from 60 to 500) gives much faster convergence. I had to recompile FEMWATER to do this because the GMS version of FEMWATER allows a maximum of 100 points. Tightening the convergence tolerance and/or turning the spline interpolation on/off seem to make little difference to the convergence rate.

I am still having problems with one node being drier than the surrounding nodes in the solution.

Further to my previous posts:

1. The FEMWATER spline interpolation option (which uses a cubic spline to interpolate the unsaturated curves, supposedly to speed convergence) can actually make convergence much worse (15 minutes compared with 2 minutes). This may be because the FEMWATER spline algorithm may not ensure that the interpolations are monotonic. The default linear interpolation is monotonic.

2. The relatively dry node in my solution I now think is actually a correct solution. I think it is due to lateral flow away from the no-flow boundary.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
 Share

×
×
  • Create New...