Solver error in V1.6

[Daniel Parras]

Hello all,

After updating to ATS v1.6, we converted an XML file that previously ran successfully in v1.4 using the python script from the git. Since the update, the simulation struggles to converge in the surface–subsurface coupled solve, causing the timestep to decrease significantly and the model to make very little progress.

The model represents a watershed-scale domain with a 3D subsurface mesh and a surface mesh generated from it. The subsurface is defined using seven vertical geologic layers, and the system simulates fully coupled variably saturated flow and pressure-based overland flow within a water-balance framework, with a PK structure similar to the priestley_taylor example.

I have attached the output log and a screenshot of the failed message for reference. Any guidance on common causes of this issue in v1.6 would be greatly appreciated.

Thank you for your time.

Best regards,
Daniel Parras

[Bo Gao]

Hi Daniel,

Can you provide your input files for v1.4 and v1.6? When you ran it successfully with v1.4, did you also start from canopy-snow-surface-subsurface coupling without spinup for initialization? Thanks.

Best,

Bo Gao

--
You received this message because you are subscribed to the Google Groups "Amanzi-ATS Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ats-users/10b66564-af18-4999-9245-372af7a8c7a6n%40googlegroups.com.

[Daniel Parras]

Hi Bo Gao,

I kept the same tree and initialization settings. We are mostly just adding layers from an exo file. 

I have attached both xml files.

Thank you,

Daniel

[Bo Gao]

Hi Daniel,

I had a quick look at the1_6.xml, I saw some wrong values like those in permeability, "NRCS-55423": "3.42-12", "NRCS-504533": "1.88-13". I am not sure whether you have changed them when you run your model and I am not sure whether the wrong values will assert errors or result in unphysical simulations. By comparing the two input files, I saw you also changed boundary conditions. I would suggest using the same PK tree, ICs, BCs, mesh, and parameters with the model successful with v1.4 and see whether it results in crash in v1.6. Besides, when we do simulations, we usually start with a steady state spinup followed by a cyclic steady state spinup. In the first spinup, snow and canopy layers are not included, but we only run coupled surface-subsurface system with constant rainfall to establish a steadystate water distribution system.

Best,

Bo Gao

To view this discussion visit https://groups.google.com/d/msgid/ats-users/9eaa2eca-76e8-4bd9-8445-72d699fb491en%40googlegroups.com.