Oscillating Maximum Head change

[Qing He]

Hi all,

Has anyone come across the oscillating Maximum head change problem? It's like for the previous iteration step the maximum head change is 100m,  and next step suddenly becomes -100m, and iteration continues looping like this. 

For some cases,  the problem can be solved by just increasing the hydraulic conductivity, but for some cases, no matter how I tuned K, the model never converges. 

Appreciate for any help!

[Jakab Andras - Gmail]

Well, this may be caused by a number of issues, but it's hard to tell what is causing it without knowledge of your particular problem. You may find some advice on how to remediate non-convergence and oscillations issues here. Please note that the recommendations are not problem specific and were prepared with the Visual MODFLOW Flex user in mind. Nevertheless, they are equally valid for any MODFLOW model regardless of the GUI being used.

--
This group was created in 2004 by Mr. C. P. Kumar, Former Scientist 'G', National Institute of Hydrology, Roorkee. Please visit his webpage at https://www.angelfire.com/nh/cpkumar/
---
You received this message because you are subscribed to the Google Groups "MODFLOW Users Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to modflow+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/modflow/883b6f40-8d1e-454b-8a73-81d2ff8d1482n%40googlegroups.com.

[Ishita Bhatnagar]

Hi Qing Hi I havent come across this problem but Id be happy to help you . Could you provide some more information like which engine are you using? and is it 2D or 3d model? how heterogenous is the basin, how many zones have you divided it into? 

Thanks and Regards

ISHITA BHATNAGAR
Research Scholar (Water Resources Engineering)
Department of Civil Engineering
Indian Institute of Technology (IIT)  Delhi
Alternate Email: cez1...@civil.iitd.ac.in
Ph no.: +91 8394834109

--

[ashutosh singh]

These kind of troubles usually happen when your solution is caught between local minima and cannot satisfy the convergence criteria.

Depending upon the solver, a larger step can be taken from this solution so that it can hop over these local minimas into another part of the solution curve where the convergence criteria is satisfied.

Using this way a global minima is easier to be found. Ofcourse the risk is that it will keep on hopping onto other parts of the curve instead of searching a solution in the smaller parts of the curve.

I had a course from Sorab Pandey who is a known name in the modflow convergence world who could explain to me in easiest words possible.

Through this email I would like to reach out to him to give these courses regularly to students and experienced groundwater modelers every year or two.

Regards

Ashutosh Singh

--

[Jonathan Quiroz]

If it is a problem with re-wetting cells I'd use modflow nwt. MF6 has the option to activate it

--

[Qing He]

Hi guys,

Thank you so much for all the useful tips. I think the most possible reason in my case may be the dry cell or the sharp gradients between grids. The model is now built for the entire continental U.S., where there should be many grids with very steep elevation changes (Rocky mountain area for example). Also, a previous steady-state map shows over many areas the water table is far below land surface ("dry cell").

There are two ways for me to solve this problem based on the Waterloo's website Jakab sends, first is to reduce the gradient of either elevation difference or the hydraulic conductivity. But my concern is whether doing so is reasonable because manually adjusting the parameters would introduce artificial error and would cause the model settings unrealistic. 

The second one is to employ NWT solver, which sounds more feasible to me. But I wonder how should I set up appropriate NWT parameters? I checked some materials but it's still unclear which parameters I should explicitly include (under_relaxation_theta, kappa,gamma, etc.), and which parameters are for linear/non-linear solutions. I'm not even sure whether I should choose linear or non-linear solver for my case. If anyone can give me further advice on this matter I'll much appreciate it.

To Ishita: Thanks a lot for your interests and kind offer to help! I'm using modflow6, with the horizontal cell size of 10km-by-10km, and aquifer depth of ~1000m. The model has only a single unconfined layer but I think it's very heterogeneous because it's for the entire CONUS. Let me know if you need more info.

[Qing He]

Hi all, 

By trying several NWT parameter settings, the drastic oscillating maximum head change has been improved. But a new problem now is very often when the maximum head change is almost meeting the convergence criteria, it bounces back to a very high value and another same "oscillating" cylce started again (see attached figure for example).

My under_relaxation parameter settings are a very stable one now (under_relaxation_theta = 0.001, kappa = 0.01, gamma = 0.01, and momentum = 0.001). I also tried to use backtrack functions but they seem to drive the non-convergence even worse.

Any further advice on this?