RE: [MODFLOW] Initial head value, Vertical Discretization, Convergence failure

[Simon Woodward]

I would think the problem is rewetting of dry cells. In you have lots of layers, this will be more of a problem, particularly if the water table crosses different layers. Layers are useful if hydraulic properties change with depth though.

 

Starting with a saturated model will help, because as the model converges cells will dry out, but not need to rewet. If you try using MODFLOW-NWT it will handle rewetting of dry cells much better.

 

Simon

 

---

From: mod...@googlegroups.com [mailto:mod...@googlegroups.com] On Behalf Of Miguel Leyes
Sent: Tuesday, 28 August 2012 00:03
To: mod...@googlegroups.com
Subject: [MODFLOW] Initial head value, Vertical Discretization, Convergence failure

Hello group,
 I would greatly appreciate your thoughtful input(s) to an issue(s) I'm facing.
 I've been working on a steady state model addressing surface water - groundwater interaction in a shallow aquifer with a complex geometry.
 The aquifer is located in a river bend, where the boundaries are the River, Upstream and Downstream constant head boundaries.
I am using Model Muse and its Wetting option where cells can become wet from the bottom and side cells.

I've read on previous posts,in the literature and Modflow 2005 manual that the Initial Head value should not affect the model results in steady state models,
but that an approximate value can be beneficial. The issue is that my simulation results actually vary a lot in the distribution of wet cells
depending on the Initial Head value.
The value that seems to me to give the best/most plausible results is that of a saturated aquifer ie: Model Top as the Inital Head.
1) Any thoughts on why the Initial Head could be causing a high variability in simulation results, when theoretically it should not for steady state models?

As I mentioned the geometry is complex, and the aquifer thickness varies from 0.1m to 4.0m.
I am using a one layer model with sub-layers as provided with Discretization in Model Muse.
My Horizontal grid spacing is a uniform 1mx1m, however I am not sure about the consequences of using 20-30 sub-layers in Vertical Discretization.
2) Is there a rule of thumb or concept that should be taken account of in regard to the Vertical Discretization?

Lastly, the simulation results I am now obtaining do seem plausible, and have a mass-balance percent discrepancy of less than 1%, however the text-file
produced still indicates that the convergence criteria was not met.
I updated my Model Muse version from 2.12 to 2.16 ,and now after a simulation the window with the run status, % discrepancy and Red/Yellow/Green lights &f aces
is not appearing anymore. Is it not part of the 2.16 version or could it be my update failed?
3) What could be behind the failure to meet convergence when the results seem realistic and the % discrepancy is below 1%?

Thank you for taking the time to read this and any input to help me understand what is going on.
Best regards,
ML

--
You received this message because you are subscribed to the Google Groups "MODFLOW Users Group" group.
To view this discussion on the web visit https://groups.google.com/d/msg/modflow/-/_6jXwGu1dsAJ.
To post to this group, send email to mod...@googlegroups.com.
To unsubscribe from this group, send email to modflow+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/modflow?hl=en.

[jing yang]

I agree with Simon.
In some cases, the initial groundwater table does affect the convergence, even for steady-state simulation, as it is the starting point of numerical calculation.
I would suggest you increase layer thickness for the case with a complex geometry

[An Ho Antonio Taylor]

 The zones change lateraly in your model? if you have very thin layers the calculation of vertical leakance become unsteable more if you using the rewetting option,  LPF takes as input the vertical hydraulic conductivity directly and computes a vertical conductance internally. The main difference (with BCF) is that it updates vertical leakance at each solver iteration for partically saturated convertible layers, The NWT SOLVER utilize the same package (identical) its call UPW, but it will handle much better than PCG2 or other

2012/8/29 jing yang <waterunc...@gmail.com>

[Richard B. Winston]

If the initial head is below the bottom of a cell in a convertible
layer, the cell will go dry immediately and dry cells are troublesome
for MODFLOW-2005. MODFLOW-NWT was developed, in part, to better deal
with that wet-dry problem. As for ModelMonitor, the program that
monitors the percent discrepancy in MODFLOW as the program is running,
it is still included in ModelMuse. It normally is in the same directory
as ModelMuse. I suggest you check "Model|MODFLOW Program Locations" and
see whether the location for ModelMonitor is specified correctly.

[Miguel Leyes]

Thank you everyone for your inputs.
The solver Modflow-NWT has worked as you suggested and the simulations
now run successfully.
To anyone doing the same change of solvers I would recommend to take a
look at the Conductivity parameters
if at first Modflow-NWT does not complete a simulation.
I also decreased the level of vertical discretization after making
sure the parameters were making sense to the NWT solver.
ML

[diana montoya]

you can use a land level like a initial head in steady state

2012/8/27 Miguel Leyes <m.le...@gmail.com>

--
Diana Maria Montoya Velilla

Ing. Sanitaria. Msc en Ingenieria

Celular 3007777248