High head error in Richard's steady case

My model have 7 layers :aquifer1(layer1~5 which discrete equally), aquitard(layer6), aquifer2(layer7).

My elevation are picture1 and 2.

I have set a reasonable recharge and conductivity.

However the head always get very low(~-1e19) in the slice1 and slice2(domain left) which show in picture3.

The budget show not convergence: Imbalace:-132.26

But the slice 3~ slice 8 get a reasonable head.

I have no idea with this error.

You can try changing Van Genuchten Parameters (or whatever option you are using to define drying curve) to something much more smooth.
In my experience, one needs really high vertical discretisation for model with actual curve for sand or silt to converge. So, I normally use Van Genuchten parameters for clay regardless of the actual material. However, if one is actually interested in the water content in the unsaturated zone, the only way is to increase discretisation.

Side note: I found that getting a decent water balance with Richard's equation requires changing error tolerance to 10^-5 from default 10^3 ("Numerical Parameters" window). I.e. even if heads look at 10^-3, water balance can still be off. Though, your model currently fails to get even to 10^-3 based on screenshots.

I fully second Durchlässigkeitsbeiwert's answers. Very likely you need to smoothen your curves. IN addition, you might also pay some attention to mesh quality which may also have a negative impact on convergence.

Thanks reply!

I gave the van genuchten parameters by the picture table(reference  Dirk Mallants).

However, cannot convergence after recharging.

My aquifer1 have gravel, sand; aquitard1 is clay; and i inactive aquifer2 now.

And i confused "increase discretisation", is it means let my aquifer1 layer be more layers?

Jimmy.

Hi Jimmy,

I am relatively new to groundwater modeling and FEFLOW, but I also understood Durchlassigkeitsbeiwert's response to mean increasing vertical discretization by increasing the number of layers. I believe you could use the 3D Layer Configurator to quickly add layers a set distance below existing slices.

For what its worth coming from a newbie, such as myself, to deal with the convergence issue, perhaps you could try blending conductivities between highly contrasting elements (i.e., in the new slices between your aquifer and aquitard near the contact, try setting the conductivity to half the difference between the conductivity of the aquitard and aquifer).

I have been trying to read up on the Van Genuchten method, would you be able to post a link to the Dirk Mallants document you mentioned?

Thanks,

[quote author=Jimmy Lin link=topic=22115.msg29349#msg29349 date=1612431882]
I gave the van genuchten parameters by the picture table(reference  Dirk Mallants).

However, cannot convergence after recharging.

My aquifer1 have gravel, sand; aquitard1 is clay; and i inactive aquifer2 now.

And i confused "increase discretisation", is it means let my aquifer1 layer be more layers?
[/quote]

I have plotted curves based on your values below.
In order to achieve convergence you need multiple points over a transition "front" on the plot.
In case of sand and gravel this would require vertical layer discretisation in the unsaturated zone in the order of centimetres (hence, the need for many more layers). In contrast, curve for clay is much more smooth, so you can get away with vertical discretisation on the order of tens of metres without numerical issues.
So, the idea is to use clay Van Genuchten parameters (but hydraulic conductivity for sand/gravel) to help ensuring that the transition in the unsaturated zone is not as abrupt. There is a presentation with useful slides on the matter at [i]http://www.feflow.info/uploads/media/Frind.pdf[/i]
Simpler approach would be to use "phreatic" setting instead of Richard's, however, this would not support any perched aquifers.


Side note: are you a part of GWF?

Hi bgourley!

Thanks u tell me add layers in FEFLOW, it help me to save lots of time .(I always add layers by coding, it need to export and import elevation data).

And the table reference:
https://www.researchgate.net/figure/1-Van-Genuchten-parameters-and-saturated-hydraulic-conductivity-K-s-for-soil-layers_tbl13_256546124

And Durchlassigkeitsbeiwert!

Thanks u so much, you help me more understand van genuchten parameters!

I think i need to look more article about unsaturated simulation and van genuchten model.

And my case is part of GWF.

Jimmy