Rapid increases in fluxes & high error

A discussion forum for users of the new HYDRUS 2D/3D. HYDRUS is a software package for simulating water, heat and solute movement in two- and three-dimensional variably saturated media. Happy Posting!
Post Reply
MAri
Posts: 8
Joined: Tue Mar 24, 2020 8:07 pm
Location: Netherlands

Rapid increases in fluxes & high error

Post by MAri » Mon Jun 15, 2020 8:15 pm

Dear sir,

I am currently building a model for a peat area with a peat layer floating on top of a water rich (>95% water) slurry layer. I need to model two transect. For the first transect I have a reasonable first estimate for a model. However, modelling the second transect I run in to some issues, eventhough I use the same approach.

My model can described shortly as followed. The transects have five layers in the following order: 1. sphagnum peat layer, 2. carex peat 3. slurry layer 4. carex peat/gyttja layer 5. sand. The FE mesh is 5cm in the top layer, reducing to 20cm in layer 2 and 3, and reducing further to 4m in the sand layer. The FE has a stretch of 5-10 in the x-direction. The transects are bounded by drainage ditches on the left and right side, giving a variable head boundary with summer and winter water levels. At the bottom a constant head boundary is used.

My first issue is with regards to the water balance error. Currently, the model of the second transect shows very large absolute and relative errors. The absolute error is within the 100s to 500s m2 while the relative error is >15% for most time steps. To reduce this error I have tried several tips I have found on the forum: I tried to reduce the spatial and temporal discretization, i tried to adjust the hCritA, I tried to adjust the iteration criteria. But nothing as of yet has worked in reducing the errors to more reasonable values.
Does anyone perhaps have any further suggestions or tips for reducing the mass balance error?

My second problem is with regards to the fluxes. Looking at the graphs especially the seepage face flux and the variable head flux worry me. Especially the seepage face flux shows sudden rapid increases in flux (from values >0.05m2/d to 1.0 m2/d). This rapid increase happens only at one time-step. The variable head only shows one such sudden increase, but this increase is extremely large -140m2/d. I think perhaps these weird fluxes tie into the water balance issue, as the error is determined based on the boundary fluxes. However, I cannot seem to find the reason behind these sudden fluxes. I tried using the 'interpolate boundary condition between time-steps' button but this has not removed the sudden high fluxes.
Why do these sudden high fluxes occur? And, mostly, how can I limit these outlier values?

If needed I can post the project/the result files.

If there are any questions please let me know.

Thank you for your time and effort!

Kind regards,
Mary
Attachments
FluxesNM.png
FluxesNM.png (62.46 KiB) Viewed 89 times

Jirka
Posts: 4961
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Rapid increases in fluxes & high error

Post by Jirka » Mon Jun 15, 2020 8:29 pm

I would need to see the project to be able to comment. J.

MAri
Posts: 8
Joined: Tue Mar 24, 2020 8:07 pm
Location: Netherlands

Re: Rapid increases in fluxes & high error

Post by MAri » Tue Jun 16, 2020 8:37 am

Dear Jirka,

Thank you for the quick response. Hereby the files as .zip. I removed the results because otherwise the file size became way too big!
If the results are needed please let me know and I'll send them by email (i think to support@hydrus?) or I can try GDrive.

Thank you for your time!

Greetings,
Mary
Attachments
HYDRUS_NM.zip
(1.43 MiB) Downloaded 4 times

Jirka
Posts: 4961
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Rapid increases in fluxes & high error

Post by Jirka » Tue Jun 16, 2020 8:52 pm

The results correctly reflect the modeling set up. You have large sudden fluxes at the seepage face and time-variable boundaries, whenever you suddenly lower or raise the water table in the ditch. Since you are using the option "Seepage face BC when the specified nodal pressure head is negative", one (or several) boundary node switches between these two BCs and you get either sudden inflow (when you raise the water table) or outflow (when you drop the water table). If you want to smooth these sudden fluxes, you should:
1. Use the option "Interpolate Time -variable pressure head in time", since then you will not get a sudden instantaneous increase/decrease in the water table, but a gradual increase/decrease. However, whenever you use this option ("Seepage face BC when the specified nodal pressure head is negative") you will always have some sudden fluxes, since at least a single boundary nodes will have to be switched from one BC to the next.
2. You should use finer discretization along the ditch boundaries (as well as in the domain). Note that currently, you have only one node at the bottom of the ditch, and the next one 70 cm higher. Also, you have quite abrupt changes from fine grid to very coarse.
To decrease the mass balance error, you will need to use smaller water content tolerance. You currently use 0.005, which is almost as having none. The recommended value is 0.001 (see the help and manual).

J.

Post Reply