Handling water balance error in wet conditions

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
mGohar
Posts: 36
Joined: Wed Jul 15, 2015 11:16 pm
Location: USA

Handling water balance error in wet conditions

Post by mGohar » Thu Oct 24, 2019 1:31 am

Hi,

I noticed that when my domain gets close to the wet condition, Hydrus seems to trigger something which results in improving the water balance error. My question is if that's true, what Hydrus does change and what condition triggers this?

WBE_T.png
WBE_T.png (19.54 KiB) Viewed 2640 times

Thanks

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

Re: Handling water balance error in wet conditions

Post by Jirka » Thu Oct 24, 2019 2:05 am

There is no special treatment of mass balances under wet and dry conditions. I would assume that this may be caused by boundary fluxes (that cause wet conditions) and which are in the denominator when calculating mass balance errors. J.

mGohar
Posts: 36
Joined: Wed Jul 15, 2015 11:16 pm
Location: USA

Re: Handling water balance error in wet conditions

Post by mGohar » Thu Oct 24, 2019 8:31 pm

Thanks for your reply Jirka, I forgot to mention that the problem is an axisymmetrical domain and water is added in every two hours with a constant flux value (for duration of ~3 min) for the entire simulation time. My first though was maybe Hydrus decreases the convergence tolerance which (I guess) you ruled it out.

mGohar
Posts: 36
Joined: Wed Jul 15, 2015 11:16 pm
Location: USA

Re: Handling water balance error in wet conditions

Post by mGohar » Sat Oct 26, 2019 12:26 am

mGohar wrote:
Thu Oct 24, 2019 8:31 pm
Thanks for your reply Jirka, I forgot to mention that the problem is an axisymmetrical domain and water is added in every two hours with a constant flux value (for duration of ~3 min) for the entire simulation time. My first though was maybe Hydrus decreases the convergence tolerance which (I guess) you ruled it out.
I would like to add that the problem description (similar influxes through the simulation run-time) may suggest that this water balance error behavior might not be due to boundary fluxes.

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

Re: Handling water balance error in wet conditions

Post by Jirka » Sat Oct 26, 2019 12:34 am

If you use recommended spatial and temporal discretization and iteration criteria, you should not have any problem with water flow mass balance errors. J.

DandD
Posts: 14
Joined: Wed Jan 15, 2020 12:11 pm
Location: Germany

Re: Handling water balance error in wet conditions

Post by DandD » Tue Feb 11, 2020 9:21 am

Hi all,
I hope it is ok to use this thread - I also have a problem with water balance error while running some more or less special conditions with Hydrus. Apologies for the long post!

My set-up is the following:
- 2D-Domain, dike shaped
- uniform soil (experimentally defined parameters, quite weird/interesting combination in my opinion...)
- initial conditions - equilibrium with 0 pressure head at bottom
- seepage phase BC at the right side ('air side') of the dike
- constant water level at the left side ('water side') -> either by constant head bc or variable head bc (I would vary water levels in future simulations)

Due to the constant head I get full saturation in the lower parts rather quickly and a constant seepage line evolves (obviously...) after some time.
I am always having trouble with big water balance errors!
The following could be observed:
1) Using the default (rather big) mesh results in lower water balance errors (still > 1%)
2) Using a finer mesh discretization leads to way higher errors (up to 20/25 %)
3) The water balance error seems to increases until water starts to flow out the seepage BC (approximately) -> then decreases again
4) Interpolating the variable pressure head in time (to get a smoother rise instead of instant rise to the defined water level) does not resolve the issue.
5) Other sediment parameters do not resolve the issue

I am assuming there is some kind of problem with the constant/variable head BC in this case, which causes the water balance errors (because there are no other boundary fluxes at the start). The uncertainties increase when using a finer discretization because there are more elements (thus more calculations and possible errors/rounding errors). (?)

So my questions would be:
a) Do you think the problem occurs due to the specified inital conditions?
b) Is there a conceptual issue?
c) Would you recommend other BCs?
d) Is the storage term part of the given water balance error/how is storage taken into account?
(and lastly...how can I resolve the issue...)

The model-file was to big to upload - it is available under (if necessary):
https://cloudstore.zih.tu-dresden.de/in ... mAwSpw9sLE

Thank you for reading!

With best regards
Frieder

DandD
Posts: 14
Joined: Wed Jan 15, 2020 12:11 pm
Location: Germany

Re: Handling water balance error in wet conditions

Post by DandD » Fri Feb 14, 2020 12:52 pm

After evaluatingat several aspects (mesh, time discretization, iteriation criteria, soil parameters, inital conditions, BC options, ...) I still can not resolve the issue (nor have I an explanation) but:
I simulated the model in a 3D-Version (1m 'width') and I do not have the water balance error issue there! So it seems like, the adaption to the 2D vertical plane results in some problem...

If you have experienced similar things or have some thoughts on the topic, I would still appreciate the input! Other than that I will just run the 3D models for now.

All the best
Frieder

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

Re: Handling water balance error in wet conditions

Post by Jirka » Sun Feb 16, 2020 12:25 am

I have never seen such a large mass balance error in HYDRUS projects. You must have hit some strange combination of various factors (spatial and temporal discretizations, iteration criteria, tolerance precision, etc) that leads to it. I cannot currently explain it. Normally, the HYDRUS projects should give you the same results no matter what units you use (i.e., m, cm, mm). However, in this case, if you simply switch to "cm", your mass balance error will go down below 0.1%. Also, if you use coarser discretization (e.g., a targeted size of 0.05 m and keep the same units, the error will drop substantially). J.

DandD
Posts: 14
Joined: Wed Jan 15, 2020 12:11 pm
Location: Germany

Re: Handling water balance error in wet conditions

Post by DandD » Mon Feb 17, 2020 7:18 am

Thanks for your reply! I never checked using other units because I want to run SlopeCube later on and therefore need the units in m (meter).
Update
Running the project in 3D does not resolve the issue. Using [m] as unit in 3D leads again to water balance errors. Changing it to [cm] resolves the issue.
My problem now is that the AddOn module SlopeCube can only be run with [m] as units...

I will check the results (pressure heads, water contents etc.) to see, if the waterbalance error is really there or if it is more an issue of display or the calculation of the error.
My idea is, that there is some problem in the code with the reference for the waterbalance calculation - e.g. if [m] is used and Hydrus would use a value with [cm] as reference in the waterbalance calculation it would automatically lead to a difference due to the unit difference (m -> cm). Do you get, what I mean?

DandD
Posts: 14
Joined: Wed Jan 15, 2020 12:11 pm
Location: Germany

Re: Handling water balance error in wet conditions

Post by DandD » Thu Feb 20, 2020 3:17 pm

Sorry for bothering again - but I checked the model results. I just added 3 observation points (x-koord. approx. 1/1.5/2 m and z-koord. approx. 0.33 m) to do a short comparison of the models when using a) unit length [m] and b) unit length [cm].
The results differ significantly for: influx & outflux (well this was obvious due to the waterbalance error displayed), pressure heads and water content (see attached EXCEL spreadsheet if necessary).
So my previous theory about a mistake in the waterbalance error calculation is incorrect I guess.

To run the SlopeCube AddOn (needs [m] as unit) I will try to figure out the best setup with reasonable results/waterbalance error...
Cheers, Frieder
Attachments
Comparison_Obs.xlsx
Comparison of the model results with [m] and [cm] as units, respectively
(1.05 MiB) Downloaded 52 times

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

Re: Handling water balance error in wet conditions

Post by Jirka » Thu Feb 20, 2020 5:01 pm

I will try to look at what causes this over the weekend. J.

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

Re: Handling water balance error in wet conditions

Post by Jirka » Mon Feb 24, 2020 2:19 am

I have found out where the problem was. If I decrease the precision criteria used in the matrix solver (ORTHOMIN), I will avoid these mass balance error problems. Please download the attached computational modules and copy them into the BIN folder, in which HYDRUS is installed (e.g., …. HYDRUS3D 3.0\Bin).

J.
Attachments
HYDRUS exes.zip
(1.21 MiB) Downloaded 54 times

DandD
Posts: 14
Joined: Wed Jan 15, 2020 12:11 pm
Location: Germany

Re: Handling water balance error in wet conditions

Post by DandD » Mon Feb 24, 2020 8:07 am

Thanks! This solution works for me. We found a similar problem/solution for the other program we use. The precision criteria was defined independently from the units used and thus resulted in this case in the error. We are planning to implement a dependency between units used and a respective precision criteria.
Thanks again for your efford in resolving the issue!
Frieder

Post Reply