Saltwaterintrusion Modellerung mit OGS

Ich schreibe derzeit meine Masterarbeit am LWI Institut für Wasserbau der TU Braunschweig. In meiner Masterarbeit geht es um den Vergleich von MODFLOW und OpenGeoSystem; dafür wurden Laborexperiment zur Salzwasserintrusion durchgeführt, welche nun numerisch reproduziert werden sollen. Bei der Modellierung mit OGS haben sich allerdings Probleme ergeben, welche unserer Vermutung nach bei den numerischen Parametern liegen. Im ersten Schritt haben wir mit Version 5.8 gearbeitet, da dies für einen Einstieg einfacher erschien, eventuell wollen wir zu einem späteren Zeitpunkt weiter Simulationen mit der aktuellen Version durchführen. Im Folgenden finden Sie eine kurze Beschreibung des Modells sowie des Problems:

Es handelt sich um ein Modell zur Salzwasserintrusion in der gesättigten Zone.
Initial Conditions: Der linke Teil ist mit Salzwasser gefüllt (head=0.45m, concentration=1), der rechte Teil mit Süßwasser (head=0.5m, concentration=0).

Boundary Conditions: Wasserstand und Konzentration am rechten oberen Rand werden konstant gehalten (head=0.5m, concentration=0), der Wasserstand am linken oberen Rand ist zeitabhängig (head1=0.45, head2=0.5, head3=0.55). Alle anderen Modell-Grenzen sind no-flow-boundaries.

Das Problem liegt darin, dass die Simulation nach dem Lesen der input files abbricht, ohne, dass eine Fehlermeldung oder ähnliches erscheint. Ich würde mich freuen, wenn Sie mir sagen können, warum die Simulation nicht funktioniert oder mir den Kontakt von jemandem geben können, der mir diesbezüglich helfen kann.

Die Dateien des Modells wollte ich im Anhang dieser E-Mail beifügen, ebenso das protocol file. Allerdings gab dann beim Versenden der E-Mail Probleme, daher finden Sie die Dateien unter folgendem Link:

https://cloudstorage.tu-braunschweig.de/getlink/fiMbFkz4VAff99NJXUJFth4A/Modell_SWI.zip

Vielen Dank für Ihr Entgegenkommen.

Mit freundlichen Grüßen

Lisa Göldel

Thanks Lisa for posting here.

@thomas.kalbacher or @wenqing.wang Can you help with saltwater intrusion on OGS-5?

@renchao.lu Is saltwater intrusion already available in OGS-6? If yes is there an example?

Dear Lisa,

I have gone through the OGS-5 files you uploaded. In my perspective, the desired simulation is doable with the ComponentTransport module in OGS-6. You would find lots of component transport examples via the link: Tests/Data/Parabolic/ComponentTransport · master · ogs / ogs · GitLab.

Best regards,
Renchao

@Lisa

If it helps with OGS v5.

You may already have read this dissertation (Walther, TU Dresden, 2013:
https://tud.qucosa.de/landing-page/?tx_dlf[id]=https%3A%2F%2Ftud.qucosa.de%2Fapi%2Fqucosa%253A28302%2Fmets

which compares OpenGeoSys version 5 with MODFLOW/SEAWAT using a numerical simulation of saltwater intrusion, based on a laboratory experiment.

In particular section 3.3, reproduced below:

"3.3 Transient Saltwater Intrusion for Unconfined Conditions

Problem Description.

This benchmark verifies the capabilities of OpenGeoSys to model transient, variable-density groundwater flow and transport against a laboratory setup and numerical Seawat simulation data (Langevin and Guo, 2006). Goswami and Clement (2007) show a saltwater intrusion experiment after Henry (1960), using a laboratory scale tank. The benchmark is an extension to the steadystate setup documented in [P2]: Walther et al. (2012b).

Methods and Model Setup.

The Goswami-Clement problem features a horizontally intruding and withdrawing saltwater front in an initially present freshwater environment. The conceptual model is depicted in figure 3.5, p. I-21, important model parameters are listed in table 3.4, p. I

Results.

Figure 3.6, p. I-22, depicts the comparison of the transient simulations between the experiment and the numerical models. Both simulation tools, Seawat and OpenGeoSys, resemble the measurements quite well in terms of temporal development, shape of the saltwater wedge, and intrusion length. Nevertheless, they also show deviations from the laboratory measurements. While Seawat overestimates intrusion length at the toe of the saltwater wedge, OpenGeoSys generally yields a better prediction or slightly underestimates intrusion length."

And here, with post by Marc Walther himself:

For OGS5 input, there are two errors:

  1. msp: only one material group is defined.
  2. mmp: $TORTUOSITY is missing.

The computation runs with the attached corrections.
SWI_correction.zip (570 Bytes)

Note: if Linux is used, please run dos2unix for the changed two files.

There are some other things wrong in material parameters or BC/IC of mass transport, which makes the linear solver of MASS_TRANSPORT diverge. Please check the input data of MASS_TRANSPORT carefully.

Thanks for your help!

@wenqing.wang, I have adopted your corrections. The simulation is running now, but it is not giving proper results. The initial conditions are set up correctly. I checked the input data as you suggested. However, unfortunately I could not solve the problem. I think boundary conditions should be correct. I have attached the results for the first time step (visualized with ParaView) and a download link for the corrected input files. It would be kind if you could take a look at it again or tell me which input file might cause the problems.

Greetings

Lisa

https://cloudstorage.tu-braunschweig.de/getlink/fiH6kGZZGuSjPyc9q4EzwNMp/SWI_aktuelleVersion.zip

I will take a look and reply you later.

I see that you use LIQUID_FLOW but the variable is assumed to be head. Except you specify the primary variable as HEAD, the computation takes the gravity term in the velocity and it leads to wrong results. Please use GROUNDWATER_FLOW as the type of process, which uses HEAD as primary variable. You can find the benchmarks of GROUNDWATER_FLOW at here as references. Once you make process type changes and you still get problem, please leave me a message here.

Hello @wenqing.wang

I changed the type of process to GROUNDWATER_FLOW, but there are still problems. It seems as if a higher concentration of salt would somehow lead to a lower density. Maybe there is something wrong with FLUID_PROPERTIES?

#FLUID_PROPERTIES
$FLUID_TYPE
LIQUID
$DENSITY
3 958.8 0 0.0428
$VISCOSITY
1 1.0e-3
#STOP

Download link for the model:
https://cloudstorage.tu-braunschweig.de/getlink/fiDSvwpUZBb4ri3QU7BHMz5n/SWI.zip

I will take a look .

The download link is expired. Please share the files again.

The download link should work now:
https://cloudstorage.tu-braunschweig.de/getlink/fiDSvwpUZBb4ri3QU7BHMz5n/SWI.zip

Density has not problem as shown in the attached figure. I do know how did you get density output because there is no density output specified in your input data . Here is a figure of density distribution by using the attached input files, in which the changes in numerical setting are just aimed to increase solution accuracy:


changed_input.zip (721 Bytes)