Björn Kaiser

Negative concentrations may be considered as a „self-diagnostic“ indicating weaknesses in the model setup.

There are several ways on dealing with negative concentrations. These different ways may be categorized in adaption and smoothing techniques. Whether adaption or smoothing techniques are preferred depends very much on the study area, individual model settings and also on the question you want to answer.

During the first attempts, I suggest prioritizing adaption techniques rather than smoothing techniques. If all attempts fail, you could try to use smoothing techniques.

Here are some (general) questions I would try to answer before I switch to smoothing techniques:

1. Is the mesh fine enough? A fine mesh is required especially at sharp interfaces of primary variables (e.g. heads or freshwater-saltwater interface) and in areas where sharp contrasts in physical rock properties characterize the hydrogeological setting.

2. Does the mesh contain bad shaped elements? Try to use the maximum interior angle of triangles instead of the Delaunay criterion violations. You may improve the mesh using mesh smoothing techniques from the Mesh Geometry Toolbar.

3. What temporal discretization are you using? I guess you use an automatic predictor-corrector time stepping scheme. That’s fine. But sometimes additional constraints by means of a growth factor between subsequent time-steps or/and of a maximum time-step size may improve your model.

4. Fully implicit schemes are generally more stable than (semi-)implicit schemes. Accordingly, you may switch from a second order-accurate semi-implicit scheme (AB/TR) to a fully implicit first-order accurate (FE/BE) scheme.

5. What error norm are you using? The maximum error norm represents a more stringent criterion compared to integral error norms given the fact that the maximum error norm focuses more on local effects within the model. In this regard you may also change the error tolerance.

Finally, if all these settings still do not give rise to the desired result or if (for instance) the mesh would contain too many elements, which exceeds the handling of your computational resources, you may think about adopting Upwinding to stabilize the numerical solution artificially. Upwinding techniques smooth steep gradients of computational findings by adding (artificial) numerical dispersion. In other words, you change the physics of your system in order to reach a numerically stable solution. Therefore, I always would be cautious when dealing with Upwinding.

You may also think about accepting negative concentrations to a certain degree. Criteria to assess whether or not a concentration is acceptable could be the magnitude of the concentration or the spatial location where the negative concentration occurs.

You may adjust the vertical exaggeration in the Projection tab of the Navigation toolbar. The faster way is using the mouse and the keyboard. Keep the shift-key pressed by using the wheel of the mouse.

The standard features of FEFLOW do not provide an automatic stop of the simulation as soon as a scalar quantity computed at a certain node is higher/lower than a threshold. A useful workaround is work with observation wells. You may assign an observation well located at the node of interest and record the evolution of temperature through time.

Yes, this is possible in a 3D View. Right-click in the 3D View. Then, choose [b]Show[/b] from the context menu and select [b]Length-Scale Display[/b].

Yes, FEFLOW allows the assignment of variable values for any kind of boundary condition. You may either assign them manually on the base of different nodal selections or you may assignments based on parameter links to external map data.

An efficient introduction into assignments is provided by our Demonstration Exercise:
http://www.feflow.com/uploads/media/feflow-booklet62_01.pdf

Corresponding demo data can be download from our website:
http://www.feflow.com/download.html?&no_cache=1

If I understand you correctly you expect heat conduction as the dominant heat driver at the margins of the basin, while free thermal convection as driven by gradients in density controls the thermal setting at the center of the basin. However, in the central part of the basin you observe a temperature which corresponds to the surface temperature instead of thermal anomalies triggered by buoyant upwelling flow.

If this is the case could you please double-check whether or not density-dependency is activated? If density-dependency is already activated could it be that external advective forces overprint thermal convection?

Hello Michael,

If you like I may have a closer look to your model. Please send us an e-mail in the FEFLOW support: mikebydhi.de@dhigroup.com

Cheers!
Björn

If you do not wish to ad-hoc postulate that water table undulations replicate exactly the topography you may assign a Seepage Face BC on the entire Slice 1. Run a couple of simulations by solving the Richards Equation and vary a homogeneous recharge systematically. All other settings should not change within the group of simulations. You will see, the lower the recharge the less water table undulations replicate the topography. In contrast, the higher the recharge the closer the approximation of the topography and the stronger your results resemble the results by early pioneering studies (Toth 1962, 1963).

If you wish that water table undulations replicate exactly the topography then apply a confined model in the Problem Settings and use a Dirichlet BC with a pressure value of 0 kPa assigned to the entire top slice. In the Data Panel you have to add a Hydraulic-head BC (Pressure): Right click on Fluid flow beneath Boundary Conditions (BC) and then on Add Parameter.

Perl, the format reads rather as:


# ID1
! Comment line for time series 1
! ...
x1 y1
x2 y2
...
xm ym
END
# ID2
! Comment line for time series 2
! ...
x1 y1
x2 y2
...
xn yn
END
# IDn
! Comment line for time series n
! ...
x1 y1
x2 y2
...
xo yo
END
END

You may find a detailed description in the help of your FEFLOW.

There are several types of Boundary Conditions (BC’s) you may use depending on (1) the problem you want to investigate and depending on (2) available data. Here is a rather general answer: If you use a fixed potential (1st kind BC) you ad-hoc impose the solution (heads) along the surface waters. Instead if you use a 3rd kind BC you rather use a reference head href, which refers to the surface water head. Then FEFLOW solves for the GW-head h. Depending whether h is smaller/larger than href you have in inflow/outflow by taking the transfer rate Phi into account: q = -Phi (href-h).

Please find more details in the FEFLOW book, Chapter 6, page 193.