psinton@aquageo.us

Use the command line with PEST...I believe you have to use the classic version of 6.  I assumed there will be a command line version for 6.1 (no classic).

Pete

It depends on the type of element (triangular prisms, cubes, etc) and which way the flow goes to or from the boundary.  Assuming triangular prisms, three nodes are needed to simulate flow through the top or bottom of the element, and you set the stage at all three nodes (each stage can be different). For the same type of element but lateral flow across one of its vertical faces, you must set the stage at four nodes, two at its top and two at its bottom.  Also note that you must also set the transfer rate at all the elements that share each 3rd type node (the default for all elements is zero).  The value of rate depends on the hydrogeology.

Pete

It depends on which tolerance type you choose.  L1 compares the tolerance to an average of all "changes" across the domain, L2 compares the tolerance to the deviation around the average change, and max compares the tolerance to the max change.  "Change" depends on what's being solved (head, solute mass, heat energy).  L1 is the least stringent and "max" is the most stringent.  Which you choose depends on the complexity of the model: more complex often require more stringent comparison.  More complex models also may require smaller tolerance value.  To help you decide, run tests on your model.  There is no better way to select the right combination of settings.

Pete 

non-convergence in a feflow model is often due to a problem with the way the model is set up.  If you are not sure what is causing the problem, a good way to identify problems is to run tests in which you set up the model with  a few inputs as possible, then add complexity until the problem shows up (or do the same in reverse: remove things until convergence improves).

After you get a model that behaves better, you can refine the solver settings.

Pete

Ideas: run a test with finer mesh, use square elements, use uniform mesh, try SAMG with user-specified timestepping, use more string error norm, etc

Certain types of problems seem to be unstable in cross-section mode (boundary condition along the bottom of a section, so you might also try a "3D" model that is set up to simulate a unit thickness cross-section.

Pete

that would have to be coded
Pete

Yes, however if the nodes that define the area across which the flow will occur are "inside" the model then the flux will be double the input value.  By "inside" I mean that flow could occur across the face of two adjacent elements.  This is easy to check in a simple model set up where you know what the answer is supposed to be.

Also, the flux may vary for each element for given K and gradient.  If you compute the average flux over several element faces and insert the average but have varying K, you may find the model computes heads that are too high or too low.  Also, if you are working with a situation in which the saturated thickness varies, then you may have to use the "integrated" version of the 2nd kind boundary condition. Please refer to the manuals and help file for more explanation for integrated 2nd kind.

By convention, inflow is negative.

Pete

It would be nice to be enable reuse of a transient flow field for uncoupled problems, especially in cases where flow field takes a long time to solve.

Kieth, First question: yes. second question: if the head at the constrained node exceeds 450 m constaint, then fixed head = 450 applied rather than flux.  The head at any of your nodes could fall below 450 with the constant flux depending on the conductivities, recharge, etc, you specified.  I agree a long-term transient may be needed.

Thomas, flux value = volumetric flow rate in L3/T divided by cross-sectional area (L2) through which flow occurs.

Pete

You should run tests to figure out the best method for solving the transport equation however shock-front capturing may work well.

If you run as a 2d model with 2 slices and one layer, you will get concentration at the top and bottom of the layer, not in the middle.  You can average the top and bottom but it would be better to simulate the system with at least three slices (two layers) with one slice in the middle.  You will get results on the middle slice directly, and the model may behave more realistically. 

Specifying 50% from top and 50% from the bottom is sensible for the system you describe.  More from one elevation implies 3d flow which implies you need more layers to simulate vertical flow more realistically.

Wells in multilayer models can be handled in two ways.  FEFLOW allows automatic set up of a 1d vertical pipe element with large K that spans all the nodes the well is open to.  In this way, more water enters the pipe element where the K is larger.  You specify the pumping node at the bottom of the 1d pipe (or at the elevation of the pump).  You can also set up individual pumping nodes with rates specified according to the layers that each pumping node touches.  You might want to constrain the pumping node(s) so that they either stop pumping or switch to a 1st-kind bc if the saturation at the node falls below 1.

Source/sink BC is difficult to use but would work as long as the element its specified is saturated.  Not recommended for the problem you describe.

Pete