Strange visualization of River nodes

[Leo]

Hi,

I am working with a rather complex PHAST model incorporating rivers and wells under unconfined conditions. I want to look at changes in dentrification due to changes in well pumping rates.

Using a "semi-transient" approach (actually steady flow), where I am assuming mean pumping rates over longer periods (years) and using the restart and head.dat files from the previous simulation, I am now at a point, where no convergence can be reached during flow calculation. I tried all kinds of combinations of numerical parameters with no "luck".

Now, I tried to slowly approach the pumping rates of the next period by a transient simulation, with pumping rates chaning slowly over the years instead of abrupt changes. Unfortunately this also doesn't work.

To get an idea of what might be a reason, I started a transient simulation for 0 years, so that some output is generated. There, I noticed in MODEL VIEWER that the nodes indicating the rivers seem to differ between the steady state and transient approaches as can be seen in the attached pictures. In the steady state case, the river "sits" on top of the water table, whereas in the transient case, it "hovers" in some areas.

As I expect the free surface and rivers to be mainly responsible for the non-convergence, I am now really curious to know, what exactly is visualized by the height of the river in model viewer. Any comments are greatly appreciated and I am also really happy about any hints to reach convergence in my case. Unfortunately, I cannot provide the input files due to confidentiality reasons.

Thanks!

[dlparkhurst]

You are not giving much to work with, and it has been a long time since I have worked on how PHAST is visualized.

I believe that the model features, including rivers, are written to the HDF file only once, and should show the estimated location of the river's head withing each column of cells overlain by the river. I'm not sure why there are multiple river nodes within one column of cells. One way would be to have multiple RIVER definitions (multiple rivers could lead to non-convergence), but perhaps there are other reasons.

I don't know if you are using Phast4Windows or not. If you are, you could export a trans.dat file. Perhaps you could send all RIVER definitions from the trans.dat file via the messages of this forum without violating too much confidentiality.

[Leo]

Sorry, I need to clarify: What is shown is the Y-Z plane, view from direction +x. There are not multiple RIVER nodes per column, there are just several rivers distributed over the model surface. Viewed from +x, they just appear on top of each other.

show the estimated location of the river's head withing each column of cells overlain by the river

Do you mean, the nodes should show the river's head as specified/interpolated by the RIVER definition? I expect them to be constant, if I don't specify them to change over time. However, according to the picture coming from the transient simulation, they appear at other z-coordinates, which I don't understand.

Thank you again for your time and effort, you're always a great help!

By the way, I am rarely using Phast4Windows, as editing the trans file is much easier and quicker for me. I normally use the program to check if model input is correct and to get an idea of the model's geometry before running simulations. Also, although I would like to, I cannot provide model excerps due to our data policy. The RIVER definitions wouldn't pose a problem, but I assume that without the rest of the model/at least the MEDIA definitions, that wouldn't be of much help.

[dlparkhurst]

Sorry, it takes me a while to get focused on the inner workings. The depiction of the river in ModelViewer should be the Z value of the specified river head.

The river definitions in the .trans.dat file are processed by the program phastinput to calculate the area that is applied to each river segment. The leakage for a segment will depend on the area, leakance, river head, and water table location (in the vertical stack of cells containing the river segment). The data is written to the phast.tmp file and processed again by phast to find the cell that contains the river head. That cell is passed to ModelViewer. Seems like ModelViewer should show the same vertical location if the river heads and grid remain the same.

[Leo]

I think, I understand. To check the "issue", I've built a much simpler example. Without changing the river's properties, you can see that there seems to be a difference in river heads in the visualization when comparing steady state and transient calculations. I don't think that the calculation itself is affected. I attached the trans file and the head.dat file for the initial heads, as well as two screenshots to show the effect mentioned above.

[dlparkhurst]

I dug into your question, and let's call it a feature of PHAST/ModelViewer.

When you run transient, the nodes that are sent to ModelViewer are the nodes that contain the head in the river; that is, the elevations that are defined in the RIVER definitions. All flow calculations occur after the data are written for ModelViewer.

When you run steady flow, the steady flow calculation occurs before the data are written for ModelViewer, and the array that originally contained the nodes for the head of the river segments gets overwritten by the nodes that contains the water table, which are the nodes that communicate with the river segments. If the nodes are higher than the river head, flow will be into the river, if lower, flow will be from the river. You could argue that the locations are okay because they show where the flow is contributed to/from the river.

The transient case is probably the way both cases should work for ModelViewer, but I am far enough away from the development that I don't want to try to change the steady-flow case at this point. It would simply change the location of the nodes in the visualization, and should not affect any flow or transport results.

[Leo]

That explains a lot, thank you!