search

MODFLOW-USGS / modflow6

USGS Modular Hydrologic Model
https://modflow6.readthedocs.io/
Other
256 stars 117 forks source link

Concentrations Blow Up in Boundary Cells #950

Closed MickeyRush closed 2 years ago

MickeyRush commented 2 years ago

I'm using MODFLOW-6's GWT model. In a couple boundary cells (17, 63, 1) and (17, 62, 1), concentrations blow up (one positive, one negative). These are confined CHD cells with thickness ~700 m on the bottom corner of the model - the CHD value for these cells (~1800 m) is about 80m greater than the top elevation of the cells (~1720 m). These cells are far from any sources and no source/sink concentrations are specified with the CHD package. Removing the CHD from the bottom layer of the model fixes the problem - is there any other way to get rid of what appears to be random noise?

conc_ts_17_63_1

conc_ts_17_62_1

head_ts_17_63_2

head_ts_17_62_2

I zipped up a 7z file to get the model to upload: model.7z.zip

I got my Linux executable here: https://github.com/MODFLOW-USGS/executables

langevin-usgs commented 2 years ago

I will plan take a look this week. Hard to say what the problem is without digging in. If you happen to gain any insight into what is the cause, please post here. I notice that you are using COMPLEX IMS settings for both flow and transport, though this model should solve relatively easily. That might be worth trying.

langevin-usgs commented 2 years ago

I made a couple of runs today. If I shut off the IMS COMPLEXITY option and comment out the dispersion package, then I do not get those large concentration changes in those two cells. This is something that we can look at further on our, but maybe this will help you move forward for now.

MickeyRush commented 2 years ago

Thanks for the tip - I'll continue to troubleshoot on my end.

langevin-usgs commented 2 years ago

I've looked at this a little bit more, but don't have any real conclusions. I am able to reproduce those strange concentrations in those cells. I suppose it could be overshoot/undershoot due to dispersion, but that seems excessive and I suspect it is something else. I don't see anything wrong in your input files. I tried to reproduce this behavior with a simple model with constant heads in a lower layer, but the simple model behaves as expected. I notice that you are using a value of 2.0 for TSMULT. This results in extremely short time step lengths, and maybe there is a precision issue due to this? If there was any way you could reproduce this with a smaller version of the model, I could run it through the debugger and see what's going on. The current version of the model is too big.

MickeyRush commented 2 years ago

Oh jeez this is embarrassing - looks like some initial heads made their way into the storage package.

langevin-usgs commented 2 years ago

We appreciate you letting us know. With a new capability like GWT we tend to think it's a bug in the code, but glad to hear that may not be the case here. Am closing this issue.

Here's a view of the rogue plume in case you're interested.

image
MickeyRush commented 2 years ago

FYI - after correcting the STO package problem and doing some runs over the weekend, I do still get concentrations tending to +-1e10 along the CHD in the bottom layer. With the XT3D_OFF setting in the DSP package, however, this does not happen. I'm going to do some more runs to see if Flow Imbalance Correction might also be able to resolve the issue.

modflow_cr.sto.zip