Callback function when there is more than one model in a simulation (eg. GWF+GWT)

[ifariasg]

This is not really a bug but more a usage question.
I'm trying to run a "simple" MF6 simulation composed of a GWF + GWT model. I'm trying to get access to the concentration in the GWT model to then update a well concentration. Thing is, I cant seem to find the way to retrieve the GWT model.

In the initialize step I do the usual:

if callback_step == Callbacks.initialize:
 print(f"\nInitializing simulation...")
 print(f"Models: {sim.model_names}")

Which prints:

Initializing simulation...
Models: ['gwf', 'gwt']

But then I use the Callback stress_period_start to retrieve the transport model like

elif callback_step == Callbacks.stress_period_start:
    gwt_model = sim.get_model("gwt")

And I get an error:

KeyError: 'Model name gwt is invalid'

I realized that sim.model_names in the Callbacks.stress_period_start now only yields ['gwf'] as a possible option.
So at this stage I'm pretty convinced the api gets access to the models in the order they are run, so at stress_period_start only the flow model is meant to be modified (?), but this begs the question, how, or with what callback function can I access the stress_period_start state of the transport model?
There is no examples that I could find that uses the API with more than one model in a simulation so I'm pretty lost.

I tried both with modflowapi 0.3.0.dev0 and the official 0.2 release.

[jlarsen-usgs]

@ifariasg

Any chance that the GWF model and the GWT model have different solution packages? If that's the case, I have an explanation of what's going on. When there are multiple solutions in MODFLOW each solution is solved individually so the GWT model would not show up in the callback when the GWF model is being solved.

If you have 2 or more solutions you could check for the GWT model first like this:

elif callback_step == Callbacks.stress_period_start:
    if "gwt" in sim.model_names:
        gwt_model = sim.get_model("gwt")

Let me know if this helps or if we need to investigate if this is an issue with the api extensions.

[ifariasg]

Hi @jlarsen-usgs, thanks for your answer.
Indeed, my case was using different solutions for each model. Now it all makes sense and I can retrieve the different models.
That said, in my case I'm trying to update the injection concentration of a well, something that is used by the GWT model, but is stored in the GWF model. So whatever callback I use to retrieve the concentration, I then have to expose the variable as a global for the callback of the GWF to "see" it.
This feels a bit hacky, but maybe cases like this, i.e. updating a variable in a model with a result from another model were meant to be dealt with differently?

I think my model is also affected by #72 so after its fix gets merged I should be able to fully test the approach