Change in default output control behavior

[samzipper]

I am new to FloPy (and Python) and am working through the tutorials on the website. For Tutorial 2, I am having an issue in which output is only saved at the first (steady-state) timestep.

In the documentation for the mfoc module, it notes that the default behavior for stress_period_data is to save head at all timesteps:

Dictionary key is a tuple with the zero-based period and step (IPEROC, ITSOC) for each print/save option list. (default is {(0,0):[‘save head’]}) ... The list is used for every stress period and time step after the (IPEROC, ITSOC) tuple until a (IPEROC, ITSOC) tuple is entered with and empty list.

However, it looks like this description is a bit dated, as the release notes for version 3.2.7 state that each timestep where output is desired must be explicitly specified:

OC package: Fixed bug when printing and saving data for select stress periods and timesteps. In previous versions, OC data was repeated until respecified.

For Tutorial 2, I am able to reproduce the imshow head contour plots by manually specifying output at the necessary timesteps:

# Output control
spd = {(0, 0): ['save head','save drawdown'],
       (1, 99): ['save head','save drawdown'],
       (2, 99): ['save head','save drawdown']}
oc = flopy.modflow.ModflowOc(mf, stress_period_data=spd,
                             compact=True)

However, I am not able to reproduce the plot of head vs time because there is only output at the three timesteps I define.

Is there a recommended solution to produce output at all timesteps, without creating a very long dictionary to pass to mfoc? For example, a flag that can be passed to stress_period_data?

Thanks for all your work creating/maintaining this excellent package!

[jtwhite79]

I tend to agree that heads should be saved at the last time step of every stress period by default. I think the save_head_every option was the easy way to do this. Maybe if stress_period_data is None, then we do this by default? I can work on this if no one has any objections.

[langevin-usgs]

As long as there are no side effects. It has been a real pain to get some simple default behavior for output control. Do we need to rethink some of this? Maybe save_head_every isn't the best argument name since we may want to include budget in this as well?

[jtwhite79]

I agree that getting good defaults for OC has been a struggle. I wasn't planning to bring back the save_head_every, but I was thinking that is stress_period_data is None, then reset to write heads at the end of each stress period. We also include budget in the default as well?

[samzipper]

Obviously I don't know the backstory here so maybe this has already been considered and rejected, but to the novice user it seems like there is value in a quick way to print all output. In the example above, something like this may be a potential approach that doesn't revive save_head_every:

spd = {(0, :): ['save head'],
       (1, :): ['save head'],
       (2, :): ['save head']}
# OR
spd = {(:,:): ['save head']}

oc = flopy.modflow.ModflowOc(mf, stress_period_data=spd,
                             compact=True)

[langevin-usgs]

I like that a lot.

[jtwhite79]

I started working on this and (I don't think) we can use (:,:) because the : isn't defined outside of the [] getitem indexer. So what should we use to indicate "all"? My first thoughs wereNone or a Slice instance, but None usually has a different meaning and a slice instance is not-so-easy to construct. But def like the idea of having an "all" indicator. Maybe string "all" would work?

[langevin-usgs]

I guess 'all' is okay. But it something just seems unsatisfying.

[jtwhite79]

So after some digging, it looks like there is an undocumented save_every optional keyword arg with a fair amount of logic (started around line 184 in mfoc.py). And I also don't like the "all" idea..so I added a new option that if stress_period_data is passed as None, you get heads and budget saved for the last time step of every stress period. I haven't changed the default value of stress_period_data yet, but I think it should set to None since that gives a "good" default behavior.

[samzipper]

Thanks for pointing that out. Just to clarify for future beginners who may stumble across this thread, the following command will save head and print budget for all timesteps:

oc = flopy.modflow.ModflowOc(mf, save_every=True, compact=True)

NB: This will override any stress_period_data input. For example in the following case, head is saved at every timestep but you don't get drawdown:

# Output control
spd = {(0, 0): ['save head','save drawdown'],
       (1, 99): ['save head','save drawdown'],
       (2, 99): ['save head','save drawdown']}
oc = flopy.modflow.ModflowOc(mf, stress_period_data=spd,
                             save_every=True, compact=True)