Help with some troubleshooting...

[d-strom]

Hi,

I've been trying to get into the pyemu-world for some time after being something of an involuntary prisoner of mixed workflows with Pyemu, GUIs, regular PEST-executables and so on. I'm still not very good with python but I manage with the assistance of documentation, AI, etc. Currently I have been working on a set up for PEST++ IES calibration which basically follows the tutorials in the GMDSI-repo. I have built and modified the underlying model mostly for a very generic test setup (don't bother too much about conceptual reason with this - the main focus is to get the workflow up and running).

I don't really know whats wrong and have been troubleshooting this attached workflow for some time now. Many times I have thought I found the reason for the erroneous behavior but it always turns out I have not... I'm now turning to the community with hope of getting some help with troubleshooting and finding out what the heck is wrong with it!?

I've attached the model and the current state of my python file.

The model is a very simple 1 layer mode. It used to have a topography but is now flat to make it even simpler. Its built with ModelMuse and the code initially process the files to make the hk-arrays etc external for better compatibility (I think) with the pyemu-workflow.

Main findings so far in the troubleshooting:

1. The IES updated parameters but didn't inform the model about the new pars. Reason: The arrays were at this point only written externally but the data was still taken from the npf-file. Solution: Flopys "set_all_data_external" method.
2. The model wasn't using real world coordinates and as such had a discrepancy between the pilot point locations and the domain. Reason: wrong handling of the SpatialReference which didnt add the real world upper left corner coordinate to the x and y. Solution: tried many different approaches, in the end the grid rotation seemed to be problematic as well and i moved the grid origin to 0,0 as well as set the grid angle to 0 for easier troubleshooting.
3. general mismatches between observations from modflow and field measurements. Solution: added a matching function to get the names correct.

A lot of errors has been dealt with and it has almost exclusively been mistakes on my end. Now I'm starting to get a bit frustrated over the things that I haven't been able to solve.

The things I cant seem to fix but I think are causing errors:

1. The kriging interpolation with pilot points doesnt seem to work. It gets stuck and no .fac-file is being built
   I have tried to use grid scale parameters instead just to progress into the calibration and see if it works then but to no avail.
2. The communication between PEST and the model seems like its not working properly. there is only one csv present in the /mult directory and its all ones.. It doesnt seem to get updated with the hk-multipliers.

I'm really at lost here. It feels like I'm close but for whatever reason I cant seem to get my model and PEST to talk when it comes to updating the pars.

I hope there is someone out there that could take a look at the attached files and see if theres anything obvious wrong with them that i havnt figured out. I also attatched my conda environment to know which versions are being used. That was also a bit of a hassle but with these settings its running at least.

Best regards!

D

Workspace.yml

PEST_Setup.zip

[d-strom]

I’ll mark this as completed for now.. I think i have found some of the more critical flaws.
For future reference it seems to be sensitive with dynamic links to the pst among other things.