Adding parameters/observations from dataframe instances?

[gmerritt123]

I am trying to build a pst file programatically, either using PstFrom or using a "blank" Pst instance (not sure what road to go down on that currently).

I'm coming at this from a non-modflow perspective having done most of my previous pst setups semi-programmatically... but now I'm trying to harness what you've provided with this awesome library. But I could use some advice!

I am initializing my parameterization programmatically because a) they're likely to be changed/modified and b) because of it being NOT modflow, I don't actually have input files that fit with the args/formats available in the add_parameters method in PstFrom, and I shouldn't need to. All I want is to:

• DONE by me: populate a dataframe of all my parameters (and their required specs --> upper bound, lower bound, transform etc etc); and

• add that dataframe to the pstFrom (or Pst) Instance so that it'll write them to the pst file appropriately.

• ... then do the same for the observations, obs value, weights, std deviation etc.

• ...then continue using all the awesome convenience provided to set other things like pest++ options etc.

• ...then write the pst file

Looking at an instance of PstFrom, I've noticed there is a public par_dfs attribute (and obs_dfs) attribute that I assume holds a list of dataframes get added when you call add_parameters/observations, then these dataframes get written to the pst when you call the write method. Can I bypass the add_parameters/observations methods by simply appending my dataframes here? And if so, what columns do I need them to have and what other attributes do I need to "manually" fill in?

All advice appreciated!

[mnfienen]

hey @gmerritt123

If your parameterization doesn't fit well into the PstFrom paradigm, I highly recommend creating your TPL and INS files externally and then look at the Pst object and in particular the from_io_files method. That will populate the dataframes for pars and obs and set up a Pst object with defaults for most values (and will populate the template and instruction file information). Then you should just need to update the forward run script and bring in the dataframe info you already assembled (bounds, etc.)

[gmerritt123]

Awesome thanks, that is where I will go :-D

[mnfienen]

sounds good! I'm going to close this discussion but def reach out if you run into issues. Good luck!