Add new possiblities for providing paramters.

[pat-schmitt]

This PR introduces two new ways to define and provide parameters for the model (in addition to the current default of using cfg.PARAMS):

1. Minimal params.cfg File: You can now pass a minimal params.cfg file to cfg.initialize. If some parameters are missing in this file, the defaults from the main params.cfg (available here) will be used automatically.
2. New run_settings.yml File, which could contain:

• Almost all parameters from params.cfg
• Values of observations used during calibration or initialization
• If any parameter is not included in run_settings.yml, it falls back to the current defaults.

These additions required significant changes in sensitive parts of the model, particularly for the second point. However, tests appear to be passing. I have also included some tests for run_settings.yml. Not every aspect is tested yet, as adding more tests would overly expand the test suite.

Points for Discussion

• Each function currently opens the run_settings.yml file independently, instead of opening it once at the start of a workflow and passing it around. While this approach may slightly reduce performance (dynamic runs are not effected by this), it maintains the flexibility needed for setting up custom workflows.

• The new run_settings.yml works similarly to the existing mb_calib logic but supports calibration for only a single flowline. For now, I’ve kept mb_calib for backward compatibility and enforce a selection between the two possibilities to avoid potential confusions. However, we could consider removing mb_calib in the future.

• When using run_settings.yml, all diagnostics are saved both to gdir.add_diagnostics and the run_settings.yml file. Maybe one or the other is enough?

• Potentially the run_settings could be used to save all used parameters in each gdir. This would avoid confusions when a default params.cfg and a 'mini-params.cfg' is used, and make it clear which actual parameter was used. However, it also means that many same numbers are saved up to 200000 times globally, where one time would also be enough...

• Tests added/passed

• Fully documented

• Entry in whats-new.rst

[pep8speaks]

Hello @pat-schmitt! Thanks for opening this PR. We checked the lines you've touched for PEP 8 issues, and found:

• In the file oggm/core/massbalance.py:

Line 1899:22: E241 multiple spaces after ','

[pat-schmitt]

I managed to fix all tests, beside of python:3.11-slim. The problem is that the package yaml is not getting installed. @TimoRoth or @fmaussion could you direct me, where I need to add the package to be installed? Alternatively, I can do all imports of yaml using try-statements.

Beside of this, the PR slowly gets into a merge-able state. However, as this touches almost all sensitive parts of OGGM, before merging we should have a meeting. I would like to discus if the way I implemented everything is a good way or if their are better ways of doing it. And we need to decide how all this is or could change the default way of handling the params in OGGM.

[fmaussion]

@pat-schmitt what is the best way to move forward with this? It's a lot to review and I'm unsure if it's useful at all for me to look into the details. One way could be to:

• write a notebook demonstrating what can be done with this that couldn't be done before
• we discuss this notebook and implementation details in a meeting, ideally with @gampnico to detect potential issues
• I have a final review and merge

Does this sound reasonable?

[pat-schmitt]

Yes, I will prepare a notebook for discussion.

[pat-schmitt]

I close this, as individual parts are added through other PRs (e.g. #1776, #1777).