Refactor regional inputs and add standalone AvaDirectory migration workflow

[chhesselbach]

Commit overview

refactor: use consistent dataUtils imports

• Replaced redundant direct helper imports with consistent dataUtils.* calls.
• Applied the same import style across PRA and AvaDirectory modules.
• Reduced duplicate imports without changing workflow behavior.

refactor: simplify regional inputs and fix PRA output handling

• Made processing boundaries and regional metadata optional and modular.
• Added automatic DEM-outline generation when no boundary is provided.
• Removed the requirement for fixed administrative column names.
• Made the forest raster optional for PRA delineation.
• Fixed Step 07 GeoJSON filenames and internal layer names.
• Fixed size-specific PRA filtering.
• Fixed duplicated praIDpraID... raster names.
• Fixed missing pathInputpra values.
• Prevented praID.tif from being copied into RELArea.

rm tests

• Removed local development tests from version control.
• Local tests remain excluded through .gitignore.
• This does not affect the model-chain runtime.

feat: add standalone AvaDirectory migration runner

• Added runAvaDirectory.py for running AvaDirectory Steps 13–15 independently.
• Added a dedicated default INI and local override support.
• Uses a single projectRoot setting.
• Automatically discovers current flat and old nested FlowPy structures.
• Normalizes outputs into flat 11_avaDirectoryData and 12_avaDirectory directories.
• Reuses completed scenario folders and processes only missing or incomplete scenarios.
• Normalizes legacy praIDpraID... filenames during collection.
• Added output validation so empty steps no longer report successful completion.
• Added the pixi run avadirectory task.

[chhesselbach]

@PaulaSp3 please confirm that all is running as it should. thanks

[PaulaSp3]

dynamic parameterization fails! Please check the minor modifications!

[chhesselbach]

dynamic parameterization fails! Please check the minor modifications!

Okay, I understand why the error occurs now. While testing the new structure from ATI PR #32 and the addition of runAutoAtesModelChain.py, everything was running, but I did not recognize that the intended separation of size classes in AvaScenario was not working correctly. Every input PRA layer contained all size classes because gSel = gBand.copy() overwrote the preceding size filter.

For runAutoAtesModelChain.py and dynamic parameterisation, keeping all PRAs together in one raster is required. The existing gSel = gBand.copy() provided this behavior, although the combined raster was misleadingly named as a size-5 output.

runAvaScenModelChain.py, however, requires real size separation: PRAs must be filtered according to their assigned size classes and processed separately.

When I removed the unconditional gSel = gBand.copy(), AvaScenario started separating the size classes correctly, but AutoATES/dynamic parameterisation lost the combined raster they require. I did not recognize this dependency at first. sorry!

Your proposed condition is a suitable quick fix:

• If assignElevSize = True, filter by the assigned size.
• If assignElevSize = False, keep all PRAs together.

This restores both workflows, although AutoATES still creates repeated or misleading size-labelled outputs.

As a cleaner long-term solution, I suggest introducing two explicit modes:

• classified for AvaScenario: split PRAs by elevation band and size.
• allSizes for AutoATES/dynamic parameterisation: create one combined PRA-area raster and one combined PRA-ID raster.

Example AvaScenario inputs:

pra030secE-1800-2000-size3-praAreaM.tif
pra030secE-1800-2000-size5-praAreaM.tif

Example AutoATES/dynamic inputs:

pra030secE-allPRA-maxSize5-praAreaM.tif
pra030secE-allPRA-maxSize3-praAreaM.tif

The existing sizeMax setting remains responsible for capping the size-dependent dynamic parameterisation, so no additional size-cap setting is needed. Including maxSize in the filename would also make the applied parameterisation limit explicit.

How do you like the idea? should i give it a try?

[chhesselbach]

@PaulaSp3 pls check again.

• added NumPy-style docstrings to the four new helper functions;
• consolidated the optional-forest handling for both path modes;
• restored the combined-PRA behavior for AutoATES/dynamic parameterisation while preserving size filtering for AvaScenario (for now easy fix! ;))

I also tested runAutoAtesModelChain.py: Step 07 now completes without the previous zero-feature error. Could you please review the latest changes again?

[PaulaSp3]

I've created an issue (#46) for the ideas you had to solve if we write different files for different classes.