Cluster network by shapes (NUTS0, NUTS1, NUTS2, NUTS3, and ADM1)

[bobbyxng]

Closes #263

Changes proposed in this Pull Request

• Implementing feature to cluster by NUTS shapes (NUTS0 to NUTS3 for EU countries and OSM administration levels 0-1 for BA, MD, UA, and XK)
• Building on PR Updating all base shapes (country_shapes, europe_shape, nuts3_shapes) #1479 and using updated nuts3_shapes.geojson
• Administration levels can be set for entire modelling region or for individual countries

Example for illustrative purposes

• All European countries set to NUTS0/country level
• DE set to NUTS3
• FR set to NUTS2
• EE set to NUTS0/country_level

Example config

clustering:
  mode: admin 
  admin:
    level: 1
    DE: 3 # optional
    FR: 2 # optional
    ES: 0 # optional
  ...

Checklist

• I tested my contribution locally and it works as intended.
• Code and workflow changes are sufficiently documented.
• Changes in configuration options are added in config/config.default.yaml.
• Changes in configuration options are documented in doc/configtables/*.csv.
• Sources of newly added data are documented in doc/data_sources.rst (Done here: Updating all base shapes (country_shapes, europe_shape, nuts3_shapes) #1479)
• A release note doc/release_notes.rst is added.

[Copilot]

PR Overview

This PR implements a new feature to cluster the network based on administrative regions using NUTS shapes, with configurable levels per country.

• Adds a new function, busmap_for_admin_regions, to map buses to administrative regions.
• Updates the clustering mode selection logic in the main clustering function.
• Extends configuration defaults with new clustering modes and administrative level settings.

Reviewed Changes

File	Description
scripts/cluster_network.py	Introduces the busmap_for_admin_regions function and modifies clustering decision logic.
config/config.default.yaml	Adds new clustering mode options and administrative level defaults.

Copilot reviewed 9 out of 9 changed files in this pull request and generated 3 comments.

[fneum]

Theoretical review first. I'll do a few practical tests now.

[bobbyxng]

Thanks for the comments @fneum !

• I have further improved the workflow by outsourcing the regionalisation to base_network so that all subsequent shapes in resources (regions_onshore_XX) also respect the NUTS boundaries. Voronoi cells are now only formed within administrative regions, if administrative clustering is selected.
• cluster_network.py only makes use of the newly generated admin_shapes.geojson to map the buses. In case no administrative clustering is selected, regionalisation in admin_shapes.geojson will default to administration level 0 / country level.

[bobbyxng]

Open edge case (work-in-progress):
At high resolutions (such as NUTS3), "holes" can form, meaning areas without a substation (see yellow spots in screenshot, reds are substations)

To do

• For these cases, proposal to merge the NUTS3/NUTSX region with a neighbour of the same NUTS2/NUTS(X-1) parent region. If multiple neighbours exist, pick the neighbour with the smallest area.

Edit: done!

[bobbyxng]

Update
I have finalised the implementation and added the following:

• For regions that do not contain a substation, implemented an algorithm: Find all neighbours of the same parent region, if no neighbour exist, allow for neighbours of other parent region (however, same country), find the neighbour with the most substations, then sort by smallest area (ascending). --> Merge with neighbour that has most substations and is the smallest in size (secondary). Algorithm is recursive. If all countries are set to level 3, this takes about 30 seconds to finish.
• Any combination of admin levels (country, admin level 1-3) is technically possible
• cluster_network: x, y coordinates are updated to internal points (Pole of Inaccessibility), so the clustered node is guaranteed to be within the largest polygon of each clustered administrative region
• London area is always clustered to level 1 (GBI), as I have observed converging issues in the past

Screenshots

• Updated x, y coordinates (yellow before/centroid; red new/PoI):
  
  

• Aggregated, merged admin regions without "holes".

• In this example, all countries are set to level 1, GR to level 2, DK, DE, GB to level 3
  

[Copilot]

PR Overview

This PR implements clustering by administrative regions using NUTS shapes, allowing different administration levels for different countries.

• Introduces new functions to create a busmap for administrative regions, update bus coordinates using the pole of inaccessibility, and select the largest polygon from multipolygons.
• Modifies the clustering logic to support an “administrative” mode and updates configuration options accordingly.

Reviewed Changes

File	Description
scripts/cluster_network.py	Adds functions for clustering by administrative regions and adjusts logic
config/config.default.yaml	Adds new configuration options for clustering including mode and admin level

Copilot reviewed 10 out of 10 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

scripts/cluster_network.py:72

• The code uses 'gpd' for geopandas functions (e.g., gpd.read_file and gpd.sjoin_nearest), but no import for geopandas is present. Please add 'import geopandas as gpd' at the top of the file.

import numpy as np

[fneum]

LGTM - Thanks a lot @bobbyxng for taking care of it!