ARCGIS

GeoSubSampler for ArcGIS Pro

ArcGIS Pro equivalent of the GeoSubSampler QGIS plugin, modelled on SGTool/ARCGIS.

File	Role
GeoSubSampler.pyt	Python toolbox. Launch GeoSubSampler GUI starts the Tkinter window as a separate process, passes it the project's point/line/polygon layers, and adds each result to the map as it is written (symbology copied from the source layer), in an Outputs or T&P group.
GeoSubSampler_GUI.py	Tkinter window - same tabs, groups and settings as the QGIS dock widget.
gss_processing.py	GUI-independent port of the plugin's processing methods. Uses the shared ../calcs package.
setup_arcgispro_env.py	Sets up the Python environment (clone of the default one plus the packages needed); run by the toolbox's Setup tool.

Install

1. In ArcGIS Pro: Catalog > Toolboxes > Add Toolbox > ARCGIS/GeoSubSampler.pyt.
2. The ArcGIS Pro Python environment needs geopandas, shapely (2.0 or newer), scikit-learn, networkx, scipy, matplotlib (and fiona for map simplification, triangle for dyke triangulation). The easiest way is to run the toolbox's Setup > Set Up Python Environment tool, then restart ArcGIS Pro. It runs setup_arcgispro_env.py from this folder, which clones the default environment, installs everything, and activates the clone; it does not change arcgispro-py3 or any other existing environment, and ArcGIS Pro must be restarted to use the new one (see the main README). To do it by hand instead: clone the default environment in the Python Package Manager, add what is missing, and activate the clone. The launch tool checks this and says what is missing.

   Note: geopandas is not included in the default ArcGIS Pro environment and has to be added before the tool will run. In Python Package Manager, clone the default environment (arcgispro-py3 cannot be modified), search for geopandas under Add Packages and install it, then activate the clone and restart ArcGIS Pro. Or from the Python Command Prompt: conda install -n <your-clone> geopandas.

   Also install the following (if not already installed): scikit-learn, networkx, scipy, matplotlib and fiona, e.g. conda install -n <your-clone> geopandas scikit-learn networkx scipy matplotlib fiona.

3. Run Launch GeoSubSampler GUI. Keep the tool running; closing the GUI window ends it.

What the window offers

The same layout as the QGIS plugin. Always visible at the top: the layer selectors, Refresh from map, Show Metadata... and the five Geology Polygon Priorities (they keep their values whichever tab is open). Below them, four tabs:

Tab	Tools
Clean	Fix Map (topology clean-up within a tolerance, with a text report), Repair (buffer 0), Densify Nodes, Thin Nodes
Points	Orientation subsampling: Stochastic, Kent with an optional Min kappa, Kent Outlier with an optional Iterative mode, Carmichael, 1o Sampling, with "Auto-fit to N pts"
Faults	Merge Segments; Length, Graph, Strat Offset, Orientation Clusters
Polygons	Rescale Map (Ignore Field and Codes, Tolerance, Diameter, Series), Simplify Map, Töpfer & Pillewizer scaling

Hover over a control for its tooltip.

Results in the map

Results are added to the first map of the project, with the symbology of the layer they were made from:

• most tools put them in an Outputs group at the top of the Contents pane, newest on top;
• Töpfer & Pillewizer scaling puts them in a T&P group at the top, set so that only one of its sublayers is visible at a time. It has one group per step, named after the ratio and the time of the run (e.g. Ratio 0.5 (14:02:31)), holding that step's points, faults and polygons in that order, so ticking a step shows the whole map at that scale. The steps of the latest run are at the top, and the newest step is shown, so the target scale is shown at the end of a run.

The fault attribute tools instead colour their result by the attribute they add: Length by line_len (graduated colours, 10 quantile classes), Graph by edge_type, Strat Offset by max_off_ds and Orientation Clusters by cluster (unique values).

Metadata

Every result gets a metadata file, <file>.shp.sgt.xml (see the main README). Show Metadata... opens a window with the history of a layer: choose it from the list (it starts with the newest result) or Browse... for a shapefile. The History tab shows when and how it was made, from which files and with which settings, back through every earlier GeoSubSampler or SGTool step; the XML tab shows the file itself.

Notes

• Fix Map (settings: Tolerance, Fill enclosed gaps, Remaining overlaps; see How Fix Map works) repairs invalid polygons, removes doubled polygons/faults, cuts holes for polygons inside polygons, closes small gaps and overlaps and matches nodes within the tolerance (metres), and adds nodes where lines cross. Outputs are <layer>_fixed shapefiles plus <layer>_fixed_report.txt. Remaining overlaps cuts any overlap left at the end out of the bigger (default), the smaller, or neither polygon. Fill enclosed gaps (on by default) fills every gap completely surrounded by polygons, of any width or size, with the neighbour sharing the longest boundary.
• Densify Nodes uses the selected polygon and/or fault layer, with a node spacing in metres; with both, new nodes are shared between them. Outputs are <layer>_dense shapefiles.
• Thin Nodes is the opposite: it removes nodes closer than a minimum spacing (metres) from the selected polygon and/or fault layer, removing a shared node from every feature that uses it. Junctions (three or more polygons, where faults meet, leave or cross boundaries or other faults) and line ends are kept, and no removal may create a self-intersection. Outputs are <layer>_thin shapefiles; progress for each round is written to the log.
• Strat Offset also writes <fault layer>_stratOffset_samples (a point layer of where the offset was read, with the offset level and the units either side); it is added to the map without copying symbology. Each fault is sampled where it meets a contact and at its two ends (an end already on a contact is not sampled twice); a fault that meets no contact is sampled at its start, middle and end.
• Feature classes in a geodatabase are exported once to a shapefile (field names shortened to 10 characters). Every button press writes its results (shapefiles, reports, Fix Map's progress file) to a new folder named after the tool and the time, e.g. Fix_map_20261105_164049, beside the source, or in the project folder if that folder is not writable - the same rules as the QGIS plugin. A series (Rescale Map, T&P scaling) shares one folder.
• "Refresh from map" re-reads the layer list after you add layers to the project.
• Layer selections and definition queries in the map are not applied: the whole data source is used.
• Add file... lets the GUI use shapefiles that are not in the map. The GUI can also be run stand-alone: python GeoSubSampler_GUI.py.