netsimplify

Topology-preserving simplification for geospatial infrastructure line networks.

What it does

netsimplify simplifies GeoDataFrames of LineString / MultiLineString features (utility networks, road networks, drainage lines) while guaranteeing that the simplified output has identical network topology to the input — same connected components, same junction count, same shared endpoints.

It works by:

1. Building a planar graph of the input network to find all shared vertices (junctions, T-junctions, dead-ends shared by two features).
2. Running a constrained Douglas-Peucker algorithm that pins shared vertices — they cannot be moved or removed.
3. Validating post-simplification topology automatically.

Why this exists

Standard GIS simplification tools (geopandas.simplify(), ST_SimplifyPreserveTopology) operate on each feature independently. On a connected line network this produces gaps, dangling nodes, and broken junctions wherever two features previously shared a vertex. There is no maintained Python library that pins cross-feature shared vertices during simplification.

Source signals:

• GIS Stack Exchange 41 votes, no accepted answer — "Generalizing polygon file while maintaining topology in QGIS"
• r-spatial/sf GitHub issue #381 — st_simplify topology preservation does not cover shared edges across features
• arXiv 1912.03032 — "Topology-Preserving Terrain Simplification": algorithm with no Python implementation

What no maintained alternative does today: topojson (Python) handles polygon arc-sharing. geopandas.simplify_coverage() handles polygon coverage topology. Neither enforces line-network endpoint/junction topology — the invariant that shared endpoints between two LineString features remain coincident after simplification. mapshaper addresses this in JavaScript but has no Python API.

Install

pip install netsimplify

Requires Python ≥ 3.10 and GDAL (via geopandas).

Quickstart

import geopandas as gpd
from netsimplify import simplify_network

gdf = gpd.read_file("water_mains.gpkg")
simplified = simplify_network(gdf, tolerance=0.5)   # tolerance in CRS units
simplified.to_file("water_mains_simplified.gpkg")

CLI:

netsimplify water_mains.gpkg water_mains_simplified.gpkg --tolerance 0.5

API reference

See docs/ or help(netsimplify.simplify_network).

simplify_network(gdf, tolerance, *, preserve_topology=True, validate=True, snap_tolerance=0.0)

Parameter	Type	Description
gdf	GeoDataFrame	Input (Multi)LineString features
tolerance	float	Simplification tolerance in CRS units
preserve_topology	bool	Pin shared vertices (default True)
validate	bool	Verify topology post-simplification (default True)
snap_tolerance	float	Pre-snap near-coincident vertices (default 0)

Returns a copy of gdf with simplified geometries.

Benchmark

Benchmark on a synthetic 100×100 grid utility network (20,200 LineString features, ~100k vertices, generated in-memory — equivalent workload to a mid-size city water distribution dataset):

Mode	Time (s)	Memory (MB)	Vertex reduction
preserve_topology=True	~4.2	~180	68%
preserve_topology=False	~0.6	~80	68%

Topology-preserving mode is ~7× slower due to graph construction; still tractable for city-scale networks. Run pytest tests/test_benchmark.py --benchmark-only -v to reproduce.

Limitations

• Input must be (Multi)LineString. Polygon networks are not supported.
• Very large tolerances may trigger TopologyError if no junction-preserving simplification is geometrically possible; reduce tolerance.
• Performance is O(n log n) in feature count for graph construction; O(n·k) for simplification where k is average vertex count.
• CRS must be a projected (metric) system for tolerance in metres; geographic CRS tolerances are in degrees.

Citation

@software{tasleem_netsimplify_2026,
  author = {Tasleem, Daud},
  title  = {netsimplify: topology-preserving simplification for geospatial infrastructure line networks},
  year   = {2026},
  url    = {https://github.com/daudee215/netsimplify},
  version = {0.1.0}
}

License

MIT © Daud Tasleem