Artefacts in output when tessellating islands with narrow channels

[jontwo]

Describe the problem

Running morphological_tessellation on some island coastlines, there were some strange artefacts in the output where the channel was <100m wide.

Input island coastlines.

Tessellation output overlaid on the coastlines, the narrow point is around 60m wide.

Steps to reproduce

import geopandas
import momepy

coastline = "path/to/coastline.shp"
polygons = "path/to/output.shp"

coast_gdf = geopandas.read_file(coastline).reset_index(drop=True)
result = momepy.morphological_tessellation(coast_gdf, segment=500, shrink=20)
result.to_file(polygons)

Versions of your packages

python=3.12.11
momepy=0.9.1
geopandas=1.1.1
libpysal=4.13.0
networkx=3.5

Your operating system

Ubuntu 22.04.5 LTS

Additional context

Having reread the doc, I can see the values we were using for segment and shrink are way off the default values. They had been working for us up to that point. When I lowered the segment value to 100, the issue went away.

So, you can probably close this issue as being user error, but could I ask for a clarification in the docs? Both segment and shrink represent a distance, but is this map units? I'm assuming it is, but it would be good to clarify it, as it could be pixels or something else.

[jontwo]

Almost forgot the test data.

coastline_repro.zip

[jontwo]

Also, there is a stray close bracket in the help string for shrink. I don't really understand this, so some clarification would be useful.

shrink: float, optional
The distance for negative buffer to generate space between adjacent polygons). By default 0.4

[jGaboardi]

Also, there is a stray close bracket in the help string for shrink. I don't really understand this, so some clarification would be useful.

That extra parenthesis is simply a typo.

[martinfleis]

Yes, this is a user error. If the distance of two edges is 60 meters, then 500 meters is simply too large.

Both segment and shrink represent a distance, but is this map units? I'm assuming it is, but it would be good to clarify it, as it could be pixels or something else.

It is always map units across the entire package but we can potentially see if some clarifications can be made.

[jontwo]

Also, there is a stray close bracket in the help string for shrink. I don't really understand this, so some clarification would be useful.

That extra parenthesis is simply a typo.

I meant that I didn't understand the help text, not the bracket ;)

[jGaboardi]

@jontwo ah, sorry for assuming! Do you have thoughts on how we can make that more clear?

[jontwo]

@jGaboardi I think the update in the PR is a big improvement! I just needed a bit more detail. Though the last sentence of segment for enclosed tessellation seems to stop prematurely (line 204).

[jGaboardi]

@jontwo -- Good point. We welcome any contribution to the docs, no matter how small. Please open a PR if you have time & desire to help us improve!