Repository navigation
Open
Description
I am using the land polygons from natural earth and get self intersections reported, however when I analyse the source geometry in QGIS it seems to be valid. I thought this could be coming from the simplification, but even if it is turned off it still reports the problem:
layer config
"land": { "minzoom": 0, "maxzoom": 8, "source": "ne_data/ne_10m_land/ne_10m_land.shp"},
"landcp": { "minzoom": 0, "maxzoom": 8, "source": "ne_data/ne_10m_land/ne_10m_land.shp", "simplify_below": 8, "simplify_level": 0.0001, "simplify_ratio": 2 },
output:
Bounding box -180, -85.06, 180, 85.06
Reading .shp land
Shapefile entity #0 type 5 is invalid. Parts:98. Reason:Geometry has invalid self-intersections. A self-intersection point was found at (-159.369, -180); method: m; operations: i/x; segment IDs {source, multi, ring, segment}: {0, 0, -1, 7989}/{0, 0, -1, 8119}... failed to correct. Reason: Multi-polygon has intersecting interiors
Shapefile entity #6 type 5 is invalid. Parts:69. Reason:Geometry has invalid self-intersections. A self-intersection point was found at (165.088, -21.8912); method: i; operations: i/u; segment IDs {source, multi, ring, segment}: {0, 56, -1, 280}/{0, 56, -1, 282}... failed to correct. Reason: Geometry has invalid self-intersections. A self-intersection point was found at (165.088, -21.8912); method: i; operations: i/u; segment IDs {source, multi, ring, segment}: {0, 56, -1, 280}/{0, 56, -1, 282}
Reading .shp landcp
Shapefile entity #0 type 5 is invalid. Parts:98. Reason:Geometry has invalid self-intersections. A self-intersection point was found at (-159.369, -180); method: m; operations: i/x; segment IDs {source, multi, ring, segment}: {0, 0, -1, 7989}/{0, 0, -1, 8119}... failed to correct. Reason: Multi-polygon has intersecting interiors
Shapefile entity #6 type 5 is invalid. Parts:69. Reason:Geometry has invalid self-intersections. A self-intersection point was found at (165.088, -21.8912); method: i; operations: i/u; segment IDs {source, multi, ring, segment}: {0, 56, -1, 280}/{0, 56, -1, 282}... failed to correct. Reason: Geometry has invalid self-intersections. A self-intersection point was found at (165.088, -21.8912); method: i; operations: i/u; segment IDs {source, multi, ring, segment}: {0, 56, -1, 280}/{0, 56, -1, 282}
input file: https://www.naturalearthdata.com/http//www.naturalearthdata.com/download/10m/physical/ne_10m_land.zip
Metadata
Metadata
Assignees
Labels
No labels
Activity
systemed commentedon Mar 30, 2022
Could be reprojection artefacts, or a difference in validity tests between JTS/GEOS (as used by QGIS) and Boost.Geometry (as used by tilemaker). They're only warnings - you can safely ignore them, but if you find any land polygons disappearing as a result then do lodge an issue.
dieterdreist commentedon Apr 5, 2022
I just found land polygons disappearing here (zooming in):

zooming close:

systemed commentedon Apr 5, 2022
Thanks. I'm not going to have a chance to look at this immediately but that's helpful.
dieterdreist commentedon Apr 5, 2022
for completeness, this is the config line for the land polygons:
"
land": { "minzoom": 0, "maxzoom": 8, "source": "ne_data/ne_10m_land/ne_10m_land.shp", "simplify_below": 8, "simplify_level": 0.0001, "simplify_ratio": 2 },as you can see I only created tiles to zoom 8, and did not simplify the z8 tiles, so what you see above is likely coming from reprojection as you suggested.
systemed commentedon Oct 3, 2022
There have been a few changes recently that might fix this - I haven't tried it out with this dataset but you might want to have a look.
dktue commentedon Oct 29, 2026
I retested this with current master (
4e0f03d) and the config from the first post (ne_10m_land, z0–8, one layer unsimplified, one simplified). The "invalid ... failed to correct" warnings for entities #0 and #6 still appear. More importantly, many output polygons are invalid in tile coordinates (crossing/overlapping edges, spikes), which is what makes renderers drop or misdraw land:The cause is that polygons are only repaired when they were simplified; quantising them onto the tile grid can make them invalid too. #929 with
"invalid_polygon_repair_scope": "all"brings the unsimplified layer down to 36 at z3 and 2 at z8 (and the simplified layer to 2 at z8). Details in #929.dktue commentedon Nov 3, 2026
Follow-up: #929 is merged now, and the option is called
"repair_polygons": true(set per layer). Rechecked on current master (2789651) with the same config as above, counting output polygons that are invalid in tile coordinates:"repair_polygons": trueSo with the option set on the land layer, most of the invalid output is gone; what remains is mostly at z0–z3, where
ne_10m_landis very coarse on the tile grid. The shipped OpenMapTiles config already sets it onwaterandocean.