{"success":true,"data":{"author":"geoBoundaries (William & Mary geoLab); Portolan mirror","bucketAccessType":"public","bucketListable":true,"bucketName":"us-west-2.opendata.source.coop","bucketPrefix":"portolan-mirrors/geoboundaries-catalog","category":"Earth Observation & Geospatial","cloudProvider":"aws","createdAt":"2026-11-18T17:37:22.720Z","description":"Every boundary in the geoBoundaries Comprehensive Global Administrative Zones (CGAZ) composite, produced by William & Mary geoLab and the geoBoundaries community, as cloud-native GeoParquet 1.1 and PMTiles. Four collections: adm0 (218 country outlines plus 20 disputed areas), adm1 (3,224 states, provinces, regions and governorates), adm2 (49,349 counties, districts, municipalities and communes) and names (52,398 named units, a sorted search index for autocomplete with shapeID joins). The levels cover the globe seamlessly; where a country has no ADM1 or ADM2 data the coarser polygon is carried up and shapeType says which level is present. Each level also has a simplified display file buffered outward so it contains the true boundary. The mirror repairs 38 double-encoded names, 1,275 invalid geometries and a missing CRS, documented in changes-from-upstream.md; per-country gbOpen releases are not included. Files are Hilbert-sorted with small row groups for range requests. Read with DuckDB, geopandas or GDAL. Licensed CC BY 4.0 by geoBoundaries; cite Runfola et al. (2020), PLoS ONE 15(4):e0231866.","displayTitle":"geoBoundaries CGAZ Global Administrative Boundaries ADM0, ADM1, ADM2 (GeoParquet, PMTiles, name index)","fuseMountBrowsable":true,"id":"5b82f6af-9a54-561d-a948-db3f1b0c1823","license":"CC-BY-4.0","licenseEvidenceUrl":"https://www.geoboundaries.org/","objectCountEstimate":4,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1371/journal.pone.0231866","region":"us-west-2","schemaCard":{"accessHints":[{"language":"python-anonymous","snippet":"import boto3\nfrom botocore import UNSIGNED\nfrom botocore.client import Config\ns3 = boto3.client('s3', region_name='us-west-2', config=Config(signature_version=UNSIGNED))\ns3.list_objects_v2(Bucket='us-west-2.opendata.source.coop', Prefix='portolan-mirrors/geoboundaries-catalog/', Delimiter='/')"},{"language":"python-local","snippet":"open('/mnt/archive/adm0/styles/adm1-units.json', 'rb')"}],"caveats":["Files are binary — use a domain-specific parser (e.g. metpy / pynexrad / pydicom) rather than text I/O."],"format":"binary","generatedAt":"2026-11-18T17:39:27.254Z","layout":{"example":"portolan-mirrors/geoboundaries-catalog/adm0/styles/adm1-units.json","pattern":"{prefix}/{seg1}/{seg2}/{seg3}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"{\n  \"version\": 8,\n  \"name\": \"First-level units per country\",\n  \"sources\": {\n    \"data\": {\n      \"type\": \"vector\",\n      \"url\": \"pmtiles://../adm0.pmtiles\"\n    }\n  },\n  \"layers\": [\n    {\n      \"id\": \"adm0-fill\",\n      \"type\": \"fill\",\n      \"source\": \"data\",\n      \"source-layer\": \"adm0\",\n      \"paint\": {\n        \"fill-color\": [\n          \"step\",\n          [\n            \"coalesce\",\n            [\n              \"get\",\n              \"adm1_units\"\n            ],\n            0\n          ],\n          \"#e7c388\",\n     ","encoding":"utf8","key":"portolan-mirrors/geoboundaries-catalog/adm0/styles/adm1-units.json","truncatedAt":512},{"bytes":"{\n  \"version\": 8,\n  \"name\": \"Outline compactness\",\n  \"sources\": {\n    \"data\": {\n      \"type\": \"vector\",\n      \"url\": \"pmtiles://../adm0.pmtiles\"\n    }\n  },\n  \"layers\": [\n    {\n      \"id\": \"adm0-fill\",\n      \"type\": \"fill\",\n      \"source\": \"data\",\n      \"source-layer\": \"adm0\",\n      \"paint\": {\n        \"fill-color\": [\n          \"step\",\n          [\n            \"coalesce\",\n            [\n              \"get\",\n              \"compactness\"\n            ],\n            0\n          ],\n          \"#f7e3c0\",\n          0.05","encoding":"utf8","key":"portolan-mirrors/geoboundaries-catalog/adm0/styles/compactness.json","truncatedAt":512}],"sampleObjects":["portolan-mirrors/geoboundaries-catalog/adm0/styles/adm1-units.json","portolan-mirrors/geoboundaries-catalog/adm0/styles/compactness.json","portolan-mirrors/geoboundaries-catalog/adm0/styles/default.json","portolan-mirrors/geoboundaries-catalog/adm0/styles/status.json"],"sidecars":[{"bytes":"# geoBoundaries CGAZ Global Administrative Boundaries\n\nEvery administrative boundary in the geoBoundaries **Comprehensive Global\nAdministrative Zones** composite, republished as cloud-native GeoParquet and\nPMTiles you can query and draw without downloading anything.\n\n[geoBoundaries](https://www.geoboundaries.org/), built by the community and\n[William & Mary geoLab](https://geolab.wm.edu/), produces the data and remains\nthe authoritative source.\n\n> **This is not a byte-for-byte mirror.** It repairs faults in the upstream\n> release — 38 double-encoded names, 1,275 invalid geometries, a missing CRS —\n> and adds products geoBoundaries does not publish: a name search index and\n> simplified display geometry. Every deviation, with how to reproduce it and\n> how to get the vanilla behaviour instead, is in\n> **[changes-from-upstream.md](changes-from-upstream.md)**.\n\n## What is here\n\nFour collections.\n\n| collection | rows | full | display | what it is |\n|---|---|---|---|---|\n| `adm0` | 218 | 104 MB | 7.5 MB | Country outlines, plus 20 disputed areas |\n| `adm1` | 3,224 | 81 MB | 13 MB | States, provinces, regions, governorates |\n| `adm2` | 49,349 | 104 MB | 42 MB | Counties, districts, municipalities, communes |\n| `names` | 52,398 | 3.9 MB | — | A name search index across all three levels |\n\nThe three boundary levels cover the globe seamlessly: every one of the 218\ncountry codes appears at every level, because where a country has no\nfirst- or second-level data CGAZ carries its coarser polygon up instead. The\n`shapeType` column says which you are actually looking at.\n\nEach boundary collection is GeoParquet 1.1 with bbox covering metadata, sorted\nalong a Hilbert curve, and cut into small row groups — 218, 202 and 193 of\nthem, holding 1, 16 and 256 rows — so a client fetches one unit's geometry in\na single range request of 250 to 325 KB instead of pulling the whole file. At\nADM0 that is one row group per country.\n\n`names` is the piece built for autocomplete: one row per named unit, sorted by\na folded form of the name, 3.9 MB in total. Load it whole in about a quarter of\na second, or resolve a typed prefix from one or two 109 KiB row groups by range\nrequest. It carries the `shapeID` that joins back to the geometry, the extent\nto fly a map to, a `geometry_ref` pointer straight at the row group holding the\ngeometry, and a count of how many other units share the same name.\n\nA PMTiles archive accompanies every collection, for drawing many units at once.\n\n**Every** boundary collection carries a **display** file: the same rows in the\nsame order in the same row groups, with simplified geometry, for drawing one\nunit's outline without fetching the full-resolution row group. One country's\noutline is about 10 KB there against 250 KB in the full file, one ADM1 unit\n37 KB against 220 KB, one ADM2 unit 151 KB against 325 KB. Canada is 528,021\ncoordinates at full resolution and 68,973 in the display file.\n\nADM2 gained its display file on 2026-11-12, reversing an earlier decision that\nwas right about the wrong quantity. ADM2 geometry is already simplified\nupstream — its median unit is 82 coordinates — so *decoding* one was never\nslow. But a browser does not decode one unit from an isolated byte range:\nhyparquet has no sub-chunk read, so fetching one ADM2 feature pulls that row\ngroup's entire geometry column chunk, and against the live catalog that was a\nmedian of 1.35 MB and 2.1 seconds, worst case 3.5 MB and 4.3 seconds. The cost\nwas bytes per row group, and it is the tail that fills a row group — the 99th\npercentile at 2,837 coordinates and the 14 units above 20,000 — not the median\n82.\n\n**The display geometry is buffered outward** by the distance it was simplified\nby, so it contains the true boundary rather than cutting inside it. That makes\nit safe to use as a spatial filter for downloading some *other* dataset: you\nover-fetch a little rather than silently dropping features along the edge.\nEvery feature at every level retains at least 99% of its original area — worst\ncases 99.95% at ADM0, 99.87% at ADM1 and 99.15% at ADM2. The c","encoding":"utf8","key":"portolan-mirrors/geoboundaries-catalog/README.md","name":"README.md","size":15745,"truncated":true},{"bytes":"{\n  \"type\": \"Catalog\",\n  \"stac_version\": \"1.1.0\",\n  \"stac_extensions\": [\n    \"https://schemas.portolan-sdi.org/portolan/v0.2.0/schema.json\"\n  ],\n  \"id\": \"geoboundaries-cgaz\",\n  \"title\": \"geoBoundaries CGAZ Global Administrative Boundaries\",\n  \"description\": \"Every administrative boundary in geoBoundaries' Comprehensive Global Administrative Zones (CGAZ) composite, mirrored as cloud-native GeoParquet and PMTiles. Three boundary collections cover the globe seamlessly: 218 country-level units, 3,224 first-level units and 49,349 second-level units, built by [William & Mary geoLab](https://geolab.wm.edu/) from country submissions clipped to US Department of State international boundaries, with disputed areas carried as polygons of their own. A fourth collection is a name search index, one row per named unit across all three levels, sorted so that a client can resolve a typed prefix with range requests instead of a download. This is a mirror; [geoBoundaries](https://www.geoboundaries.org/) is the authoritative source. Read the [agent guide](AGENTS.md) before you query it.\",\n  \"updated\": \"2026-11-12T00:00:00Z\",\n  \"links\": [\n    {\n      \"rel\": \"self\",\n      \"href\": \"https://data.source.coop/portolan-mirrors/geoboundaries-catalog/catalog.json\",\n      \"type\": \"application/json\",\n      \"title\": \"geoBoundaries CGAZ Global Administrative Boundaries\"\n    },\n    {\n      \"rel\": \"root\",\n      \"href\": \"./catalog.json\",\n      \"type\": \"application/json\",\n      \"title\": \"geoBoundaries CGAZ Global Administrative Boundaries\"\n    },\n    {\n      \"rel\": \"describedby\",\n      \"href\": \"./README.md\",\n      \"type\": \"text/markdown\",\n      \"title\": \"Catalog README\"\n    },\n    {\n      \"rel\": \"agents\",\n      \"href\": \"./AGENTS.md\",\n      \"type\": \"text/markdown\",\n      \"title\": \"Catalog agent guide\"\n    },\n    {\n      \"rel\": \"vcs\",\n      \"href\": \"https://github.com/portolan-mirrors/geoboundaries-catalog\",\n      \"title\": \"Source repository\"\n    },\n    {\n      \"rel\": \"issues\",\n      \"href\": \"https://github.com/portolan-mirrors/geoboundaries-catalog/issues\",\n      \"title\": \"Issue tracker\"\n    },\n    {\n      \"rel\": \"via\",\n      \"href\": \"https://www.geoboundaries.org/globalDownloads.html\",\n      \"type\": \"text/html\",\n      \"title\": \"geoBoundaries CGAZ downloads (upstream source)\"\n    },\n    {\n      \"rel\": \"cite-as\",\n      \"href\": \"https://doi.org/10.1371/journal.pone.0231866\",\n      \"type\": \"text/html\",\n      \"title\": \"Runfola et al. (2020), geoBoundaries: A global database of political administrative boundaries, PLoS ONE 15(4): e0231866\"\n    },\n    {\n      \"rel\": \"license\",\n      \"href\": \"https://creativecommons.org/licenses/by/4.0/\",\n      \"type\": \"text/html\",\n      \"title\": \"Creative Commons Attribution 4.0 International\"\n    },\n    {\n      \"rel\": \"child\",\n      \"href\": \"./adm0/collection.json\",\n      \"type\": \"application/json\",\n      \"title\": \"CGAZ ADM0 - Country Boundaries\"\n    },\n    {\n      \"rel\": \"child\",\n      \"href\": \"./adm1/collection.json\",\n      \"type\": \"application/json\",\n      \"title\": \"CGAZ ADM1 - First-Level Administrative Boundaries\"\n    },\n    {\n      \"rel\": \"child\",\n      \"href\": \"./adm2/collection.json\",\n      \"type\": \"application/json\",\n      \"title\": \"CGAZ ADM2 - Second-Level Administrative Boundaries\"\n    },\n    {\n      \"rel\": \"child\",\n      \"href\": \"./names/collection.json\",\n      \"type\": \"application/json\",\n      \"title\": \"CGAZ Name Search Index\"\n    }\n  ],\n  \"providers\": [\n    {\n      \"name\": \"geoBoundaries, William & Mary geoLab\",\n      \"description\": \"The geoBoundaries Global Database of Political Administrative Boundaries, built by the community and William & Mary geoLab. CGAZ composites country submissions against US Department of State international boundaries.\",\n      \"roles\": [\n        \"producer\",\n        \"licensor\"\n      ],\n      \"url\": \"https://www.geoboundaries.org/\"\n    },\n    {\n      \"name\": \"Portolan Mirrors\",\n      \"description\": \"Republishes the CGAZ release as cloud-native GeoParquet and PMTiles. Adds no boundaries and changes no geometry.\",\n      \"roles\": [\n        \"processor\",\n        \"host\"\n      ],\n      \"url\": \"http","encoding":"utf8","key":"portolan-mirrors/geoboundaries-catalog/catalog.json","name":"catalog.json","size":4163,"truncated":true},{"bytes":"# CGAZ ADM0 — Country Boundaries\n\nThe 218 country-level polygons of the geoBoundaries **Comprehensive Global\nAdministrative Zones** composite: 198 recognized countries and 20 disputed\nareas, as GeoParquet you can query in place and PMTiles you can draw.\n\nThis is a **mirror**. [geoBoundaries](https://www.geoboundaries.org/) is the\nproducer and the authoritative source.\n\n| | |\n|---|---|\n| Rows | 218 (198 `ADM0`, 20 `DISP`) |\n| GeoParquet | 104 MB, 218 row groups of 1 row, zstd, bbox covering |\n| PMTiles | 7.1 MB, zoom 0-6 |\n| CRS | CRS84 (EPSG:4326), as published |\n| Vertices | 9,807,040 as published, a mean of 44,986 per country (The upstream GeoPackage holds 10,101,960; this catalog publishes fewer because `ST_MakeValid` drops degenerate vertices from the geometry it repairs) |\n\n## What these boundaries are, and are not\n\nCGAZ outlines come from the **US Department of State Large Scale International\nBoundaries** file, dissolved to one polygon per ISO code. geoBoundaries\ndescribes the composite as \"clipped to international boundaries (US Department\nof State), with gaps filled between borders\", and says \"disputed areas are\nremoved and replaced with polygons following US Department of State\ndefinitions\".\n\nSo these are **not** the boundaries a country claims for itself. geoBoundaries'\nper-country releases do represent each nation \"as they would represent\nthemselves\", and two of those can overlap; these cannot. If sovereignty matters\nto your use, this is the wrong product.\n\nDependent territories are merged into their parent state, so Greenland is part\nof the Denmark polygon and has no row of its own. 218 rows is not a list of 218\nsovereign states.\n\nUnlike the ADM1 and ADM2 levels, **this level is not simplified** — the vertex\nreduction in the upstream build applies to the lower levels only. That is why\n218 country polygons weigh as much as all 49,349 municipalities. Use the\nPMTiles archive to draw outlines; the GeoParquet is for analysis.\n\n> **Not a byte-for-byte mirror.** This catalog repairs faults in the upstream\n> release and adds products geoBoundaries does not publish. Every deviation,\n> with how to reproduce it and how to get the vanilla behaviour, is in\n> [changes-from-upstream.md](../changes-from-upstream.md).\n\n## The display file\n\n`cgaz_adm0_display.parquet` holds **the same rows in the same order with the\nsame row group boundaries** as the data file, with geometry simplified and\nthen buffered outward by the same distance. Row *i* and row group *g* mean the\nsame unit in both, so one byte-range plan serves either. It is 7.5 MB against\nthe data file's 104 MB, and one country's outline costs about 11 KB there\nagainst 210 KB in the data file.\n\nUse it to draw an outline, and to pre-filter another dataset to this unit. The\noutward bias is deliberate: a filter shape that sits inside the true boundary\nsilently drops features along the edge, and over-fetching is cheap. Every\nfeature retains at least 99.95% of its original area, the worst case being\nKiribati at 99.9535%.\n\nDo **not** measure area from it, do not use it to decide which side of a\nborder a point falls on, and do not publish it as a boundary: neighbouring\nunits overlap in the margin. For any of that, read the data file. The full\nrule and the latitude caveat on a buffer measured in degrees are in the\n[catalog agent guide](../AGENTS.md).\n\n## Known issues\n\n- **No `shapeID` column.** ADM1 and ADM2 have one; CGAZ ADM0 does not publish\n  it. `shapeGroup` is the join key here, and it is unique.\n- **20 rows are disputed areas, not countries.** Filter on\n  `shapeType = 'ADM0'` for the 198 recognized ones.\n- **`shapeGroup` is not an ISO code for those 20 rows.** They carry `111`\n  through `129`, except Western Sahara which uses `ESH`.\n- **87 of the 218 geometries were not OGC-valid upstream, and are repaired\n  here.** The published file carries `ST_MakeValid` output for those 87 rows\n  and the other 131 byte for byte; 0 rows fail `ST_IsValid` now. At most\n  0.003 km² of area was introduced on any row.\n- **No validity dates.** geoBoundaries records a year represented per","encoding":"utf8","key":"portolan-mirrors/geoboundaries-catalog/adm0/README.md","name":"README.md","size":6701,"truncated":true}],"sourceObjectKey":"portolan-mirrors/geoboundaries-catalog/adm0/styles/adm1-units.json","starterCell":"with open('/mnt/archive/adm0/styles/adm1-units.json', 'rb') as fh:\n    head = fh.read(1024)\nprint(len(head), head[:64])","starterSamplePath":"adm0/styles/adm1-units.json","totalObjects":"4 top-level prefixes"},"summary":"Every boundary in the geoBoundaries Comprehensive Global Administrative Zones (CGAZ) composite, produced by William & Mary geoLab and the geoBoundaries community, as cloud-native…","title":"geoBoundaries CGAZ Global Administrative Boundaries ADM0, ADM1, ADM2 (GeoParquet, PMTiles, name index)","updatedAt":"2026-11-18T19:10:15.830Z","visibility":"public","websiteUrl":"https://www.geoboundaries.org/"}}