Portolan NL — Cloud-Native Dutch Geodata
A proof-of-concept cloud-native catalog of Dutch government geodata, inspired by PDOK (Publieke Dienstverlening Op de Kaart). The same commitment to open, standards-based geodata disclosure, built on modern cloud-native formats: GeoParquet for analytics, PMTiles for visualization, and STAC for metadata.
Published on Source Cooperative.
Data Providers
Organized by the Dutch government institution that produces the data:
Collections
Getting Started
Use AI to explore the data. Point Claude Code or Gemini CLI at any collection's llms.txt and ask it to query the data, build interactive maps, or generate charts. Every collection includes an llms.txt with field descriptions, query examples, and usage context.
Browse and download. Navigate to any collection above to find GeoParquet files for analytics and PMTiles for instant map visualization. All files are on Source Cooperative and can be accessed directly via HTTP — no account or API key required.
Query directly with DuckDB. All GeoParquet files work with remote HTTP access:
Formats
- GeoParquet — Analytics-ready columnar format. Every collection has a
.parquetfile with spatial indexing (Hilbert sort, bbox covering columns). Query remotely with DuckDB, Python (GeoPandas), R, or any Parquet reader. - PMTiles — Single-file vector tile archives for instant map visualization. No tile server needed — works directly from object storage via HTTP range requests.
- STAC — SpatioTemporal Asset Catalog metadata in
collection.jsonfiles. Machine-readable, standardized, and browsable.
Coordinate Systems
Most datasets use EPSG:28992 (RD New / Amersfoort), the Dutch national coordinate system — coordinates are in meters. CBS INSPIRE datasets use EPSG:3035 (ETRS89-LAEA), the pan-European standard. The INSPIRE Buildings collection uses EPSG:4258 (ETRS89). All PMTiles are reprojected to WGS84 for web map display.
Complementing PDOK
This catalog demonstrates how PDOK's existing geodata services could be extended with cloud-native formats. Instead of running WFS/WMS servers, data can be served as GeoParquet for scalable analytics and PMTiles for instant map visualization — both work directly from object storage via HTTP range requests, with no server infrastructure. Adding STAC metadata and llms.txt files makes the data discoverable and queryable by AI agents, opening geodata to a much broader audience.
AI / Agent Friendly
Every collection includes an llms.txt file with field descriptions, query examples, and usage context — designed for AI agents to understand and work with the data effectively. Built with Portolan, a framework for cloud-native geodata infrastructure.
License
Most data is published under CC0 1.0 (public domain). Some collections use CC-BY-4.0. See individual collection metadata for details.
Status
This is a proof-of-concept. It is not affiliated with or endorsed by PDOK or any Dutch government organization. Data is sourced from publicly available PDOK services and ArcGIS Feature Services and converted to cloud-native formats.
Contact
Chris Holmes — cholmes@9eo.org
Published with Portolan