Spatial Data Infrastructure Portal Seabandl

Independent project developed as a proof-of-concept for a regional GIS working group in West Java, exploring how a shared metadata standard could reduce duplicate data collection across agencies like DBMPR and BAPPEDA.

Regional agencies store spatial data across disconnected formats (shapefiles, spreadsheets, ad-hoc databases) with no shared schema or metadata standard, making cross-department data reuse slow and error-prone.

- Build a single portal for dataset discovery and metadata search - Enforce schema and CRS validation before publishing - Reduce duplicate data collection across agencies

Started by auditing existing dataset formats from three sample agencies, then designed a normalized PostGIS schema with metadata fields aligned to ISO 19115. Built a validation pipeline that checks geometry, CRS, and attribute completeness on upload before allowing publication.

Iterative development with weekly validation testing against real sample datasets. Metadata structure benchmarked against ISO 19115 and referenced against SEA-BANDL's data modeling approach.

Sample administrative boundary, land use, and infrastructure datasets from open government portals, plus synthetic test datasets for edge-case validation.

React frontend communicating with a Node.js/Express API, backed by PostGIS for spatial storage and GeoServer for WMS/WFS tile serving. Validation runs as a background job queue before datasets are marked publishable.

Handling inconsistent CRS declarations in legacy shapefiles required building a fallback detection step using geometry bounds heuristics rather than trusting embedded metadata.

A working prototype validating and serving 3 sample agency datasets with automated CRS correction and metadata search.

Demonstrated a reusable pattern for reducing manual QA time on spatial dataset onboarding, cutting validation checks from a manual multi-hour process to under 5 minutes per dataset.

Real-world spatial data is messier than academic datasets — metadata standards only help if validation is enforced at ingestion, not left to documentation alone.

Prototype only tested against a limited set of vector formats; raster dataset support and authentication/role management were not implemented.

Add raster dataset support, agency-level access control, and an API for third-party GIS tools to query the catalog directly.