G-NAF — Geocoded National Address File
Australia's authoritative address file, used as the entry point to the land spine. Coverage, update cadence, licence and the queries the agent runs against it.
First data connectors are live
G-NAF, cadastral parcels, Geoscience Australia bushfire boundaries, SILO climate and DEA Coastlines are loaded as parquet on R2 and queryable by the agent.
The first batch of connectors is loaded and queryable. Each one is a pipeline that fetches the source release, harmonises it, and writes partitioned parquet to R2 where the sandbox can read it with DuckDB.
Live now:
- G-NAF — quarterly release, partitioned by state, ~15.3 million addresses.
- Cadastral parcels — parcel geometry for the states with open cadastre.
- Geoscience Australia bushfire boundaries — national historical fire extents.
- SILO — gridded daily climate observations from 1889, aggregated to the metrics we need.
- DEA Coastlines — annual shoreline positions since 1988.
What we learned loading them
Format variance is worse than volume. G-NAF ships as pipe-delimited text with a relational schema across a dozen tables. SILO is NetCDF. DEA Coastlines is vector. Cadastre varies by jurisdiction in both format and attribute naming, and the same conceptual field is called four different things across four states.
The harmonisation layer ended up being most of the work, and most of the value. A query that spans addresses, parcels and climate grids should not need to know that one of them came from a 1990s fixed-width export.
Next
NSW flood studies, CMIP6 downscaled projections, NGER emissions and the ACCU register. Flood is the one that will take longest, for reasons we wrote about here.
Related
See it run on your portfolio
Zenancy is in private preview with Group 2 reporters and their advisers.
Request access