LIST open data, Tasmania - cloud-native mirror

GeoParquet and Cloud Optimized GeoTIFF copies of the Land Tasmania LIST open data bulk downloads (https://listdata.thelist.tas.gov.au/opendata/data/), with the original zips alongside and a manifest describing every file.

Data: (c) State of Tasmania, Creative Commons Attribution 3.0 Australia. Code that built this: https://github.com/hypertidy/thelist-opendata (MIT).

Attribution

Land Tasmania asks for this form, and it applies to everything here whether you use the GeoParquet, the COGs or the original zips:

"<dataset name>" from theLIST © State of Tasmania

for example "LIST Parcels" from theLIST © State of Tasmania. Land Tasmania does not endorse this mirror or any use of it. Some datasets are published through the LIST on behalf of other custodians (TasWater, TasNetworks, EPA Tasmania, Natural Resources and Environment Tasmania and others); the readme.txt inside each zip in raw/ names the custodian and any additional conditions, and the .shp.xml beside it is the ESRI metadata record.

What is here

The source is a plain directory of about 1700 zips, nearly all holding one layer three ways (FileGDB, shapefile, MapInfo TAB). Most products are split by the 29 local government areas (LGAs); some are statewide; a few are both. This mirror converts each layer once, from the FileGDB where there is one, and keeps the zip.

Sizes as of the first release: 13 GB GeoParquet, 7 GB COG, 183 GB raw.

Reading it

Everything is readable in place. The GeoParquet is GeoParquet 1.1 with a covering bbox column and rows sorted by bbox, so spatial predicates prune row groups.

DuckDB, one LGA-split product as a single table:

GDAL / ogr:

R:

Python:

Set AWS_NO_SIGN_REQUEST=YES for anonymous S3 access with GDAL.

What conversion did and did not do

• Field names and types are the FileGDB's (Int16, DateTime and so on), not the shapefile's truncated names. Where a zip has no usable FileGDB the shapefile was used; converted.csv records which.
• Geometries are promoted to MULTI types. FileGDB curves (CurvePolygon, MultiSurface) are linearised, since GeoParquet has no curve types. M ordinates are dropped (one layer); Z is kept.
• The CRS is the source CRS, EPSG:28355 (GDA94 / MGA zone 55) nearly everywhere. Where the source carried an unidentified ESRI WKT naming MGA55, EPSG:28355 was assigned; nothing was reprojected.
• Coded-value field domains (code -> description lookups) are not carried into Parquet. The lookups are in the FileGDBs in raw/.
• Raster FileGDBs (the NCH soil and enterprise-suitability grids) come through without their colour tables, but their attribute tables are in parquet/. Byte class grids have nearest-neighbour overviews.
• The LGA boundary that every per-LGA zip ships (municipality_<lga>) is converted once, under geoparquet/municipality/.
• Features that straddle LGA boundaries appear in both LGAs' files. The _union.vrt is a union, not a dissolve; deduplicating is left to the reader (for LIST products LIST_GUID or UFI are the stable keys).
• Nothing was cleaned, clipped, simplified or edited.

Known issues in the source, as found

• A dozen zips carry an empty FileGDB (system tables only) beside a full shapefile; the shapefile was used.
• LIST_LAND_USE_2013_BRS_GLAMORGAN_SPRING_BAY: the FileGDB has 0 features, the shapefile 9900.
• NCH_ES_WHEAT_STATEWIDE.zip contains a barley scenario FileGDB.
• About 40 of 2700 FileGDB/shapefile pairs differ by a few features.

Details and the exact files are in manifest/layers.csv and in the code repository's README.

Provenance and updates

Every file traces to a source URL, its size, mtime and HTTP etag at the time of conversion (manifest/index.csv, manifest/converted.csv). A re-run of the pipeline against the live listing redoes only what changed upstream. First release: crawled 2026-09-10, converted 2026-09-17.