GDAL/OGR for Remote Geospatial Data

GDAL/OGR is the foundational library for reading and writing geospatial vector and raster formats. Its virtual filesystem (VSI) layer makes it possible to read remote files over HTTP or S3 without downloading them first.

The VSI Virtual Filesystem: Why Prefixes Matter

GDAL was designed around local file paths. To extend it to remote storage, GDAL uses a virtual filesystem (VSI) layer that intercepts path strings and transparently handles the I/O. Any GDAL tool (ogrinfo, ogr2ogr, Python GDAL, and even DuckDB's bundled GDAL via ST_Read()) can read remote files simply by prepending the right prefix.

Without VSI prefixes, you would have to download the entire file before GDAL could inspect or convert it — which is impractical for large files and impossible for selective layer reads.

Core VSI Prefixes

/vsicurl/ vs /vsis3/ — When to Use Each

Use /vsicurl/ when:

• The file is publicly accessible via an HTTPS URL
• You are reading from any S3-compatible endpoint using its public HTTP URL (e.g., https://s3-west.nrp-nautilus.io/bucket/file.gdb)
• You do not have or need S3 credentials

Use /vsis3/ when:

• You have S3 credentials configured and the bucket requires authentication
• You need to write output directly to S3 (only /vsis3/ supports write)
• The file is in a private bucket, or you want efficient authenticated range requests via the S3 API

The key differences:

• /vsicurl/ is read-only; /vsis3/ supports both read and write
• /vsicurl/ requires a full HTTPS URL; /vsis3/ takes bucket/path (no protocol or hostname)
• /vsis3/ requires credentials and optionally a custom endpoint

Converting Between URL Styles

# Public HTTPS URL → vsicurl
https://s3-west.nrp-nautilus.io/bucket/path/data.gdb
→ /vsicurl/https://s3-west.nrp-nautilus.io/bucket/path/data.gdb

# s3:// URL → vsis3 (authenticated) or vsicurl (public endpoint)
s3://bucket/path/data.gdb
→ /vsis3/bucket/path/data.gdb                                        (authenticated S3)
→ /vsicurl/https://s3-west.nrp-nautilus.io/bucket/path/data.gdb     (public HTTP access)

S3 Credentials for /vsis3/

export AWS_ACCESS_KEY_ID=...
export AWS_SECRET_ACCESS_KEY=...
export AWS_S3_ENDPOINT=rook-ceph-rgw-nautiluss3.rook   # internal k8s endpoint
export AWS_HTTPS=NO
export AWS_VIRTUAL_HOSTING=FALSE

For public S3 endpoints, /vsicurl/ with the full HTTPS URL requires no configuration.

Choosing the Right Tool

Key principle: Use GDAL for format conversion and inspection; use DuckDB for analytics on Parquet. See the duckdb-spatial skill for DuckDB usage.

GDAL Parquet Driver Availability

GDAL is often not built with the Parquet (Arrow) driver. Check before assuming:

ogrinfo --formats | grep -i parquet
ogrinfo --formats | grep -i arrow

No output means GDAL cannot read or write Parquet. The ghcr.io/osgeo/gdal:ubuntu-full-latest Docker image includes the Parquet driver. Most system installs (apt install gdal-bin) do not.

When GDAL lacks Parquet support, use DuckDB's native read_parquet() for reading and DuckDB COPY ... TO 's3://...' for writing.

OGR: Inspecting Remote Data

List layers in a multi-layer source

ogrinfo /vsicurl/https://s3-west.nrp-nautilus.io/public-padus/raw/PADUS4_1Geodatabase.gdb

Get layer schema and feature count

# Summary only — essential for remote files (without -so, ogrinfo dumps every feature)
ogrinfo -so /vsicurl/https://example.com/data.gdb LayerName

# All layers summary
ogrinfo -so -al /vsicurl/https://example.com/data.gdb

Count features in a layer

ogrinfo -so /vsicurl/https://example.com/data.gdb LayerName | grep "Feature Count"

OGR: Format Conversion

GDB layer → GeoParquet

ogr2ogr -f Parquet output.parquet /vsicurl/https://example.com/data.gdb "LayerName"

GDB layer → GeoJSONSeq

ogr2ogr -f GeoJSONSeq output.geojsonl /vsicurl/https://example.com/data.gdb "LayerName" -progress

Write directly to S3

# Read from S3, write to S3 (requires credentials set via AWS_* env vars)
ogr2ogr -f Parquet /vsis3/bucket/output.parquet /vsis3/bucket/raw/source.gdb "LayerName" -progress

Stream to tippecanoe for PMTiles

ogr2ogr -f GeoJSONSeq /vsistdout/ /vsis3/bucket/data.gdb "LayerName" \
  | tippecanoe -o output.pmtiles -l layername \
    --drop-densest-as-needed --extend-zooms-if-still-dropping --force

Geometry Edge Cases

Curved geometries (MULTISURFACE, CURVEPOLYGON)

Some GDB files contain curved geometry types that downstream tools cannot handle. Linearize during conversion:

ogr2ogr -f GPKG \
  -nlt CONVERT_TO_LINEAR \
  -nlt PROMOTE_TO_MULTI \
  output.gpkg input.gdb "LayerName"

• -nlt CONVERT_TO_LINEAR — converts CircularString, CompoundCurve, CurvePolygon, MultiSurface to linear equivalents
• -nlt PROMOTE_TO_MULTI — ensures consistent multi-geometry types

For large files with mixed curved/linear geometry, a two-step pipeline (GDB → GPKG → Parquet) is more reliable than a single pass.

FID issues

Some formats (GDB especially) include an FID column that can cause Arrow type mapping issues:

ogr2ogr -f Parquet output.parquet input.gdb "LayerName" -unsetFid

Common Pitfalls

1. Forgetting -so with remote ogrinfo — Without summary-only mode, ogrinfo tries to dump every feature over the network.
2. Assuming GDAL has Parquet support — Check ogrinfo --formats | grep -i parquet. Most system installs lack it; use DuckDB instead.
3. Using /vsicurl/ when you need to write — /vsicurl/ is read-only. Use /vsis3/ for output to S3.
4. Passing s3:// URLs to GDAL tools — GDAL does not understand s3:// protocol strings. Convert to /vsis3/bucket/path or /vsicurl/https://endpoint/bucket/path.
5. Curved geometries breaking downstream tools — Always check with ogrinfo and linearize GDB sources if MULTISURFACE or CURVEPOLYGON types are present.