Download location
GitHub Releases publishes zstd-compressed aircraft_diffs_YYYY-MM-DD.parquet files under monthly YYYY-MM releases. No authentication or quota applies. Release notes provide row counts and SHA-256 checksums.
Quick start
import duckdb
url = "https://github.com/Sky-Power-Services/adsbiq-data/releases/download/2026-06/aircraft_diffs_2026-06-29.parquet"
duckdb.sql(f"SELECT hex, count(*) AS changes FROM read_parquet('{url}') GROUP BY hex LIMIT 20").show()Schema and reconstruction
Rows are append-only state diffs keyed by hex and ordered by ts. Rows with is_snapshot=true contain full state; other rows populate only changed fields. Partition by aircraft, order by timestamp, and forward-fill to reconstruct state.
Core columns include ts, hex, flight, lat, lon, alt_baro, gs, track, squawk, type plus the published readsb field set. Added nullable columns are backward-compatible evolution.
Efficiency, security, and compatibility
Use DuckDB projection and predicate pushdown or Polars scan_parquet(); do not load full days unnecessarily. Pin checksums for reproducible research. Compatible with DuckDB, Polars, pandas/PyArrow, and Spark. Dataset format version: 1. License: ODbL-1.0; attribute the “ADSB IQ feeder network.”
Troubleshooting
- 404: confirm monthly tag and UTC filename.
- Unexpected nulls: ordered forward-fill after the latest snapshot.
- Memory pressure: project and filter in the scan.
See open-data release details.