djgersh/dragtrack-ops-data

DragTrack Ops Data

Snapshot of the DRAGTRACK_ROOT data tree produced by the DragTrack thermospheric density pipeline.

Snapshot date: 2026-08-20

Reference

Gershman (2026), NASA-TM/20260006171 — Enabling Near-Real-Time Thermospheric Density Monitoring.

Contents

Per-NORAD tree (bundled as tarballs — 57k+ files each; extract after download):

Path	Extract to	Content
tles_raw.tar	<root>/tles/raw/	Per-NORAD raw element rows from Space-Track (v6 schema)
tles_daily.tar	<root>/tles/daily/	Per-NORAD daily-averaged rows with Picone drag acceleration
tles_meta.tar	<root>/tles/meta/	Per-NORAD ingest state (last epoch, last processed)

Solver + catalog + forcing (individual files):

Path	Content
manifest.parquet	Fleet-wide state (one row per NORAD)
catalog/ref_catalog_lowalt_{YYYY}.parquet	27-day rolling vdot-vs-forcing fits
fleet/year=YYYY/part.parquet	Per-year fleet median-element pre-filter
solver/v1/joint_trf_2014_2026_v1.pkl	Daily joint-TRF solver output (per-day a_d + Tinf, per-sat B*)
solver/subdaily/subdaily_2014_2026_6h_v1.pkl	6h subdaily refinement
forcing/f107.p, omni_sw.p, softG.p, euvs.p	Latest forcing pickles (F10.7, solar wind, geomagnetic, EUVS)
forcing/goes19_euvs_avg1d.nc	GOES-19 EUVS archive (NetCDF)

Not included

• tles/tle{YYYY}.txt — raw Space-Track dumps (28 GB, redownload from space-track.org with your own credentials, or dragtrack-hourly)
• solver/dict/, solver/obs_load_cache/ — solver scratch (regenerated)
• reports/ — product JSONs and plots (regenerated hourly by the pipeline)
• j70_lut/ — J70 density lookup table (bundled in the pipeline repo at assets/j70/j70_density_lookup_tx.pkl)
• Historical snapshots (*_TIMESTAMPZ.p, *_TIMESTAMPZ.pkl) — only the latest of each pkl is included

Usage

Download + extract:

export DRAGTRACK_ROOT=/srv/dragtrack/data
mkdir -p $DRAGTRACK_ROOT

# 1. Download everything
huggingface-cli download djgersh/dragtrack-ops-data --repo-type dataset \
    --local-dir $DRAGTRACK_ROOT

# 2. Extract the per-NORAD tarballs
cd $DRAGTRACK_ROOT
mkdir -p tles/raw tles/daily tles/meta
tar -xf tles_raw.tar   -C tles/raw
tar -xf tles_daily.tar -C tles/daily
tar -xf tles_meta.tar  -C tles/meta
rm tles_raw.tar tles_daily.tar tles_meta.tar   # optional — save 32 GB

Then set up the pipeline:

git clone https://github.com/dangershman/dragtrack-solver dragtrack_ops
cd dragtrack_ops
pip install -e '.[dev]'
export SPACETRACK_USER=<your-user>
export SPACETRACK_PASS=<your-pass>

# Live-edge update from this snapshot
dragtrack-hourly                        # pulls fresh TLEs from Space-Track
dragtrack-solve --stage forcing         # refresh forcing pickles
dragtrack-solve --stage rho             # emit product

See DEPLOYMENT_GUIDE.md in the pipeline repo for the full stage list and scheduler wiring.

Provenance

Solver pkls carry embedded provenance (git commit, source hash, calibration constant C_CAL, per-year catalog metadata). Local paths in provenance JSONs have been scrubbed to <DRAGTRACK_ROOT>.

Every forcing pickle includes source URL, fetch timestamp, and processor git commit in its atomic envelope.

Contact

Daniel Gershman — daniel.j.gershman@nasa.gov