Paper → data-source inventory (repoint-to-dex campaign)
Built 2026-07-07. Read-only sweep of the 19 shipped papers (web/src/pages/articles/*.astro)
mapping each to the data sources it draws from, tagging each source DEXED (backs an existing dex
kind) or UNDEXED. Purpose: sequence the "papers source their data from a dex" standardization.
Three buckets, per the plan:
- Already dexed — repoint is near-pure plumbing.
- Undexed but dexable — dex the source first, then repoint.
- Undexed and not dexable — model output or not dex-shaped; needs a substitute-or-accept call.
Fully-dexed papers (bucket 1 — easy repoint pilots, no new dex needed)
Every source already backs a dex:
cascadia-swarm-apr2026— usgs_earthquake[eq]reno-earthquake-swarm— usgs_earthquake[eq](reuses cascadia workspace)kermadec-doublet-apr2026— usgs_earthquake[eq]sanriku-m66-may2026— usgs_earthquake[eq]lunar-syzygy-seismicity— usgs_earthquake[eq](lunar phase = offline ephemeris, not a source)earthquake-lights-104-years— usgs_earthquake[eq]+ nuforc[nuforc]tornado-alley-east— spc_tornado_history[tor]fireball-surge-2026— CNEOS fireballs[neo]
8 of 19. All seismic, plus tornado-alley and the CNEOS fireball paper.
Undexed-source rollup (bucket 2/3 — sorted by leverage)
The unit of work is the SOURCE. One dex can repoint many papers.
- WSPR / wsprnet (wspr.live) — 6 papers (corrected 2026-07-07 from 8;
sky-anomaliesuses NUFORC not WSPR, andsolar-watch-april-2026only references a WSPR number from another paper, it analyzes none): wspr-aircraft-detection, wspr-solar-cycle-bug, wspr-tornado-precursor, wspr-tornado-v2-historical, wspr-tornado-v3-general, wspr-tornado-v4-decoupled. MEASURED. Biggest single unlock. SHAPE SETTLED 2026-07-07 (Mike): two spot structures, no station Locationdex (every spot self-carries tx/rx lat/lon in the raw). (1)wspr_spot_censusmonth-grain Yeardex — BUILT (data/year_storehouse/, 221 month slots over the 200 GB raw on /mnt/ursa; carry-both: 5 MB day×band aggregate as working surface + raw partition as archive pointer). Serves wspr-solar-cycle-bug. (2)wspr_event_windowevent-window Eventdex — BUILT (data/event_storehouse/wspr_event_window/, spine-parquet; 65 slots = 17 outbreaks + 44 controls- 3 aircraft corridors + 1 storm-season, deduped across papers; slot = a paper's event window, carve-both: cell×15min×band SNR aggregate working surface + raw partition pointer; 61 carved from the 2011-2025 raw, 4 in the un-backfilled 2026 tail pending WS-4). Serves the 5 event papers. Both carved from the same /mnt/ursa raw; no fetching.
- DSCOVR solar wind (Bz/speed) — 4 papers: bz-dst-lead-time, starlink-thermosphere-storm-pilot, skinwalker-ranch-seven-layers, solar-watch-april-2026. MEASURED. Spine UNSURE (single L1 spacecraft continuous feed; possibly a one-monitor spine).
- SILSO sunspots — 3 papers: starlink-thermosphere-storm-pilot, wspr-solar-cycle-bug, skinwalker-ranch-seven-layers. MEASURED, NO-SPINE (single global daily index).
- GOES X-ray flux — 3 papers: starlink-thermosphere-storm-pilot, skinwalker-ranch-seven-layers, solar-watch-april-2026. MEASURED. Spine UNSURE (satellite instrument feed).
- NOAA space-weather Kp — 2 papers: sky-anomalies, skinwalker-ranch-seven-layers. MEASURED, UNSURE.
- OpenSky + ADS-B (adsb.fi) aircraft — 2 papers: wspr-aircraft-detection, skinwalker-ranch-seven-layers. MEASURED, SPINE (aircraft/flight roster).
- Safecast radiation — 1 paper: skinwalker-ranch-seven-layers. MEASURED, SPINE (device roster).
- AMS fireballs — 0 shipped papers (corrected 2026-07-07). Earlier draft mislabeled
skinwalker's bolide layer as AMS; skinwalker actually uses NASA CNEOS Fireballs (objective
government-sensor bolides), which is already in the
entrydex (cneos_fireballs). AMS is an inactive stub (0 rows, needs an API key, issue #94) and is used by no shipped paper. Not a campaign item; if ever wanted it's a new-source fetch, not a dex build. - SPC convective outlook — 1 paper: skinwalker-ranch-seven-layers. MODEL (forecast) → bucket 3.
- open_meteo + open_meteo_aqi — 1 paper: skinwalker-ranch-seven-layers. MODEL, held-out → bucket 3.
- era5_cloud — 1 paper: skinwalker-ranch-seven-layers. MODEL reanalysis, held-out → bucket 3.
Notes
- Bucket 3 is nearly empty for shipped papers. The only model sources in use (SPC convective
outlook, open_meteo, era5_cloud) all live in one paper —
skinwalker-ranch-seven-layers— a kitchen-sink null baseline (~28 sources) where they are context layers, already held out of the measured-reality tests. So almost every shipped paper is either already dexed or dexable. - skinwalker is the outlier. It alone carries most of the undexed sources AND all three model sources. It is a baseline, not a single-source paper; repointing it is the hardest and least valuable move. Deprioritize / handle last.
- The solar/space-weather undexed sources (DSCOVR solar wind, SILSO, GOES X-ray, Kp) co-occur in the same ~4 solar papers — a natural second batch after WSPR.
- 3 compound landing articles have no single dedicated workspace (
sky-anomalies,solar-watch-april-2026,reno-earthquake-swarm); inventoried from cited sources. nasa_neoslug matched by identity to theneo/entrydex (CNEOS CAD + fireballs).