Method & data governance

Data sources

• Kp (planetary K-index): NOAA SWPC 1-minute feed and 3-hour forecasts, fetched hourly
• Dst (storm-time ring current): Kyoto World Data Center, hourly; final→provisional→realtime by release status
• Solar flares: NOAA GOES X-ray flux, daily peak and class
• Schumann resonance: Tomsk State University spectrograms (frequency/amplitude/quality), archived hourly

Governance rules (mandatory)

• Missing = missing: Kyoto sentinel values (3999/9999) become null with a quality_flag at ingest — no interpolation, no forward-fill, API passes null through
• Range validation: Kp∈[0,9], Dst∈[-2500,2500]; violations fail the gate
• Freshness: Kp lag >6h or SR >26h fails the quality gate
• Provenance: every record carries a source field; Dst labeled final/provisional/realtime
• Immutable archive: JSONL is append-only

Derived quantities (public definitions)

• G-scale (G0-G5): mapped from 3-hourly Kp peaks per the NOAA scale (Kp5→G1 … Kp9→G4/G5)
• Active day: daily Kp_max ≥ 5 (the grouping used in diary correlation stats)
• dst_min_daily: daily minimum of hourly Dst

Personal correlation statistics

Each diary entry automatically carries that day's Kp/Dst snapshot. The Mirror view groups your metric means by "active vs quiet days" and shows sample sizes; any group with n<5 is explicitly labeled "insufficient to conclude". We compute correlation; we never claim causation.