Methodology

How raw spots become propagation intelligence. This section describes the pipeline from ingested observations to the signatures and statistics published on this site.

The short version: 13+ billion raw spots go in, 156+ million quality-filtered, solar-joined signatures come out. Every number on this site traces back to a specific query against that corpus.

---

The Pipeline

Raw Spots (bronze)
    ↓
  Quality Filter (balloon removal, ground-wave filter, AGC outliers)
    ↓
  Solar Join (SFI, Kp, SSN at 3-hour resolution)
    ↓
  Grid Resolution (callsign → Maidenhead grid via callsign_grid)
    ↓
Signatures (silver / gold)
    ↓
  Aggregation (by band, path, time window)
    ↓
Published Reports (this site)

Each step is deterministic and reproducible from the raw data. Nothing is estimated or interpolated except the Maidenhead grid-to-latitude/longitude conversion for distance calculations.

---

Pages

• Signatures — What a signature is: aggregated float4 vectors combining geography, time, frequency, and solar context. How 10.8B raw spots become 93.3M signatures.
• Data Quality — What gets filtered and why: balloon callsigns, ground-wave spots, RBN AGC outliers, grid normalization.

---

Bronze → Silver → Gold

Layer	Table(s)	Rows	Content
Bronze	wspr.bronze, rbn.bronze, contest.bronze, pskr.bronze	13.18B+	Raw ingested spots, minimal transformation
Silver	wspr.silver	4.4B	Float4 embeddings from CUDA signature engine
Gold	wspr.signatures_v2_terrestrial, rbn.signatures, contest.signatures	156.4M	Quality-filtered, solar-joined, ground-wave excluded

Solar data: solar.bronze (GFZ Potsdam, 2000–present, ~1-day lag) for historical training; wspr.live_conditions (NOAA SWPC, 15-min updates) for current-conditions reporting.