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: 14+ 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.

Bronze → Gold¶

Layer	Table(s)	Rows	Content
Bronze	`wspr.bronze`, `pskr.bronze`, `rbn.bronze`, `contest.bronze`	22.4B	Raw ingested spots, minimal transformation
Gold	`wspr.signatures_v2_terrestrial`, `rbn.signatures`, `contest.signatures`, `pskr.signatures`	175.1M	Quality-filtered, solar-joined, ground-wave excluded

There are two layers, not three. A silver layer was documented for a long time and never carried the pipeline: wspr.silver was found holding zero rows and retired on 2026-09-22. Every signatures table above is built from bronze directly.

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.

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 (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 12.7B raw spots become 93.6M signatures.
Data Quality — What gets filtered and why: balloon callsigns, ground-wave spots, RBN AGC outliers, grid normalization.