How it works · Offshore Analytics

Offshore Analytics is a four-stage pipeline. Every stage is auditable in code — the cadences and thresholds shown below are the actual values in production, not aspirational targets.

  Sentinel-5P ─┐
  GHGSat*      │           ┌─ Attribution ─┐        ┌─ Alerts (email / webhook)
  MethaneSAT*  ├─ Ingest ──┤   25 km / ±6h ├── IME ─┤─ Reports (OGMP · EPA · CSRD)
  EMIT*        │           └─ conf ≥ 0.60  ┘        └─ Audit log + evidence
  Carbon Map.* ─┘
     * referenced; awaiting operator / partner feed

Ingest

cadence · 30 min

A scheduled worker pages the Copernicus Data Space Ecosystem STAC API for new Sentinel-5P L2 CH₄ granules, filters to the bounding boxes of monitored facilities, and stores retrieval metadata in the detections table. Higher-resolution sensors (MethaneSAT, GHGSat, EMIT, Carbon Mapper) attach to the same schema as their feeds come online.

Attribute

cadence · on ingest

Every detection is scored against candidate facilities within a 25 km radius and ±6 h overpass window. Confidence scoring combines distance, plume centroid vs. wind back-trajectory, and asset type. Detections above a 0.60 confidence threshold are attributed; below it they remain unattributed and visible for analyst review.

Quantify

cadence · 60 min

Emission rate is estimated using the Integrated Mass Enhancement (IME) method with effective wind scaling U_eff = 1.4 · U10 (Varon et al. 2018). Wind speed and direction at 10 m are pulled from ERA5 reanalysis (via Open-Meteo archive) on the 25 km grid at hourly cadence, cached per facility-hour. Reported 1-σ uncertainty is ±50% — an honest reflection of the IME method, not a marketing number.

Deliver

cadence · real-time

Attributed and quantified events fan out to (a) the operational dashboard with severity buckets, (b) email / webhook alerts scoped by facility and threshold, (c) the reports engine which builds OGMP 2.0, EPA Subpart W, EU Methane Regulation and CSRD ESRS E1 outputs from the same evidence chain, and (d) the audit log with per-event provenance.

Deeper reading

Methodology — IME quantification, attribution scoring, GWP conversion, uncertainty.
Data sources — every satellite and dataset we ingest or reference.
Validated case studies — end-to-end pipeline against public, verified super-emitter events.
Live pipeline status — last ingest, last quantification, 24 h detection counts.