Methodology

How it works

We continuously ingest public methane-sensing satellite observations over monitored oil, gas and industrial facilities, attribute detected plumes to the most likely source using facility geometry and prevailing meteorology, quantify emission rate using accepted plume-based inverse methods, and reconcile measured mass against operator-declared inventories.

The output is delivered as an operational dashboard, event-scoped alerts, and audit-ready reports aligned to OGMP 2.0, EPA Subpart W, the EU Methane Regulation and CSRD ESRS E1 — all backed by an immutable evidence chain from raw granule to reported number.

Scientific basis

Detection uses TROPOMI XCH₄ column enhancements over trailing-window local baselines, following common practice in the peer-reviewed literature (Lauvaux et al. 2022, Science). Quantification uses plume-based inverse methods of the Integrated Mass Enhancement (IME) family (Varon et al. 2018, Atmospheric Measurement Techniques) coupled with ECMWF-class reanalysis winds. Uncertainty ranges are consistent with the blind-test results reported in Sherwin et al. 2024 (Nature).

Our specific thresholds, scoring weights, calibration and validation results are proprietary and available to customers and partners under NDA.

Standards alignment

• IPCC AR6 WG1 — fossil-methane GWP factors used for CO₂e conversion.
• OGMP 2.0 — Level 4 (source-level) · UNEP IMEO
• EPA GHGRP Subpart W — 2024 Waste Emissions Charge · US EPA
• EU Methane Regulation — Article 12 verification · European Commission
• CSRD ESRS E1-6 — Category 1 Scope 1 (MB) · EFRAG

Coverage

Vendor-neutral multi-sensor fusion is Offshore Analytics' operating principle: every plume seen by more than one sensor within a wind-aware tolerance window becomes a single fused observation with per-sensor confidence, so no single provider gates coverage or credibility. The numbers below are live from our production database.

Detection floor is the sustained emission rate the sensor can reliably detect under favourable conditions (peer-reviewed sources cited in the sensor registry). Median latency is the P50 elapsed time from satellite overpass to alert creation, over the trailing 90 days. Observation counts include all attributed detections in the window.

Independent verification

Every regulator-grade report Offshore Analytics generates is SHA-256 hashed, its manifest published to a public registry, and (best-effort) anchored in the Sigstore Rekor transparency log. A verifier who receives a report file can recompute the hash and open /verify/<sha256> to confirm — without contacting us — the framework, facility scope, reporting period, code version and Rekor log index recorded at generation time.

• Content hash — SHA-256 of the exact bytes the regulator receives (XML, JSON, or CSV). Any post-hoc edit breaks it.
• Builder version — pinned in compliance-export.ts; recorded on every manifest so an auditor can reproduce a historical export from the same code.
• Code git SHA — captured at build time; ties the manifest to a specific commit in the deployed pipeline.
• Rekor log index — public, append-only Merkle-tree position from the Sigstore transparency log. Anyone can view it on search.sigstore.dev.

False-positive & false-negative rates

We run a nightly adversarial suite against a global panel of non-emitter controls (water bodies, salt flats, wetlands, urban surfaces) and score ground-truth misses against published controlled-release campaigns. Any detection that lands inside a non-emitter control radius counts as a false positive; any ground-truth event with no matching detection within 25 km / 6 h counts as a miss. Target: FP < 2% per sensor.

Metrics load from the live database. The first monthly window populates after the adversarial suite runs against a live ingest.

Method versions are stamped on every row so an auditor can reproduce a historical window from the same code. FN rate is reported as "—" when we have no ground-truth events for that sensor in the window (missing data is not counted as a zero miss).

Data sources

Offshore Analytics is deliberately vendor-neutral: we integrate every public methane satellite we can access and reference the commercial ones our customers already license, so no single provider gates coverage.

Satellite methane sensors

LIVE means the source is ingested in production. REFERENCED means the source is in the pipeline schema and public research from it informs our comparisons, but automated ingest is not yet live — typically pending operator/partner access or paid tasking budget.

Meteorology

• ECMWF-class global reanalysis for near-surface wind speed and direction, used as the transport wind in quantification.

Inventories and asset registries

• Climate TRACE — global methane-emitting asset catalogue. Public / CC-BY.
• Operator-declared inventories — customer-uploaded, per-facility, per reporting window. Reconciled against measured mass.
• UNEP IMEO MARS — Methane Alert and Response System super-emitter notifications. Public.

Subprocessors

Cloud hosting (Cloudflare Workers), managed database and authentication (Supabase). No customer data is sent to satellite operators or reanalysis providers — those APIs are one-way ingests. See the Privacy Policy for the current list.