Kaliabe Flare Ledger

Methodology & confidence

The Flare Ledger is built on the principle that honest uncertainty is a feature, not a weakness. Every verdict states what was detected, the floor below which this satellite can no longer resolve a flare, and a confidence score. We never present a derived estimate as a regulatory finding.

Independence note. The Kaliabe Flare Ledger is an independent, open-data project. It is not affiliated with, endorsed by, or reporting for the Nigerian Upstream Petroleum Regulatory Commission (NUPRC), NNPC, or any operator. Penalty figures are derived estimates from the published regulatory basis and are not official assessments.

Method version KFL-M 1.3 (2026-10-17). New in 1.3: population around every site (GRID3), power-gap estimates, a live NASA FIRMS thermal check with an under-report flag, and ACLED security context — see §7a. Changes in 1.2 from 1.1: penalty basis corrected to the Flare Gas (Prevention of Waste and Pollution) Regulations, 2018 (S.I. No. 9 of 2018) with both Reg. 13 tiers; the GFMR data vintage is labelled; emissions ratios updated to the June 2026 tracker (and unburned methane now reported as CH₄ mass); the Niger Delta flag uses the nine-state boundary instead of a rectangle; every threshold is named, explicit and listed in §8.

1. Detection layer — NOAA VIIRS Nightfire

Flares are detected by NOAA VIIRS Nightfire (VNF), run by the Earth Observation Group at the Colorado School of Mines / Payne Institute. VIIRS sees infrared heat at night and is sensitive only to hot emitters. Gas flares burn hotter than almost any other terrestrial source, so readings with temperatures above ~1,400 K are treated as flare candidates (biomass burning and power plants burn cooler).

Detection floor. VIIRS resolves flare heat, not gas molecules directly. Small or intermittent flares fall below the instrument's resolution or are obscured by cloud — the Niger Delta is cloudy for much of the year. A site with no detection in a given year is therefore labelled unverified / below floor, not "not flaring".

2. Volume estimation — World Bank GFMR calibration

Radiant heat is converted to a gas-flow volume using the calibration published by the World Bank's Global Flaring and Methane Reduction (GFMR) partnership, which anchors satellite radiance to country-level gas-flaring statistics. The volumes on these pages are GFMR's per-flare-site estimates; the underlying method is documented by the World Bank (and builds on the VIIRS Nightfire product and Schuit et al. 2023 for methane hot-spot detection principles).

Data vintage. The seed is the Flare Volume Estimates by Individual Flare Location 2012–2025 workbook released with the World Bank's Global Gas Flaring Tracker of 23 June 2026. That release uses the new GFMR / Payne Institute methodology: three VIIRS satellites (Suomi NPP, NOAA-20, NOAA-21) instead of one, multi-year flare location tracking and a revised volume calibration. The whole 2012–2025 series was restated, so figures will not match pre-2026 editions. We checked this: the workbook's global 2025 total is 166.6 bcm, matching the tracker's 167 bcm, and its 2024 total of 156.9 bcm matches the tracker's restated 157 bcm (the 2025 edition said 151). The vintage is stamped on every API response (meta.vintage) and every brief.

3. Attribution to operators

Each flare site is assigned to a presumed operating company by geometric / radiative association with field boundaries and operator records in the GFMR metadata. Where no reliable operator is available a site is marked Unattributed. Attribution is indicative, not a legal determination.

4. Verdicts and confidence

VerdictMeaningBase confidence
Verified flareA positive flare volume was detected in the latest VIIRS/GFMR record.0.55–0.98
UnverifiedNo volume in the latest record — below detection floor or cloud-obscured.0.4

Confidence is raised by data coverage — the share of years in the record with a positive detection (zero-volume years count as non-detections, not coverage) — and by a named operator for the site; it is capped at 0.98 because VIIRS volume↔radiance calibration carries intrinsic uncertainty from gas composition and combustion efficiency. Correlation is not perfect; treat every verdict as a direction, not an exact quantity.

Trend

Each site's trend compares the mean flared volume of its most recent three years with the mean of up to five years before them (with fewer than four years of record, the last year is compared with the first). Changes within ±5% are stable; beyond that, rising or falling. A site with no baseline volume but recent detections is new (no percentage is defined); a site with no detections in either window is inactive.

Significance. A direction is only claimed when the change is larger than detection noise: the window means must differ by at least 0.5 million m³ a year, and each window must have at least two detected years (or every year, if the window is shorter). Otherwise the trend is labelled not significant with the reason, not given a direction. Many sites that look as if they are “falling” have simply dropped below the detection floor, and below the floor is not the same as not flaring.

5. Penalty basis — Flare Gas Regulations 2018, two tiers

The governing instrument is the Flare Gas (Prevention of Waste and Pollution) Regulations, 2018: S.I. No. 9 of 2018, Federal Republic of Nigeria Official Gazette No. 88, Vol. 105, 9 July 2018, made on 5 July 2018 under s.9 of the Petroleum Act and s.5 of the Associated Gas Reinjection Act. (Earlier versions of this page cited “NUPRC 2022 Regulations”. That was wrong.)

Regulation 13 sets a tiered flare payment on the volume of gas flared, routine or non-routine, per 28.317 standard cubic metres (1,000 scf):

TierApplies where the OML or marginal field producesRateLedger use
Reg. 13(1)10,000 barrels of oil per day or moreUS$2.00 per 1,000 scfUpper bound
Reg. 13(2)less than 10,000 barrels of oil per dayUS$0.50 per 1,000 scfLower bound

Assumption, stated plainly: we do not hold production rates, and the regulation sets $2.00 (≥10,000 bbl/d) or $0.50 (<10,000 bbl/d). We therefore cannot tell which tier applies to a site, so every penalty is shown at both tiers. We never publish a single figure, and we never use the midpoint. The tier depends on production in the whole lease or field, not on one flare.

Penalty (US$) = volume (m³) ÷ 28.317 × rate. Example: Oso flared 308.37 million m³ in 2025, which gives US$21.78 m at the $2.00 tier and US$5.44 m at the $0.50 tier.

Exemptions in the Regulations. We cannot see these from satellite data, so we do not deduct them, and a real liability may be lower:

Not modelled: the Reg. 19 additional payment of US$2.50 per 1,000 scf for specified non-compliance. Cumulative penalties for 2012–2025 are notional, because the Regulations were made in July 2018. ₦ figures use an indicative FX snapshot. Actual assessments are made by NUPRC and are not reproduced here.

6. Unburned methane & CO₂e

Flaring converts gas to CO₂ but also lets some methane escape unburned, and methane is a far stronger greenhouse gas. We apply the World Bank Global Gas Flaring Tracker, June 2026 ratios (data year 2025): 167 bcm flared ≈ 429 MtCO₂e, which includes 50 MtCO₂e of unburned methane (1.8 Mt CH₄ by mass). Per bcm that is 2.569 MtCO₂e and 0.0108 Mt CH₄. We show CH₄ as mass. Earlier versions labelled a CO₂e quantity “t CH₄” and used the 2025 tracker's ratios (151 bcm ≈ 389 MtCO₂e incl. 46 MtCO₂e methane); both are superseded. These are global average ratios, so treat them as order-of-magnitude estimates, not site-level combustion modelling.

7. Decision-layer derivations method KFL-M 1.2 · 2026-10-17

These are Kaliabe derivations from the GFMR series. They add framing, not data.

The full method, with every constant and its source, is kept in the repository at docs/METHODOLOGY.md. Every decision brief is stamped with this method version and the data vintage.

7a. Opportunity and live enrichment method KFL-M 1.3 · 2026-10-17

These turn a flare record into a development brief: the flared gas as fuel, the people nearby as demand, and a live and a security check as current ground truth. Every figure here is an estimate on the assumptions shown.

8. Niger Delta classification

“Niger Delta” means the union of the nine Niger Delta states: Abia, Akwa Ibom, Bayelsa, Cross River, Delta, Edo, Imo, Ondo and Rivers. The boundaries come from the geoBoundaries ADM1 layer this site already ships, and each site is classified by point-in-polygon (scripts/classify_niger_delta.py). This replaces the old study rectangle (4.5–8.5° E, 4.5–7.0° N), which misclassified offshore fields such as Oso. There are three classes:

The web layer is simplified, so sites within a few hundred metres of the coastline can fall on either side of it. GFMR's own ONSHORE/OFFSHORE field is shown separately and is not overwritten.

9. Thresholds and their defaults

Every filter threshold lives in one named set, THRESHOLDS in web/public/js/ledger-core.js. The Worker, the map, the watchlist and this table all read from it.

ThresholdDefaultWhat it does and why
Loading from ledger-core.js…

10. Limitations

← Back to the ledger