Methane Copilot

Inspiration

Methane is one of the fastest, cheapest climate wins available — a single well-pad leak can do more near-term warming damage than months of a facility's CO2 output, yet most operators have no real-time way to see it, price it, or act on it. UNEP's MethaneSAT program proved the detection side is possible from orbit. We wanted to close the loop: take that satellite signal and turn it into something an operator or regulator actually acts on the same day — not a PDF report six weeks later.

What it does

Methane Copilot takes a live methane plume detection and walks it through four stages: Detect → Bill → Heal → Credit.

Source: pinpoints the leak's origin (facility, equipment, coordinates) from satellite spectral data, with a confidence score. Bill: converts the leak into real-time financial exposure — lost product value, regulatory penalty accrual, and climate damage cost, ticking up by the hour. Heal: dispatches a repair crew, tracks the fix through a live lifecycle status. Credit: once repaired, calculates the avoided emissions and mints them as carbon credits, pre-matched against buyers. How we built it

The backend is Python/FastAPI with an ML model handling source attribution and emissions estimation, built on real satellite plume data forked from UNEP's MethaneSAT program. The frontend is a mission-control style dashboard designed in Stitch and wired up as a click-through demo across all four pipeline stages plus supporting instrument views (satellite constellation status, spectral analysis, raw sensor telemetry).

Challenges we ran into

Turning a satellite detection into a defensible dollar figure in real time meant reconciling several different valuation models (commodity loss, statutory penalty, social cost of carbon) into one live number without it feeling arbitrary. Wiring a multi-stage dashboard so each screen's state (confidence score, accrual clock, dispatch status) stays consistent as you move through the pipeline was its own puzzle.

Accomplishments that we're proud of

Getting the full leak-to-credit loop working end to end — not just detection, but the financial and market-settlement layers that actually make a leak worth fixing fast. Grounding it in real UNEP satellite data instead of a synthetic demo dataset.

What we learned

The hardest part of climate tech isn't sensing the problem — it's making the economic case land instantly for the person who has to act on it. A confidence percentage doesn't move anyone; a dollar-per-hour ticker does.

What's next for Methane Copilot

Live integration with more satellite constellations for faster re-fly confirmation, automated dispatch to third-party field service providers, and formal MRV (measurement, reporting, verification) partnerships so the minted credits meet registry standards for real trading.