Finding methane plumes in the vast ocean using satellite innovation

In July 2025, we ran a focused four week Research and Development (R&D) sprint to test how open-access satellite imagery could be used to detect methane emissions from offshore oil and gas infrastructure, which is a notoriously hard-to-detect, under-monitored source of climate pollution. This post shares what we set out to do, what we built, what we learned, and where we plan to go next.

Why We Did It

Methane is around 80x more potent than CO₂, making it one of the most urgent greenhouse gases to address. It’s also one of the most actionable climate pollutants. Unlike CO₂, methane breaks down quickly in the atmosphere, which means that stopping emissions can yield relatively rapid climate benefits.

In recent years, methane monitoring has advanced rapidly, especially for land-based sources. Satellite detection efforts like IMEO and Carbon Mapper are rapidly expanding global coverage, but these systems are still early in their development and global coverage remains uneven.

Offshore emissions are even harder to observe. Oil and gas infrastructure is often remotely located, public data is limited, and the detection techniques that work well over land tend to break down over water. As a result, offshore methane remains one of the least visible and least accounted for sources of climate pollution.

Part of the challenge is scale: only the largest emissions are visible from space using today’s open-access tools, and dedicated methane satellites don’t yet offer broad coverage. Even under ideal conditions, satellites like Sentinel-2 and Landsat can only detect very large methane releases, typically thousands of kilograms per hour. Specialized instruments like NASA’s EMIT or Carbon Mapper’s Tanager can detect smaller plumes with greater accuracy, but their observations are limited in scope and frequency.

That’s why we set out to answer the question: Can open-access satellite imagery reliably detect offshore super-emitters? If so, we can start closing one of the most important observational gaps in global climate monitoring while we await more comprehensive coverage from dedicated methane sensors to come fully online.

What We Built

To better understand how effectively open satellite data could be used to detect methane over water, we built a lightweight detection pipeline to adapt existing land-based methane detection approaches to the much trickier ocean environment.

Detecting methane with satellites relies on the fact that methane absorbs specific wavelengths of non-visible light. Using sensors like Sentinel-2 and Landsat, we can measure these methane-sensitive wavelengths to reveal plumes. This technique has been used effectively over land, but deploying it over the ocean presents major obstacles:

Sensor Limitations: Open-access satellites like Sentinel-2 and Landsat weren’t built for trace gas detection, though they have shown promise for detecting super emitters. Their relatively coarse spectral resolution makes it difficult to capture small or faint plumes, especially over the ocean. More sensitive instruments like EMIT and TROPOMI offer better detection capabilities, but either lack frequent coverage or are too coarse to spot individual plumes.

Unstable Background: Over land, you can compare multiple satellite passes to detect anomalies. Ocean waters are constantly changing—choppy and inconsistent between passes—so we must detect plumes in a single snapshot, which is much more challenging.

The Dark Ocean Problem: Oceans are dark, absorbing much of the light needed for methane detection, while land typically reflects light back to the sensor for stronger signals. For methane to be detectable over water, the ocean needs to reflect light back toward the satellite to create a bright surface over which methane is visible. That only happens when sunlight hits the water at just the right angle, creating a phenomenon known as “sun glint.” So to detect methane over the ocean, we first needed to solve a different problem – modeling the intensity of sun glint across the entire ocean. That work is described below.

A Key Technical Advancement: Mapping Sun Glint

We modeled global sun glint conditions using satellite and solar angle data to calculate when and where reflectance off the ocean surface would be strong enough to support methane detection. We developed mathematical models to predict optimal detection conditions worldwide by:

• Calculating glint alpha: We measured the difference between the satellite’s viewing angle and the “golden angle” where sunlight reflects perfectly off the ocean surface into the sensor.
• Mapping global coverage: We identified where and when throughout the year sufficient sun glint occurs for methane detection (areas with glint alpha less than 20 degrees).
• Refining glint scenes with pixel data: We paired our scene-level models of glint alpha with within-scene measurements of pixel brightness to confirm whether glint produced a usable signal near potential methane sources. This helped us prioritize scenes and specific pixels most likely to yield detectable methane plumes.
• Strategic targeting: We overlayed these optimal conditions with known offshore oil and gas infrastructure to focus our detection efforts

Our analysis revealed that equatorial regions provide more consistent sun glint coverage throughout the year, while high latitudes like the North Sea might only yield a handful of viable scenes each year.

The Complete Detection Pipeline

Once we identified where and when methane detection was most feasible, we built a prototype pipeline to test it. The goal was to move from theoretical detection conditions to actual methane plume candidates that could be reviewed and verified. Here’s how it worked:

Target Selection: We used locations from the Methane Alert Response System (MARS) database to focus on known methane-emitting infrastructure, and combined that with our sun glint predictions to identify probable detection opportunities.

Sentinel-2 Scene Filtering: For each target, we filtered Sentinel-2 imagery based on cloud cover, wind speed, and glint quality. This helped eliminate scenes where atmospheric conditions, sun angle, or surface disruption would make plumes undetectable.

Methane Raster Generation: Using the shortwave infrared bands from Sentinel-2, we calculated spectral indices sensitive to methane absorption using Varon et. al.’s Multiband Single Pass methodology [2021]. These indices help visualize the subtle signals that may indicate the presence of a plume.

Polygon Extraction: We converted high-signal regions in the methane rasters into vector polygons representing potential plumes, then exported them for review.Human-in-the-Loop Verification: A custom verification tool allowed us to visually inspect candidate detections, distinguishing real methane plumes from false positives like oil slicks, ship wakes, and cloud shadows.

Early Results

We tested the pipeline on two known methane-emitting infrastructure sites: the Zap-C platform in the Gulf of Mexico, and a platform in the Gulf of Thailand known to be a chronic methane emitter. In both regions, we focused on scenes that other researchers have determined to contain signs of methane emissions.

In the Gulf of Mexico, we highlighted one especially promising case: a methane plume that was invisible in standard RGB imagery but appeared clearly in the methane-sensitive bands. The plume aligned with wind direction, matched known infrastructure, and passed all filtering steps, providing a strong signal and useful validation of the pipeline.

In Thailand, we reviewed scenes where emissions were suspected, but early results were harder to interpret. One candidate plume aligned spatially with offshore infrastructure and showed up in the methane index, but it was difficult to rule out confounding factors like haze, sun glint artifacts, or unrelated operational activity.

The bottom line: offshore methane detection is possible with Sentinel-2 data under the right conditions, but the results tend to be noisy. Reliable detection depends as much on expert review and contextual data as it does on spectral signatures in the imagery. We look forward to expanding this work through additional case studies and longer time series, finding more examples, testing our assumptions, and refining the approach.

What We Learned

Technical Challenges Remain

Signal-to-Noise Issues: Ocean scenes are inherently noisy, with oil slicks, ship wakes, and surface disturbances creating false positives that can mimic methane signatures.

Coverage Limitations: Even in optimal locations like the Gulf of Mexico, usable imagery might only be available 36 times per year due to sun glint requirements, cloud cover, and orbital constraints. This means that short-lived, one-time ephemeral methane releases are highly unlikely to be captured in imagery, and the opportunity primarily applies to chronic pollution.

Expert Knowledge Gap: Interpreting methane detection rasters requires specialized expertise. Partnering with academic researchers and methane monitoring experts could help us advance this work.

The Needle-in-Haystack Challenge

As one team member noted, reviewing detection imagery is “a needle in a haystack situation where there’s a lot of hay and the hay all looks kind of like needles.” While individual image review isn’t extremely time-intensive, the sheer volume of potential scenes makes comprehensive monitoring challenging without better filtering or methods to amplify the fairly weak signals.

The Path Forward

This sprint helped us understand where the boundaries are and where the opportunities might be for advancing offshore methane detection. The work represents more than a technical achievement — it’s about bringing transparency to one of climate change’s most critical blind spots. Offshore methane emissions happen far from public view, but satellite monitoring can change that dynamic.

Our next steps include expanding satellite coverage with Landsat and EMIT imagery, developing advanced signal processing techniques to reduce false positives, and detecting “flaring gaps” where infrastructure stops burning methane and may be venting it instead. We’re also planning a Brazil feasibility study as part of a funded COP30 report examining fugitive methane plumes in Brazilian waters.

If developed further, this approach could help regulators target large methane sources for enforcement, enable watchdog groups to hold companies accountable, provide NGOs with credible evidence for campaigns, and give communities visibility into pollution that affects their health and climate.

This pilot embodies the creative, rigorous, and public-serving approach that makes SkyTruth a unique force for planetary transparency. We’ve proven that offshore methane detection is technically feasible, identified the key challenges that limit scaling, and built the foundational tools for a comprehensive monitoring system.

The ocean may be vast and its methane emissions hidden, but they’re no longer invisible.

If you’re interested in partnering to improve these models or funding expanded global analysis, let’s talk. Together, we can turn satellite data into meaningful climate action — one plume at a time.