The importance of reducing methane emissions is underestimated and understudied. Tracking methane emissions is hard due to gas leaking in many places along the long processing and distribution chain. Luckily, the newest technology allows us to track the emissions from space. Satellites armed with advanced sensors and AI are flagging invisible methane leaks that were once impossible to detect. In the era where cutting emissions is more urgent than ever, technology that exposes ‘super-emitter’ leaks from oil wells, gas pipelines, coal mines, and landfills is key to stopping climate change.
Why Methane Matters?
Methane has a global warming potential higher than that of carbon dioxide on the 20-year scale. It traps heat more effectively than CO2, making it a main contributor to the short-term climate change effect. Moreover, like other greenhouse gases, methane accelerates the climate feedback loop, which in return releases more greenhouse gases. If CO2 is a slow buildup of warmth, methane is a sudden fever. If one CO2 molecule wraps Earth in a single blanket, one methane molecule is like throwing on 80 blankets. It breaks down in the atmosphere way faster than carbon dioxide, but is more harmful in the short term. That is why scientists see reducing methane emissions as the fastest way to slow down global warming. The problem is that methane is invisible and odorless. Leaks can go undetected for years and are related to many different industries, from oil and gas to decomposing trash in landfills.
How Can Satellites “See” Methane?
The issue with methane being transparent is that it cannot be discovered on a simple satellite image. However, imaging spectrometer sensors that are used in space measure what wavelengths of light are absorbed as the sunlight bounces off the Earth. Every gas has a unique light ‘fingerprint’. Methane absorbs sunlight in specific infrared bands that are invisible to our eyes but easily measured by a spectrometer. Thanks to spectrometers, satellites can detect methane by looking for the shadow of the infrared glow on the Earth below, which is specific for CH4. There is a broad range of sensors that can either analyze broad areas or zoom in on fine details, using them together helps create a more precise understanding of the location and size of methane plumes on Earth.
Using AI to Contextualize the Data
Sensors from the satellites help create billions of data points every day, but it is Artificial Intelligence that helps understand that data. Machine learning models excel at pattern recognition in huge datasets. For methane detection, these models are trained to distinguish between the subtle signal of a plume from the background noise and from lookalike artifacts. There are multiple models that are trained depending on the data provided. One example is a two-step ML pipeline for Sentinel-5P data that first uses a convolutional neural network (CNN) to scan for anything plume-shaped, then a support vector machine (SVM) classifier double-checks those candidates to remove false detections. It proved extremely effective in 2021, discovering almost 3000 methane plumes around the world. Importantly, AI does this in almost real time, which is key in enforcing leak repairs.
Detecting the plumes is just the first step on the path to reducing emissions and tracking them. AI can also help trace them to their sources. This includes linking a specific plume to an infrastructure such as well, pipeline, or facility on the ground that’s responsible for it. Using the newest models, companies like Google are able to create infrastructure mappings that analyze high-resolution satellite images (such as Google Earth) and map them to the equipment around the world. Cross-referencing these maps with the maps of the methane plumes makes it easier to find specific culprits responsible for the emissions. This step is crucial as it transforms a vague pollution signal into an actionable insight that can be as specific as ‘valve #5 at pipeline X’ is leaking - fix needed.
Real-World Impact
Finding the methane leak from space is a tremendous achievement, but the real impact comes from actually fixing it. So, how is this used in practice? UN Environment Programme is one example of applying data and technology to create impact using the Methane Alert and Response System (MARS). It continuously tracks the data coming from the satellites and alerts governments and companies when a major plume is detected. Unfortunately, even though in a year, MARS sent 1200 alerts, only 12 were actioned on and fixed, meaning only 1% of all leaks were removed. This shows that while the technology has been rapidly improving, policy is lagging. As Inger Andersen (UNEP’s director) put it bluntly, governments and companies need to ‘stop paying lip service’ and start fixing the leaks when they are alerted.
Still, there are a few notable success stories. Many alerts were actioned, most notably in Algeria, Azerbaijan, Nigeria, and the U.S.. In certain countries, new regulations, such as a methane fee in the U.S., give companies strong financial incentives to use an alerting system and act on the leaks. Azerbaijan’s state oil company, for example, said it identified 400 leaks via satellite monitoring and is working to fix them. And globally, there’s a push to make all this data public. Certain nonprofit initiatives, such as the Climate TRACE coalition and Carbon Mapper data portal, are aggregating satellite emissions data and sharing it openly. In the ideal scenario, there would be radical transparency, where everyone can see who is emitting, which then puts pressure on the emitter to fix the issue. A good example is the situation in Turkmenistan in 2022. After satellite data revealed huge methane plumes, it turned out the country was the single largest source of super-emitter leaks. This news helped gather international assistance to fix its aging infrastructure.
Challenges Ahead
As much as the progress in the areas is impressive, we are far from perfect technology. For example, satellites can’t see through thick clouds, which means methane detection needs clear skies. This makes the tracking especially hard in tropical and high-altitude areas. There is research being done on the ways satellite sensors can complement other techniques, such as on the ground sensors, but the perfect solution is not there yet. Other challenges, such as having enough satellites and 24/7 coverage of the entire Earth in a financially viable manner, are still waiting to be resolved. Finally, even with perfect technology, the final solution still requires boots on the ground work to fix the issues and AI cannot do that for us. Luckily, as we overcome these challenges, the value of the technology itself will only increase.