Machine Learning for Methane Detection and Quantification from Space - A survey

TL;DR

The surveyed work analyzes space-based methane detection in the SWIR, contrasting traditional physics-based retrieval and emission-rate estimation with modern ML approaches for plume segmentation and rate estimation. It covers current and future multispectral/hyperspectral sensors, reviews physically-based radiative transfer and CO$_2$ proxy retrieval methods, and catalogs ML architectures (notably CNN/U-net and transformers) used on diverse datasets, many of which rely on simulated plumes. It highlights the data scarcity, cross-sensor comparability issues, and reliance on synthetic data as major barriers, while identifying potential solutions in explainable AI, ensemble methods, and physics-informed ML. The work also discusses policy drivers (EU methane strategy) that may boost data availability and automation, and it points toward future trends favoring higher spectral resolution and large, benchmark datasets for robust ML-based methane detection and quantification.

Abstract

Methane ($CH_4$) is a potent anthropogenic greenhouse gas, contributing 86 times more to global warming than Carbon Dioxide ($CO_2$) over 20 years, and it also acts as an air pollutant. Given its high radiative forcing potential and relatively short atmospheric lifetime (9$\pm$1 years), methane has important implications for climate change, therefore, cutting methane emissions is crucial for effective climate change mitigation. This work expands existing information on operational methane point source detection sensors in the Short-Wave Infrared (SWIR) bands. It reviews the state-of-the-art for traditional as well as Machine Learning (ML) approaches. The architecture and data used in such ML models will be discussed separately for methane plume segmentation and emission rate estimation. Traditionally, experts rely on labor-intensive manually adjusted methods for methane detection. However, ML approaches offer greater scalability. Our analysis reveals that ML models outperform traditional methods, particularly those based on convolutional neural networks (CNN), which are based on the U-net and transformer architectures. These ML models extract valuable information from methane-sensitive spectral data, enabling a more accurate detection. Challenges arise when comparing these methods due to variations in data, sensor specifications, and evaluation metrics. To address this, we discuss existing datasets and metrics, providing an overview of available resources and identifying open research problems. Finally, we explore potential future advances in ML, emphasizing approaches for model comparability, large dataset creation, and the European Union's forthcoming methane strategy.

Figures (4)

• Figure 1: Overview of existing methane monitoring systems with categorization according to gsd and categorization into multi- and hyperspectral missions.
• Figure 2: This image by guo2020cloud depicts the U-net architecture of ronneberger2015u under a CC-BY license. The U-net architecture has been used in different forms for recent methane plume segmentation approaches. The architecture uses different convolution sizes, reduces the image size first (encoder) to reduce the feature space, and decodes the image (decoder) back to its original shape, additionally providing skip connections for more effective learning.
• Figure 3: Architecture used by schuit2023automated under the CC-BY 4.0 license
• Figure 4: Architecture based on cnn and transformer elements used by kumar2023methanemapper for plume detection and segmentation under the CC-BY 4.0 License. $f_{mc}$ describes potential ch4 feature maps, $f_{comb}$ is the combined output of the Backbones using different spectral ranges, $f_Z$ is the output of two mlp from $f_{comb}$, $f_e$ is the feature encoded $f_Z$.