Remote Sensing and AI for Monitoring Methane Emissions

Methane emissions are a significant contributor to global warming, and accurate monitoring is crucial for compliance with emission abatement regulations. This paper presents a novel approach to detecting methane emissions using satellite imagery and advanced AI algorithms. The study focuses on some well-studied regions for methane detection, including the Permian Basin, Hassi Messaoud, and Korpezhe. High-resolution imagery acquired from the European Copernicus satellites is employed, and an algorithm inspired by the principles of the multi-band multi-pass (MBMP) method is implemented, focusing on enhancing the distinction between methane-containing pixels and background pixels. The study introduces an ensemble AI framework comprising two complementary models: a machine learning classifier that utilizes textural and spatial feature analysis, and a CNN-based classifier trained on curated image patches. This dual-strategy approach enhances the robustness of the methane detection pipeline by efficiently distinguishing genuine methane signals from artifacts across multiple spectral bands, particularly in B12, B11, and B8a. The results demonstrate the effectiveness of this method, providing detailed spatial and spectral data that enhance the accuracy of emission detection. This research contributes to the energy industry by offering a reliable and scalable method for monitoring methane emissions, supporting regulatory compliance and environmental sustainability efforts.