Use of artificial intelligence and machine learning for automated detection of methane emissions on satellite images

Abstract

The aim of this study was to analyse modern machine learning methods and their integration into the processing of satellite data for the automated detection of methane emission sources. The research examined such methods as convolutional neural networks (CNN), support vector machines (SVM), random forest (RF), and k-nearest neighbours (KNN). The main findings demonstrated the efficiency of CNNs in the automated analysis of satellite images, particularly their ability to detect spatial patterns characteristic of methane emission sources with high accuracy and resilience to noise. It was established that the implementation of this technology for monitoring in Azerbaijan confirmed its capability to promptly identify anthropogenic threats based on satellite imagery. Meanwhile, the SVM method, implemented in Python, achieved a classification result with a 90% probability in favour of methane emission presence. At the same time, the R-based programme successfully classified images into the categories “methane emissions” and “no methane emissions”, effectively enabling the localisation of emission sources. The random forest method also proved effective in identifying methane sources using spectral characteristics, vegetation indices, and temperature data satellite images. The R programme displayed the geographic distribution of methane sources, while the Python-based code validated the method’s effectiveness by processing an image of the city of Baku, identifying potential methane sources even at low concentrations. In turn, the KNN method showed potential in classifying pixels by spatial and spectral features, allowing for rapid detection of hazardous zones in new satellite images. The obtained results affirm the feasibility of integrating machine learning methods into satellite monitoring systems to ensure accurate, prompt, and automated detection of methane emission sources, which is a significant step towards enhancing environmental safety and effective natural resource management in Azerbaijan and beyond.