A Data Analytics Approach for Unraveling the Complexity of Methane Emissions: A Permian Basin Study

Summary Methane emissions pose significant environmental challenges, particularly in regions with extensive oil and gas operations. In this study, we present a comprehensive, data-driven approach to analyze and predict methane enhancements in the Permian Basin, a major US hydrocarbon-producing basin. Methane enhancement refers to “the increase in methane concentration above the baseline background level” (Dlugokencky et al. 2009). Despite advances in satellite remote sensing, accurately characterizing methane enhancements in oil and gas regions remains challenging due to the available data’s coarse spatial and temporal resolution. In this study, we address this gap by integrating satellite-retrieved methane concentrations with detailed operational data from the Permian Basin, facilitating a better understanding of the complex interactions influencing methane emissions. Building on previous studies (e.g., Bian et al. 2023), our work further refines the analytical framework for regional methane emission estimation. We employ Sentinel-5P satellite data, alongside oil and gas operational data—including well counts, production volumes, and facility proximities—to capture regional emission patterns. We begin with a descriptive analysis of methane enhancement data attributed to different operators based on their geographical distribution across the basin. Next, our approach utilizes supervised learning algorithms—namely, random forest regression and classification—and unsupervised clustering techniques—hierarchical density-based spatial clustering of applications with noise (HDBSCAN) and K-means++—to help predict methane enhancement levels quantitatively, offering insights into influential features contributing to methane emissions. Finally, impurity-based feature importance and SHapley Additive exPlanations (SHAP) values are used to evaluate these models’ predictive power and interpretability, decoding the “black-box” nature and enabling an in-depth understanding of the factors driving methane enhancements. Our analysis reveals several key insights. Feature importance evaluation highlights that wind speed, month, and gas production are the primary drivers of methane enhancements. Furthermore, unsupervised clustering using HDBSCAN and K-means++ uncovers spatially distinct clusters corresponding to different operational and geological zones. We not only explore the complex dynamics of methane emissions in the Permian Basin in this study but also set a foundation for future investigations to refine our comprehension and prediction capabilities of methane emissions in oil and gas regions.