Optimization of data collection and analysis methods to reduce the impact of oil and gas industry upstream activities on the environment: a focus on SCOPE 1 emissions

As the oil and gas industry faces growing pressure to mitigate its climate impact, this thesis develops a robust, data-driven framework to estimate direct (Scope 1) greenhouse gas (GHG) emissions from upstream treatment plants. The study leverages administrative datasets that require significant preprocessing to extract meaningful insights into key emission sources. Fuel gas consumption, identified as the dominant contributor to emissions, served as the primary modeling focus due to its critical importance and the availability of detailed data. To account for heterogeneity across treatment facilities, unsupervised clustering techniques—particularly a random forest algorithm—were applied, resulting in the identification of four distinct operational groups. Principal Component Analysis (PCA) was used to address issues arising from highly correlated variables, improving interpretability and model accuracy. These clusters informed a split-panel regression approach, enabling the development of models tailored to specific operational behaviors. Model robustness was assessed using diagnostics such as Cook’s distance and hierarchical bootstrapping, ensuring adaptability to newly introduced treatment plants while maintaining internal homogeneity within clusters. Following the successful validation of fuel gas models, the study expanded to include other emission sources, such as flaring and venting. To quantify uncertainty and prioritize key variables, a Global Sensitivity Analysis (GSA) was conducted. This method allowed for the systematic evaluation of input variability under both uncorrelated and correlated assumptions, using a two-step procedure. The approach produced comprehensive emission estimates and associated uncertainty ranges for selected treatment plants. By integrating clustering, regression, and sensitivity analysis, the thesis offers a versatile and scalable methodology for estimating upstream GHG emissions. This framework supports ongoing monitoring and provides a foundation for future regulatory reporting, scenario planning, and emission-reduction strategies across diverse operational settings in the oil and gas sector.