Vehicle-Based Methane Detection, Source Attribution, and Quantification in the Upstream Oil and Gas Sector: Method Overview and Controlled Release Validation

Abstract Vehicle-based measurement systems occupy an important niche in the multiscale methane measurement landscape between component-level technologies and aerial-based methods. Performance evaluations are often limited to one dimension (e.g., quantification), while methodological opacity limits interpretation and intercomparisons between technologies. Here, we develop a vehicle-based measurement unit employing repeated site-perimeter transects and a processing framework featuring a novel weighted-likelihood algorithm for automated source attribution. Using controlled release experiments at two dedicated testing facilities, we characterize system performance across three dimensions─detection, attribution, and quantification. The system achieved a 94.5% true-positive detection rate with no false positives and a 90% detection limit as low as 0.078 kg/h with five repeat transects. Attribution accuracy was high at the source-group level, reaching 99.7% in single-source releases and 86% in multisource scenarios. Quantification was unbiased for low-to-moderate complexity releases (2 source groups; slope = 0.99, R2 = 0.81) but overestimated under highly complex conditions (5 source groups; slope = 1.45, R2 = 0.91). These results provide a transparent performance benchmark, demonstrating that these systems are well-suited to survey the large populations of low-to-moderate-complexity sites (e.g., single wells and batteries), as a screening layer in LDAR applications, and to quantify the low end of emission distributions in inventories, reducing reliance on generic emission factors.