Advanced Leak Detection Methods for Belowground Natural Gas Pipeline Leaks: Evaluation under Diverse Environmental and Operational Conditions

Advanced leak detection (ALD) methods are increasingly employed to enhance the safety of natural gas (NG) infrastructure and reduce methane (CH₄) emissions. While these technologies have proven effective for detecting aboveground leaks, their effectiveness in detecting belowground leaks remains limited due to the complex behavior of subsurface gas migration and the diffuse nature of surface and atmospheric plume presentation. This study presents a comprehensive experimental evaluation of ALD methods, including walking, driving, and simulated unmanned aerial vehicle (UAV_sim)-based surveys, conducted under controlled conditions at leak rates of 0.5, 5, and 10 slpm (0.334, 3.34, and 6.67 g/min). Experiments were carried out over two years at the Methane Emissions Technology Evaluation Center (METEC) and the U.S. Air Force Academy's Field Engineering and Readiness Laboratory (FERL), encompassing a range of environmental and operational conditions, including gas composition, soil moisture and permeability, urban geometry, sloped terrain (5% grade), and snow-covered surfaces. The probability of detection (POD) was used as the primary performance metric to quantify detection success under each scenario. Results showed that POD varied widely from 0 to 100% for the same detection method depending on site conditions. Walking surveys consistently outperformed other methods, achieving POD values greater than 90% across most conditions and downwind distances, and demonstrating strong reliability even in snow, moist soils, and sloped terrain. Driving and UAV-based surveys exhibited high PODs (up to 100%) in dry, open environments but experienced reductions exceeding 80 pp (percentage points) under snow-covered surfaces and urban conditions. For instance, gas composition had a measurable effect, with Denver-Julesburg (DJ) and Permian Basin gases improving POD for walking surveys but reducing it for driving and UAV-based surveys relative to distribution-grade gas, as heavier hydrocarbons alter subsurface migration and emission patterns. Similarly, low soil permeability reduced POD by up to 36 pp for driving surveys, while moist soils led to reductions of up to 79 pp for UAV-based surveys. In urban environments, POD for UAV-based surveys declined by 88 pp at 15 m downwind and failed to detect any atmospheric CH₄ concentrations above the detection threshold under snow-covered conditions. These findings underscore the strong dependence of ALD performance on environmental and operational variables and provide critical insights for refining detection protocols and optimizing survey strategies in complex, real-world conditions.