Methane-mitigation policies increasingly rely on simulation tools to design and evaluate Leak Detection and Repair (LDAR) programs, yet the extent to which these models reproduce real-world emissions remains unclear. This study provides the first direct, data-driven evaluation of 2 widely used open-source LDAR simulators, Fugitive Emissions Abatement Simulation Toolkit (FEAST) and the Leak Detection and Repair Simulator (LDAR-Sim), using a comprehensive regulatory dataset from the British Columbia Energy Regulator (BCER) covering 2020–2023. To ensure methodological alignment, simulations were configured to match empirical survey frequencies, optical gas imaging (OGI)-detection conditions, and BCER-calibrated leak generation parameters. Annual methane emissions from each model were compared with observed OGI-detectable emissions, and model behavior was further assessed through stratified Monte Carlo simulations, bootstrap aggregation, and targeted sensitivity analyses. Both models systematically underestimated annual emissions and did not reproduce the non-monotonic interannual patterns observed in the BCER records, especially the 2021 peak. The sensitivity analyses indicate that the realism of the repair process, including the timing of repairs and the occurrence of incomplete or delayed repairs, is the primary factor governing model accuracy. This influence is stronger than that of detection thresholds or survey frequency. LDAR-Sim demonstrated greater robustness because it samples repair delays from empirical distributions, while FEAST exhibited higher volatility when subjected to realistic repair behavior. These findings highlight key structural limitations in current LDAR simulation frameworks and underscore the need for improved representation of repair compliance. The results provide a transparent benchmark for LDAR model evaluation and offer guidance for enhancing simulation fidelity to support methane-mitigation policy design.
Wang et al. (Thu,) studied this question.