Comparison of simulated and observed methane plumes at oil and gas sites in the Permian Basin using advanced dispersion, coupled mesoscale-LES atmospheric modeling, and scientific machine learning

Identifying the source and rate of emissions from oil and gas (O&G) facilities, across large geographic areas with dense infrastructure, is extremely challenging due to multiple reasons: a) emissions might be intermittent; b) sensing is sparse and multi-modal; c) atmospheric forcing is often not measured at an O&G site, and if such measurements are available they are likely incomplete and very sparse; d) computing atmospheric forcing at resolutions required for facility-scale plume simulations (e.g., between 1 m to 10 m) is extremely challenging; e) the physics involved in computing the atmospheric forcing, turbulence, and dispersion, at such scales, is quite complex, and is often implemented in heavily-parameterized models; and f) lack of high-resolution, accurate and pertinent plume observations that could be used for validation are scarce. In this talk, we highlight these challenges, and report on our progress to alleviate them by using domain knowledge, advanced atmospheric and dispersion models, and scientific machine learning methods (SciML). Specifically, we compare the performance of high-resolution Eulerian and Lagrangian stochastic particle models to validate dispersion at select O&G sites in the Permian Basin. We highlight the common challenges of plume simulation under different atmospheric turbulence and spatial and temporal scales by comparing them to aircraft observations. We use two different techniques to compute the atmospheric forcing at facility-scales. These are: a) coupled mesoscale-Large-Eddy Simulation (LES) winds through the Weather Research and Forecasting (WRF) model; and b) SciML, which improves the resolution of operational High Resolution Rapid Refresh (HRRR) wind fields (provided by NOAA) while preserving conservation laws. Finally, we comment on the challenges to be addressed for deploying a reliable, timely, and cost-efficient emission detection and estimation framework within an operational setting.