A generalized Large Eddy Simulation (LES) based framework for quantifying methane emissions from oil & gas facilities: An alternative to controlled-release field tests

A robust, end-to-end, and high-performance-computing (HPC) enabled utility is developed that fuses spatial and temporal information acquired from a matrix of point sensors installed at an arbitrary oil & gas (O&G) facility. Using this input information, it sets up and performs LES of atmospheric boundary layers (ABLs) above surfaces and under atmospheric conditions mimicking real O&G sites. One objective is to establish a repeatable and computationally efficient methodology that complements existing field experiments, e.g. tests performed under the single-blind ADED protocol (Bell et al., ES&T 2023). Our tool presents a viable alternative for real O&G sites, for which controlled-released testing data (CRTD) is generally absent. Anthropogenic methane discharge into the atmosphere at O&G sites account for a large fraction of global greenhouse gas (GHG) emissions. Although initiatives requiring annualized GHG estimates like OGMP-2.0 have existed for a while, the recently finalized rule by the EPA under subpart OOOOb of the New Source Performance Standards has introduced stricter guidelines. These require more active monitoring at higher frequencies. Continuous monitoring (CM) platforms consisting of multiple fixed-point sensors and coupled anemometers are an attractive solution for ensuring the proposed benchmarks. The sensor output from such CM matrices are used in conjunction with ‘quantification’ algorithms to create a dynamic local picture of the O&G facility, including constructing actionable insights like source localization and overall site-level profiles of GHG emissions. Integrating  time-varying wind characteristics and pollutant concentrations from the sensor matrix with physical insights about the local structure of complex atmospheric surface layers provides a first-principles based route for constructing accurate ‘quantification’ models. The development of next-generation pollutant dispersion models is constrained by the limited availability of CRTD against which the dispersion models can be validated. Existing CRTD offers at best a partial representation of conditions encountered at real sites, which themselves could vary significantly from one site to another.  Our novel general-purpose eddy-resolving utility attempts to bridge the aforementioned gap. Methane is modeled as a passive tracer and the discrete point sources are approximated as 3D Gaussian functions. Undulations in surface topography are added by importing a site-specific digital elevation model via an immersed-boundary method, whereas thermal stratification of the ABL is modeled using the Boussinesq approximation. The surface layer is approximated by the Monin-Obukhov similarity theory (MOST) and the dynamic Smagorinsky model captures the unresolved motions. A discussion on the accuracy of the simulation tool is presented after providing an overview of the computational utility. This includes a comparison between the predictions by the LES studies and the dispersion of methane as measured by the point-sensor matrix from an illustrative set of known single and multiple-release tests peformed during the 2023 ADED campaign. Dispersion profiles estimated by the LES that replicate the external conditions observed at a couple of real O&G sites in the Permian Basin and the Appalachian region are explored. The efficacy of the traditional dispersion modeling route that invokes the Gaussian Plume approximation is examined in the near-field region.