Abstract:

Methane emissions from oil and gas (O&G) extraction processes pose a significant economic burden due to losses in product value, equipment damage, maintenance issues, and safety risks, which can result in increased operational costs. Fortunately, mitigating emissions through infrastructure maintenance and fugitive methane capture offer a cost-effective solution and are feasible with current technology, bringing short-term benefits to the environment. To effectively identify high-emission areas, reliable leak detection and estimation methods are essential. Hyperspectral satellites offer a promising approach to monitor methane plumes at regional scales, leveraging high spatial resolution and frequent revisit times without requiring access to observation sites. Recently, significant efforts have been devoted to the development of machine learning plume detection (U-plume) and estimation tools (integrated methane mass enhancement, IME) to process satellite data, using large eddy leak simulations (LES) as a proxy to generate satellite pseudo-observations to be used as training data. However, current methods have been primarily designed for inland regions, and their application in offshore O&G extraction areas is essential for reducing emissions and ensuring the long-term viability of these operations. Notably, under reporting of emissions is a significant concern, exemplified by the Gulf of Mexico (GOM), the largest offshore O&G extraction basin in the US. Our research addresses the lack of offshore data representation in current tools, by expanding the IME, improving on machine learning detection methods, and incorporating plume data from LES using offshore meteorological data from the NASA Satellite Coastal and Oceanic Atmospheric Pollution Experiment-II field campaign. Specifically, 40 hours of WRF-LES at a 25 m resolution for a 9 x 9 km2 x 3,000 m domain were added to our plume analysis dataset, with 20 of these hours generated with meteorological data from the NOAA IGRA2 database at four different sites along the Gulf Coast, and 20 combined hours of WRF-LES with input ozonesonde measurements at four O&G platforms in the GOM, with different local ocean depths. Our results allow the evaluation of errors in source rate estimation of offshore leaks when the effective wind speed curve parameters are generated using inland plumes only, by comparing them to the same estimates using our updated "offshore curve", which we fit from plumes with influence from the marine boundary layer. Our overall goal is to develop the tools needed to assist the monitoring of offshore O&G extraction areas by expanding the accuracy and applicability of current techniques to support more effective methane emission mitigation strategies, especially for instruments with publicly available data, like NASA's Earth Surface Mineral Dust Source Investigation (EMIT) and Carbon Mapper's Tanager-1.

Speaker Bio:

 Pedro is a 5th year PhD student and Graduate Research Assistant in Mechanical Engineering at the University of Colorado Boulder. His research, conducted in collaboration with the Henze Group and the Carbon Cycle Greenhouse Gases at the NOAA Global Monitoring Laboratory through the Cooperative Institute for Research in the Environmental Sciences (CIRES), leverages machine learning, turbulence modeling, and statistical analysis to better understand methane emissions from oil and gas extraction in offshore regions. Before pivoting to environmental data science and large-scale flows, Pedro specialized in complex fluids, earning master's degrees in Materials Engineering (2022) and Condensed Matter Physics (2019) for his work on polymers and liquid crystals.