Modeling Methane Emissions from a Natural Gas Field in Appalachia, USA
Abstract
We investigated the meteorological controls of CH4 emissions from natural gas fields using a data analytics approach and developed a user-friendly predictive model with the dominant drivers. Time-series (15 min) data were collected during 2019-2020 with an eddy-covariance tower located in Westover, West Virginia, representing Marcellus shale wells. We isolated the CH4 emissions upwind to the gas field and filtered the data to account for nighttime low turbulence conditions. Principal component analysis and factor analysis, alongside Pearsons correlation matrix, were used to investigate the underlying grouping and interrelation of CH4 emissions with the different driving variables. Linkages with the meteorological variables were reliably estimated with partial least squares regression models by appropriately resolving multicollinearity in the data matrix. CH4 emissions were most strongly linked with the ambient atmospheric concentration of CH4 (CCH4), followed by wind speed (WS). The linkages further confirmed that the measured CH4 emissions did not comprise of any noteworthy biogenic emissions. Multivariate linear regression model with a bootstrap Monte Carlo resampling procedure (1000 iterations) indicated wind speed and CCH4 as the dominant and statistically significant successful predictors of the CH4 fluxes (n = 5037; p-value < 0.0001; R2 = 0.79) from the natural gas field. The findings of the study would help reduce the cost of field surveys to estimate leaky CH4 emissions from the natural gas fields in Appalachia and beyond.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFMGC25N0804Z