Scalable estimation of speciated volatile organic compound composition in the upstream natural gas sector
Methane super-emitters from upstream oil and gas infrastructure are prevalent; local air quality may be impacted by co-emitted non-methane volatile organic compounds (NMVOCs), including benzene and other hazardous air pollutants. Risk assessment has been constrained by the uncertainty of NMVOC emiss...
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2025-01-01
|
| Series: | Environmental Research Communications |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2515-7620/adcd49 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Methane super-emitters from upstream oil and gas infrastructure are prevalent; local air quality may be impacted by co-emitted non-methane volatile organic compounds (NMVOCs), including benzene and other hazardous air pollutants. Risk assessment has been constrained by the uncertainty of NMVOC emissions, which can be estimated from methane emissions data and NMVOC:methane molar ratios. However, publicly available gas composition data is scattered and geographically limited, pointing to the need for harmonized data and scalable modeling for sources without available composition data. To address this gap, we compiled 4,525 non-flash and 3,487 flash upstream gas composition analyses from air permit applications spanning 11 of the top natural gas-producing states; the sampled facilities represent 46% of reported upstream methane emissions. We developed NMVOC:methane molar ratio models incorporating source basin, industry segment, equipment, and unspeciated composition data and select models via leave-one-out cross-validation. Benzene was detected in 97% of non-flash gas samples (mean; standard deviation [SD]: 230; 294 ppm) and >99% of flash gas samples (mean; SD: 1,785; 3,099 ppm). Our models of non-flash benzene:methane molar ratios empirically demonstrated low bias (normalized mean error <10%) and moderate precision (24%–45% variance explained) when infrastructure and unspeciated gas composition (methane and C6+ fraction) data were available, with lower precision when facility-specific unspeciated composition data is unavailable. Uncertainty can be expressed via appropriately-calibrated prediction intervals from quantile models. Modeling results were similar, albeit with higher bias, for flash gas. The observed variation in gas composition illustrates the importance of localized gas composition estimates for risk assessment of super-emitters. Our gas composition database and models can facilitate scalable risk assessment of NMVOC emissions, especially for non-flash gas sources with available localized, unspeciated gas composition data. |
|---|---|
| ISSN: | 2515-7620 |