The air quality impacts of pre-operational hydraulic fracturing activities

Abstract

Hydraulic fracturing (fracking) is a short phase in unconventional oil and natural gas (O&G) development. Before fracking there is a lengthy period of preparation, which can represent a significant proportion of the well lifecycle. Extensive infrastructure is delivered onto site, leading to increased volumes of heavy traffic, energy generation and construction work on site. Termed the "pre-operational" period, this is rarely investigated as air quality evaluations typically focus on the extraction phase. In this work we quantify the change in air pollution during pre-operational activities at a shale gas exploration site near Kirby Misperton, North Yorkshire, England. Baseline air quality measurements were made two years prior to any shale gas activity and were used as a training dataset for random forest (RF) machine learning models. The models allowed for a comparison between observed air quality during the pre-operational phase and a counterfactual business as usual (BAU) prediction. During the pre-operational phase a significant deviation from the BAU scenario was observed. This was characterised by significant enhancements in NO_x and a concurrent reduction in O₃, caused by extensive additional vehicle movements and the presence of combustion sources such as generators on the well pad. During the pre-operational period NO_x increased by 274 % and O₃ decreased by 29 % when compared to BAU model values. There was also an increase in primary emissions of NO₂ during the pre-operational phase which may have implications for the attainment of ambient air quality standards in the local surroundings. Unconventional O&G development remains under discussion as a potential option for improving the security of supply of domestic energy, tensioned however against significant environmental impacts. Here we demonstrate that the preparative work needed to begin fracking elevates air pollution in its own right, a further potential disbenefit that should be considered.

Keywords: Baseline monitoring; Fracking; Machine learning; NO(x); Oil and gas emissions; Random Forest.