Application of nonparametric regression and statistical testing to identify the impact of oil and natural gas development on local air quality

- Air monitoring data were analyzed from one background and two downwind sites.
- Kernel regression was used to identify directional trends in concentration.
- A block bootstrap test for trend significance accounted for autocorrelation.
- A statistically significant regional impact was identified at the background site.
- The downwind sites exhibited statistically significant impacts from local sources.

The objective of the current work was to develop a statistical method and associated tool to evaluate the impact of oil and natural gas exploration and production activities on local air quality. Nonparametric regression of pollutant concentrations on wind direction was combined with bootstrap hypothesis testing to provide statistical inference regarding the existence of a local/regional air quality impact. The block bootstrap method was employed to address the effect of autocorrelation on test significance. The method was applied to short-term air monitoring data collected at three sites within Pennsylvania's Allegheny National Forest. All of the measured pollutant concentrations were well below the National Ambient Air Quality Standards, so the usual criteria and methods for data analysis were not sufficient. Using advanced directional analysis methods, test results were first applied to verify the existence of a regional impact at a background site. Next the impact of an oil field on local NOx and SO2 concentrations at a second monitoring site was identified after removal of the regional effect. Analysis of a third site also revealed air quality impacts from nearby areas with a high density of oil and gas wells. All results and conclusions were quantified in terms of statistical significance level for the associated inferences. The proposed method can be used to formulate hypotheses and verify conclusions regarding oil and gas well impacts on air quality and support better-informed decisions for their management and regulation.