Source apportionment of gaseous pollutants in oil and gas extraction areas: A comparison between positive matrix factorization and self-organizing maps approaches

Emissions from oil and gas extraction and processing may represent a significant source of atmospheric pollution, with potential hazards for human health, the environment, and the economy on a local to global scale. In this work, twelve atmospheric pollutants were quantified at hourly time resolution over eight months near the Val d'Agri Oil Center (COVA), the largest onshore oil and natural gas extraction and primary processing facility in Europe, located in a semi-rural area of southern Italy. Temporal trends and average concentration values indicate that few exceedances of regulated species were observed during the investigated period. Two multivariate statistical techniques were employed for source apportionment: the well-established Positive Matrix Factorization (PMF) and the neural-network based Self-Organizing Map (SOM) approaches were used, and their results were compared. Both PMF and SOM produced a six-source solution, identifying the main common factors responsible for air quality near COVA, including: photooxidative processes for the production of secondary gaseous species, traffic, and high-temperature COVA operations, in particular gas turbines. Interestingly, although the overlap of some of the sources emerged, likely in association with high-frequency environmental conditions, output differences emerged in the two cases. SOM succeeded in capturing highly episodic and seasonally affected sources (Claus process, gas flaring), while PMF discriminated primary and secondary production of NO_x and achieved the description of fugitive emissions from extraction wells. Overall, the double approach applied provides a comprehensive description of emission sources, demonstrating that this kind of parallel source apportionment approach, when possible, would be of great benefit to environmental studies.