Tracking Industrial Emissions and Odor Nuisance through Integrated Modeling and Citizen Reporting

The management of air quality in residential areas adjacent to large industrial hubs requires addressing two distinct yet overlapping challenges: monitoring pollutants with health implications and mitigating odor nuisances that significantly degrade quality of life. This study presents a multidisciplinary, integrated system designed to track, quantify and attribute these atmospheric impacts in one of Europe’s largest coastal petrochemical complexes. In the industrial area of Syracuse Province (Sicily, Italy), the emissions from refineries and port activities are a persistent source of both health concerns and community complaints. The NOSE (Network for Odour SEnsitivity) system has been operational since 2019 across the municipalities of Melilli, Priolo, Augusta and Siracusa, enabling citizens to report, via a dedicated web-app, the intensity and specific characteristics of odor episodes. In this framework, we developed an experiment based on three integrated pillars: a network of air quality and meteorological monitoring stations, the GRAMM-GRAL Lagrangian dispersion model and the data collected by the NOSE system. To address the frequent underestimation of the emissions in standard inventories, a Bayesian inversion framework was implemented to optimize prior emission estimates of benzene (C₆H₆), toluene (C₇H₈) and hydrogen sulphide (H₂S). Given the limitations of Lagrangian models in representing the photochemistry of complex volatile organic compounds, C₆H₆ and H₂S were used as conservative tracers and proxies for highly odorant non-methane hydrocarbon mixtures typically emitted by refinery processes.
Our findings demonstrate that the inversion procedure substantially improved dispersion model performance. The use of posterior emissions reduced the average Root Mean Square Error across all stations from 1.69 to 0.78 µg m^-3 for C₆H₆, from 2.46 to 0.76 µg m^-3 for C₇H₈, and from 8.1 to 0.81 µg m^-3 for H₂S. Correspondingly, the average Pearson correlation coefficient increased from 0.25 to 0.67 for C₆H₆ and C₇H₈, and from near-zero values to 0.45 for H₂S. Finally, we compared forward simulations using posterior emissions with spatio-temporal clusters of odor nuisance reports submitted by citizens. These results suggest that two major coastal refineries are the primary contributors to regulated pollutant concentrations and citizen-reported odor impacts. This integrated system, which combines citizen reporting, Lagrangian dispersion modeling and Bayesian inversion, provides local authorities with a powerful tool for identifying high-impact sources and developing targeted strategies for health protection and odor mitigation.