Distribution-based modeling of hourly PM10 exceedance risk in Tabriz, Iran

Particulate matter (PM) pollution remains a critical environmental and health issue. This study applies seven classical probability distribution functions (PDFs) alongside Gaussian mixture models (GMMs) to hourly $P{M}_{10}$ concentration data collected in 2021 at an urban-traffic site (RasteKuche) and an industrial site (Rahahan). Model fitting using maximum likelihood estimation and evaluation via the Kolmogorov - Smirnov test and Akaike information criterion highlight distinct site-specific behaviours. At the urban site, seasonal variability is well captured by Gumbel, log-normal, and gamma distributions, with winter exhibiting the highest $P{M}_{10}$ exceedance probability (0.51) and a short average return period of 1.94 hours, indicating frequent exceedance of the WHO interim target of 50 $\mathrm{\mu g}/m^3$. Achieving compliance during winter would necessitate emission reductions over 64%. In contrast, the industrial site exhibits complex, multimodal $P{M}_{10}$ distributions poorly represented by classical PDFs; instead, 4- to 5-component GMMs capture these patterns effectively. Seasonal exceedance probabilities at this site peak in spring (0.23) and summer (0.22), requiring moderate emission reductions (up to 37%). These findings emphasize the urgency of seasonally tailored emission control strategies and demonstrate the utility of GMMs for risk assessment in heterogeneous industrial environments. This integrated probabilistic framework provides robust, locally calibrated benchmarks for air quality management in Tabriz.