This study proposes a Bayesian binomial regression model for monthly counts of extreme ozone levels. We implement a hierarchical modeling framework, involving different hierarchies of mixed effects, to best explain pollution data provided by ARPA Lombardia. We consider two different daily thresholds, providing two associated datasets counting the number of days above the threshold for each month. The data are collected over multiple years through a wide network of monitoring stations. The model we propose accounts for both spatial structure and time-varying covariates in a convenient and flexible structure. We apply the same Bayesian model to the two datasets on extreme ozone pollution in Northern Italy, in particular focusing on potential differences in the posterior inference.