Modelling threshold exceedances of air pollution concentrations via non-homogeneous Poisson process with multiple change-points

A Bayesian multiple change-point model is proposed to analyse violations of air quality standards by pollutants such as nitrogen oxides (NO2 and NO) and carbon monoxide (CO). Threshold exceedance occurrences are modelled by a step rate Poisson process fitted after short-range correlations in the exceedance data are removed via declusterisation. The change-points are identified, and the rate function is estimated, using a reversible jump MCMC algorithm adapted from Green (1995). This technique is applied to the daily concentration data collected in Leeds, UK (1993–2009). Results are validated by running the MCMC estimator on the posterior-replicated data. Findings are discussed in the context of the past environmental actions and events. The proposed methodology may be useful for the air quality management by providing quantitative means to measure the efficacy of pollution control programmes.

► Reversible jump Markov chain Monte Carlo (MCMC) method due to Green (1995) is adapted to threshold violations of air pollutants. ► Analysis is based on a simple Poisson (step rate) model with no prior restrictions. ► Improved preprocessing is used to tackle missing data, seasonality and correlations. ► Results are assessed and validated via both classical and Bayesian statistical tools. ► MCMC performance is cross-validated using simulation techniques.