This study analyzes air quality data in the Taranto municipal area. This is a high environmental risk region being characterized by the massive presence of industrial sites with elevated environmental impact activities. We focus on three pollutants formed by combustion processes and related to meteorological conditions, namely particulate matter, sulphur dioxide, and nitrogen dioxide. Preliminary analysis involved addressing several data problems. First of all an imputation technique was considered to cope with the large number of missing data. Missing data imputation was addressed by a leave-one-out procedure based on the recursive Bayesian estimation and prediction of spatial linear mixed effects (LME) models enriched by a time-recursive prior structure. Secondly, a unique daily weather database at the city level was obtained combining data from three stations, characterized by gaps and unreliable measurements. Spatio-temporal modeling of the multivariate normalized daily pollution data was then performed within a Bayesian hierarchical framework, including time varying weather covariates and a semiparametric spatial covariance structure. Daily estimates of the pollutants’ concentration surfaces allow us to identify areas of higher concentration (hot spots), possibly related to specific anthropic activities.
A multivariate approach to the analysis of air quality in a high environmental risk area
POLLICE, Alessio
Methodology
;
2010-01-01
Abstract
This study analyzes air quality data in the Taranto municipal area. This is a high environmental risk region being characterized by the massive presence of industrial sites with elevated environmental impact activities. We focus on three pollutants formed by combustion processes and related to meteorological conditions, namely particulate matter, sulphur dioxide, and nitrogen dioxide. Preliminary analysis involved addressing several data problems. First of all an imputation technique was considered to cope with the large number of missing data. Missing data imputation was addressed by a leave-one-out procedure based on the recursive Bayesian estimation and prediction of spatial linear mixed effects (LME) models enriched by a time-recursive prior structure. Secondly, a unique daily weather database at the city level was obtained combining data from three stations, characterized by gaps and unreliable measurements. Spatio-temporal modeling of the multivariate normalized daily pollution data was then performed within a Bayesian hierarchical framework, including time varying weather covariates and a semiparametric spatial covariance structure. Daily estimates of the pollutants’ concentration surfaces allow us to identify areas of higher concentration (hot spots), possibly related to specific anthropic activities.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
