Open Conference Systems, 50th Scientific meeting of the Italian Statistical Society

Font Size: 
Multivariate statistical emulators for city-level air quality management
Francesco Finazzi, Yoana Borisova, Marian E Scott, Alan Hills, Michela Cameletti

Last modified: 2018-05-22


Directive 2008/50/EC of the European Union regulates air quality in terms of pollutant concentration thresholds not to be exceeded at relevant monitoring sites. In order to understand what drives pollutant concentrations and to predict probabilities of compliance, environment protection agencies often make use of physical models called simulators. Given a set of drivers such as emissions and meteorological conditions, simulators are able to predict pollutant concentrations across space and over time. Due to the complexity of the physical model, however, the computational burden is usually high. In this work, we present an emulator which aims at replacing the simulator when probabilities of compliance at the monitoring sites are to be computed and when actions able to reduce the observed pollutant concentrations are to be defined. The emulator is based on a multivariate spatial model able to handle missing data. The spatial model is estimated using the simulator output obtained on the basis of a design of experiment. As a case study, results for the city of Aberdeen (UK) are provided.