Analytical determination and classification of pollutant concentration fields using air pollution monitoring network data

During the last years, an important number of episodes with peak nitrogen dioxide levels occurred in the Paris region. Modelling air pollution is necessary to predict future episodes and to take decisions for the protection of populations. Our study proposes a methodology for urban air pollution analysis, as a preliminary stage of its modelling. This paper focuses on two major points: the estimation of pollutant concentration fields using measures from a monitoring network, and the definition of a reduced number of concentration spatial patterns. Interpolation by thin plate splines is proposed and the approximation quality is estimated by leave-one-out cross validation tests. Analysis of large numbers of pollutant fields is difficult: cluster analysis of spatial distributions is proposed. For this study, Ward’s algorithm is selected and applied to AIRPARIF nitrogen dioxide data during peak episodes from January 1993 to December 1997. Finally, some results and conclusions are presented.