Artificial neural networks prediction of PM10 in the Milan area

Abstract

Spectral analysis is used to investigate the troposheric ozone formation and decomposition processes. Although ozone is a reactive, secondary air pollutant, a similar method of examination is used for elemental carbon which can be applied to ozone time series. To detect the reason for extreme high ozone concentrations in summer, the data was divided into low-frequency seasonal component and high-frequency component. The latter has been evaluated by using the corresponding power density spectrum. It is shown that meteorological large- and synoptic-scale fluctuations affect the ozone concentrations at all monitoring locations. In the data set of 1993-1995 this influence contributed peaks at cycle durations of 16-35 day peaks in the power density spectrum estimated from the ozone data. The level of the seasonal component at a particular site depends on the level of local influence by traffic and industry emissions. The power density spectrum of the mean early morning ozone concentration (2-4a.m.) at a measurement site situated 324m elevation is very similar to the power spectra of the ozone data at all other monitoring locations in 1995. This affirms the dominance of spatial homogenous fluctuations and indicates a relative uniform residual layer over a large region. Copyright (C) 2000 Elsevier Science Ltd.