A Neural Prediction Model for the Maximum Daily Ozone Concentration

Tropospheric Ozone has been recognised as a harmful pollutant since it accumulates when the natural equilibrium between ozone and nitrogen dioxide is altered due to the presence of air pollutants (e.g. volatile compounds) in presence of high values of the solar radiation and with the contribution of not favourable meteo-climatic conditions (low wind speed, opaque cloud cover etc) [1]. In this paper we present some results concerning the modelling of Ozone time series in an area characterised by a high density of petrochemical plants. In order to handle the difficulties which arise in identifying non-linear systems with traditional techniques, neural networks are used to build 1-day-ahead prediction models.