Linear and regressive stochastic models for prediction of daily maximum ozone values at Mexico City atmosphere

We developed a procedure to forecast, with 2 or 3 hours, the daily maximum of surface ozone concentrations. It involves the adjustment of Autoregressi ve Integrated and Moving Average (ARIMA) models to daily ozone maximum conc entrations at 10 monitoring atmospheric stations in Mexico City during one- year period. A one-day forecast is made and it is adjusted with the meteoro logical and solar radiation information acquired during the first 3 hours b efore the occurrence of the maximum value. The relative importance for fore casting of the history of the process and of the meteorological conditions is evaluated. Finally an estimate of the daily probability of exceeding a g iven ozone level is made.