STATISTICAL FORECAST MODELS FOR DAILY AIR PARTICULATE IRON AND LEAD CONCENTRATIONS FOR MADRID, SPAIN

Statistical models that estimate the daily concentrations of air-particulate Fe and Pb (which are linked to natural and anthropogenic sources, respectively) in a semi-urban area of Madrid, Spain, have been specified. State-space modeling, coupled with Kalman filtering, and Box-Jenkins (transfer function and ARIMA) modeling has been used. Wind speed, which is related to the atmospheric diffusion, has been selected as an exogenous variable for the state-space and transfer function models. The adaptive state-space models do not seem to perform significantly better than the transfer function models but clearly do better than ARIMA models. One-day-lagged wind speed and the persistence were the most important inputs for the proposed Fe and Pb concentration models. However, the predictions of daily Fe particle concentration appears to be better than those of Pb. The performance of the adaptive state-space models suggests that this kind of modeling can be suitable for particulate matter air quality forecasting.