A stochastic model to forecast lead pollutant

Abstract

A transfer function model with the average daily wind speed as single external input has been used to forecast the 24 h lead concentrations recorded at a suburban «reference point». The model has been calibrated to minimize the inaccuracies observed between calculated and observed concentrations. The improved model has been validated with satisfactory results. The model has also been fitted to the recorded data at additional urban sampler points operating during short sampling campaigns. The comparison between the coefficients obtained at different urban sampler points revealed the possibility of forecasting lead by a simple method: the forecast value is obtained multiplying the forecast concentration at the «reference point» by a factor representing the ratioR between the average concentrations at the urban points and the average concentration recorded at the «reference point». This empirical method was validated with satisfactory results.