Author(s)

P. Perez

Abstract

Prediction of maximum of 24-hour average of PM10 concentrations in Santiago, Chile P. Perez Departamento de Fisica Universidad de Santiago de Chile Abstract According to the value of the 24-hour average of PM10 concentration calculated every hour, three levels of increasing potential harm for the population of Santiago, Chile are defined: alert between 195 and 239, pre-emergency between 240 and 329 and emergency above 330 ^g/nf. We have developed a neural network based model that uses past values of PM10 concentrations plus forecasted values of meteorological variables as input in order to predict the level reached by the maximum of the 24-hour average of PM10 concentration on the next day. We have adjusted the parameters of the model using 1994 data and tested it with 1995 data. By comparing the results of the model with actual data we obtain an overall efficiency of 85%. Correct predictions of high level cases may be increased by giving more weight to the sample cases, g

Keywords