Author(s)

J. Hooyberghs, C. Mensink, G. Dumont, F. Fierens & O. Brasseur

Abstract

In this paper we summarize a survey that was made to determine which variables are most relevant as input data for short-term PM10 forecasting based on a neural network model. Since the health impact of airborne particulate matter is becoming a topic of increasing interest, this study was performed as a first step towards the design of an operational system that can inform the media when the PM10 concentration is expected to exceed a given level of concern. The research is based on ambient PM10 measurements from different monitoring sites in Belgium during the period 1997-2001. Besides these data, we used ECMWFforecasts of meteorological parameters. This parameter set includes standard (gr

Keywords