A neural network forecast for daily average PM10 concentrations in Belgium

Over the past years, the health impact of airborne particulate matter (PM) has become a very topical subject. In the environmental sciences a lot of research effort goes towards the understanding of the PM phenomenon and the ability to forecast ambient PM concentrations. In this paper, we describe the development of a neural network tool to forecast the daily average PM10 concentrations in Belgium one day ahead. This research is based upon measurements from ten monitoring sites during the period 1997-2001 and upon ECMWF simulations of meteorological parameters. The most important input variable found was the boundary layer height. A model based on this parameter currently operational online serves to monitor the daily average threshold of 100 μg m−3. By extending the model with other input parameters we were able to increase the performance only slightly. This brings us to the conclusion that day-to-day fluctuations of PM10 concentrations in Belgian urban areas are to a large extent driven by meteorological conditions and to a lesser extend by changes in anthropogenic sources.