Description of air pollution by means of pattern recognition. Part 2

Abstract

Based on meteorological observations and hourly measurements of chemical constituents at various locations in the city of Schiedam in the Rhine-mouth area near Rotterdam, The Netherlands, a learning machine has been constructed for the description and prediction of complaint situations for polluted atmospheres. It is demonstrated that a seven-parameter model may classify about 80% of the complaint and between 50 and 60% of the non-complaint situations correctly. Sometimes a prediction of a complaint situation up to 6 h in the future is possible, but different meteorological conditions may hamper the predictive ability of the learning machine.

The conditional probability describing the burden on the population in the area runs parallel with the reaction of the human nervous system to an exposure of increasingly concentrated noxious smelling air in test panel experiments. It is shown that the seriousness of the burden influences the total number of complaints filed within a certain period of time.