PM-10 Forecasting Using Neural Networks Model

PM-10 is one of major air pollutants which affect on human health. Since PM-10 comes from various emission sources and its level of concentration is largely dependent on meteorological and geographical factors of the local region, the forecasting of PM-10 concentration is of great interest to protect daily human health. In this study, the dependent variables on PM-10 concentration were derived from the correlation analysis between PM-10 and meteorological as well as environmental factors based on the observations at the monitoring stations. Using the potential variables on the PM-10 level, the neural network model was developed and tested. The root mean square errors of the prediction in test runs were 0.064 to 0.077 and the test results implied that the system could be used in real forecasting within 10% error rates.