Keywords: ozone forecasting, neural networks, air pollution, megacities, tropospheric ozone, air quality, Brazil, meteorological variables
Ground–level ozone prediction using a neural network model based on meteorological variables and appiled to the metropolitan area of São Paulo
A neural network model to predict ozone concentration in the São Paulo Metropolitan Area was developed, based on average values of meteorological variables in the morning (8:00–12:00 hr) and afternoon (13:00–17:00 hr) periods. Outputs are the maximum and average ozone concentrations in the afternoon (12:00–17:00 hr). The correlation coefficient between computed and measured values was 0.82 and 0.88 for the maximum and average ozone concentration, respectively. The model presented good performance as a prediction tool for the maximum ozone concentration. For prediction periods from 1 to 5 days 0 to 23% failures (95% confidence) were obtained.