Time Series Predictive Model Application to Air Pollution Assessment

Physical polluting agents monitoring and control is a relevant problem to be considered in all areas where human activities take place. Air pollution, acoustical noise, electromagnetic fields, etc., should be carefully assessed in order to protect human health. Regarding air pollution, the importance of developing proper mathematical models, able to fit observed data and predict future behavior of pollutants is obvious. Among all the possible approaches, regression methods seem to be feasible when a large dataset is available and the trend and eventual periodicities can be evaluated. In this paper, a Time Series Analysis model is developed and applied to hourly CO concentrations in the urban site of San Nicolas de los Garza, Nuevo Leon, Mexico. The calibration made on one year dataset will show a 24 hours seasonal effect and a quite stable trend. The validation on two different periods, not used in the calibration phase, will exploit quite different results, showing that the general slope of the data is quite good reconstructed, while the local oscillation are difficult to be predicted.