ARIMA models application to air pollution data in Monterrey, Mexico

The environmental impact of physical and chemical pollution is one of the most relevant problems nowadays. In order to monitor the chemical air pollutants concentrations, field measurements can be performed. Data recorded in strategical areas can be used to calibrate predictive models, useful to produce predictions for following periods and to help administrations in protecting people from high and risky exposure. In this paper, the ARIMA (AutoRegressive Integrated Moving Average) technique is applied to a dataset of CO concentrations in San Nicolas de Garza, one of the twelve municipalities of the Metropolitan Area of Monterrey. The results will show that the hourly data allow to build a model that gives a reliable prediction on a short time range. The main shortcoming of the ARIMA models, in fact, is that they adopt data measured in previous periods to build the future forecasts, thus the prediction cannot be extended to any future time period. The specific models presented in this paper can be used to predict maximum 24 hours in the future, resulting in a compromise between precision and forecast horizon. This could be very useful for policy maker, in order to apply extraordinary counter measures for keeping the pollution below the regulation thresholds.