ARIMA models for the study of air pollution time series

Abstract

 Time series of nitrogen oxides, NOx, monthly concentrations between 2014 and 2019 in the boundary layer of air in the city of Salta are analyzed at 15 monitoring sites. The objective is to characterize the temporal evolution of pollution and generate forecasts using ARIMA models. The Box and Jenkins methodology is implemented in three stages: identification, estimation and va­lidation. Identification involves analyzing the trend, heteroscedasticity, seasonality and stationarity of the series. Estimation consists of fitting ARIMA(p,d,q)xARIMA(P,D,Q)s models to the series, considering the presence of seasonality. Finally, validation compares the values predicted by the model with the experimental data. The results reveal that the complexity of the ARIMA models for NOx is adequately related to the magnitude of pollution. Simple models such as ARIMA(0,0,0)x(0,1,1)12 fit well to NOx series of sites with low pollution, while sites with higher pollution, require more complex models.

In conclusion, ARIMA models are useful tools for analyzing and predicting air pollution. The choice of the appropriate model depends on the specific characteristics of each series.