Modeling of the PM10 pollutant in Monterrey, Nuevo León, Mexico, with ARIMA, transfer functions and GARCH models

Particles of matter smaller than 10 μm (PM10) are significant pollutants due to their impact on health and the environment. This study analyzed daily PM10 data from the Obispado station (Monterrey, Nuevo León, Mexico) provided by the INECC, covering the period from 1997 to 2014. Climatological variables such as precipitation, wind speed (at 2 and 10 m), and atmospheric pressure, obtained from NASA’s POWER project, were included. Data were partitioned into training and testing sets using an 80-20% split, employing 3798 daily observations for training (2002 to 2012) and 950 observations for testing (2012 to 2014). Descriptive analysis and time series decomposition were performed to identify trends and seasonality. ARIMA models, univariate and with transfer functions incorporating meteorological variables, were applied. Augmented Dickey-Fuller and Ljung-Box tests validated the stationarity and residual independence assumptions. The ARIMA (1,1,1)(0,0,0)[365] model with transfer functions outperformed the univariate model. A significant relationship between meteorological variables and PM10 predictions was identified, supporting their use for short-term forecasting (≤ 10 days). Future studies should consider applying multivariate models with additional predictors and geostatistical approaches to improve spatiotemporal characterization.