-Physical and chemical pollutions are a key problem in populated urban areas. The long term monitoring of air pollutants concentrations is a very helpful aid for policy maker, to control the exposure and to keep track of the slope of the data. When and where measurements are not possible, predictive models can help in these issues. Among the several possible techniques, “AutoRegressive Integrated Moving Average” (ARIMA) models are a good choice when a sufficiently large database of measurements is available. In this paper, the authors use the CO concentrations measured in San Nicolas de Garza, in the Metropolitan Area of Monterrey, Mexico, to calibrate and implement two different models. Both the models will provide reliable predictions on a short time range, since they use in input the data measured in close past periods. For this reason, the ARIMA models presented here can provide predictions to maximum 24 hours forward the last measured data. 24, in fact, is the lag that maximizes the autocorrelation of the data and thus it is the seasonality implemented in the models. Finally, the authors will present a validation (comparison with data not used in the calibration) of the models in four different days along the year, showing that the models are not affected by overfitting effects and the results are good also on data not used during the model parameters tuning.

Prediction of CO concentrations in monterrey, Mexico, by means of ARIMA models

Guarnaccia, Claudio
;
Mancini, Simona;Quartieri, Joseph;
2018-01-01

Abstract

-Physical and chemical pollutions are a key problem in populated urban areas. The long term monitoring of air pollutants concentrations is a very helpful aid for policy maker, to control the exposure and to keep track of the slope of the data. When and where measurements are not possible, predictive models can help in these issues. Among the several possible techniques, “AutoRegressive Integrated Moving Average” (ARIMA) models are a good choice when a sufficiently large database of measurements is available. In this paper, the authors use the CO concentrations measured in San Nicolas de Garza, in the Metropolitan Area of Monterrey, Mexico, to calibrate and implement two different models. Both the models will provide reliable predictions on a short time range, since they use in input the data measured in close past periods. For this reason, the ARIMA models presented here can provide predictions to maximum 24 hours forward the last measured data. 24, in fact, is the lag that maximizes the autocorrelation of the data and thus it is the seasonality implemented in the models. Finally, the authors will present a validation (comparison with data not used in the calibration) of the models in four different days along the year, showing that the models are not affected by overfitting effects and the results are good also on data not used during the model parameters tuning.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11386/4724323
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact