Development of time series models for various pollutants in Bangalore city using the Akaike information criterion

Abstract

Pollution levels in developing countries, such as India, have become a major source of health problems. They need to be monitored and controlled. Bangalore, one of the major cities in India, faces a huge amount of pollution. Due to the dire need to control these pollutants, a sound mathematical modeling approach needs to be created for forecasting, controlling and monitoring. One such approach is time series modeling. The current work addresses a time series model that has been developed for the major pollutants in Bangalore city. These pollutants include PM10, PM2.5, NOx and SO2. The models used vary from AR (autoregressive), ARMA (autoregressive moving average) and ARIMA (autoregressive integrated moving average) for modeling air pollution in Bangalore city. Additionally, the selection of the best models was based on the Akaike Information Criterion, p-value and Box-Pierce test. Various steps were followed to build the model, which included identification of missing and extreme values followed by creating an appropriate imputing method and then identification of time series models using autocorrelation and partial autocorrelation plots to obtain various time series models. The best time series models were chosen based on the Akaike Information criterion (AIC) and various other statistical tests.

Keywords

• time series
• air pollution
• akaike information criterion
• arima
• statistics