Forecasting air quality index data with autoregressive integrated moving average models

Abstract

Air pollution arises from several sources, encompassing industrial, transportation, and home activities, and carries significant implica- tions for environmental health. High population mobility in a place, such as Jakarta, might exacerbate air pollution. In 2021, Jakarta, des- ignated as the Special Capital Region, had the highest population den- sity in Indonesia, with 15,978 individuals per square kilometer (km²). IQAir reports that Jakarta frequently places among the cities with the most unfavorable air quality globally.  In 2021, Jakarta was identified as the most polluted city in Indonesia, while Indonesia was placed 17th out of 118 countries for having the poorest air quality. Hence, the Jakarta Environmental Agency has formulated an Air Pollution Control Strategy till 2030 to diminish the proportion of lethal pollu- tion levels. Given the significance of air pollution’s detrimental effects on health, it is imperative to consistently regulate and oversee air pol- lution, including forecasting. This study utilizes the forecasting of the Air Quality Index (AQI) in Jakarta at air quality monitoring stations DKI1, DKI2, DKI3, DKI4, and DKI5. The Air Quality Index (AQI) data for Jakarta were obtained from the Jakarta Open Data portal spanning the years 2010 to 2021.  The ARIMA model was utilized to process this data. The generated models were assessed for error levels using the parameters Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Ab- solute Percentage Error (MAPE). This study produced a total of 25 ARIMA models to forecast the levels of air quality index (AQI) con- taminants. The levels of PM₁₀, SO₂, CO, O₃, and NO₂ at the five stations were determined to be highly accurate, accurate, and quite accurate, with Mean Absolute Percentage Error (MAPE) values rang- ing from 8% to 43%.