Seasonal Time Series Modelling and Forecasting of Particulate Matter (PM2.5) Datasets

Abstract

The results of man-made and natural releases of hazardous pollutants like particulate matter (PM_2.5) have serious effects on humanity. The prolonged proximity of such particles causes numerous critical and long-term ailments. Therefore, this study is used to model and forecast Nigerian yearly mean particulate matter to have an insight into future values. The methods used are descriptive statistics and Seasonal Autoregressive Integrated Moving Average (SARIMA) models. The results from the descriptive statistics of Nigerian yearly mean particulate matter (PM_2.5) signified the mean is 63.77 and the standard deviation is 7.84. The time plot showed that the series exhibited secular and seasonal variations. The series is stationary at the initial difference, as demonstrated by applying the technique of the Augmented Dickey-Fuller test. The tentative SARIMA model was determined using autocorrelation and partial autocorrelation function plots. The values of the Schwarz and Akaike information criteria (AIC) were used to select the optimal model after estimation with the ordinary least squares technique. The adequacy of the SARIMA(1,1,1)x(1,1,1)₁₂ model was determined based on residual autocorrelation, partial autocorrelation function plots and Modified Box-Pierce (Ljung-Box) Chi-Square Statistic. The SARIMA(1,1,1)x(1,1,1)₁₂ model forecast for Nigerian yearly mean particulate matter for 19 years signified a seasonal fluctuating movement from 2020 to 2039 and fluctuation within the interval 27.735 – 81.659respectively for the next 19 years. Conclusively, the Nigerian yearly mean particulate matter from 1990 to 2019 and the forecast from 2020 to 2039 are above the World Health Organisation (WHO) standard of 10 annual mean. Therefore, the high level of yearly mean particulate matter is alarming and this is expected to negatively affect the health of in Nigerian over the years