Applying exponential state space smoothing model to short term prediction of NO₂

Predicting air pollutant level has been important aspect as part of air quality management. A time series model exponential state space smoothing (ESSS) method was employed to short-term predict traffic-related pollutant, nitrogen dioxide (NO₂) during January 2013. Compared with autoregression (AR) and autoregressive integrated moving average (ARIMA) the ESSS model performed better with R² 0.673 respectively. The performance was also consistent for prediction over days in terms of R². For correlation between prediction and observation, the R² ranged from 0.4 to 0.6, showing that ESSS model has exceptional performances compared to AR and ARIMA. Hence, ESSS has potential to be applied as part of air quality management for daily air quality warning purposes.