Forecasting air pollution index in Hong Kong

Abstract

The Environmental Protection Department (EPD) of Hong Kong Government started the reporting of the Air Pollution Index (API) and its forecast in June 1995. API is calculated by comparing the concentrations of five pollutants, namely, nitrogen dioxide, sulphur dioxide, ozone, carbon monoxide, respirable suspended particulate, measured at EPD's eleven air quality monitoring stations with respect to the Hong Kong Air Quality Objectives. The Hong Kong API is a 5B2B index system with nine sub-indices being modeled separately by stratified multiple regression models, recommended by a consultancy study carried out in the early 90's. The EPD has been reporting to the public the daily API and the forecast of the following day for three land use categories: Urban, New Development and Industrial. In this paper, we give a brief overview of the API system of Hong Kong, in particular, our discussion focuses on the methodology used for modeling and forecasting of individual air quality sets. The main disadvantage of the present method is that it assumes model parameters can only vary discontinuously in a very limited pre-defined values (i.e. in stratification) over time even though the system is highly non-stationary. The present study demonstrates that the forecasting performance of the present method can be improved upon using more advanced time series modeling and forecasting methods based on time variable parameter version of time series regression model developed in recent years.