Dynamical and statistical relationships among air pollutant species and meteorological factors at an urban site in Hong Kong

Abstract

The aims of the study are to explore the temporal variations of urban air pollutant species and meteorological parameters, as well as the interactions among them with time. As a typical urban area, the Central-Western district is chosen as the study site where an air quality monitoring unit as, an integral part of the territory-wide network, has been set up by the Environmental Protection Department since mid eighties. Hourly data in the year 1994 are extracted. They include nitrogen oxides, ozone, wind speed and direction, and global solar radiation flux. Since most of the time series are nonstationary and exhibit different characters from month to month, they were segmented on monthly basis. Recursive time-varying filtering algorithms are applied to every segment of the air pollutant series, and confirm the existence of diurnal patterns in these species for all months. While natural and anthropogenic sources are responsible for the patterns in ozone, anthropogenic sources alone are the main contributors to the patterns in nitrogen oxides. Diurnal patterns are also found in the meridional wind component in non-summer periods, but it is believed to be a consequence from land-sea thermal contrast. Zonal wind component in the same periods, on the other hand, is found negatively correlated with the ozone level, supporting speculation that the nitrogen oxides are not imported from abroad but emitted locally. It is also found that in January the underlying trend in solar flux resembles to that in nitrogen dioxides, suggesting the prolonged existence of nitrogen trioxides sufficient enough to be photolysed. Cross-correlation analysis also confirmed that nitrogen oxides and ozone are negatively correlated. Therefore two types of time series models were devised in an attempt to mimic temporal evolution of intertwining chemical transformations between ozone and nitrogen dioxide. The first type is the threshold autoregressive (TAR) family, which consists of a system of piecewise linear transfer functions subjected to one or more constraints, or threshold values, in observation data. The second type is the multiple time series family, which includes a member called vector autoregressive system (VAR). VAR consists of a system of linear equations with autoregressive parameters of a time-lag written as a matrix for that time-lag. In general, VAR is advantageous over TAR in several aspects, including computation cost, consistency with the fundamentals in atmospheric chemistry and ability to reveal causal relationships between the species at all time scales. Diagnostic checking, however, shows that both types of the models are not adequate. Inclusion of other parameters such as concentrations of volatile organic compounds, dew points, etc., may improve the modelling and possibly forecasting performance.

Date of Award	16 Oct 2000
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Chung Leung Johnny CHAN (Supervisor)