The influence of meteorology on winter carbon monoxide and suspended particulate matter pollution in Christchurch City

Abstract

Christchurch City has long been known for its air pollution, consisting mainly of smoke emissions from domestic home heating appliances. The first known press report of smoke nuisance recorded in the city was published on August 9, 1869. Since then, a large number of events have occurred that have lead to the development of Environment Canterbury's Air Quality Chapter of the Proposed Natural Resources Regional Plan. Several studies have linked PM10 concentration to several health effects. CO is another contaminant of concern as concentrations have exceeded guidelines on a significant number of occasions. It can be exceedingly dangerous to human health if exposure is sufficient. Christchurch is located within a topographical depression, which complements the meteorological conditions required for high air pollution. The purpose of this study was to determine the role of meteorology as to the concentration of CO and PM10, as opposed to emission sources.

Environment Canterbury supplied the CO, PM10 and ground-based meteorological data used during this project. Scatter graphs were created using these data, comparing both contaminants with each meteorological variable. From these comparisons it was decided that wind speed, temperature difference and 10 metre temperature would be used within the models. Regression models for CO and PM10 were then created using these three meteorological variables. Only those data observed between 17:01 and 22:59 were incorporated rather than the full 24-hours, as it is between these times that high emissions coincide with inversion conditions. Additionally, the dataset was divided into two sections, 1988-95 and 1996-99. The former set was used to develop the models, while the latter was used in order to test them.

It was found that by incorporating the three main variables, just over half the data could be explained. Both the CO and PM10 models produced r² values of 54.4% and 55% respectively. In testing the models, r² values of 65.6% and 54% were obtained for CO and PM10 respectively. This indicates that meteorology, particularly these variables, play a significant role in whether these contaminants exist in certain concentrations. The models would be made more robust if variables such as rainfall and cloud-cover were incorporated as well human behaviour. Residual data show the extent to which these other factors may be contributing. At present each model is underestimating by varying degrees, depending on the concentration. Prediction accuracy is greater where concentration is low, while accuracy decreases as concentration increases. This is a product of there being a lack data and therefore, a lack of higher-concentration events. The fall in accuracy is accentuated by the behaviour of those who use domestic fires for heating. Although fires are lit during the evening, they were allowed to eventually go out, bringing concentration down. Moreover, the number of fires in Christchurch is insufficient for producing more extreme concentration events. This is an inherent limitation of the models and cannot be readily rectified without a larger number of observations and ironically, more fires.