Air Quality Forecasting: A Review and Comparison of the Approaches used Internationally and in Australia

Poor air quality poses a risk to human health, particularly to those people suffering cardiovascular illness. In this paper, we survey a range of techniques used in Australia and internationally to forecast air quality. The levels of technology used in Australia appear to be comparable to those used internationally and the accuracy of the best Australian forecast results is comparable to the best internationally. Based on this survey, we recommend that future work in Australia should be directed towards unifying the methods of displaying and reporting the forecast results. Secondly, techniques need to be developed to blend different types of forecasts using fuzzy neural network techniques. Thirdly, we need to consider the use of ensemble forecasts and model-data fusion to reduce forecast uncertainty and to provide an estimate of forecast error. The implementation of such a major program would require a nation-wide, coordinated research effort to develop new models and to provide improved methodologies of assessment of the impacts of pollution (such as directly predicting rates of mortality); to develop strategies for reduction of pollution; and to strengthen the interface between air quality science and those charged with the responsibility for air quality regulation and downstream clients.