Abstract for presentation at 14th IUAPPA World Congress

The accurate prediction of short-term concentrations at time scales such as 3 to 10 minutes is becoming increasingly important across a range of applications. These include the modelling of odour events, short-term exposure to air toxics from industrial sites and roadways, biosecurity hazards, and modelling for health risk assessments.
Power-law and peak-to-mean approaches have been found to be reasonably reliable for predicting the peak event in a year (such as the highest 10-minute average), but although based on statistical results, these approaches provide little or no information about the statistics of short-term events. For example, it is not obvious how these approaches can be applied to predict the 10th highest annual 10-minute concentration. Nor can they provide statistics on the short-term concentrations that are likely in any particular period given model predictions for, say, 1-hour averages.
This paper describes a method for obtaining more detailed statistical information about the short-term peaks. It starts from an analysis of 5 and 10-minute monitoring data in terms of the statistics of peak-to-mean ratios for each hour of the year. This demonstrates that the distribution of peak-to-mean ratios is independent of the mean concentration. Via a theoretically-based pseudo time series generator, this result is parameterised to develop a simple statistical model for the pdfs of short-term peaks.