A switching poisson process model for high concentrations in short-range atmospheric dispersion

High concentrations of a pollutant dispersing in a turbulent atmosphere may be described in terms of the times at which concentration exceeds a high threshold and the values it reaches at those times. A stochastic model based on a switching Poisson process is proposed to account for both aspects, extending an earlier model of Mole et al. (1995, Environmetrics 6, 595–606), which described only the magnitudes of high concentrations. The model is fitted by maximum likelihood and is shown to be capable of capturing the broad features of extreme concentrations in a series of atmospheric dispersion experiments. Evidence is found that in some cases parameters of the model vary with time, and it is argued that this lends support to an explanation of the variability of extreme concentrations based on a meandering plume hypothesis.