Stochastic Simulation of Atmospheric Trajectories

Abstract

Methods are presented for generating an ensemble of synthetic atmospheric trajectories. These include methods for a set of independent trajectories, and methods for a correlated set of sequential trajectories. The models incorporate first-order autocorrelation of successive zonal and meridional displacement, and in the case of the sequential set, correlations between trajectories due to spatial patterns of air flow. The model for independent trajectories is applied and validated using three years (1976?78) of observed forward trajectories originating from southeast Michigan. The synthetic trajectories are statistically consistent with observed wind fields, and result in large-scale diffusion which is comparable to existing estimates. Limitations in the assumption of spatial homogeneity of wind statistics are explored using backward trajectories arriving at six locations in eastern North America during July 1978. The applications highlight the problems which may arise when observed trajectories are used in atmospheric transport modeling, and support the use of synthetic trajectories, at least in a complementary analysis, to obtain a more accurate long-term representation. The procedure for correlated trajectories is not implemented; however, it may prove useful for transport problems of shorter time scales where wind persistence is important.