Particle simulations of dispersion using observed meandering and turbulence

A Lagrangian stochastic particle model driven by observed winds from a network of 13 sonic anemometers is used to simulate the transport of contaminates due to meandering of the mean wind vector and diffusion by turbulence. The turbulence and the meandering motions are extracted from the observed velocity variances using a variable averaging window width. Such partitioning enables determination of the separate contributions from turbulence and meandering to the total dispersion. The turbulence is described by a Markov Chain Monte Carlo process based on the Langevin equation using the observed turbulence variances. The meandering motions, not the turbulence, are primarily responsible for the 1-h averaged horizontal dispersion as measured by the travel time dependence of the particle position variances. As a result, the 1-h averaged horizontal concentration patterns are often characterized by streaks and multi-modal distributions. Time series of concentration at a fixed location are highly nonstationary even when the 1-h averaged spatial distribution is close to Gaussian. The results show that meandering dominates the travel-time dependence of the horizontal dispersion under all atmospheric conditions: weak and strong winds, and unstable and stable stratification.