Lagrangian stochastic dispersion modeling

 

     A Lagrangian stochastic (LS) model calculates the paths of a large number of individual particles as they travel with the local wind field.  Rather than calculating the wind field exactly, its statistical properties (mean and variances) are prescribed and the values acting on an individual particle at any time are selected from a Gaussian distribution of random deviates.  The basic assumption made is that the particles have only limited memory of their previous state.

     If the ground is perfectly reflective, i.e. no deposition of the tracer occurs, particle velocity is reversed and its position reflected whenever it passes below an imaginary plane, typically located at surface roughness length Zo.  To calculate concentration at a point in space, the total residence time of particles crossing through a volume of space surrounding the point is determined.

 

Advantages of the Lagrangian stochastic approach

 

     LS models have several advantages over their Gaussian and Eulerian counterparts, as well as at least one disadvantage.  LS models are more physically valid than Gaussian models, which do not incorporate wind shear or inhomogeneity of the turbulence field, and which can be shown to be equivalent to K-theory models with constant diffusivity.  They do not require artificial diffusivity, as do the Eulerian models for convective transport.  They can be released from an arbitrarily complex source region, by approximating the source as a collection of evenly distributed point sources; alternatively, they may be released from the sensor and be allowed to travel backward in time to determine what their source region might have been . Both deposition of tracer to the underlying surface and other time-dependent phenomenon such as radioactive decay can be handled in a natural way. 

 

Its disadvantages

 

     The most significant disadvantage of LS models is their computation time requirements, which can be several orders of magnitude larger than those required to solve algebraically reduced Gaussian models or even Eulerian models.  Because the calculation of concentration involves an averaging of particle residence time over a volume of space, through which many of the particles released in an experiment may not pass, a large number of particles may be required to generate an estimate of concentration with sufficient accuracy.  Further, if the collector volume is increased to capture more particles, the concentration will represent an average over a larger volume and be less representative of the target point.  To circumvent this problem in the backward-in-time mode used for area sources,  WindTrax automatically generates touchdown catalogs that can be used whenever the atmospheric conditions are similar.