Off the Grid: The use of Gaussian mixture models with Lagrangian transport and dispersion models for density estimation and feature identification.

Abstract:

Lagrangian particle atmospheric transport and dispersion models simulate the dispersion of passive tracers in the atmosphere. At the most basic level, model output consists of the position of computational particles and the amount of mass they represent. In order to obtain concentration values, this information is then converted to a mass distribution via density estimation. To date, density estimation is performed with a non-parametric method so that output consists of gridded concentration data on grids usually defined at the start of the simulation. Here we introduce the use of Gaussian mixture models, GMM, for density estimation. The approach obviates the need for a predefined concentration grid and can significantly reduce the number of computational particles needed in the simulation. We show forecast verification of the concentrations produced using this approach with the HYSPLIT model and compare to the use of a histogram or bin-counting method and kernel density estimation. We also explore the use of the mixture model for automatic identification of features in a complex plume such as is produced by a large volcanic eruption. We conclude that use of a mixture model has potential to be very beneficial for some dispersion modeling applications.