Modelling the Concentration Distributions of Aerosol Puffs Using Artificial Neural Networks

Abstract

A neural network model was developed to predict the short-term (<150 s) concentration distributions of aerosols released from point sources over very short time periods (approximately 2 s). The model was based on data from field experiments covering a wide range of meteorological conditions. The study focused on relative dispersion about the puff centroid, with puff/cloud meander and large-scale gusts not being considered. The artificial neural network (ANN) model included explicitly a number of meteorological and turbulence parameters, and was compared with predictions from two Gaussian-based puff models to the measurements of four independent trials representing different stability conditions. The performance of the neural network model was comparable (in stable conditions) or better (in unstable and neutral conditions) than these two models when high concentration predictions were considered. Simulations of concentration distributions under different stability conditions were also generated using the developed neural network model, with the result that Gaussian distributions provided good descriptors for puff dispersion in the downwind and crosswind directions, and for particles close to the centroid in the vertical when dealing with short dispersion times.

Artificial Neural Networks for Radiation Dose Prediction in Nuclear Emergencies-Preliminary Investigations

“…Cao et al [6] applied ANN in prediction of short-term concentration distributions of aerosols released from point sources. He concluded that the performance of the neural network model was comparable or better than predictions from two Gaussian-based puff models.…”

Origin of atmospheric aerosols at the Pierre Auger Observatory using studies of air mass trajectories in South America

“…The HYSPLIT model, developed by the Air Resources Laboratory, NOAA, 6 is a complete system designed to support a wide range of simulations related to regional or long-range transport and dispersion of airborne particles (Martet et al, 2009). It is possible to compute simple trajectories for complex dispersion and deposition simulations using either puff or particle approaches within a Lagrangian framework (Cao et al, 2010;De Vito et al, 2009). In this work, HYSPLIT will be used to get backward/forward trajectories by tracking air masses backward/forward in time.…”

Dispersion Coefficients for Gaussian Puff Models

“…The data used for ANN and multiple non-linear regression models of dispersion coefficients were the same as those we used for the direct ANN modelling of concentration dispersion (Cao et al 2010). Briefly, data were collected during week-long trials conducted at Valcartier, Quebec, in all four seasons, on a 2 km 2 level plain.…”

“…The choice of the release location depended on the wind direction at the time. Details of the meteorological parameters measured, and the instrumentation used, can be found in Cao et al (2010).…”

“…To establish statistical significance, however, in the simulation of the dispersion of particles, a large number of particles (of the order of 10 5 , according to De Hann and Rotach 1998;De Haan 1999) are typically required, with the concomitant requirement for computing resources. The concentration distribution of a transient (instantaneously released) puff, within a short time after release, can be approximated by a Gaussian shape (DeVito et al 2009;Cao et al 2010). Compared with LS particle models, Gaussian puff models are much simpler to use, and in operation take far less computer time.…”