Improvement Of Lagrangian Stochastic Aerosol Modeland Simulation Of Non-gaussian Turbulence

Particle dispersion in the air is of increasing concern to gene flow and PM10(particle diameter10μm) for management of natural resources and air quality.Modeling aerosol dispersion focused on the particle trajectories within and abovecanopy. As macroscopic observation proved wind turbulence is a deterministic processin statistics, researchers succeeded in simulating Gaussian turbulence by methods thatwere classified in Eulerian and Lagrangian frameworks. This dissertation aims toexplore the simulation of non-Gaussian turbulence in the air on the basis of one existingLagrangian stochastic (LS) aerosol model.Several publications have dealt with both experimental and modeling aspects ofnon-Gaussian turbulence. Even though such models showed less capability ofthatsimulation than Gaussian one, there summarized the well-mixed conditioncriteria bywhich the validity of novel models are verified, where coefficients are selected formaintaining the stable distribution of fluctuation velocities.Modeling approaches of this study provided solutions to make PDFs of simulated3-D turbulence congruity with that of wind data that was collected from fields atUniversity of Illinois at Urbana-Champaign (UIUC)of United Statesby10Hzanemometer, and was transformed into the turbulence with30-min scale obeyingprevious experience of observations and experiments.Referred LS model was advantageous in statistical features of mean and variance,and had behavior of history effects in friction velocity using Markov chain. To addsuitable skewness and kurtosis, this thesis introduced a Pearson IV random term andincorporated it into the dynamic random-walk model. Experiments calibrated relatedparameters and made sure the accuracy, stability and computational efficiency of thatterm for fitting the well-mixed condition criteria.The major improvements over traditional LS models are that on average simulatedturbulent data featured in better accuracy of skewness and kurtosis, while mean,variance and friction velocities of that data maintained same performance as before,which indicated closerpattern similarity between simulated and measured windturbulence above canopy.