Title: Modelling of externally mixed particles in the atmosphere
Ecole des Ponts ParisTech - Coriolis building - Caquot amphitheater.
6-8 avenue Blaise Pascal, Cité Descartes,
Champs-sur-Marne FR-77455 Marne la Vallée
- M. DUPONT Sylvain, Research Director, INRA (referee)
- Mme CALMET Isabelle, Lecturer, Ecole Centrale de Nantes (referee)
- M. DROBINSKI Philippe, Research Director, Ecole Polytechnique (reviewer)
- M. PERKINS Richard, Professor, Ecole Centrale de Lyon (reviewer)
- M. LACOME Jean-Marc, Doctor, INERIS (reviewer)
- M. CARISSIMO Bertrand, Lecturer, CEREA (thesis supervisor)
An experimental program has been designed in order to study pollutants dispersion at a complex site with a focus on stable conditions, which are still challenging for numerical modelling. This experimental program is being conducted at the SIRTA site in a southern suburb of Paris and consists in measuring, in near field, the turbulence and the pollutants dispersion. The aim of this program is to characterize the fine structure of turbulence and associated dispersion through high temporal and spatial resolution measurements. Then, these measurements allow to validate and improve the performance of CFD simulation for turbulence and dispersion in a heterogeneous field. The instrumental set up includes 12 ultrasonic anemometers measuring continuously wind velocity and temperature at 10 Hz, and 6 photo-ionization detectors (PIDs) measuring gas concentration at 50 Hz during tracer tests. Intensive observations periods (IOPs) with gas releases have been performed since March 2012.
First of all, a detailed study of flow on the site is made, because it must be characterised and properly simulated before attempting to simulate the pollutants dispersion. This study is based on two years of continuous measurements and on measurements performed during IOPs. Turbulence strong anisotropy in the surface layer is characterized by calculating variances, integral length scales and power spectra of the three wind velocity components. Propagation of turbulent structures between sensors has been characterized with velocity correlations. Energy spectra show several slopes in different frequency regions. Also, data analyses show impact of terrain heterogeneity on the measurements. The forest to the north of experimental field modifies wind velocity and direction for a large northerly sector. It induces a strong directional wind shear and a wind deceleration below the forest height. Numerical simulations are carried out using the CFD code, Code_Saturne, in RANS mode with a standard k-ε closure adapted for atmospheric flows and a canopy model for the forest. These simulations are shown to reproduce correctly the characteristics of the mean flow on the measurements site, especially the impact of the forest for different wind directions, in both neutral and stable stratification. Simulation results also show the directional wind shear and the turbulent kinetic energy increase induced by the forest. A sensitivity study has been made for various values of forest density and shows that the typical features of canopy flow become more pronounced as canopy density increases.
Pollutant dispersion study is made for several IOPs. Concentration data analysis shows a consistency with previous measurements made in a near-source region where the plume scale is smaller than the large-scale turbulence eddies. Concentration fluctuations are characterized through concentration time series, histogram and statistical analysis. The internal subrange can be observed in the concentration spectra. Next, pollutant dispersion is modelled by transport equations for concentration and its variance. The mean concentrations show a good agreement with measurements in values for all the IOPs studied, except that the position of the concentration peak depends on the accuracy of simulated wind rotation below the forestheight. The concentration fluctuations obtained from simulations seem to be affected significantly by the condition at the source and the modelling of the dissipation term. A sensitivity study to the parametrisation is then presented.