No description available.

La surveillance de la qualité de l'air est actuellement effectuée avec des mesures de concentration et à partir d'outils de modélisation de la dispersion atmosphérique. Ces modèles numériques évaluent les concentrations des polluants avec une résolution spatio-temporelle plus fine que les mesures. Néanmoins, les estimations fournies par ces modèles sont moins précises que les mesures. Dans ce projet de recherche, nous avons étudié les approches de couplage d'échelles et d'assimilation de données pour améliorer les estimations fournies par le modèle de dispersion atmosphérique SIRANE, dédié à l'échelle urbaine. L'approche de couplage d'échelles consiste à déterminer les conditions aux limites d'une simulation à partir d'une autre simulation à plus grande échelle. Au cours de ce travail de thèse, nous avons analysé trois méthodes afin de coupler le modèle urbain SIRANE et le modèle à méso-échelle CHIMERE. Cette étude montre que ces méthodes permettent potentiellement d'estimer la qualité de l'air à l'échelle urbaine de manière plus satisfaisante que les modèles à méso-échelle (utilisés seuls). Cependant, elles n'améliorent pas forcément la modélisation des conditions aux limites d'une simulation à l'échelle urbaine et les estimations fournies par celles-ci. Cela est a priori lié au fait que les estimations fournies par le modèle CHIMERE ne sont pas suffisamment satisfaisantes sur notre cas d'étude. Il est néanmoins possible que ces méthodes améliorent les résultats à l'échelle urbaine en utilisant une simulation à l'échelle régionale de meilleure qualité. L'approche d'assimilation de données consiste à combiner les mesures et les données modélisées afin de déterminer la meilleure estimation de l'état d'un système. Durant cette thèse, nous avons étudié trois méthodes d'assimilation de données : la méthode de débiaisement, la méthode que nous avons nommée modulation de la contribution des sources et la méthode Best Linear Unbiased Estimator. Cette étude indique que ces méthodes permettent globalement d'améliorer les estimations fournies par le modèle SIRANE. L'étude de sensibilité vis-à-vis du nombre de mesures utilisées lors de l'assimilation de données indique qu'en général, plus ce nombre est élevé plus les résultats sont satisfaisants. Enfin, les résultats montrent que les performances statistiques associées à ces trois méthodes d'assimilation de données sont globalement comparables entre elles sur notre cas d'étude. / Air quality monitoring is currently carried out with concentration measurements and with atmospheric dispersion modeling tools. These numerical models evaluate pollutant concentrations with a finer spatio-temporal resolution than measurements. Nevertheless, the estimates provided by these models are less accurate than measurements. In this research project, we studied multiscale coupling and data assimilation approaches to improve the estimates provided by the SIRANE atmospheric dispersion model, dedicated to the urban scale. The multiscale coupling approach consists in determining the boundary conditions of a simulation from another simulation on a larger scale. In this thesis work, we analyzed three methods for coupling the SIRANE model with the CHIMERE mesoscale model. This study shows that these methods can potentially estimate the air quality at the urban scale more satisfactorily than the mesoscale models (used alone). However, they do not necessarily improve the modeling of the boundary conditions of a simulation at the urban scale and the estimates provided by them. This is a priori due to the fact that the estimates provided by the CHIMERE model are not sufficiently good on our case study. It is possible, however, that these methods improve the results at the urban scale by using a better simulation at the regional scale. The data assimilation approach consists of combining the measurements and the modelled data to determine the best estimate of the system state. During this thesis, we studied three data assimilation methods : the unbiased method, the method that we called source apportionment modulation, and the Best Linear Unbiased Estimator method. This study indicates that these methods generally improve the estimates provided by the SIRANE model. The sensitivity study on the number of measurements used during the data assimilation indicates that, in general, higher is this number, more satisfactory are the results. Finally, the results show that the statistical performances associated with these three data assimilation methods are globally comparable on our case study.

This work is a multi-scale effort to confront the rapidly evolving threat of biological weapons attacks through improved bioaerosol surveillance, detection, and response capabilities. The effects of bioaerosol release characteristics, transport in the atmospheric surface layer, and implications for bioaerosol sampler design and real-time detection were studied to develop risk assessment and modeling tools to enhance our ability to respond to biological weapons attacks. A simple convection-diffusion-sedimentation model was formulated and used to simulate atmospheric bioaerosol dispersion. Model predictions suggest particles smaller than 60 micrometers in aerodynamic diameter (AD) are likely to be transported several kilometers from the source. A five fold increase in effective mass collection rate, a significant bioaerosol detection advantage, is projected for samplers designed to collect particles larger than the traditional limit of 10 micrometers AD when such particles are present in the source distribution. A family of dynamically scaled wetted-wall bioaerosol sampling cyclones (WWC) was studied to provide bioaerosol sampling capability under various threat scenarios. The effects of sampling environment, i.e. air conditions, and air flow rate on liquid recovery rate and response time were systematically studied. The discovery of a critical liquid input rate parameter enabled the description of all data with self-similar relationships. Empirical correlations were then integrated into system control algorithms to maintain microfluidic liquid output rates ideally suited for advanced biological detection technologies. Autonomous ambient air sampling with an output rate of 25 microliters per minute was achieved with open-loop control. This liquid output rate corresponds to a concentration rate on the order of 2,000,000, a substantial increase with respect to other commercially available bioaerosol samplers. Modeling of the WWC was performed to investigate the underlying physics of liquid recovery. The set of conservative equations governing multiphase heat and mass transfer within the WWC were formulated and solved numerically. Approximate solutions were derived for the special cases of adiabatic and isothermal conditions. The heat and mass transfer models were then used to supplement empirical correlations. The resulting semi-empirical models offer enhanced control over liquid concentration factor and further enable the WWC to be deployed as an autonomous bioaerosol sampler. / text

Les ISDND produisent du biogaz par fermentation des déchets organiques. Le biogaz principalement composé de CH4 et CO2 représente un enjeu environnemental majeur. Cette étude propose un outil d’estimation des émissions surfaciques de biogaz d’une ISDND par modélisation de la dispersion atmosphérique d’un gaz traceur, en l’occurrence, le méthane. Les dynamiques spatiales et temporelles des concentrations en CH4 et en COV ont été suivies sur et dans l’entourage de l’ISDND d’étude (Séché Environnement) en fonction des conditions météorologiques. Les résultats montrent des faibles teneurs atmosphériques en COV et en CH4 sur le site d’étude validant une faible émission de ces espèces. Les COV mesurés sont émis par différentes sources dont la contribution est plus importante que celle de l’ISDND et aucun COV ne constitue un traceur de biogaz émis par le site. En revanche, l’ISDND apparait comme source principale du CH4 détecté. Deux méthodes sont proposées pour estimer les émissions surfaciques de méthane en utilisant un modèle de dispersion atmosphérique Gaussien ADMS, validé par comparaison des teneurs atmosphériques en méthane mesurées et modélisées et leur dynamique temporelle. La première méthode repose sur une approche inverse et la deuxième est une approche statistique par régression. Les émissions de CH4 sont estimées pour la période d’exposition de la diode laser aux émissions du site pour 4 scénarios météorologiques types identifiés par une classification hiérarchique. Les résultats valident l’influence des paramètres météorologiques, surtout de la stabilité de l’atmosphère, sur la dispersion atmosphérique et les émissions surfaciques en méthane. / Landfill sites produce biogas by degradation of biodegradable organic matter. Biogas mainly composed of CH4 and CO2 represents a major environmental challenge. This study propose a method to estimate biogas surface emissions in landfill sites using atmospheric dispersion modeling of a tracer gas, in this case, methane. The spatial and temporal dynamics of CH4 and VOC concentrations have been followed on the studied landfill site (Séché Environnement) for several weather conditions. Measurement results show low atmospheric VOC and CH4 concentrations on the studied landfill site which validates low emissions of these compounds. Detected VOC are emitted by different sources, excluding the landfill site. The contribution of these sources on VOC concentrations is more important than that of the landfill site and no VOC could be identified as tracer of biogas emitted by landfill site. However, CH4 is emitted by the landfill site, its principal source. Two methods are proposed to estimate methane surface emissions using a Gaussian atmospheric dispersion model ADMS. Gaussian model is validated by comparison of the temporal dynamics and atmospheric concentrations of methane measured on the site and those modeled. The first method is based on an inverse approach and the second one is a statistical regression approach. CH4 emissions are estimated for the exposure period of the laser diode to the site emissions and for 4 weather scenarios identified by a hierarchical classification. Results validate the influence of meteorological parameters, especially the stability of the atmosphere, on the atmospheric dispersion and methane surface emissions.

No description available.

In the event of a radioactive release in an urban environment the consequent response mustbe swift and precise. As soon as first responders have correct information, they can make anaccurate risk assessment. However, if the position, release rate and time of the radioactiverelease is unknown it is hard to know how the pollutant will spread. This thesis aims to testa model which approximates these three unknowns using weather data (wind and rain) as wellas measurement data collected at sensors placed around an urban environment. An atmospheric dispersion model based on an existing Reynolds Averaged Navier-Stokes modelis set up in two geometries of different complexity to create forward mode synthetic depositiondata and adjoint mode concentration fields resulting from a fixed dry deposition velocity andscavenging effect for wet deposition. Variations of time- and space-dependent rainfall is simu-lated. The resulting data is used in an existing optimization model, where a parameter studyis conducted regarding regularization coefficients. This thesis shows that the optimization model accurately estimates position and its approximaterelease rate of a 2D geometry of radioactive releases using a logarithmic optimization approach,and fail to do so using a linear optimization approach. The logarithmic optimization model alsoapproximately estimates position and release rate in a 3D geometry. Regularization parametersshould be within the range of 0.1 and 1.2 depending on rain. More rain requires smallerparameters and will estimate a lower release rate. Time-dependent rainfall is shown to have amajor negative effect on simulation time.iii