Dans les systèmes de vidéosurveillance, les algorithmes de vision assistée par ordinateur ont joué un rôle crucial pour la détection d’événements liés à la sûreté et la sécurité publique. Par ailleurs, l’incapacité de ces systèmes à gérer plusieurs scènes de foule est une lacune bien connue. Dans cette thèse, nous avons développé des algorithmes adaptés à certaines difficultés rencontrées dans des séquences vidéo liées à des environnements de foule d’une ampleur significative comme les aéroports, les centres commerciaux, les rencontres sportives etc. Nous avons adopté différentes approches en effectuant d’abord une analyse globale du mouvement dans les régions d’intérêt de chaque image afin d’obtenir des informations sur les comportements multimodaux de la foule sous forme de structures spatio-temporelles complexes. Ces structures ont ensuite été utilisées pour détecter des événements de surveillance inhabituels au sein-même de la foule. Pour réaliser nos expériences, nous nous sommes principalement appuyés sur trois ensembles de données qui ont suscité notre réflexion. Les résultats reflètent à la fois la qualité et les défauts de ces approches. Nous avons également développé une distance pseudo-euclidienne.Pour démontrer son utilité, une méthodologie qui lui est propre a été utilisée pour la détection de plusieurs événements de surveillance standards issus de la base TRECVID2008. Certains résultats montrent la robustesse de cette méthodologie tandis que d’autres soulignent la difficulté du problème. Les principaux défis portent, entre autres, sur le flux massif de personnes, l’importance de l’occlusion, la réflexion, les ombres, les fluctuations, les variations de la taille de la cible, etc. Cependant, nos idées et nos expériences de ces problèmes d’ordre pratique ont été particulièrement utiles. De plus, cette thèse développe un algorithme permettant de suivre une cible individuelle dans le cadre de plusieurs scènes de foule. Les séquences vidéo de la base de PETS2009 Benchmark ont été prises en compte pour évaluer les performances de cet algorithme. Si on analyse ses avantages et ses inconvénients, celui-ci fait toujours preuve d’une grande exactitude et sensibilité vis-à-vis des effets de variationde la lumière, ce qui atteste de sa grande efficacité même lorsque la luminosité baisse, que la cible entre ou sort d’une zone d’ombre ou en cas de lueur soudaine. / Computer vision algorithms have played a vital role in video surveillance systems to detect surveillance events for public safety and security. Even so, a common demerit among these systems is their unfitness to handle divers crowded scenes. In this thesis, we have developed algorithms which accommodate some of the challenges encountered in videos of crowded environments (e.g., airports, malls, sporting events) to a certain degree. We have adopted approaches by first performing a global-level motion analysis within each frame’s region of interest that provides the knowledge of crowd’s multi-modal behaviors in the form of complex spatiotemporal structures. These structures are then employed in the detection of unusual surveillance events occurred in the crowds. To conduct experiments, we have heavily relied on three thought-provoking datasets. The results reflect some unique global excellences of the approaches. We have also developed a pseudo Euclidian distance. To show its usage, a methodology based on it has been employed in the detection of various usual surveillance events from theTRECVID2008. Some results report the robustness of the methodology, while the rest gives evidence of the difficulty of the problem at hand. Big challenges include, but are not limited to, massive population flow, heavy occlusion, reflection, shadow, fluctuation, varying target sizes, etc. Notwithstanding, we have got much useful insights and experience to the practical problems. In addition, the thesis explores an individual target tracking algorithm within miscellaneous crowded scenes. Video sequences from the PETS2009 Benchmark data have been used to evaluate its performance. Viewing its pros and cons, the algorithm is still highly accurate and its sensitivity to the effects of diversity in noise and lighting, which ascertains its high-quality performance on disappearances, targets moving in and out of the shadow, and flashes of light.

MBR process consists of a suspended growth biological reactor combined with a membrane unit. The widespread of this system for waste water treatment is contained by membrane fouling, which is strongly influenced by three factors: biomass characteristics, operating conditions and membrane characteristics. Fouling control techniques mainly include low-flux operation (sub-critical flux operation) and/or high-shear slug flow aeration in submerged. configuration. Based on the concept of the critical flux (Jo), the flux-step method has been developed to more fully characterise transmembrane pressure (TMP) behaviour during constant-fluxoperation. A zero rate of TMP increase was never attained during the trial, such that no critical flux, in its strictest definition, could be defined in this study for a submerged MBRs challenged with real and simulant sewage. Under similar operating conditions, Jc was obtained around 18 and 10 L.m-2.h-1 for a submerged MBR fed by real and synthetic sewage respectively. Three TMP-based parameters have been defined, all indicating the same flux value at which fouling starts to be more significant (the weak form of Jo). Results from factorial experimental designs revealed the relative effect of MLSS levels, aeration rate and membrane pore size on J, The MLSS effect on Jc was generally around double that of the aeration effect. The calculation of mean sub-critical values for the different TMP-based parameters suggest lower short-term fouling resistance for large pore sized membranes. A direct comparison between the two MBR configurations revealed a greater J, for the submerged compared to the SS MBR (22 and 11 L.m-2.h-1 respectively) under similar hydraulic conditions. The fluid hydrodynamics has been studied for both configurations, leading to an accurate calculation of shear at the membrane surface in SS MBR and to the determination of the minimum gas velocity required for Taylor bubble formation in submerged MBR (around 0.1 m.s-1). Finally, the effect of operating conditions such as process configuration, feed nature, and aeration type on biomass characteristics has been assessed and link to membrane fouling. Key words: Fouling, MBR, critical flux, process configuration, biomass characterisation.

Cette thèse contribue à l’économie de la migration. Ce domaine de recherche analyse les raisons pour lesquelles les individus migrent, et quelles sont les conséquences pour les sociétés d’origine et d’accueil. Dans ce travail, nous nous concentrons sur les déterminants de la mobilité internationale. Plus particulièrement, nous cherchons à savoir dans quelle mesure la mondialisation détermine la migration internationale. Plusieurs définitions du terme de mondialisation existent. Il s’agit d’un processus d’intégration mondiale des marchés mondiaux pouvant être caractérisé par une normalisation progressive des réglementations nationales et par un accroissement de la concurrence entre agents économiques. Ce processus résulte de la libéralisation du commerce et des flux de facteurs de production, et d’une amélioration des technologies de transport et de communication. Plus précisément, nous étudions (i) la façon dont la migration internationale interagit avec les autres flux internationaux, à savoir le commerce international et les investissements directs étrangers, et (ii) la façon dont les décisions individuelles de migrer sont affectées par l’intégration croissante des économies. / This thesis contributes to the economics of migration. This field of research analyses why people move and the consequences for origin and host societies. Our work focuses on the determinants of the international mobility of people, particularly on how globalisation forces shape international migration. Various definitions of the term globalisation exist. Here, it is defined as a global integration process of world markets that can be characterised by a progressive standardisation of national regulations and increasing competition among economic agents. This process results from trade and factor flow liberalisation, as well as from an improvement of communication and transport technologies. More precisely, we study (i) how international migration interacts with other international flows, namely trade and foreign direct investment, and (ii) how migration decisions are impacted by the growing integration of economies.

本論文主要通過衛星遙感觀測資料和海洋數值模式的方法，來研究南海海洋上層對颱風西馬侖（2006）的回應過程，以及南海水平方向的渦度熱通量的年變化過程。 / 通過衛星海表溫度資料和氣候態海洋溫度資料反演颱風西馬侖引起的混合層加深的問題。反演結果顯示，2006年11月3日，對應海表溫度降低了4.4度，混合層則由颱風前的43.2米加深了104.5米，該結果與一維混合層模型(GOTM)的類比結果一致。此外，颱風引起的海表面溫度梯度可用來計算斜壓地轉流場和渦度。負渦度顯示了反氣旋斜壓迴圈在混合層底部最強，在50米水深處地轉流速可達到0.2米每秒。2006年11月3日，颱風西馬侖在最大的海表溫度降低的附近，向西南方向轉彎，此時行進速度比較緩慢(1.7 米每秒)進而導致在亞臨界條件弗勞德數(颱風的平移速度與第一斜壓相速度的比值)為0.6，在颱風尾區因缺少慣性重力波從而促進了海表溫度冷卻和混合層加深。通過比較Argo浮標觀測資料和氣候態資料的溫度剖面，混合層加深程度估算的誤差在10米以內。 / 三維海洋數值模型ROMS 用來研究颱風期間海洋物理動力和生態回應。海表溫度類比值同衛星觀測值比對得到的相關係數高達84%以上，表明ROMS基本上可以模擬在颱風期間南海的海表面溫度變化情況。但是深入研究發現由於垂向混合強度不夠，模式結果低估了海表面溫度冷卻和混合層加深，混合層深度被低估。通過增加波致混合效應(Bv)改進KPP混合方案，可以提高海表面溫度冷卻和混合層加深的模擬精度。類比結果顯示在颱風尾區，葉綠素的大規模爆發則發生在颱風經過的1周以後。透光層葉綠素濃度由颱風前的0.1 mg m⁻³ 增加到11月9日的1.9 mg m⁻³。衛星觀測顯示在颱風尾區，葉綠素濃度在11月16日仍高達0.85 mg m⁻³。 / 高度計的海表面高度異常值與海表面溫度的衛星觀測資料可以用來計算南海渦度熱通量。南海表層渦度熱通量的年變化趨勢表明，在南海西邊的熱通量表現出北向輸送特徵，其強度甚至與黑潮延伸區域的強熱通量相當。渦熱通量在冬季最強，熱量由南海南部流入，並且從呂宋海峽流出，甚至通過臺灣海峽進入東海海域。冬季熱通量的輻合帶主要分佈在南海東部近呂宋海峽附近海域和越南東南海域，而夏季主要集中在越南東南海域。研究表明，冬季和夏季的海表面渦度熱通量對海洋上層熱收支平衡的調整有顯著影響。 / This dissertation focuses on the investigation of the upper ocean response to typhoon Cimaron (2006) and annual variations of horizontal eddy fluxes in the South China Sea (SCS) through the methods both of satellite remote sensing and numerical ocean modeling. / The mixed layer deepening induced by typhoon Cimaron is derived based on satellite observed sea surface temperature (SST) and climatological temperature profiles in the SCS. Corresponding to the SST drop of 4.4ÅC on November 3, 2006, the mixed-layer deepened by 104.5 m relative to the undisturbed depth of 43.2 m, which is consistent with the simulation results from the one-dimensional mixed-layer model (GOTM). Furthermore, baroclinic geostrophic velocity and vorticity are calculated from the surface temperature gradient caused by the typhoon. The negative vorticity, associated with the typhoon cooling, indicated an anti-cyclonic baroclinic circulation strongest at the base of the mixed-layer, and at the depth of 50 m, the geostrophic speed reached as high as 0.2 m s⁻¹. Typhoon Cimaron proceeded slowly (1.7 m s⁻¹) when it was making a southwestward turn on November 3, 2006, resulting in a subcritical condition with a Froude number (the ratio of typhoon translation speed to first baroclinic mode speed) of 0.6 around the maximum SST drop location and facilitating high SST cooling and mixed-layer deepening due to absence of inertial-gravity waves in the wake of the typhoon. Comparison of Argo buoy data with the climatological temperature suggests that the average uncertainty in the mixed-layer deepening estimation caused by the difference between Argo and climatological temperature profiles is less than 10 m. / The physical dynamic and biological responses to typhoon Cimaron are investigated through a three-dimensional ocean model, the Regional Ocean Modeling System (ROMS). The correlation between simulated sea surface temperature (SST) and the satellite observations is over 84%, which indicates ROMS can generally simulate the sea surface temperature in the South China Sea during typhoon process. However, detailed analysis shows that the ROMS model underestimates the sea surface temperature cooling and mixed layer deepening because of insufficient mixing in the modeling. The wave-induced mixing term (Bv) added into the nonlocal K-Profile Parameterization (KPP) scheme can increase the simulation accuracy of surface temperature cooling and mixed layer depth deepening in response to the typhoon forcing. The simulation results show that the blooming of phytoplankton in the wake of storm appeared one week later after typhoon’s passage. The concentration of chlorophyll is 0.1 mg m⁻³ at pre-typhoon time and increase to 1.9 mg m⁻³ on November 9. Satellite Observation indicates the concentration of chlorophyll in wake of typhoon Cimaron was also in a high value of 0.85 mg m⁻³ on November 16. / The eddy heat flux in the SCS is derived from the satellite data including the altimeter surface height anomalies and optimally interpolated sea surface temperature. The long term heat flux shows a northward heat transport on the west side of the SCS, comparable to that in other strong flux regions such as the Kuroshio extension. The eddy flux becomes the strongest in winter with the inflow flux in the south and the outflow through the Luzon, and the eddy heat can flux through the Taiwan Strait into the East China Sea. The convergence of the flux indicates that heat accumulation in the eastern SCS close to Luzon Strait in winter and also to southeast of Vietnam in winter and summer. The eddy heat flux is more significant in adjusting the ocean upper layer heat budget flux in winter and summer. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Sun, Yujuan. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2013. / Includes bibliographical references (leaves 101-111). / Abstracts also in Chinese. / Abstract --- p.ii / 摘要 --- p.v / Acknowledgements --- p.vii / Table of Contents --- p.viii / List of Tables --- p.x / List of figures --- p.xi / Chapter 1. --- Introduction --- p.1 / Chapter 1.1 --- Ocean responses to typhoons (or tropical storms and hurricanes) --- p.4 / Chapter 1.2 --- Eddy Heat Transport --- p.9 / Chapter 2. --- The one-dimension remote sensing model (the mixed-layer deepening) --- p.11 / Chapter 2.1 --- Introduction --- p.11 / Chapter 2.2 --- Data and Methodology --- p.11 / Chapter 2.3 --- Results --- p.16 / Chapter 2.4 --- One-dimensional mixed-layer model --- p.17 / Chapter 2.5 --- Discussions --- p.20 / Chapter 2.5.1 --- Horizontal baroclinic pressure gradient and vorticity --- p.20 / Chapter 2.5.2 --- Effects of subsurface temperature variation on the mixed-layer deepening --- p.22 / Chapter 2.6 --- Summary --- p.28 / Chapter 3. --- Three-dimensional numerical ocean model --- p.29 / Chapter 3.1 --- Model description --- p.29 / Chapter 3.1.1 --- Physical model --- p.30 / Chapter 3.1.2 --- Biological model --- p.34 / Chapter 3.2 --- Model setting --- p.40 / Chapter 3.2.1 --- Initial and lateral boundary conditions --- p.41 / Chapter 3.2.3 --- Bio-module setting --- p.47 / Chapter 3.3 --- Model result validation --- p.51 / Chapter 3.3.1 --- Satellite observations --- p.51 / Chapter 3.3.2 --- Validations of observations and simulations --- p.54 / Chapter 3.4 --- Model results analysis --- p.56 / Chapter 3.4.1 --- Ocean temperature --- p.56 / Chapter 3.4.2 --- Ocean current --- p.62 / Chapter 3.4.3 --- bio-results of simulation --- p.70 / Chapter 3.4.4 --- Effect of the wave-induced mixing --- p.77 / Chapter 3.5 --- summary --- p.83 / Chapter 4. --- Annual Variations of Horizontal Eddy Heat Flux in the South China Sea --- p.84 / Chapter 4.1 --- Introduction --- p.84 / Chapter 4.2 --- Methodology and data --- p.86 / Chapter 4.3 --- Results --- p.89 / Chapter 4.4 --- Summary --- p.96 / Chapter 5. --- Conclusion --- p.97 / Chapter 6. --- Future work --- p.100 / Bibliography --- p.101

De nos jours, dans des domaines tels que la sécurité et la santé, une forte demande consiste à pouvoir analyser le comportement des personnes en s'appuyant notamment sur l'analyse faciale. Dans cette thèse, nous explorons de nouvelles approches à destination de systèmes d’acquisition peu contraints. Plus spécifiquement, nous nous intéressons à l'analyse des expressions faciales en présence de variation d'intensité et de variations de pose du visage. Notre première contribution s'intéresse à la caractérisation précise des variations d'intensité des expressions faciales. Nous proposons un descripteur innovant appelé LMP qui s'appuie sur les propriétés physiques déformables du visage afin de conserver uniquement les directions principales du mouvement facial induit par les expressions. La particularité principale de notre travail est de pouvoir caractériser à la fois les micro et les macro expressions, en utilisant le même système d'analyse. Notre deuxième contribution concerne la prise en compte des variations de pose. Souvent, une étape de normalisation est employée afin d'obtenir une invariance aux transformations géométriques. Cependant, ces méthodes sont utilisées sans connaître leur impact sur les expressions faciales. Pour cela, nous proposons un système d'acquisition innovant appelé SNaP-2DFe. Ce système permet de capturer simultanément un visage dans un plan fixe et dans un plan mobile. Grâce à cela, nous fournissons une connaissance du visage à reconstruire malgré les occultations induites par les rotations de la tête. Nous montrons que les récentes méthodes de normalisation ne sont pas parfaitement adaptées pour l'analyse des expressions faciales. / Facial expression recognition has attracted great interest over the past decade in wide application areas, such as human behavior analysis, e-health and marketing. In this thesis we explore a new approach to step forward towards in-the-wild expression recognition. Special attention has been paid to encode respectively small/large facial expression amplitudes, and to analyze facial expressions in presence of varying head pose. The first challenge addressed concerns varying facial expression amplitudes. We propose an innovative motion descriptor called LMP. This descriptor takes into account mechanical facial skin deformation properties. When extracting motion information from the face, the unified approach deals with inconsistencies and noise, caused by face characteristics. The main originality of our approach is a unified approach for both micro and macro expression recognition, with the same facial recognition framework. The second challenge addressed concerns important head pose variations. In facial expression analysis, the face registration step must ensure that minimal deformation appears. Registration techniques must be used with care in presence of unconstrained head pose as facial texture transformations apply. Hence, it is valuable to estimate the impact of alignment-related induced noise on the global recognition performance. For this, we propose a new database, called SNaP-2DFe, allowing to study the impact of head pose and intra-facial occlusions on expression recognition approaches. We prove that the usage of face registration approach does not seem adequate for preserving the features encoding facial expression deformations.

L'objectif de ce travail est l'étude du procédé de pervaporation et plus particulièrement la compréhension des mécanismes de transfert de masse et de chaleur dans une membrane en polymère qualifiée de dense. Une meilleures compréhension des mécanismes permettrait de lever les verrous limitant le développement de ce procédé, comme les faibles flux de matière ainsi que l'origine et la quantité de chaleur nécessaire au transfert à travers la membrane. Pour cela, nous avons développé un dispositif expérimental qui permet de mesurer en simultané les densités de flux de matière et de chaleur. La configuration frontale statique de perméation du dispositif mis en place donne accès au profil de température du liquide d'alimentation. Ces données de température ont permis de calculer les densités de flux de chaleur engagées durant les expériences de pervaporation grâce à un calcul par méthode inverse couplé à une simulation STAR CCM+. La densité de flux de matière est mesurée par une nouvelle méthode. La nouvelle méthode utilise un capteur de pression situé dans le réservoir d'alimentation permettant de mesurer en continu la densité de flux de matière synchronisée avec la mesure des températures. Afin de simplifier au maximum les contraintes expérimentales, nous nous sommes restreints à la perméation de composés purs. La corrélation des deux flux mesurés nous a conduit à observer que la quantité de chaleur prise au fluide en amont pour pervaporer une unité de masse de liquide pur est inférieure à la quantité de chaleur nécessaire pour vaporiser ce même liquide. Elle représente 50 % de celle-ci dans le cas de l’eau et seulement 25 % dans le cas de l’éthanol. / The aim of this work is to study the pervaporation process and specifically to understand the mass and heat transport mechanisms in a dense polymeric membrane. A better understanding of these mechanisms would make it possible to improve the limiting parameters for the development of this process, such as the low mass fluxes as well as the origin and the quantity of heat required for transport through the membrane. In order to answer these questions, we have developed an experimental setup that allows simultaneous measurement of mass flux and heat flux density. The dead-end permeation of the setup developed gives access to the temperature profile of the liquid feed. These temperature data make possible the estimation of the heat flux densities engaged during the pervaporation experiments by means of an inverse computation coupled with a STAR CCM + simulation. The mass flux is measured by a new method in addition to the gravimetric method used as a reference. The new method uses a pressure sensor located in the feed tank to continuously measure the mass flux with 1Hz raw acquisition frequency synchronized with the temperature measurement. In order to simplify the experimental constraints, we applied only permeation of pure liquids. The correlation of the two fluxes (mass and heat density) measured led us to observe that the amount of heat taken to the feed side to pervaporate a unit mass of pure liquid is less than the amount of heat required to vaporize the same liquid. It represents 50% of it in the case of water and only 25% in the case of the ethanol.

Graduation date: 2000

Streamline simulators have received increased attention because of their ability to effectively handle multimillion cell detailed geologic models and large simulation models. The efficiency of streamline simulation has relied primarily on their ability to take large timesteps with fewer pressure solutions within an IMPES formulation. However, unlike conventional finite-difference simulators, no clear guidelines are currently available for the choice of timestep for pressure and velocity updates. That is why we need largely an uncontrolled approximation, either managed by engineering judgment or by potentially time-consuming timestep size sensitivity studies early in a project. This will clearly lead us to the lack of understanding of numerical stability and error estimates during the solution. This research presents a novel approach for timestep selection during streamline simulation that is based on three elements. First, we reformulate the equations to be solved by a streamline simulator to include all of the three-dimensional flux terms - both aligned with and transverse to the flow directions. These transverse flux terms are totally neglected within the existing streamline simulation formulations. Second, we propose a simple grid-based corrector algorithm to update the saturation to account for the transverse flux. Third, we provide a discrete CFL (Courant-Friedrich-Levy) formulation for the corrector step that leads to a mechanism to ensure numerical stability via the choice of a stable timestep for pressure updates. This discrete CFL formulation now provides us with the same tools for timestep control as are available within conventional reservoir simulators. We demonstrate the validity and utility of our approach using a series of numerical experiments in homogeneous and heterogeneous ¼ five-spot patterns at various mobility ratios. For these numerical experiments, we pay particular attention to favorable mobility ratio displacements, as they are known to be challenging to streamline simulation. Our results clearly demonstrate the impact of the transverse flux correction on the accuracy of the solution and on the appropriate choice of timestep, across a range of mobility ratios. The proposed approach eliminates much of the subjectivity associated with streamline simulation, and provides a basis for automatic control of pressure timestep within full field streamline applications.

Operators, especially those managing production from deepwater reservoirs, are striving to produce hydrocarbons at higher and higher rates without exposing the wells to completion failure risk. To avoid screen failures, recent studies have favored gravel pack (GP) and high rate water pack (HRWP) completions over high-permeability fracturing (HPF), known in the vernacular as a frac&pack (FP) for very high rate wells. While a properly designed GP completion may prevent sand production, it does not stop formation fines migration, and, over time, fines accumulation in the GP will lead to increasing completion skin. Although, and not always, the skin can be removed by acidizing, it is not practical to perform repeated acid treatments on deepwater wells, particularly those with subsea wellheads, and the alternative has been to subject the completion to increasingly high drawdown, accepting a high skin effect. A far better solution is to use a HPF completion. Of course the execution of a successful HPF is not a trivial exercise, and frequently, there is a steep learning curve for such a practice. This work explains the importance to HPF completions of the well trajectory through the interval to be hydraulically fractured, for production, not execution, reasons. A new model quantifies the effect of the well inclination on the connectivity between the fracture and the well via perforations. Guidelines based on the maximum target production rate, including forecasts of multiphase flow, are provided to size the HPF completion to avoid common completion failures that may result from high fluid rate and/or fines movement. Skin model will be developed for both vertical and deviated wells. Once the HPF is properly designed and executed, the operators should end up with a long term low skin good completion quality well. The well will be safely produced at the maximum flow rates, with no need for well surveillance and monitoring.

There is a need for a robust and accurate technique to measure ammonia (NH3) emissions from animal feeding operations (AFOs) to obtain emission inventories and to develop abatement strategies. Seasonal studies were conducted to measure NH3 emissions from open-lot and free-stall dairies in central Texas since summer of 2003. Ammonia emission flux (EFl) was measured using an isolation flux chambers (FC) protocol from ground level area sources (GLAS) and converted to emission factor (EF) to potentially develop source specific NH3 emission control strategies. The GLAS including open-lots, free-stall barns, separated solids, primary and secondary lagoons and milking parlor were sampled to estimate NH3 emissions. In the first study, assessment of summer and winter data from the open-lot dairy indicated that overall NH3 EFs were 11.6 ±7.1 kg NH3 year-1 head-1 for the summer and 6.2 ±3.7 kg NH3 year-1 head-1 for the winter season. The estimated annual NH3 EF was 9.4 ±5.7 kg NH3 year-1 head-1 for this open-lot dairy. The estimated NH3 emission factor for winter was nearly 47% lower than summer EF. Open-lot corrals (~63%) in summer and (~95%) in winter were the highest contributors to NH3 emissions for the open-lot dairy. In the second study, the EFs for the free-stall dairy were determined to be 11.1 ±4.9 kg NH3 year-1 head-1 for summer season and 4.7± 4.9 kg NH3 year-1 head-1 for winter season. The estimated annual NH3 EF was 8.4 ±4.9 kg NH3 year-1 head-1 for this free-stall dairy. In winter, composted manure and free-stalls contributed nearly 73% to the total NH3 emissions for the dairy. However in summer, approximately 65% of overall NH3 emissions were contributed by two lagoons at the dairy. The overall differences between winter and summer NH3 emissions from the dairies were due to ambient temperature variations and loading rates of manure on GLAS. There was spatial variation of NH3 emissions from the open-lot earthen corrals due to variable animal density within different divisions of the open-lot. This spatial variability was attributed to dispirit manure loading within these areas.