Le travail présenté ici a pour objet l’étude de la zone de recirculation qui se développe à l’aval d’un élargissement brusque ayant lieu dans un canal à surface libre, avec une attention particulière portée sur la longueur de recirculation. Ce travail consiste en une approche combinée expérimentale et numérique. L’analyse dimensionnelle ainsi que les travaux préliminaires antérieurs à cette thèse montrent que la longueur de recirculation adimensionnée dépend de 3 paramètres que sont : le nombre de frottement S, la hauteur d’eau adimensionnée par la taille de l’élargissement h/d et enfin le rapport géométrique de l’élargissement Rb. Cependant, faire varier dans les expériences S ou h/d sans affecter l’autre paramètre s’avère être une tâche très délicate qui a été négligée dans les études précédentes. En utilisant cette même approche, les résultats présentés ici font état d’une forme de cloche très inattendue de la courbe : L/d= f (S). Ces résultats sont en fort désaccord avec ceux de la littérature. Afin d’améliorer notre approche et de faire varier S et h/d indépendamment, une campagne de modélisation numérique 3D est menée et prouve que L/d dépend en fait des deux paramètres considérés S et h/d et que la forme de cloche résulte des influences opposées de ces deux paramètres. De plus, l’analyse de la couche de mélange qui prend place entre la zone de recirculation et l’écoulement principal, mesurée expérimentalement pour 4 écoulements à différentes valeurs de S et h/d montre que la longueur de la zone de recirculation est gouvernée par le confinement latéral dû à la paroi latérale et à la taille des cellules turbulentes advectée le long de la couche de mélange. Pour aller plus loin, les bilans de quantité de mouvement et d’énergie à l’échelle de l’écoulement dans son ensemble montrent que i) la force de cisaillement exercée le long de la couche de mélange est négligeable par rapport aux autres forces mises en jeu et ii) que la réelle signification de S est de quantifier l’intensité du frottement du fond à l’échelle de l’écoulement global intervenant dans ces bilans. Les différents régimes d’écoulement qui peuvent être rencontrés dépendent donc: i) selon la valeur du nombre de frottement S l’écoulement peut être frictionnel (S élevé) ou non-frictionnel (S faible) et ii) selon la valeur de la hauteur d’eau adimensionnelle, l’écoulement peut être confiné verticalement (faible valeur de h/d) ou non-confiné (forte valeur de h/d). Une corrélation empirique permettant d’estimer la longueur de la zone de recirculation L/d en fonction des paramètres S, Rb and h/d est finalement obtenue. Elle s’avère être en bon accord avec les calculs numériques et les mesures expérimentales. / The present research focuses on the recirculation zone developing downstream of an expansion in a shallow open-channel flow with a specific attention on its length. The work consists of combined experimental and numerical approaches. The dimensional analysis and previous studies permit to express the dimensionless recirculation length as a function of 3 parameters: the friction number S, the ratio between the water depth and the expansion step h/d and the geometrical aspect ratio Rb. Nevertheless, varying either S or h/d on the experimental set-up without affecting the other is a complicated task which was not performed by previous studies. Following this approach permitted to obtain an unexpected bell shape for the L/d=f(S) curve, differing form the literature results. In order to improve the approach and vary S and h/d independently, a 3D numerical campaign was performed and proved that L/d actually depends both on S and h/d parameters and that the bell shape is in fact the consequence of the opposite influence of both parameters. Moreover, the precise experimental analysis of themixing layer at the frontier between the main flow and the recirculation for flows with different S and h/d values showed that the recirculation length is governed by the lateral confinement due to the reattachment wall and the size of the eddies present in the mixing layer. Hence, an integral approach is adopted, using balances of momentum and of energy at the whole flow scale, showing: i) that the shear force exerted along the mixing layer is negligible compared to the other forces and ii) that the meaning of S parameter is to quantify the intensity of the bottom friction of the whole flow on these balance. The following regimes can thus be encountered: i) according to the bed friction S values, the flow can be non-frictional (small S) or frictional (large S) and ii) according to the relative water depth, the flow can be vertically unconfined (large h/d) or confined (small h/d). An empirical correlation permitting to estimate the recirculation length L/d as a function of S, Rb and h/d is finally obtained and appears to fairly fit the numerical calculations and experimental measurements.

[pt] Neste trabalho três métodos foram utilizados para estudar o escoamento turbulento em regiões de florestas. No primeiro método, a influência da vegetação no escoamento foi modelada através da adição de termos fontes nas equações de quantidade de movimento, energia cinética turbulenta e sua taxa de dissipação. No segundo, a vegetação foi considerada um meio poroso homogêneo. Finalmente, a camada do dossel foi representada por modelos 3-D de árvores, consideradas como obstáculos individuais. As equações foram resolvidas através do modelo de turbulência k −E padrão com o programa comercial FLUENT 6.2.16. As previsões dos perfis verticais da velocidade do vento médio, da intensidade da turbulência e dos tensores de Reynolds, foram comparadas com dados de experimentos de túnel de vento. Os resultados preditos dos perfis verticais da velocidade média e intensidade da turbulência, na primeira e na segunda metodologias, apresentaram boa concordância com os valores experimentais, porém, foram observadas discrepâncias nos perfis modelados do tensor de Reynolds. Entretanto, qualitativamente, a modelagem consegue capturar o comportamento físico do tensor de Reynolds no interior de florestas. Uma possível explicação para este fato, é que o modelo considera a isotropia para a viscosidade turbulenta, implicando na incapacidade de prever qualquer forte anisotropia do campo turbulento. Na terceira metodologia, as previsões dos perfis verticais de velocidade média e intensidade da turbulência apresentaram discrepâncias em relação às medições. Porém, os perfis verticais do tensor de Reynolds apresentaram boa concordância. Todos os perfis verticais da velocidade média apresentaram um ponto de inflexão na interface vegetação-atmosfera, característico de uma camada de mistura. Nas duas primeiras metodologias, este padrão foi confirmado nos perfis de tangente hiperbólica de uma camada de mistura. / [en] This work investigates different procedures in order to study the turbulent flow over the scale model of a forest region. Initially, the canopy flow was modeled by using source terms in the momentum, turbulent kinetic energy and its dissipation rate equations. After that, the forest canopy was considered a homogeneous porous medium. In the last step, the canopy boundary layer was modeled by artificial 3-D tree models. This was done by using the standard k−E turbulence model with the FLUENT commercial program. The modeled profiles of mean velocity, turbulence intensity and Reynolds stress were compared against data from wind tunnel experiments. In the two first methodologies, the model predictions of the vertical profiles of the wind speed and turbulence intensity showed good agreement with the experimental data. It was found that predictions of the Reynolds tensor were sensitive to the parameterization scheme of the standard k −E model. However, qualitatively, the model was capable of predicting the physical behavior of the Reynolds stress tensor in the canopy flow. A possible explanation for this behavior is the omission of any anisotropic eddy-viscosity effects within the k - E modelling approach. When it was considered the tree array, the model predictions for the wind speed and turbulence intensity were less satisfactory. However, it was found that the predicted results of the Reynolds stress tensor agreed well with the measured data. All the vertical profiles of the mean velocity contained an inflection point, something which is a necessary criterion for the mixing layer flow. In the tree array, the modeled results failed to the capture this behavior of the canopy flow. In the 2-D numerical simulations, it was found the characteristic hyperbolic tangent profile of a mixing layer.

Une modélisation d'ordre réduit basée sur le partitionnement des données (cluster-based reduced-order modelling ou CROM) est développée pour identifier de manière non supervisée des mécanismes d'interaction non linéaires. La connaissance de ces mécanismes permet de pronostiquer la formation d’événements souhaitables ou non. L’approche proposée adopteun point de vue probabiliste en mettant à profit la linéarité de l’équation d’évolution de probabilité qui tient cependant compte d'éventuelles actions non linéaires des actionneurs. Le cadre est appliqué à l’attracteur de Lorenz, aux données numériques de la couche de mélange, à la turbulence tridimensionnelle du sillage d’un corps non profilé, d’un train, et aux données expérimentales d’un moteur à combustion.Pour ces exemples, le CROM permettait l'identification des quasi-attracteurs par exemple les deux régimes d’écoulement de la couche de mélange ou les états bimodaux du corps Ahmed . Les transitions principales entre ces quasi-attracteurs sont caractérisées par des regroupements de données appelé « flipper cluster ». L'identification de ces « flipper cluster » peut servir pour le contrôle des écoulements en utilisant le partitionnement des données obtenues par exemple de l'évolution temporelle de la traînée ou de la portance.Un contrôle en boucle fermé basé sur la CROM est appliqué à un écoulement le long d'une rampe courbée en vue de diminuer les extensions de la zone de recirculation par rapport à la meilleure excitation périodique en boucle ouverte. L'actionneur est mis en marche en fonction des regroupements préalablement observés. Le résultat est comparé à l’ensemble des lois de contrôle définies par toutes les combinaisons possibles des « on » et « of » par les regroupements de données. Bien quele contrôle basé sur la CROM ne permet pas de réduire la zone de recirculation par rapport à la réduction maximale en boucle ouverte, 28 % de l'apport d 'énergie nécessaire et 81 % pour une loi de contrôle particulière peuvent être économisé. / A cluster-based reduced-order modeling strategy is developed for the unsupervised identification of nonlinear flow mechanisms and precursors to desirable or undesirable events. The proposed approach assumes a probabilistic viewpoint taking advantage of the linearity of the evolution equation for the probability while including nonlinear actuation dynamics.The framework is applied to the Lorenz attractor, numerical data of the spatially evolving mixing layer, the three-dimensional turbulent wake of a bluf body, of a train, and experimental data of a combustion engine.For these examples, CROM has been shown to identify quasi-attractors such as the two shedding regimes of the mixing layer or the bimodal states of the Ahmed body; main transition processes between those quasiattractors are characterized by branching regions or flipper cluster; desirable phase space regions and possible actuation mechanisms areindicated by analysis of cluster features like drag and lift forces which can be further exploited for control purposes.In particular, a CROM-based feedback control is applied to a separating flow over a smooth ramp to examine whether the recirculation area can be diminished compared to the best open-loop periodic excitation by turning the actuation on or of depending on the applicable cluster. The CROMbased control is compared to the complete set of control laws defined byall possible combinations of 'on' and 'of' for the given set of clusters.While the recirculation area cannot be further decreased compared to the best open-loop forcing, a similar size can be achieved for 28% (CROMbased control) or 81% (one particular control law) savings in the control input energy.

As aplicações aeroespaciais estão frequentemente associadas a escoamentos compressíveis com altíssimos números de Reynolds. No entanto, existem no contexto aeroespacial importantes aplicações que envolvem escoamentos compressíveis a Reynolds relativamente baixos. Entre eles se destacam o escoamento em pás de turbina a gás e ao redor de dispositivos de alta sustentação como eslates e flapes em grandes ângulos de ataque. Pode-se destacar também o processo de combustão supersônica que está intimamente ligado e é fortemente beneficiado pelo presente estudo. Nas aplicações aerodinâmicas em baixos números de Reynolds frequentemente uma parcela significativa do escoamento se apresenta no regime de transição para turbulência, ou nos estágios iniciais do escoamento turbulento. O objetivo do presente projeto é a simulação numérica direta de escoamentos compressíveis transicionais com desenvolvimento de um código para simulação em três dimensões de escoamentos alto subsônicos. O escoamento a ser estudado no projeto é a evolução linear e não linear de trens de onda e pacotes de onda em uma camada de mistura compressível. A solução das equações de Navier-Stokes é obtida através do método das diferenças finitas. As derivadas espaciais são resolvidas através de um método compacto de sexta ordem, enquanto que as derivadas temporais são resolvidas através do método de Runge-Kutta de quarta ordem. Os métodos de aproximação foram modificados para trabalhar com malhas não uniformes visando refinar a malha em pontos em que o fenômeno ocorre e, consequentemente, reduzir o custo computacional. A investigação numérica inicia-se com a análise da taxa de amplificação dos trens de ondas fortemente modulados em regime linear. Os resultados obtidos foram comparados favoravelmente com a teoria linear. Os testes foram estendidos para a análise não linear, e consequentemente, foi possível reproduzir os fenômenos clássicos de instabilidade hidrodinâmica através da evolução dos trens de ondas oblíquos. / Aerospace applications are frequently associated with compressible flows at relatively high Reynolds number. Nevertheless important applications involve compressible flows at relatively low Reynolds number in the aerospace context. Among them, the flow on gas turbine blades and high lift devices such as slats and flaps at high angle of attack are particulary important. Besides, progress in aeroespace research is dependent on developing more efficient propulsion systems. In aerodynamic applications at low Reynolds number, often a substancial portion of the flow is in the transition regime, or in the initial stages of a turbulent flow. The objective of the present study is the Direct Numerical Simulation of three-dimensional transition of compressible flows in a mixing layer. Inspired on the worked devoted to modulated waves, the current work investigates the linear and nonlinear temporal evolution of wavetrains in this phenomenon. The Navier-Stokes equations were solved with a sixth-order compact finite-difference schemes. The time integration was performed by a fourth-order Runge-Kutta scheme. Moreover, the methods to solve the spatial derivatives were modified to work with non-uniform grids. This technique was implemented with the objective to improve the resolution of the grid where the phenomenon occurs and to reduce the computational cost. The numerical investigation starts with an analysis of the growth rate of the wavetrains in linear regime to verify the numerical code. The results compared favourably with linear theory. Tests were also performed in the nonlinear regime to simulate the oblique wavetrains and it was possible to reproduce the classical hydrodynamic instability phenomena.

<p>A new theory for predicting the windprofile over a canopy has been evaluated. The theory was first presented by Harman and Finnigan (2007). The theory relies on the forming of a mixing-layer above the canopy, due to different mean wind in and above the canopy. Characteristics from both mixing-layer and Monin Obukhov similarity theory have been used to develop the governingequations that give the wind profile. The theory has been used to calculate wind profiles for sixdifferent atmospheric stabilities. In order to evaluate the theory, profiles from the theory have beencompared to measurements from Jädraås forest, Sweden. Profiles from Monin Obukhov similarity theory were also used for comparison.In general the mixing-layer theory gives better results than Monin Obukhov similarity theory. Agreement with measurements is good in neutral conditions, but fails when the atmospheric stability is altered, especially in convective conditions. This is believed to be due to the canopy lacking in thickness. The mean wind speed is systematically underestimated and this is also believed to be caused by insufficient thickness of the canopy. A correction for this behaviour is proposed. The theory gives higher values of the mean wind speed in convective conditions with the correction and the calculated values of mean wind speed are closer to the measurements.</p>

As aplicações aeroespaciais estão frequentemente associadas a escoamentos compressíveis com altíssimos números de Reynolds. No entanto, existem no contexto aeroespacial importantes aplicações que envolvem escoamentos compressíveis a Reynolds relativamente baixos. Entre eles se destacam o escoamento em pás de turbina a gás e ao redor de dispositivos de alta sustentação como eslates e flapes em grandes ângulos de ataque. Pode-se destacar também o processo de combustão supersônica que está intimamente ligado e é fortemente beneficiado pelo presente estudo. Nas aplicações aerodinâmicas em baixos números de Reynolds frequentemente uma parcela significativa do escoamento se apresenta no regime de transição para turbulência, ou nos estágios iniciais do escoamento turbulento. O objetivo do presente projeto é a simulação numérica direta de escoamentos compressíveis transicionais com desenvolvimento de um código para simulação em três dimensões de escoamentos alto subsônicos. O escoamento a ser estudado no projeto é a evolução linear e não linear de trens de onda e pacotes de onda em uma camada de mistura compressível. A solução das equações de Navier-Stokes é obtida através do método das diferenças finitas. As derivadas espaciais são resolvidas através de um método compacto de sexta ordem, enquanto que as derivadas temporais são resolvidas através do método de Runge-Kutta de quarta ordem. Os métodos de aproximação foram modificados para trabalhar com malhas não uniformes visando refinar a malha em pontos em que o fenômeno ocorre e, consequentemente, reduzir o custo computacional. A investigação numérica inicia-se com a análise da taxa de amplificação dos trens de ondas fortemente modulados em regime linear. Os resultados obtidos foram comparados favoravelmente com a teoria linear. Os testes foram estendidos para a análise não linear, e consequentemente, foi possível reproduzir os fenômenos clássicos de instabilidade hidrodinâmica através da evolução dos trens de ondas oblíquos. / Aerospace applications are frequently associated with compressible flows at relatively high Reynolds number. Nevertheless important applications involve compressible flows at relatively low Reynolds number in the aerospace context. Among them, the flow on gas turbine blades and high lift devices such as slats and flaps at high angle of attack are particulary important. Besides, progress in aeroespace research is dependent on developing more efficient propulsion systems. In aerodynamic applications at low Reynolds number, often a substancial portion of the flow is in the transition regime, or in the initial stages of a turbulent flow. The objective of the present study is the Direct Numerical Simulation of three-dimensional transition of compressible flows in a mixing layer. Inspired on the worked devoted to modulated waves, the current work investigates the linear and nonlinear temporal evolution of wavetrains in this phenomenon. The Navier-Stokes equations were solved with a sixth-order compact finite-difference schemes. The time integration was performed by a fourth-order Runge-Kutta scheme. Moreover, the methods to solve the spatial derivatives were modified to work with non-uniform grids. This technique was implemented with the objective to improve the resolution of the grid where the phenomenon occurs and to reduce the computational cost. The numerical investigation starts with an analysis of the growth rate of the wavetrains in linear regime to verify the numerical code. The results compared favourably with linear theory. Tests were also performed in the nonlinear regime to simulate the oblique wavetrains and it was possible to reproduce the classical hydrodynamic instability phenomena.

Depuis les travaux pionniers de Brown et Roshko portant sur les effets des variations de masse volumique au sein de l’écoulement de couche de mélange, plusieurs autres études tant théoriques, expérimentales ou numériques se sont attelées à étudier finement cet écoulement. Les motivations sont d’ordre pratiques (industrie de la chimie, l’aérodynamique, la combustion . . .) ou alors purement théoriques (rôle des structures cohérentes, instabilités secondaires. . .). Ces études se sont intéressées, entre autres, aux effets de compressibilité et/ou de masse volumique variable. A notre connaissance, les effets des variations de viscosité dans la configuration de couche de mélange sont peu abordés dans la littérature. L’objectif de ces travaux de recherche est l’exploration théorique et numérique de l’écoulement de couche de mélange temporelle à viscosité variable, plus particulièrement durant sa phase initiale de développement. D’un point de vu numérique, les équations de Navier-Stokes sont résolues,en formulation faiblement compressible, au moyen du solveur CHOC-WAVES, basé sur le schéma WENO. L’approche DNS est justifiée par l’absence, dans la littérature, de modèles de sous-maille capables de prendre en compte les effets de la viscosité variable. Les équations de transport des différentes grandeurs moyennes et fluctuantes en un point et en chaque échelle (bilan d’énergie cinétique) sont réécrites en formulations incompressible et à viscosité variable. Des termes supplémentaires, engendrés par les variations spatio-temporelles de la viscosité, apparaissent dans ces équations. Celles-ci sont utilisées comme outil, afin d’explorer l’écoulement de couche de mélange et d’étudier le développement de la turbulence dans un milieu hétérogène. Les rapports de viscosité simulés sont Rv = [1−18]. Les résultats numériques montrent que l’épaisseur de la zone de mélange δθ évolue plus rapidement lorsque le rapport de viscosité Rv est élevé. De même, les gradients verticaux de la vitesse longitudinale sont amplifiés par les gradients de viscosité, un gain de près de 60%, par rapport aux valeurs initiales, est observé. La production de l’énergie cinétique turbulente est également amplifiée.L’évolution temporelle des fluctuations des vitesse est accélérée, celles-ci sont augmentées de près de 120% par rapport à l’écoulement à viscosité constante. Le régime autosimilaire du tenseur de Reynolds est atteint plus rapidement par l’écoulement à viscosité variable et l’isotropie des fluctuations de vitesse est améliorée. / Since the pioneering work of Brown and Roshko on the effects of density variations within the mixed layer flow, several other theoretical, experimental and numerical studies harnessed to finely investigate this flow. The motivations are of practical order (chemical industry, aerodynamics, combustion. . .) or purely theoretical (the role of coherent structures,secondary instabilities). These studies have focused on, among others, the effects of compressibility and/or variable density. To our knowledge, the effects of viscosity variations in the mixing layer configuration are not discussed in the literature. The objective of this researchis the theoretical and numerical exploration of the variable viscosity temporal mixedlayer flow, especially during its initial phase of development. From a numerical viewpoint, the Navier-Stokes equations are solved in weakly compressible formulation, using the solver CHOC-WAVES, based on WENO scheme. The DNS approach is justified by the absence in the literature of subgrid models that account for the effects of variable viscosity. The transport equations of different mean and fluctuating quantities at a point and each scale (scale-by-scale energy budget) are rewritten in incompressible and variable-viscosity formulation. Additional terms, generated by the spatial and temporal variations of viscosity occur in these equations. These are used as a tool to explore the mixed layer flow and study the development of turbulence in a heterogeneous environment. The simulated viscosity ratios are Rv = [1 − 18]. The numerical results show that the mixing layer thickness δθ growsfaster when the viscosity ratio Rv is high. The vertical gradients of the longitudinal mean velocity are amplified by the viscosity gradients, a gain of almost 60 %, compared to initial values was observed. The production of turbulent kinetic energy is also amplified. The temporal evolution of the velocity fluctuations is accelerated, they are increased to nearly 120 % with respect to the constant viscosity flow. The self-similar regime of the Reynolds tensor is reached more quickly by the variable viscosity flow and the isotropy of the velocity fluctuations is improved.

A new theory for predicting the windprofile over a canopy has been evaluated. The theory was first presented by Harman and Finnigan (2007). The theory relies on the forming of a mixing-layer above the canopy, due to different mean wind in and above the canopy. Characteristics from both mixing-layer and Monin Obukhov similarity theory have been used to develop the governingequations that give the wind profile. The theory has been used to calculate wind profiles for sixdifferent atmospheric stabilities. In order to evaluate the theory, profiles from the theory have beencompared to measurements from Jädraås forest, Sweden. Profiles from Monin Obukhov similarity theory were also used for comparison.In general the mixing-layer theory gives better results than Monin Obukhov similarity theory. Agreement with measurements is good in neutral conditions, but fails when the atmospheric stability is altered, especially in convective conditions. This is believed to be due to the canopy lacking in thickness. The mean wind speed is systematically underestimated and this is also believed to be caused by insufficient thickness of the canopy. A correction for this behaviour is proposed. The theory gives higher values of the mean wind speed in convective conditions with the correction and the calculated values of mean wind speed are closer to the measurements.

L’objectif de cette thèse est d’étudier les caractéristiques physiques et biogéochimiques de l’habitat épipélagique (0-200 m), ses variabilités spatio-temporelles et les processus qui les gouvernent dans le Golfe de Tadjourah (Djibouti). L’analyse spectrale singulière (SSA) et la fonction empirique orthogonale (EOF) sont appliquées à deux jeux des données satellitales dérivées du radiomètre AVHRR_MetopA et des capteurs de la couleur de l’océan (Modis et Meris). Cette analyse statistique montre que les variabilités de la température de la surface de la mer (SST) et de la concentration de la chlorophylle a (CHLa) sont essentiellement expliquées par les cycles annuels et semi-annuels. Le cycle annuel de la SST montre l’alternance des eaux chaudes d’avril à octobre et des eaux froides de novembre à mars. Le cycle semi-annuel indique une légère baisse de la SST entre juillet et aout, particulièrement à l’ouest du golfe. Pour la CHLa, la variabilité est entièrement représentée par le cycle annuel. Celui-ci indique l’enrichissement des eaux du large avec un fort gradient côte-large de juillet à novembre et une tendance inverse de décembre à juin. En outre, l’analyse spectrale singulière multi-canal (M-SSA) et la fonction de corrélation croisée avec fenêtre de 120 jours, appliquées à l’ensemble des paramètres océanique (SST et CHLa) et atmosphériques (Vent, Température et humidité spécifique de l’air) révèlent que le cycle annuel de la SST est relié aux flux de chaleur à l’interface air-mer. En revanche, le refroidissement des eaux durant juillet-aout associé au pic de la CHLa, a été attribué au phénomène d’upwelling. Dans la seconde partie, afin d’élucider l’influence océanique sur la SST et la CHLa, les structures thermohalines and biogéochemiques de la couche supérieure (0-200 m) sont étudiées à l’aide des données collectées durant juillet-aout 2013, septembre 2013 et Février 2014. Les résultats montrent qu’en juillet-aout, la couche superficielle se composait d’une couche de mélange (CM) s’étendant sur environ 20-30 m de profondeur, suivie d’une thermocline localisée entre 30 and 50 m. La CM était réduite à l’ouest et au sud-est du golfe où le gradient thermique et la CHLa étaient plus élevées proche de la surface. En septembre, cette stratification persistait mais la CM était plus chaude et salée. En Février, la CM s’étendait sur environ 120 m de profondeur et la thermocline était moins prononcée. La comparaison des courants mesurés avec les courants de dérive d’ekman et les courants géostrophiques ont révélé que les structures thermohalines et biogéochimiques sont influencées par les vents de moussons du Sud-Ouest (MSO) et du Nord-Est (MNE). Les MSO qui soufflent de juin à aout, déplacent les eaux de surface du Golfe de Tadjourah vers le Golfe d’Aden et induisent la remontée des eaux profondes à l’ouest et l’intrusion par le nord-est des eaux salée de la thermocline. Celle-ci se rapproche de la surface particulièrement à l’ouest où elle se mélange avec les eaux de surface. En revanche, les vents de moussons du nord-est (MNE) prédominants de novembre à mai, emmènent les eaux froides vers le Golfe de Tadjourah. Le mélange convectif profond épaissit la CM. Cette thèse montre que les vents de moussons et leur renverse saisonnière jouent un rôle crucial dans la stratification de la colonne d’eau et que la topographie du bassin influence et module leurs effets. Durant l’été, la forme en cuvette du bassin et la pente continentale plus allongée à l’ouest favorise l’upwelling à l’ouest du golfe où les anomalies de la SST et de fortes concentrations de la CHLa ont été observées. / The objective of this thesis is to study the physical and biogeochemical characteristics of the pelagic habitat (0-200m), its spatio-temporal variability and the processes that govern them in the Gulf of Tadjourah (Djibouti).Singular spectrum analysis (SSA) and empirical orthogonal function (EOF) were applied to two data sets derived from AVHRR_MetopA radiometry and sea colour sensors (Modis and Meris). These statistical analysis shows that the variability of sea surface temperature (SST) and chlorophyll a concentration (CHLa) are mainly explained by annual and semi-annual cycles. The annual cycle of SST consists of an alternation between warmer (April to October) and cooler (November to March) surface water. The semi-annual cycle shows a slight drop of SST between July and August, particularly in the west of the gulf. For the CHLa, the variability is fully dominated by the annual cycle indicating summer enrichment of seashore water (July-November) with a strong seaward gradient which is reversed from October to May. Multichannel spectrum analysis (M-SSA) and cross correlation function analysis applied to the oceanic (SST, CHL-a) and atmospheric parameters (wind speed, air temperature and humidity) showed that the annual cycle of SST is connected to heat flux at the air-sea interface, while the summer drop of SST and its associated CHL-a increase were attributed to upwelling. In the second part, in order to elucidate this oceanic influence on SST and CHL-a, the vertical thermohaline and biogeochemical structures of the upper layer (0-200 m) were studied using high-resolution hydrographic data collected in July-August 2013, September 2013 and February 2014. During summer, the superficial layer consisted of the mixed layer (ML) extending to a depth of about 20-30 m, followed by the thermocline located between 30 and 50 m depth. The ML was thicker in the west and the southeast where the thermal gradient and chlorophyll a concentrations were particularly high. During September, this stratification persisted but the ML became warmer and saltier and the thermocline moved slightly deeper. In February, the ML extended to about 120 m, and the thermocline was less pronounced. A comparison of the directly measured currents to the wind induced Ekman currents and to geostrophic current profiles revealed that the thermohaline and the biogeochemical features were related to the southeast and northeast monsoon winds (SWM & NEM). Between June and August, the SWM drives surface water from the Gulf of Tadjourah to the Gulf of Aden and thus induces the westward intrusion of high salinity thermocline water from the Gulf of Aden. This near surface flow mixes surface waters in the extreme west of the Gulf of Tadjourah. In contrast, the NEM which blow from September to May, bring cold water toward the Gulf of Tadjourah and thickens the ML through convective mixing. Our thesis shows that the monsoon winds and their seasonal reversal play a crucial role in the stratification of the water column, but that their effects are additionally influenced by basin topography. During summer the bowl-shape of the basin and its elongated slope in the west enhance the upwelling in this area where negative sea surface temperature anomalies and high chlorophyll a concentrations were observed.

Les cavités latérales sont des zones mortes à surface libre situées sur le côté d’un écoulement fluvial ou côtier. Les vitesses caractéristiques au sein de la cavité étant beaucoup plus faibles que celles de l’écoulement, une couche de mélange se développe à l’interface entre ces deux régions. Cette couche de mélange peut alors transférer de la quantité de mouvement de l’écoulement vers la cavité et ainsi mettre en mouvement la cavité et peut aussi transférer de la masse entre les deux régions, telle une pollution venant de l’écoulement amont. L’étude de cette thèse a alors consisté à étudier les caractéristiques de la couche de mélange, qui est rendue spécifique par le fait qu’elle se développe entre deux coins géométriques formés par l’intersection entre les parois de la cavité et celles de l’écoulement principal. Nous avons alors pu identifier l’origine et l’alternance des mouvements de fluide dans la direction transverse: de la cavité vers l’écoulement et inversement. Concernant la mise en mouvement de la cavité, le choix a été fait de considérer un écoulement principal fixé et de modifier l’extension de la cavité dans la direction perpendiculaire à l’écoulement, passant ainsi d’une cavité rectangulaire alignée avec l’écoulement principal à une cavité allongée dans le sens opposé. La mesure de champ de vitesse par PIV 2D a alors montré une forte évolution de la forme de l’écoulement à mesure que la géométrie de la cavité évolue : un système avec deux cellules alignées dans le sens de l’écoulement à un système à une seule cellule, puis un système à deux cellules et enfin un système complexe 3D ont ainsi été observés pour une cavité de plus en plus allongée. Ensuite, une modification du dispositif expérimental a permis de mesurer de deux façons différentes le transport de scalaire de l’écoulement principal vers la cavité, de comprendre les processus associés à ce transfert et enfin de quantifier cette capacité de transfert pour différents écoulements principaux et différentes géométries de cavités. Nous avons notamment montré que la géométrie de la cavité a peu d’effet alors que le nombre de Reynolds et la profondeur d’eau normalisée ont un effet majeur sur cette capacité de transfert de masse entre les deux régions. / Lateral cavities are free-surface dead-zones located on the side of a fluvial or coastal main flow. As the typical velocities are much larger in the main flow than in the cavity, a mixing layer appears at the interface between both regions. This mixing layer is able to transfer between the main flow and the cavity momentum which then sets the fluid in the cavity in motion and also passive scalar, such as a pollution coming from upstream. The objective of this work was then to investigate the characteristics of the mixing layer, which specificity comes from the fact that it is constrained between the upstream and downstream geometrical corners. It was possible to observe the origin and alternation of the transversal fluid motions: from the cavity towards the main flow and conversely. Regarding the motion in the cavity, the choice was made to keep a constant main flow and to measure the 2D horizontal velocity field using PIV as the extension of the cavity increases. The flow pattern then passes from a 2-cell patterns aligned in the direction of the main flow to a single-cell pattern, then a 2-cells patterns aligned along the direction perpendicular to the main flow and finally a complex 3D pattern for the widest cavity. Then a modification of the experimental set-up permitted to investigate the passive scalar exchanges from the main stream towards the cavity. It was possible to understand the processes responsible for such transfer and to quantify the transfer capacity. The analysis dimensional revealed that in the present subcritical, smooth simplified geometry cavity, the three parameters possible responsible for the modification of the transfer capacity are the geometrical aspect ratio of the cavity, the Reynolds number of the main flow and finally the normalized water depth. It was then shown that the impact of the cavity geometry remains negligible but that the Reynolds number and the normalized water do impact this passive scalar transfer capacity.