Dissolved organic matter (DOM) is a central component of the global carbon cycle. Thus, small changes to the amount of DOM imported, processed and produced within lakes can have a large effect on regional carbon budgets. In addition to being a vital energy source at the base of the aquatic food web, DOM is physico-chemically reactive. However, identifying and understanding the controls of DOM processing has remained challenging due to the complex composition of DOM. DOM comprises a mixture of decomposition by-products of terrestrial origin as well as newly synthesized material from in situ production. DOM compounds form gradients of reactivity to biogeochemical processes, such as photodegradation, biodegradation, and flocculation, and they perform a suite of functions in aquatic systems. The overarching goal of this thesis was to investigate controls of DOM processing in Swedish lakes. We do this in two ways: 1) by characterizing the molecular-level composition of DOM in lakes, and 2) by investigating interactions between very labile and relatively recalcitrant DOM. The first three chapters utilize ultrahigh resolution mass spectrometry to show that the detailed chemical composition of DOM varies along a hydrology gradient, and secondarily along a temperature gradient that co-varies with agriculture and nutrients. Next, we illustrate the coherence between molecular-level characteristics and bulk optical characteristics. Together, these studies suggest that protein-like fluorescence, aliphatic compounds, and N-containing compounds are either resistant to degradation or tightly cycled in the system, and thus persist at long water residence times. The most oxidized compounds, such as vascular plant-derived polyphenolic compounds, are abundant in areas with high precipitation and are lost with increasing water residence time. Vascular plant-derived polyphenolic compounds were most strongly related to DOM with high apparent molecular weight, suggesting that hydrophobic interactions drive aggregate formation. Furthermore, the association of high molecular weight DOM with polyphenolic compounds suggests that aggregates are hotspots of reactivity in aquatic systems. Finally, we find no indication that the addition of labile organic matter enhances the biodegradation of less reactive DOM. Thus, we suggest that in freshwaters, intrinsic molecular properties, such as the basic structural features of compounds, dominate over extrinsic factors.

Depuis le succès croissant des systèmes mobiles au cours des années 1990, les nouvelles technologies sans fil ont été développées afin de répondre à la demande croissante de services multimédias de haute qualité avec des taux d'erreur les plus faibles possibles. Un moyen intéressant pour améliorer les performances et obtenir de meilleurs taux de transmission consiste à combiner l'utilisation de plusieurs diversités avec un accès de multiplexage dans le cadre des systèmes MIMO. L'utilisation de techniques de sur-échantillonnage, d'étalement et de multiplexage, et de diversités supplémentaires conduit à des signaux multidimensionnels, au niveau de la réception, qui satisfont des modèles tensoriels. Cette thèse propose une nouvelle approche tensorielle basée sur un codage spatio-temporel tensoriel (TST) pour les systèmes de communication sans fil MIMO. Les signaux reçus par plusieurs antennes forment un tenseur d'ordre quatre qui satisfait un nouveau modèle tensoriel, modèle PARATUCK-(2,4) (PT-(2,4)). Une analyse de performance est réalisée pour le système TST ainsi que pour un système spatio-temporel-fréquentiel (STF) récemment proposé dans la littérature, avec l'obtention du gain maximum de diversité dans le cas d'un canal à évanouissement plat. Un système de transmission basé sur le codage TST est proposé pour les systèmes MIMO avec plusieurs utilisateurs. Une nouvelle décomposition tensorielle est introduite, appelée PT-(N1,N). Cette thèse établit les conditions d'unicité du modèle PT-(N1,N). À partir de ces résultats, différents récepteurs semi-aveugles sont proposés pour une estimation conjointe des symboles transmis et du canal, pour les systèmes TST et STF.

Potentiel stock de carbone dans les horizons minéraux des sols, sources de nutriments et vecteurs de polluants dans les écosystèmes aquatiques, les matières organiques dissoutes (MOD), leurs origines, et leurs transferts des sols à la rivière sont encore mal connus. Cette thèse a pour objectif de caractériser la variabilité temporelle et spatiale de la quantité et de la qualité des MOD des sols et des rivières et d'en identifier les facteurs de contrôle. Ce travail s'appuie sur le bassin versant expérimental de Kervidy-Naizin (Morbihan, Observatoire de Recherche en Environnement AgrHys). Ce site de 5 km² bénéficie de 13 années de suivi journalier des concentrations en carbone organique dissous (COD) dans la rivière et hautes fréquences des paramètres hydro-climatiques (débit, niveaux piézométriques, climat). Un suivi au moins bimensuel de la quantité et de la qualité des MOD de sols hydromorphes soumis à différents usages (culture, prairie, bois) et de rivières a été réalisé sur 2 cycles hydrologiques. L'exploitation des données long-terme a permis de proposer un schéma complet et cohérent des processus hydrologiques contrôlant les exports de MOD, leur relais dans le temps (en proposant une définition objective des saisons hydrologiques) et leur importance relative variable suivant le contexte climatique de l'année. L'exportation hivernale des MOD produites dans les sols pendant l'été contrôle la dynamique intra-annuelle décroissante des concentrations en COD des eaux de rivière. L'effet compensatoire de ces mécanismes de production et d'exportation explique la relative stabilité des concentrations moyennes annuelles de la rivière. La caractérisation des MOD par spectroscopie de fluorescence 3D, couplée à la modélisation PARAFAC (parallel factor analysis) des matrices d'excitation-émission générées, a permis d'identifier des exportations de MOD d'origine agricole, faiblement aromatiques, protéiques et dérivées de l'activité microbienne. Ces MOD sont exportées saisonnièrement pendant les périodes d'inter-crue des hivers très humides, par remontée de la nappe dans les sols cultivés des versants, et ponctuellement au printemps, lors de crues suivant l'épandage d'effluents d'élevage sur les sols du bassin. L'organisation spatiale des dynamiques temporelles de la quantité et de la qualité des MOD des sols suggère un contrôle hydro-pédo-topographique des différences rapportées. Les MOD faiblement aromatiques des horizons organo-minéraux sont préférentiellement exportées depuis les sols situés dans les pentes de versant. Les MOD exportées depuis les horizons minéraux sont partiellement compensées par les apports de MOD provenant des sols situés en amont. L'instauration de conditions réductrices dans les sols de bas de versant, produit des MOD aromatiques uniquement exportées en crue. La composition et la position dans le bassin des MOD des sols agricoles les rendent donc sensibles au processus d'exportation. La question du devenir de ces MOD est posée. / The role of dissolved organic matter (DOM) as carbon storage in mineral soil horizons and the impacts of DOM on aquatic ecosystems, either as a source of nutrients, or a vector of pollutants, raise the need to understand its origin, and the mechanisms linked to its transport from soils to stream. This work aims to characterize the temporal and spatial dynamics of the amount and the quality of DOM in soil and stream water, and to identify the controlling factors. It is based on the Kervidy-Naizin catchment of 5 km² (Morbihan ; Environment Research Observatory AgrHyS). This site benefits from 13 years of daily stream dissolved organic carbon (DOC) concentrations and high frequency of hydro-climatic parameters (discharge, groundwater table depths, and meteorological variables). DOM concentration and composition of soil water under contrasted land uses (cropland, grassland, or woodland) and stream water were assessed at least fortnightly during 2 hydrological cycles. A comprehensive and consistent pattern of hydrological processes that control the DOM exports, their changes in time (by providing an objective definition of hydrological seasons), and the changes of their relative importance according to the climatic context of the year was proposed. The exports of DOM produced in the soils during summer control the intra-annual decrease of stream DOC concentrations. This suggests a compensatory mechanism whereby increased winter exports can balance the summer production of DOM resulting in a relatively uniform mean annual DOC concentration. The DOM composition of soil and stream water was determined by fluorescence spectroscopy combined with PARAFAC (parallel factor analysis) modeling built from excitation-emission matrices. Exports from agricultural sources characterized by low aromatic, protein-like, microbial-derived DOM were reported. This DOM is transported to the stream seasonally during inter-storm periods of very wet winters, by rising of the water table in arable hillslope soils, and occasionally in spring, during storm events that follow the animal manure application on catchment soils. The spatial organization of temporal dynamics of the DOM concentration and composition of soil waters suggested that they were controlled by combined effects of hydrology, soil properties and topography. DOM with fewer aromatic moieties is preferentially exported from organo-mineral horizons of hillslope soils. DOM loss from mineral soil horizons are partially balanced by DOM inputs from upper soils. The onset of reducing conditions in downslope soils produces aromatic DOM that is transported to the stream only during storm events. Hence, DOM from agricultural soils is more susceptible being transport to stream due to its composition and its location within the catchment. The question of the fate of these DOM in aquatic ecosystems is raised.

Les réservoirs hydroélectriques contribuent à la remise en suspension du mercure (Hg) inorganique terrestre et de la matière organique suite à l’inondation des terres lors de la construction d’un barrage. Le développement de zones anoxiques dû à l’augmentation de la colonne d’eau et à la dégradation de matière organique inondée est une condition propice pour la méthylation du Hg par les communautés bactériennes aquatiques, le rendant neurotoxique et bioamplifiable. Les résultats de cette étude, conduite dans le complexe de réservoirs hydroélectriques de la rivière Romaine, démontrent que le pourcentage du mercure total (THg) étant du méthylmercure (MeHg) retrouvé dans les zones inondées est en moyenne 10 fois plus élevé que dans les systèmes naturels avoisinants et dans la portion de la rivière en amont de la série de barrages. Les concentrations en MeHg demeurent toutefois relativement faibles comparativement à d’autres réservoirs au Québec et ailleurs dans le monde. Des patrons de l’amont vers l’aval ont pu être observés, où le MeHg augmente de façon concomitante au dioxyde de carbone (CO2) à travers les réservoirs, les deux ayant des valeurs plus élevées au printemps, suite à la fonte des glaces, qu’en été, suggérant qu’une production sous la glace est probable due à une anoxie partielle avant que le retrait des glaces ne revienne oxygéner la colonne d’eau. De plus, les concentrations de mercure total ainsi que les patrons des composantes de matière organique dissoute et du carbone organique dissous demeurent très peu variables entre les sites inondés et naturels. Cela suggère que les réservoirs hydroélectriques agissent plutôt comme réacteurs de transformation du mercure inorganique terrestre en méthylmercure que comme mobilisateurs de mercure nouvellement importé dans le système, et que les processus opérant des centaines de kilomètres en amont et des mois auparavant sont importants pour les dynamiques de carbone et de mercure observées dans ce système hautement connecté. / Following the flooding of soil during the construction of hydroelectric reservoirs, terrestrial inorganic mercury (Hg) and organic matter typically increase in concentrations in the aquatic environment. The anoxia developed due to the deepening of the water column and the degradation of organic matter due to aquatic microbial communities create new conditions in the system that are prone to the methylation of mercury into its neurotoxic and bioaccumulative form : methylmercury (MeHg). Results from this study, led in the hydroelectric complex of reservoirs of the Romaine River, showed that the percentage of total mercury which is methylmercury in the flooded area is on average 10 times higher than in surrounding natural systems from the watershed and in the upstream section of the river. MeHg concentrations in the Romaine River, however, remain relatively low when compared to other reservoirs in Quebec and worldwide. Upstream to downstream patterns have been observed, where MeHg increases along with carbon dioxide (CO2) throughout the series of reservoirs and where both showed higher values in June than in August, following the snowmelt and the melting of the ice cover. Concentrations were lower in the summer campaign, suggesting there was under-ice production due to the development of partial anoxia before the water column is mixed and oxygenated in spring. Moreover, the distribution of total mercury concentrations, dissolved organic carbon and dissolved organic matter components remained stable between flooded and natural sites. This suggests that hydroelectric reservoirs act as reactors for the transformation of terrestrial inorganic Hg in MeHg rather than as mobilizers of newly imported Hg and that processes occurring several kilometers upstream and months prior to sampling affect the carbon and Hg dynamics of this inter-connected system.

La matière organique dissoute (MOD) joue un rôle clé dans le fonctionnement des rivières à titre de vecteur de carbone organique et d’énergie, tout en contribuant au transport de nutriments, de métaux et de divers contaminants. Elle est principalement dégradée à travers la minéralisation photochimique et la respiration bactérienne. La vitesse à laquelle se produit cette dégradation est influencée à la fois par la composition chimique et moléculaire de la MOD et par les conditions environnementales dans lesquelles elle est transformée. Toutefois, les études évaluant simultanément ces effets intrinsèques et extrinsèques sont rares en milieu fluvial et la plupart se concentrent sur un seul des processus de dégradation. De plus, les résultats de ces expériences sont rarement transposés à une colonne d’eau entière, ce qui limite l’évaluation de la réactivité écosystémique en rivière. Durant l’été 2019, nous avons échantillonné 40 sites le long d’un transect de 207 kilomètres sur le fleuve Saint-Laurent, une grande rivière tempérée dans laquelle s’écoulent côte-à-côte deux masses d’eau contrastées en termes de propriétés chimiques et physiques. En laboratoire, nous avons évalué la bio- et la photo-réactivité de la MOD au moyen d’incubations et d'expositions sous un simulateur solaire pour ensuite transposer ces données à l’écosystème en estimant pour tous les sites échantillonnés des taux surfaciques de dégradation in situ à partir de profils lumineux et bathymétriques. Nos résultats indiquent que la dégradation in situ de la MOD est essentiellement déterminée par la dégradation biologique. Cette dernière est généralement un à deux ordres de grandeur plus importante que la dégradation photochimique. Les taux surfaciques totaux de dégradation (calculés comme la somme des taux de dégradation photochimique et biologique) vont de 36,7 à 892,1 mg C m-2 j-1. Les taux surfaciques de dégradation photochimique sont principalement influencés par la photoréactivité de la MOD. En revanche, nous avons identifié un pool relativement constant de MOD bioréactive qui semble indépendant de la concentration initiale, de la composition ou des conditions environnementales. Des taux disproportionnellement élevés de biodégradation (2,5 à 4 fois la moyenne) ont été observés pour quelques sites peu profonds et situés près d’effluents urbains ou d’îles. Ces taux sont probablement stimulés par une combinaison de concentrations élevées en nutriments et une proportion plus importante de composantes protéiniques dans la MOD. Nos travaux fournissent l’une des rares démonstration expérimentale de la dominance de la dégradation biologique dans les grandes rivières, mais révèlent que la dégradation photochimique peut être d’importance comparable dans certaines zones de faible profondeur plus fortement connectées au milieu terrestre. En somme, ces observations suggèrent l’existence de points de contrôle écosystémiques pour la réactivité de la MOD dans les grandes rivières et laissent entrevoir de nouveaux questionnements quant à leur rôle dans l’export et le transport de matière terrestre. / Dissolved organic matter (DOM) is central to rivers functioning as it is a complex mixture composed of vast amounts of organic carbon and energy, also acting as a vector for nutrients, metals and various contaminants. Both the chemical and molecular composition of the DOM and the environmental conditions in which it is processed can influence its rate of removal from the water column, mainly through photochemical mineralization and bacterial respiration. Studies evaluating these intrinsic and extrinsic aspects simultaneously, particularly on both the biological and photochemical degradation are rare, particularly in large fluvial systems. During the summer of 2019, we sampled a 207 km transect of the St. Lawrence (SLR), a large temperate river in which flows two strikingly distinct water masses in terms of origin as well as chemical and physical properties. We then assessed DOM bio- and photo-reactivity at 40 sites along the river through a series of standardized incubations and exposure to simulated sunlight, and then used water irradiance and morphometric profiles to estimate in situ areal rates of processing across the river. We found that the in situ reactivity was mostly driven by biological degradation, which was typically one to two order of magnitude higher than photochemical degradation. Total daily processing (calculated as the sum of photochemical and biological degradation) across the whole water column ranged from 36.7 to 892.1 mg C m-2 d-1. In situ photochemical degradation was largely driven by DOM photoreactivity. In contrast, we found a relatively constant baseline pool of biolabile DOM that appeared to be independent from changes in concentration and environmental conditions. Disproportionately high biodegradation rates (2.5-4x the average) were found in a few shallower sites, near effluents or islands, potentially driven by a combination of local increases in nutrient concentration and compositional changes in the proportion of protein-like DOM. We provide rare experimental evidence for the often-assumed dominance of biodegradation over photodegradation in large rivers, but also emphasize that photodegradation can locally be just as important in terrestrially connected areas of low depth. In turn, these observations hint at the presence of ecosystem control points in large and heterogenous rivers with contrasting sources and composition of DOM, opening new research perspectives into the role of large rivers in the export and retention of terrestrial matter.

CoordenaÃÃo de AperfeiÃoamento de NÃvel Superior / Em diversas aplicaÃÃes do processamento de sinais em sistemas de comunicaÃÃo sem-fio, o sinal recebido à de natureza multidimensional, possuindo uma estrutura algÃbrica multilinear. Neste contexto, a decomposiÃÃo tensorial PARAFAC tem sido utilizada em vÃrios trabalhos ao longo dos Ãltimos seis anos. Observa-se, entretanto, que decomposiÃÃes tensoriais generalizadas sÃo necessÃrias para modelar uma classe mais ampla de sistemas de comunicaÃÃo, caracterizada pela presenÃa de estruturas de transmissÃo mais complexas, por modelos de canal mais realistas, e por tÃcnicas de processamento de sinais mais eficientes no receptor. Esta tese investiga novas abordagens tensorias e suas aplicaÃÃes em modelagem de sistemas MIMO, equalizaÃÃo, separaÃÃo de sinais e estimaÃÃo paramÃtrica de canal. Inicialmente, duas novas decomposiÃÃes tensoriais (PARAFAC em blocos com restriÃÃes e CONFAC) sÃo desenvolvidas e estudadas em termos de identificabilidade. Em uma segunda parte do trabalho, novas aplicaÃÃes destas decomposiÃÃes tensoriais sÃo propostas. A decomposiÃÃo PARAFAC em blocos com restriÃÃes à aplicada, primeiramente, Âa modelagem unificada de sistemassuperamostrados, DS-CDMA e OFDM, com aplicaÃÃo em equalizaÃÃo multiusuÃria. Em seguida, esta decomposiÃÃo à utilizada na modelagem de sistemas de transmissÃo MIMO com espalhamento espaÃo-temporal e detecÃÃo conjunta. Em seguida, a decomposiÃÃo CONFAC à explorada na concepÃÃo de uma nova arquitetura generalizada de transmissÃo MIMO/CDMA que combina diversidade e multiplexagem. As propriedades de unicidade desta decomposiÃÃo permitem o uso do processamento nÃo-supervisionado no receptor, visando a reconstruÃÃo dos sinais transmitidos e a estimaÃÃo do canal. Na terceira e Ãltima parte deste trabalho, explora-se a decomposiÃÃo PARAFAC no contexto de duas aplicaÃÃes diferentes. Na primeira, uma nova estrutura de transmissÃo espaÃo-temporal-freqÃencial à proposta para sistemas MIMO multiportadora. A segunda aplicaÃÃo consiste em um novo estimador paramÃtrico para canais multipercursos. / In several signal processing applications for wireless communications, the received signal is multidimensional in nature and may exhibit a multilinear algebraic structure. In this context, the PARAFAC tensor decomposition has been the subject of several works in the past six years. However, generalized tensor decompositions are necessary for covering a wider class of wireless communication systems with more complex transmission structures, more realistic channel models and more efficient receiver signal processing. This thesis investigates tensor modeling approaches for multiple-antenna systems, channel equalization, signal separation and parametric channel estimation. New tensor decompositions, namely, the block-constrained PARAFAC and CONFAC decompositions, are developed and studied in terms of identifiability. First, the block-constrained PARAFAC decomposition is applied for a uniÂed tensor modeling of oversampled, DS-CDMA and OFDM systems with application to blind multiuser equalization. This decomposition is also used for modeling multiple-antenna (MIMO) transmission systems with block space-time spreading and blind detection, which generalizes previous tensor-based MIMO transmission models. The CONFAC decomposition is then exploited for designing new MIMO-CDMA transmission schemes combining spatial diversity and multiplexing. Blind symbol/code/channel recovery is discussed from the uniqueness properties of this decomposition. This thesis also studies new applications of third-order PARAFAC decomposition. A new space-time-frequency spreading system is proposed for multicarrier multiple-access systems, where this decomposition is used as a joint spreading and multiplexing tool at the transmitter using tridimensional spreading code with trilinear structure. Finally, we present a PARAFAC modeling approach for the parametric estimation of SIMO and MIMO multipath wireless channels with time-varying structure.

Dissolved organic matter (DOM) is the largest pool of organic carbon in aquatic systems and an important component of the global carbon cycle. Large amounts of DOM are decomposed within lakes, resulting in fluxes of CO2 and CH4 to the atmosphere. Therefore, there is a considerable interest in understanding the controls of DOM decomposition in freshwaters. There is evidence that in lakes intrinsic controls related to DOM composition are of primary importance, yet our knowledge about molecular drivers of DOM degradation is limited. This thesis addresses the link between chemical composition and reactivity of lake DOM by applying an experimental approach, molecular-level DOM characterization, and kinetic modeling of DOM decay. The first study shows that photoinduced transformations and partial removal of colored aromatic components of DOM have profound effects on DOM degradation kinetics, mediated by the shifts in the relative share of rapidly and slowly degrading DOM fractions. Two following studies estimate exponential decay coefficients for each individual molecular formula identified within bulk DOM. A continuous distribution of exponential decay coefficients is found within bulk DOM, which directly corroborates the central and previously empirically untested assumption behind the reactivity continuum model of DOM decay. Further, individual decay rates are evaluated in connection to specific molecular properties. On average, highly unsaturated and phenolic compounds appear to be more persistent than compounds with higher aromatic content (plant polyphenols and polycondensed aromatics), and aliphatic compounds demonstrate the highest decay rates. The reactivity of aromatics additionally increases with increasing nominal oxidation state of carbon. Molecular analysis further indicates that increasing reactivity of DOM after UV exposure is caused by disintegration of supramolecular complexes. Study IV shows that changes in relative proportion of terrestrial versus algal DOM control degradability of DOM through seasons. Under ice, when algal-derived DOM is maximally depleted, DOM degradation potential converges to similarly low levels, regardless of lake type (productive or humic), and bacterial respiration primarily relies on terrestrial carbon. This suggests a general pattern of baseline metabolism across boreal lakes. I conclude that DOM is a dynamic reactivity continuum and a tight link exists between DOM behavior and compositional properties.

Une large classe de systèmes physiques peut être représentée à l'aide du modèle de Volterra. Il a notamment été montré que tout système non-linéaire, invariant dans le temps et à mémoire évanouissante peut être représenté par un modèle de Volterra de mémoire et d¤ordre finis. Ce modèle est donc particulièrement attrayant pour les besoins de modélisation et d'identification de systèmes non-linéaires. Un des atouts majeurs du modèle de Volterra est la linéarité par rapport à ses paramètres, c¤est à dire les coefficients de ses noyaux. Cette caractéristique permet d'étendre à ce modèle certains résultats établis pour l'identification des modèles linéaires. Il est à noter que le modèle de Volterra peut, par ailleurs, être vu comme une extension naturelle de la notion de réponse impulsionnelle des systèmes linéaires aux systèmes non-linéaires. Toutefois, certaines limitations sont à circonvenir: un nombre de paramètres qui peut être très élevé et un mauvais conditionnement de la matrice des moments de l'entrée intervenant dans l¤estimation du modèle au sens de l¤erreur quadratique moyenne minimale (EQMM). Il est à noter que ce mauvais conditionnement est aussi à l¤origine de la lenteur de convergence des algorithmes adaptatifs de type LMS (Least Mean Squares). Cette thèse traite principalement de ces deux questions. Les solutions apportées sont essentiellement basées sur la notion d'orthogonalité. D'une part, l'orthogonalité est envisagée vis à vis de la structure du modèle en développant les noyaux de Volterra sur une base orthogonale de fonctions rationnelles. Ce développement est d'autant plus parcimonieux que la base est bien choisie. Pour ce faire, nous avons développé de nouveaux outils d'optimisation des bases de Laguerre et BFOR (Base de Fonctions Orthonormales Rationnelles) pour la représentation des noyaux de Volterra. D'autre part, l'orthogonalité est envisagée en rapport avec les signaux d'entrée. En exploitant les propriétés statistiques de l¤entrée, des bases de polynômes orthogonaux multivariables ont été construites. Les paramètres du modèle de Volterra développé sur de telles bases sont alors estimés sans aucune inversion matricielle, ce qui simplifie significativement l¤estimation paramétrique au sens EQMM. L¤orthogonalisation des signaux d¤entrée a aussi été envisagée via une procédure de Gram-Schmidt. Dans un contexte adaptatif, il en résulte une accélération de la convergence des algorithmes de type LMS sans un surcoût de calcul excessif. Certains systèmes physiques peuvent être représentés à l¤aide d¤un modèle de Volterra simplifié, à faible complexité paramétrique, tel que le modèle de Hammerstein et celui de Wiener. C¤est le cas d¤un canal de communication représentant l'accès à un réseau sans fil via une fibre optique. Nous montrons notamment que les liaisons montante et descendante de ce canal peuvent respectivement être représentées par un modèle de Wiener et par un modèle de Hammerstein. Dans le cas mono-capteur, en utilisant un précodage de la séquence d'entrée, nous développons une solution permettant de réaliser l'estimation conjointe du canal de transmission et des symboles transmis de manière semiaveugle. Il est à noter que, dans le cas de la liaison montante, une configuration multi-capteurs peut aussi être envisagée. Pour une telle configuration, grâce à un précodage spécifique de la séquence d¤entrée, nous exploitons la diversité spatiale introduite par les capteurs et la diversité temporelle de sorte à obtenir une représentation tensorielle du signal reçu. En appliquant la technique de décomposition tensorielle dite PARAFAC, nous réalisons l'estimation conjointe du canal et des symboles émis de manière aveugle. Mots clés: Modélisation, Identification, Bases orthogonales, Base de Laguerre, Base de fonctions orthonormales rationnelles, Polynômes orthogonaux, Optimisation de pôles, Réduction de complexité, Egalisation, Modèle de Volterra, Modèle de Wiener, Modèle de Hammerstein, Décomposition PARAFAC.

Optical parameters measured via absorption spectroscopy and high-resolution fluorescence spectroscopy were used to characterize dissolved organic matter (DOM) in the springshed of Kings Bay, a spring-fed estuary located on Florida's Springs Coast. Over the past 40 years, springs supplying groundwater to Kings Bay have shown an increase in nitrate concentration. The overall goal of this project was to fingerprint wells and spring sites with elevated nitrogen concentrations using CDOM optical properties and establish relationships between nutrient and optical parameters. Samples were obtained from various sites: springs, Kings Bay surface (KBS), wells, coastal waters in and at the mouth of Crystal River (Coast) and lakes and rivers (LNR), during dry and wet seasons. The relationships between the environmental parameters and traditional optical parameters which provide insight into source characteristics were analyzed. Excitation emission matrix spectroscopy (EEMS) provided information about the concentration and chemical nature of organic matter in the study area. CDOM optical properties combined with salinity clearly separated the sources of fixed nitrogen in the Bay. Northern springs with elevated dissolved inorganic nitrogen (DIN) concentration had lower salinities and showed a presence of protein peaks. CDOM concentration was negatively correlated with total nitrogen (TN) and DIN, which suggests that these are subjected to anthropogenic influences. Humic peaks dominated the composition of the southern springs. CDOM concentrations were much higher than in the northern springs and there was a positive correlation between CDOM and both TN and DIN. These findings suggest that the fixed nitrogen in the southern springs is naturally occurring organic matter and the low concentrations may partially be a result of subsurface mixing of saltwater and freshwater in the aquifer. Thus, hypothesis testing showed that there was a significant difference between northern and southern springs Hypothesis testing also showed that there is a significant and unexpected positive relationship between CDOM and salinity studying Kings Bay, which is due to the low CDOM concentration in the springs discharging fresh water. This unique dataset also determined that the intercept of the mixing line was significantly different form zero. This indicates that CDOM is present and detectable at very low concentrations. Parallel Factor Analysis (PARAFAC) was used to evaluate CDOM composition from excitation emission matrix spectra (EEMs) and five components were identified: two humic, two marine humic, and one protein-like. The marine-like components, peak M, were produced in the marine environment and in meteoric groundwater. The study found a unique groundwater marker for coastal regions. Northern Kings Bay sites were characterized by a protein-like component, which has been associated with wastewater. Additional optical and environmental parameters were used in discriminate analysis, which successfully identified the CDOM markers for both natural and anthropogenic sources of nutrients in the environment. It is vital to improve the analysis of water, nutrients, and carbon from groundwater discharge into the coastal zone. Elevated DIN concentrations in groundwater are a widespread problem in Florida and over the past 30 years many spring waters have shown an increase in DIN concentrations. Nutrient discharge into delicate coastal areas can lead to ecological concerns. Investigating CDOM and nutrient distribution together can be a beneficial tool that can help differentiate sources from riverine/lacustrine, estuarine, marine, groundwater, and sewage impacted categories.

Banco do Nordeste do Brasil / A avaliaÃÃo educacional dispÃe de mÃtodos para a obtenÃÃo de dados que podem ser Ãteis para avaliar grupos de indivÃduos (alunos, professores, administradores, tÃcnicos e outros), projetos, produtos e materiais, instituiÃÃes e sistemas educacionais, nos seus diversos nÃveis e competÃncias. No campo da educaÃÃo em engenharia, os processos avaliativos podem ajudar os gestores a tomarem decisÃes e a realizarem mudanÃas em cursos de graduaÃÃo. Esta tese investiga de forma inÃdita uma nova abordagem para a anÃlise e interpretaÃÃo de dados no campo da educaÃÃo em engenharia com Ãnfase no processo de avaliaÃÃo, levando em consideraÃÃo dois aspectos de modo integrado: a) a percepÃÃo/opiniÃo dos estudantes sobre o contexto/ambiente educacional (Learning Context - LC) e b) os resultados/rendimentos obtidos pelos mesmos discentes (Learning Outcomes - LO). Para a realizaÃÃo desta pesquisa, foram coletados dados de estudantes do curso de graduaÃÃo em Engenharia de TeleinformÃtica (ETI) do Centro de Tecnologia (CT) da Universidade Federal do Cearà (UFC). Os dados de LC foram coletados a partir da aplicaÃÃo do instrumento SEEQ (Studentâs Evaluation of Educational Quality) da metodologia SETE (Student Evaluate Teaching Effetivecness). Os dados de LO foram coletados a partir das informaÃÃes dos resultados de desempenho da aprendizagem dos mesmos discentes. Na realizaÃÃo do processamento da informaÃÃo dos dados matriciais e tensoriais obtidos, foram utilizadas duas ferramentas matemÃticas: a decomposiÃÃo bilinear, por meio da AnÃlise de Componentes Principais (Principal Component Analysis - PCA) e a decomposiÃÃo multilinear tensorial por meio da AnÃlise de Fatores Paralelos (Parallel Factor Analysis - PARAFAC). Os resultados obtidos permitem identificar caracterÃsticas comuns e semelhanÃas em componentes curriculares, tanto em termos da percepÃÃo quanto do desempenho dos estudantes. Os modelos PCA e PARAFAC tambÃm demonstraram um potencial significativo para extrair informaÃÃes de dados relacionados com variÃveis latentes em contextos educativos. / Educational evaluation provides methods to obtain data that can be useful for evaluating groups of individuals (students, teachers, administrators, technicians and others), projects, products and materials, educational institutions and systems at different levels and skills. In engineering education, evaluation processes can help managers to make decisions and changes in undergraduate courses. This thesis investigates in unprecedented way a new approach to the analysis and interpretation of data in the field of engineering education with emphasis in the evaluation process, taking into account two aspects in an integrated manner: a) perception / opinion of students about the context / educational environment (Learning Context - LC) and b) the results / income earned by the same students (Learning outcomes - LO). For this research, we collected data related to undergraduate students in Teleinformatics Engineering (TEI), at Technology Center (CT) of the Federal University of Cearà (UFC). LC data were collected from the application of SEEQ (Studentâs Evaluation of Educational Quality) instrument of SETE (Student Teaching Evaluate Effetivecness) methodology. The LO data was collected from the information of the performance of the studentsâ learning outcomes. Carrying out the information processing of the obtained tensor and matrix data, we have used two mathematical tools: the bilinear decomposition, called Principal Component Analysis - PCA decomposition and the multilinear tensor decomposition by Parallel Factor Analysis - PARAFAC. The results allow us to identify common features and similarities in curriculum components, both in terms of perception as the performance of students. The PCA and PARAFAC models also showed significant potential to extract data information related to latent variables in educational settings.