Este estudo teve como objetivo avaliar a partida de reator anaeróbio de fluxo ascendente e manta de lodo (UASB), em escala real, no tratamento de esgoto sanitário, sob condições hidráulicas desfavoráveis. O reator UASB faz parte do sistema de tratamento da ETE Água Vermelha, e é seguido de um biofiltro aerado submerso e de uma unidade de desinfecção por ultravioleta. Foi utilizado metade do volume do reator (117 \'M POT.3\'), possibilitando menor tempo de detenção hidráulica (TDH). A pesquisa foi dividida em duas fases, uma sem a adição de inóculo (fase I) e outra com inoculação (fase II). Na fase I, foram aplicados vazão de 480 \'M POT.3\'/dia e carga orgânica volumétrica (COV) de 1,74 kgDQO/\'M POT.3\'.dia com TDH de 6 horas. Para a fase II, estes valores foram de 600 \'M POT.3\'/dia, 2,88 kgDQO/\'M POT.3\'.dia e 5 horas. Os resultados mostraram que as excessivas velocidades ascensionais não permitiram que o reator tivesse desempenho satisfatório. Ocorreu remoção de matéria orgânica e sólidos suspensos apenas na fase II, com exceção dos sólidos suspensos voláteis, que apresentaram remoção nas duas fases. Os lançamentos indevidos de lodo de fossas à que a ETE está submetida, provocaram aumento de sólidos dissolvidos e alcalinidade no efluente. Os leitos de lodo formados nas fases I e II possuíam concentrações semelhantes de sólidos, porém o segundo apresentou maior eficiência de tratamento. Na fase I a remoção de matéria orgânica foi pequena e não ocorreu a formação da manta de lodo. Na fase II ocorreu o provável início da estabilização do processo, aproximadamente 60 dias após a inoculação, a partir de quando a remoção de matéria orgânica dissolvida foi considerada constante e foi observada a formação da manta de lodo. O descarte de 8,5 \'M POT.3\' de lodo pode ser feito a cada 30 dias. / The objective of this study was to evaluate the start-up of a full-scale upflow anaerobic sludge blanket (UASB) reactor, treating domestic sewage, under unfavorable hydraulic conditions. The UASB reactor is part of the wastewater treatment plant of Água Vermelha (São Carlos, SP, Brazil), and it is followed by a submerged aerated biofilter and an ultraviolet radiation disinfection unit. Half of the reactor volume was used (117 \'M POT.3\'), making it possible to work with a low hydraulic retention time (HRT). The research was divided in two phases, one without reactor seeding (phase I), and another when the reactor was seeded (phase II). At phase I, a flow rate of 480 \'M POT.3\'/d and a volumetric loading rate of 1,74 kgDQO/\'M POT.3\'.d were applied, with 6 hours of HRT. At phase II, this values were 600 \'M POT.3\'/d; 2,88 kgDQO/\'M POT.3\'d and 5 hours. The results showed that excessive upflow velocities prevented satisfactory performance of the reactor. Removal efficiencies of COD, BOD and TSS were obtained only in phase II, except for VSS, which showed removal in both stages. Unauthorized inputs of septic tanks sludge caused increase in dissolved solids and alkalinity in the effluent. The sludge beds obtained in phases I and II had similar solids concentration; however, the second presented better treatment efficiency. At phase I, organic matter removal efficiency was low and the blanket sludge was not observed. At phase II, the process likely stabilized 60 days after inoculation (operations\' day 240). From then on, dissolved organic matter digestion was constant and the sludge blanket was observed. The removal of the excess sludge can be done every other 30 days.

Este estudo teve como objetivo avaliar a partida de reator anaeróbio de fluxo ascendente e manta de lodo (UASB), em escala real, no tratamento de esgoto sanitário, sob condições hidráulicas desfavoráveis. O reator UASB faz parte do sistema de tratamento da ETE Água Vermelha, e é seguido de um biofiltro aerado submerso e de uma unidade de desinfecção por ultravioleta. Foi utilizado metade do volume do reator (117 \'M POT.3\'), possibilitando menor tempo de detenção hidráulica (TDH). A pesquisa foi dividida em duas fases, uma sem a adição de inóculo (fase I) e outra com inoculação (fase II). Na fase I, foram aplicados vazão de 480 \'M POT.3\'/dia e carga orgânica volumétrica (COV) de 1,74 kgDQO/\'M POT.3\'.dia com TDH de 6 horas. Para a fase II, estes valores foram de 600 \'M POT.3\'/dia, 2,88 kgDQO/\'M POT.3\'.dia e 5 horas. Os resultados mostraram que as excessivas velocidades ascensionais não permitiram que o reator tivesse desempenho satisfatório. Ocorreu remoção de matéria orgânica e sólidos suspensos apenas na fase II, com exceção dos sólidos suspensos voláteis, que apresentaram remoção nas duas fases. Os lançamentos indevidos de lodo de fossas à que a ETE está submetida, provocaram aumento de sólidos dissolvidos e alcalinidade no efluente. Os leitos de lodo formados nas fases I e II possuíam concentrações semelhantes de sólidos, porém o segundo apresentou maior eficiência de tratamento. Na fase I a remoção de matéria orgânica foi pequena e não ocorreu a formação da manta de lodo. Na fase II ocorreu o provável início da estabilização do processo, aproximadamente 60 dias após a inoculação, a partir de quando a remoção de matéria orgânica dissolvida foi considerada constante e foi observada a formação da manta de lodo. O descarte de 8,5 \'M POT.3\' de lodo pode ser feito a cada 30 dias. / The objective of this study was to evaluate the start-up of a full-scale upflow anaerobic sludge blanket (UASB) reactor, treating domestic sewage, under unfavorable hydraulic conditions. The UASB reactor is part of the wastewater treatment plant of Água Vermelha (São Carlos, SP, Brazil), and it is followed by a submerged aerated biofilter and an ultraviolet radiation disinfection unit. Half of the reactor volume was used (117 \'M POT.3\'), making it possible to work with a low hydraulic retention time (HRT). The research was divided in two phases, one without reactor seeding (phase I), and another when the reactor was seeded (phase II). At phase I, a flow rate of 480 \'M POT.3\'/d and a volumetric loading rate of 1,74 kgDQO/\'M POT.3\'.d were applied, with 6 hours of HRT. At phase II, this values were 600 \'M POT.3\'/d; 2,88 kgDQO/\'M POT.3\'d and 5 hours. The results showed that excessive upflow velocities prevented satisfactory performance of the reactor. Removal efficiencies of COD, BOD and TSS were obtained only in phase II, except for VSS, which showed removal in both stages. Unauthorized inputs of septic tanks sludge caused increase in dissolved solids and alkalinity in the effluent. The sludge beds obtained in phases I and II had similar solids concentration; however, the second presented better treatment efficiency. At phase I, organic matter removal efficiency was low and the blanket sludge was not observed. At phase II, the process likely stabilized 60 days after inoculation (operations\' day 240). From then on, dissolved organic matter digestion was constant and the sludge blanket was observed. The removal of the excess sludge can be done every other 30 days.

French vertical-flow constructed wetlands (VFCW) directly treating raw wastewater are known to perform well on for SS, COD and nitrification. They are also known to robustly cope with hydraulic overloads during rainfall events. Although numerous systems have been installed in areas equipped with a combined sewer, the limits of stormwater acceptance remain ill-defined and need to be improved. Looking at the various VFCW designs and usages reported in the literature, it is difficult to draw any consensus on their hydraulic limits. Consequently, designing VFCW to accept hydraulic overloads is a complex task, as local context strongly impacts inlet flows produced during rainfall events. Dynamic models appear a requisite for filter design in such cases. Numerical CW models have essentially focused on horizontal flow, with few attempting to study VFCW dynamics which are more commonly tackled via mechanistic models. Although mechanistic models are powerful tools for describing processes within the VFCW, they are generally too complicated to be readily used by designers. The choice between detailed description and easy handling will depend on the modelling aims. If the aim is a global design tool, simplified models offer a good alternative. However, the simplified models geared to studying VFCW dynamics are extremely reduced. They are easy-handling for design and well-adapted to specific purposes (combined sewer overflow -CSO- treatment) but not necessarily to VFCW treating combined sewer wastewater, where long-term infiltration rates vary significantly. Consequently, this PhD thesis work focused on developing a simplified hydraulic model of VFCW to guide designers through the process of adapting VFCW systems to treat domestic wastewater in both dry and rain events. The simplified model makes it possible to link (i) hydraulics, by simulation of ponding time variations, (ii) biological performances, by establishing "dysfunction alerts" based on treatment performance assessment and variations in online N forms effluent from the young VFCW. These "dysfunction alerts" plot the maximal hydraulic load that a filter can accept without compromising its biological activity. The simplified model was used to model long-term hydraulics in the VFCW (i) to analyze the impact of local context and filter design on hydraulic overload acceptance (using "dysfunction alerts" and bypass discharges) and (ii) to propose VFCW designs for accepting hydraulic overload in different contexts. The modelling demonstrates that VFCW can limit days with bypass discharges to less than 20 times per year without jeopardizing filter performances. Moreover, the most problematic scenario on stormwater treatment remains a watershed with high imperviousness coefficient and low slope under a Bretagne-type climate, demonstrating that the filter is more sensitive to periodicity and duration than to intensity of rainfall events.

Les filtres plantés de roseaux à écoulement vertical (FPR-V) pour le traitement des eaux usées domestiques sont relativement bien développés en France et permettent de réaliser un traitement poussé de la matière carbonée et la nitrification. La robustesse de cette filière réside également dans sa capacité à accepter des surcharges hydrauliques en temps de pluie. Cependant ces limites d’acceptation hydraulique ne sont pas bien définies et demandent à être optimisées. La conception des FPR-V pour accepter des surcharges hydrauliques est un travail complexe car le contexte local affecte fortement les débits d’eau entrants dans la station lors d’un événement pluvieux. Dans de tels cas, la conception de filtres demande l’utilisation de modèles dynamiques. Ces modèles s’appuient le plus souvent sur une approche mécanistique et sont à l'origine conçus et utilisés pour la recherche. Cependant, si ces modèles sont des outils puissants pour décrire en détail la dynamique du FPR-V, ils sont généralement trop complexes à manipuler pour des utilisateurs non experts. Choisir entre une description détaillée et une manipulation facile dépendra des objectifs de modélisation. Quand des objectifs de dimensionnement global sont visés par les concepteurs, l'utilisation de modèles simplifiés semble une bonne alternative. Les modèles simplifiés sont faciles à utiliser pour la conception de FPR-V mais ils sont peu nombreux. Cependant ils sont adaptés à des configurations spécifiques liées au traitement des surverses de déversoirs d’orage, ce qui n’est pas le cas pour les FPR-V traitant des eaux usées domestiques. En effet, pour ce type de FPR-V les vitesses d’infiltration du système varient considérablement. En conséquence, ce travail de thèse a pour objectif de développer un modèle hydraulique simplifié des FPR-V qui permettra de guider les concepteurs dans le processus d’adaptation des FPR-V pour traiter les eaux usées domestiques des périodes de temps sec et de temps de pluie. Le modèle simplifié permet de relier (i) l’hydraulique du filtre, en simulant le temps de noyage à la surface, et (ii) les performances biologiques, en établissant des «alertes de dysfonctionnement », basées sur l'évaluation des performances épuratoires et sur les variations des formes azotées à la sortie du filtre mesurées en continu. Les «alertes de dysfonctionnement» représentent la charge hydraulique maximale qu'un filtre peut accepter sans compromettre son activité biologique. Le modèle simplifié a été utilisé pour la modélisation hydraulique du FPR-V à long terme (i) pour analyser l’impact du contexte local et la conception du filtre dans l’acceptation de surcharges hydrauliques. Grâce à cet analyse, (ii) on peut proposer les dimensionnements de FPR-V qui arrivent à mieux gérer l’acceptation des surcharges hydrauliques. / French vertical-flow constructed wetlands (VFCW) directly treating raw wastewater are known to perform well on for SS, COD and nitrification. They are also known to robustly cope with hydraulic overloads during rainfall events. Although numerous systems have been installed in areas equipped with a combined sewer, the limits of stormwater acceptance remain ill-defined and need to be improved. Looking at the various VFCW designs and usages reported in the literature, it is difficult to draw any consensus on their hydraulic limits. Consequently, designing VFCW to accept hydraulic overloads is a complex task, as local context strongly impacts inlet flows produced during rainfall events. Dynamic models appear a requisite for filter design in such cases. Numerical CW models have essentially focused on horizontal flow, with few attempting to study VFCW dynamics which are more commonly tackled via mechanistic models. Although mechanistic models are powerful tools for describing processes within the VFCW, they are generally too complicated to be readily used by designers. The choice between detailed description and easy handling will depend on the modelling aims. If the aim is a global design tool, simplified models offer a good alternative. However, the simplified models geared to studying VFCW dynamics are extremely reduced. They are easy-handling for design and well-adapted to specific purposes (combined sewer overflow -CSO- treatment) but not necessarily to VFCW treating combined sewer wastewater, where long-term infiltration rates vary significantly. Consequently, this PhD thesis work focused on developing a simplified hydraulic model of VFCW to guide designers through the process of adapting VFCW systems to treat domestic wastewater in both dry and rain events. The simplified model makes it possible to link (i) hydraulics, by simulation of ponding time variations, (ii) biological performances, by establishing “dysfunction alerts” based on treatment performance assessment and variations in online N forms effluent from the young VFCW. These “dysfunction alerts” plot the maximal hydraulic load that a filter can accept without compromising its biological activity. The simplified model was used to model long-term hydraulics in the VFCW (i) to analyze the impact of local context and filter design on hydraulic overload acceptance (using “dysfunction alerts” and bypass discharges) and (ii) to propose VFCW designs for accepting hydraulic overload in different contexts. The modelling demonstrates that VFCW can limit days with bypass discharges to less than 20 times per year without jeopardizing filter performances. Moreover, the most problematic scenario on stormwater treatment remains a watershed with high imperviousness coefficient and low slope under a Bretagne-type climate, demonstrating that the filter is more sensitive to periodicity and duration than to intensity of rainfall events.