The research described in the thesis details the investigation of build-up and wash-off process kinetics of Polycyclic Aromatic Hydrocarbons (PAHs) and heavy metals in urban areas. It also discusses the design and development of a rainfall simulator as an important research tool to ensure homogeneity and reduce the large number of variables that are usually inherent to urban water quality research. The rainfall simulator was used to collect runoff samples from three study areas, each with different land uses. The study areas consisted of sites with typical residential, industrial and commercial characteristics in the region. Build-up and wash-off samples were collected at each of the three sites. The collected samples were analysed for a number of chemical and physico-chemical parameters. In addition to this, eight heavy metal elements and 16 priority listed PAHs were analysed in five different particle size fractions of the build-up and wash-off samples. The data generated from the testing of the samples were evaluated using multivariate analysis, which reduced the complexity involved in determining the relative importance of a single parameter in urban water quality. Consequently, variables and processes influencing loadings and concentrations of PAHs and heavy metals in urban stormwater runoff from paved surfaces at any given time were identified and quantified using Principal Component Analysis (PCA). Furthermore, the process kinetics found were validated using a multivariate modelling approach and Partial Least Square (PLS) regression, which confirmed the transferability of chemical processes in urban water quality. Fine particles were dominant in both the build-up and wash-off samples from the three sites. This was mirrored in the heavy metal and PAH concentrations at the three sites, which were significantly higher in particles between 0.45-75μm than in any other fraction. Thus, the larger surface area and electrostatic charge of fine particles were favourable in sorbing PAHs and heavy metals. However, factors such as soil composition, total organic carbon (TOC), the presence of Fe and Mn-oxides and pH of the stormwater were all found to be important in partitioning of the metals and PAHs into different fractions. Additionally, PAHs were consistently found in concentrations above their aqueous solubility, which was attributed to colloidal organic particles being able to increase the dissolved fraction of PAHs. Hence, chemical and physico-chemical parameters played a significant role in the distribution of PAHs and heavy metals in urban stormwater. More importantly, the research showed the wide range of factors that distribute metals and PAHs in an urban environment. Furthermore, it indicated the need for monitoring these parameters in urban areas to ensure that urban stormwater management measures are effective in improving water quality. The build-up and wash-off process kinetics identified using PCA at the respective land uses were predicted using PLS and it was found that the transferability of the governing processes were high even though the PAHs and metal concentrations and loads were highly influenced by the source strength at each site. The increased transferability of fundamental concepts in urban water quality could have significant implications in urban stormwater management. This is primarily attributed to common urban water quality mitigation strategies relying on studies based on physical concepts and processes derived from water quantity studies, which are difficult to transfer between catchments. Hence, a more holistic approach incorporating chemical processes compared to the current piecemeal solutions could significantly improve the protection of key environmental values in a region. Furthermore, urban water quantity mitigation measures are generally designed to reduce the impacts of high-flow events. This research suggests that fairly frequent occurring rainfall events, such as 1-year design rainfall events, could carry significant heavy metal and PAH concentrations in both particulate and dissolved fractions. Hence, structural measures, designed to decrease quantity and quality impact on receiving waters during 10 or 20-year Average Recurrence Interval (ARI) events could be inefficient in removing the majority of PAHs and heavy metals being washed off during more frequent events. The understanding of physical and chemical processes in urban stormwater management could potentially lead to significant improvements in pollutant removal techniques which in turn could lead to significant socio-economic advantages. This project can serve as a baseline study for urban water quality investigations in terms of adopting new methodology and data analysis.

Accurate and reliable estimations are the most important factors for the development of efficient stormwater pollutant mitigation strategies. Modelling is the primary tool used for such estimations. The general architecture of typical modelling approaches is to replicate pollutant processes along with hydrologic processes on catchment surfaces. However, due to the lack of understanding of these pollutant processes and the underlying physical parameters, the estimations are subjected to gross errors. Furthermore, the essential requirement of model calibration leads to significant data and resource requirements. This underlines the necessity for simplified and robust stormwater pollutant estimation procedures. The research described in this thesis primarily details the extensive knowledge developed on pollutant build-up and wash-off processes. Knowledge on both build-up and wash-off were generated by in-depth field investigations conducted on residential road and roof surfaces. Additionally, the research describes the use of a rainfall simulator as a tool in urban water quality research. The rainfall simulator was used to collect runoff samples from small-plot surfaces. The use of a rainfall simulator reduced the number of variables which are common to pollutant wash-off. Pollutant build-up on road and roof surfaces was found to be rapid during the initial time period and the rate reduced when the antecedent dry days increase becoming asymptote to a constant value. However, build-up on roofs was gradual when compared to road surfaces where the build-up on the first two days was 66% of the total build-up. Though the variations were different, it was possible to develop a common replication equation in the form of a power function for build-up for the two surface types with a as a multiplication coefficient and b as a power coefficient. However, the values for the two build-up equation coefficients, a, and b were different in each case. It was understood that the power coefficient b varies only with the surface type. The multiplication coefficient varies with a range of parameters including land-use and traffic volume. Additionally, the build-up observed on road surfaces was highly dynamic. It was found that pollutant re-distribution occurs with finer particles being removed from the surface thus allowing coarser particles to build up. This process results in changes to the particle size composition of build-up. However, little evidence was noted of re-distribution of pollutants on roof surfaces. Furthermore, the particulate pollutants in both road and roof surfaces were high in adsorption capacity. More than 50% of the road and more than 60% of the roof surface particulates were finer than 100 μm which increases the capacity to adsorb other pollutants such as heavy metals and hydrocarbons. In addition, the samples contained a significant amount of DOC which would enhance the solubility of other pollutants. The wash-off investigations on road and roof surfaces showed a high concentration of solid pollutants during the initial part of events. This confirmed the occurrence of the 'first flush' phenomenon. The observed wash-off patterns for road and roof surfaces were able to be mathematically replicated using an exponential equation. The exponential equation proposed is a modified version of an equation proposed in past research. The modification was primarily in terms of an additional parameter referred to as the 'capacity factor' (CF). CF defines the rainfall's ability to mobilise solid pollutants from a given surface. It was noted that CF varies with rainfall intensity, particle size distribution and surface characteristics. Additional to the mathematical replication of wash-off, analysis further focused on understanding the physical processes governing wash-off. For this, both particle size distribution and physicochemical parameters of wash-off pollutants were analysed. It was noted that there is little variation in the particle size distribution of particulates in wash-off with rainfall intensity and duration. This suggested that particle size is not an influential parameter in wash-off. It is hypothesised that the particulate density and adhesion to road surfaces are the primary criteria that govern wash-off. Additionally, significantly high pollutant contribution from roof surfaces was noted. This justifies the significance of roof surfaces as an urban pollutant source particularly in the case of first flush. This dissertation further describes a procedure to translate the knowledge created on pollutant build-up and wash-off processes using small-plots to urban catchment scale. This leads to a simple and robust urban water quality estimation tool. Due to its basic architecture, the estimation tool is referred to as a 'translation procedure'. It is designed to operate without a calibration process which would require a large amount of data. This is done by using the pollutant nature of the catchment in terms of buildup and wash-off processes as the basis of measurements. Therefore, the translation procedure is an extension of the current estimation techniques which are typically complex and resource consuming. The use of a translation procedure is simple and based on the graphical estimation of parameters and tabular form of calculations. The translation procedure developed is particularly accurate in estimating water quality in the initial part of runoff events.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2015-05-07T14:06:12Z No. of bitstreams: 2 Dissertação - Luciana Inácio Pauletti - 2012.pdf: 2287630 bytes, checksum: 392dd39420f8918a2239eaf29eea5ee0 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2015-05-07T14:08:33Z (GMT) No. of bitstreams: 2 Dissertação - Luciana Inácio Pauletti - 2012.pdf: 2287630 bytes, checksum: 392dd39420f8918a2239eaf29eea5ee0 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2015-05-07T14:08:33Z (GMT). No. of bitstreams: 2 Dissertação - Luciana Inácio Pauletti - 2012.pdf: 2287630 bytes, checksum: 392dd39420f8918a2239eaf29eea5ee0 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2012-09-21 / The stormwater runoff in urban areas is an important source of non-point pollutants. This study aimed to quantify the build-up and wash-off of pollutants, particularly metals, water runoff from a paved surface urban Goiânia. Studies to estimate the processes of build-up and wash-off are still incipient in Brazil. We evaluated twelve plots paved with size of 3m2, located in a public parking lot, where they were analyzed on build-up and wash-off in June and July 2012. To wash-off was used a rainfall simulator in three intensities, 69 mm/h, 109 mm/h and 180 mm/h. Quantitation of accumulation in sampled surfaces determined load per m2 to obtain the coefficient build-up as well as for the analysis of the influence of background dried days before sampling campaigns. The build-up was adjusted at four possible equations for use in models water quality: (1) Linear, (2) Power (3) Exponential, and (4) Michaelis-Menton by obtaining the best fit. The water quality parameters analyzed were used for calculation of Average Concentration Event (CME), building polutogramas, coefficient of wash-off and correlation analyzes. The accumulation rate obtained was, on average, 8.53 g/m2dia, and more days dry, the greater the accumulation of pollutants. Levels of metals were found in all analyzed plots for all simulated events in different periods. There sharp and direct correlation between variables, particularly suspended solids and metals. The intensity of rain on the wash-off was clear, since the simulated rain intensity shifted with greater dirtier. Parameters were obtained for modeling the quality of rainwater. / O presente trabalho teve como objetivo quantificar o acúmulo (build-up) e carreamento (wash-off) de poluentes, principalmente metais, na água do escoamento de uma superfície pavimentada urbana de Goiânia. Os estudos para estimativa dos processos de build-up e wash-off são ainda insipientes no Brasil. Foram avaliadas doze parcelas experimentais asfaltadas, com dimensão de 3m2, situadas em um estacionamento público, onde foram realizadas as determinações de build-up e wash-off nos meses de junho e julho de 2012. Para wash-off foi utilizado um simulador de chuvas em três intensidades, 69 mm/h, 109 mm/h e 180 mm/h. A quantificação do acúmulo nas superfícies amostradas determinou a carga por m2 para obtenção do coeficiente de build-up, bem como para a análise da influência dos dias secos antecedentes antes das campanhas amostrais. O build-up foi ajustado nas quatro equações: (1)Linear, (2)Potencial, (3)Exponencial, e (4)Michaelis-Menton, com a qual se obteve o melhor ajuste. As análises de qualidade de água de escoamento permitiram o cálculo da Concentração Média do Evento (CME), construção de polutogramas, determinação do coeficiente de wash-off e análises de correlação. A taxa de acúmulo obtida foi, em média, de 8,53g/m2dia, sendo que quanto mais dias secos antecedentes à coleta, maior o acúmulo de poluentes. Foram encontrados metais (cromo, cádmio, cobre, chumbo, níquel, zinco, manganês e ferro) em todas as parcelas experimentais para todos os eventos simulados nos diferentes períodos. Verificou-se correlação direta e acentuada entre as variáveis avaliadas, principalmente entre sólidos suspensos e metais. O efeito da intensidade de chuva sobre o wash-off ficou evidente, visto que as chuvas simuladas com maior intensidade deslocaram mais poluentes. Neste estudo foram obtidos parâmetros para a modelagem da qualidade das águas pluviais.

Urban pollutants have been identified as a significant source of environmental pollution, posing a risk to human health, the environment, and are toxic to flora and fauna. Highways are recognised as one of the key sources of pollution, from both vehicles, and surrounding infrastructure. A number of studies have investigated accumulation of sediment and the associated pollutants on highways, and the runoff generated as a result of rainfall. Car parks share many potential contributory sources of pollutants with highways, but there is a lack of studies regarding car parks, despite them being identified as a significant percentage of urban land use. A series of experiments were undertaken in order to develop an understanding of the characteristics of car park sediment. The physical and chemical characteristics of sediment were analysed at different stages throughout the drainage system. Firstly on the sediment accumulated on car park surfaces, followed by that mobilised and transported into a channel drain during simulated rainfall events. Finally, potential treatment of pollutants within sediment in a channel was quantified. The physical and chemical characteristics of car park sediment was shown to be similar during both the accumulation (build-up) and wash-off stages, suggesting that the accumulated sediment is generally mobilised and transported to the channel drain (wash-off). Furthermore, both the physical and chemical characteristics of the sediment were shown to be similar to those found on highways, thorough the build-up and wash-off phases. Finally, potential treatment of organic pollutants (PAH) by biodegradation was demonstrated, but not comprehensively proved.

Submitted by Erika Demachki (erikademachki@gmail.com) on 2014-09-22T21:00:54Z No. of bitstreams: 2 Dissertação EMANOELLE PEREIRA DA SILVA.pdf: 6638194 bytes, checksum: d7aa7d2837e9a72171e815ff0dcd4137 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2014-09-23T15:26:34Z (GMT) No. of bitstreams: 2 Dissertação EMANOELLE PEREIRA DA SILVA.pdf: 6638194 bytes, checksum: d7aa7d2837e9a72171e815ff0dcd4137 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2014-09-23T15:26:34Z (GMT). No. of bitstreams: 2 Dissertação EMANOELLE PEREIRA DA SILVA.pdf: 6638194 bytes, checksum: d7aa7d2837e9a72171e815ff0dcd4137 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2014-03-14 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This work aimed to quantify the build-up and wash-off pollutants in urban asphalt surfaces, mainly metals and solids, correlating the pollutant load with dry days, traffic load and rainfall intensity. The relationship between metals and particle size has been also evaluated. To do this, three streets with different traffic load as study area in Goiânia City were selected. The sediments samples were collected in six plots of 0,5 m² located in the central track of the streets during 14 days, collecting samples on 1°, 2°, 3°, 5°, 7° e 14° dry days by the vacuuming and sweeping method. For water sampling two rainfall intensities were selected to be simulated, through a simulated rainfall over an area of 3m², being on low and other high intensity, respectively 25 mm/h and 115 mm/h both with 15 minutes duration. Were simulated three events on the same experimental plot in witch study area and collected 5 water samples during witch rainfall event, which were analyzed by concentration of suspend and dissolved solids, DQO, conductivity and the metals copper, chromium, lead, cadmium, zinc, iron and manganese, determined by Atomic Absorption Spectroscopy. The results showed that build-up has a strong direct relationship with dray days and an inverse relationship with traffic load. For metals in sediments was verified that the higher concentrations are on particles < 63μm, and follow in general the concentration order Fe > Mn > Cu > Pb > Zn > Cd > Cr. The runoff water quality was observed that all the parameters analyzed, except chromium, were detected on selected study areas. The influence of rainfall intensity on wash-off pollutants was observed that this has fundamental importance, so that higher intensity wash-off a higher mass of solids and metals. / O presente trabalho teve por objetivo quantificar o acúmulo e carreamento de poluentes em superfícies asfálticas urbanas, principalmente de metais e sólidos, correlacionando-os com o número de dias secos antecedentes, fluxo de veículos e intensidade da chuva. Avaliou-se também a relação entre os metais e a granulometria das partículas. Para tal, foram selecionadas três ruas com diferentes tráfegos de veículos como área de estudo na cidade de Goiânia. As amostras de sedimento foram coletadas em seis parcelas de 0,5 m², localizadas na faixa central das ruas durante 14 dias, sendo coletadas amostras no 1°, 2°, 3°, 5°, 7° e 14° dia seco pelo método de aspiração e escovação. Para a coleta de água foram selecionadas duas intensidades de chuva a serem simuladas, por meio de um simulador de chuvas sobre uma área de 3 m², sendo uma de baixa e uma alta, respectivamente 25 mm/h e 115 mm/h ambas com duração de 15 minutos. Foram simulados três eventos em uma mesma parcela experimental em cada área de estudo e coletadas 5 amostras de água durante cada evento de chuva, as quais foram analisadas quanto às concentrações de sólidos suspensos e dissolvidos, DQO, condutividade, e os metais cobre, cromo, chumbo, cádmio, zinco, ferro e manganês, determinados por Espectrofotometria de Absorção Atômica. Os resultados mostram que o acúmulo de sedimentos tem uma relação direta com os dias secos antecedentes e uma relação inversa com o fluxo de veículos. Para os metais no sedimento, verificou-se que as maiores concentrações estão preferencialmente nas partículas < 63 μm, e que seguem em geral a ordem de concentração Fe > Mn > Cu > Pb > Zn > Cd > Cr. Da qualidade da água escoada, observou-se que todos os parâmetros analisados, exceto o cromo, foram detectados na água das três áreas de estudo. Da influência da chuva no carreamento de poluentes, observou-se que esta tem importância fundamental, de forma que quanto maior a intensidade maior a massa carreada de sólidos e metais.

La maîtrise de la contamination générée par temps de pluie en milieu urbain constitue un enjeu environnemental important pour limiter la dégradation des milieux aquatiques superficiels. Les outils de modélisation traditionnelle utilisés pour estimer les flux de polluants dans les eaux de ruissellement sont jugés insuffisants dans leur capacité à reproduire les dynamiques des polluants à l’exutoire. Cela est souvent lié au manque de connaissances précises sur les processus en jeu d’une part, et d’autre part aux difficultés d’acquérir des bases de données représentatives et en continu sur des sites réels. Cette thèse a donc pour objectif d’améliorer l’état de la modélisation de la qualité. Elle vise en particulier le développement d’un outil de modélisation conceptuelle de la qualité des eaux de ruissellement à l’échelle du quartier, à partir d’une compréhension approfondie des processus d'accumulation et de lessivage. La simulation des pollutogrammes de matières en suspension (MES) à l’avaloir du bassin versant routier avec les modèles conceptuels d’accumulation-lessivage montre la faible performance des modèles pour estimer les dynamiques d’émissions de MES pour des longues périodes ; la variabilité du processus d’accumulation est le responsable principal de l’inadéquation de ces modèles. L’évaluation de la contribution des retombées atmosphériques sèches à la contamination des eaux de ruissellement en hydrocarbures aromatiques polycycliques (HAPs) et métaux montrent que l’atmosphère ne joue qu’un rôle très mineur dans la contamination des eaux de ruissellement par ces substances. Ainsi le couplage des modèles atmosphériques, qui ne tiennent pas compte des émissions directes liées au trafic, avec les modèles de qualité de l’eau, ne semble pas très pertinent dans l’objectif d’améliorer la prédiction de la contamination des eaux pluviales à l’exutoire. L’investigation à la micro-échelle du mécanisme de lessivage montre que les particules fines sont les plus susceptibles d’être mobilisées par le ruissellement. Cette étude a été menée en utilisant un simulateur de pluie innovant qui présente les avantages d’être mobile et léger, et la possibilité d’avoir des enregistrements en ligne du débit et de la turbidité. Les nouvelles connaissances acquises sur les processus soulignent une grande variabilité qui remet en cause l'intérêt de leur modélisation avec des approches déterministes. Ces connaissances sont intégrées à l’échelle du quartier pour développer un outil de modélisation conceptuelle basé sur une approche stochastique d'estimation de la concentration moyenne de MES et des paramètres de qualité. Le modèle développé est intégré dans le modèle hydrologique URBS. L’application de ce modèle permet d’intégrer la variabilité spatiale et temporelle des émissions en distinguant les contributions de chaque occupation du sol. Les résultats sont prometteurs en termes d’estimation des niveaux de concentration de MES à l’exutoire du bassin versant et de réplication du comportement général de la dynamique de MES, cependant des améliorations peuvent être envisagés pour consolider l’approche et améliorer ses prédictions. La comparaison de ce modèle avec des approches de modélisation empirique globale, conceptuelle semi-distribuée et physique distribuée, montre qu’en termes de pouvoir prédictif et de fiabilité, l’approche URBS-stochastique en parallèle avec l’approche de modélisation physique distribuée sont les plus performantes. En termes de simplicité d’implémentation et d’ajustement entre les observations et les simulations, les approches de modélisation empirique globale et conceptuelle semi-distribuée sont les plus puissantes. A l’issue de cette comparaison, il est clair qu’il n’existe pas un modèle parfait qui couvre toutes les caractéristiques de la modélisation de la qualité des eaux de ruissellement. Le choix de l’approche de modélisation la plus appropriée doit se faire en fonction des objectifs attendus par le modélisateur / Urban runoff contamination is recognized as a major source of the deterioration of the quality of surface water. Commonly used stormwater quality models have poor performance in predicting the pollutant dynamics at the surface outlet, mainly due to the lack of precise knowledge on the governing processes and the difficulties of acquiring representative and continuous databases on real sites. The main purpose of this Ph.D. thesis is to improve the state of stormwater quality modeling. It aims in particular to develop a conceptual modeling tool for stormwater quality prediction at the scale of a city district catchment, based on a deep understanding of the build-up and the wash-off. The application of commonly used stormwater build-up/wash-off models to simulate the dynamics of total suspended solids (TSS) at the outlet of the road catchment suggests that the models poorly replicate the temporal variability of the TSS concentrations unless short periods are considered. The unpredictable nature of the accumulation is largely responsible for the model failure. The evaluation of the contribution of atmospheric dry deposition to stormwater loads for polycyclic aromatic hydrocarbons (PAHs) and metals shows that atmospheric deposition is not a major source of contaminants in stormwater runoff. Thus, linking the air and water compartment in a modeling chain to have more accurate estimates of pollutant loads in stormwater runoff may not be relevant unless the direct traffic emissions are accounted for. The investigation of the wash-off process on elementary surfaces shows that the fine particles are the most likely to be mobilized and transported during a rainfall event. Stormwater samples were collected for this study using an innovative rainfall simulator that allows continuous, on-site monitoring of instantaneous flow and turbidity measurements and that can be easily transported and used on real sites. The new knowledge acquired on the build-up and wash-off processes underlines the great variability of these processes and calls into question their modeling with deterministic approaches. Hence, this knowledge is incorporated into developing a new conceptual stormwater quality model based on the stochastic drawing of event mean concentrations (EMC) of TSS and water quality parameters. The model is integrated within the hydrological model URBS. The application of this approach accounts for the spatial and temporal variability of pollutant emissions by distinguishing the contributions of each land use separately. The obtained results are promising in terms of estimating the concentration levels of TSS at the outlet of the city district catchment and replicating the general behavior of the TSS dynamics. However, improvements can be envisaged to consolidate the approach and improve its predictions. Comparison of this model with global empirical, semi-distributed conceptual and distributed physical modeling approaches shows that in terms of predictive power and stability, the stochastic-URBS and the physically distributed approaches are the most efficient. However, in terms of ease of implementation and best fit between observations and simulations, the global empirical and semi-distributed conceptual modeling approaches are the most powerful. This comparison shows that the perfect model that covers all aspects of stormwater quality modeling does not exist. The choice of the most appropriate modeling approach should mainly be driven by modeling objectives

Le développement des réseaux séparatifs entraîne le transfert fréquent de polluants urbains vers les milieux récepteurs (plans d’eau, rivières, etc.). La compréhension des processus de production et de lessivage des polluants dans le milieu urbain est pourtant incomplète à l’heure actuelle. Afin de répondre aux questions liées à la gestion des eaux urbaines, l’amélioration des connaissances des processus physiques est nécessaire, tant au niveau des surfaces urbaines que les réseaux d'assainissement. Pour cela, la modélisation du transfert hydrologique des polluants en milieu urbain peut être un outil précieux.Cette thèse a pour objectif de développer et d'analyser des modèles distribués à base physique pour simuler les flux de polluants routiers (Matières En Suspension (MES), Hydrocarbures, Métaux) dans un environnement urbain. Elle s'inscrit dans le cadre du projet ANR "Trafipollu" et bénéficie des résultats expérimentaux mis en œuvre dans ce projet pour la calibration et validation des modèles utilisés. Le travail de thèse s’articule autour de deux échelles de modélisation : l’échelle locale et l’échelle du quartier.A l'échelle locale, le code FullSWOF (volumes finis, schéma numérique d'ordre 2) couplé au modèle d’érosion d'Hairsine and Rose (1992a; 1992b) et des données géographiques très détaillées (résolution spatiale centimétrique) ont été utilisés et adaptés afin d'améliorer nos connaissances des processus physiques du lessivage des polluants sur les surfaces urbaines. La comparaison aux mesures en continu permet d’évaluer la performance d’une modélisation physique pour représenter les variations spatiales et temporelles des processus de transferts des polluants sur les surfaces urbaines. Les analyses des résultats obtenus permettent de constater la prédominance des effets d'arrachement liés à la pluie sur les processus d'entrainement par l'advection sur la majeure partie du bassin versant routier. L’utilisation d’un modèle d’érosion pour modéliser le transport particulaire en zone urbaine est une innovation importante de cette thèse.A l’échelle du quartier, la deuxième étape du travail consiste à coupler séquentiellement le modèle TREX (Velleux, England, et al., 2008) avec le modèle CANOE (Alison, 2005), nommé "TRENOE" plateforme. En changeant différentes options de mise en œuvre et de configurations du modèle, l’adaptation de la précision numérique et l’utilisation de données détaillées d’occupation du sol semblent être les facteurs clés pour une telle modélisation. Par ailleurs, ce couplage a montré des problèmes de fond tels que la modélisation du schéma numérique des flux en surface (seulement dans 4 directions), ainsi que l'utilisation de l'équation USLE pour simuler l'érosion en milieu urbain, ne comprenant pas d’impact des gouttes de pluie pour la modélisation.Pour remédier à ces défauts, la plateforme opensource LISEM-SWMM est développée en couplant le modèle LISEM (De Roo, Wesseling, et al., 1996), modèle d’érosion développé initialement pour le milieu naturel, et le modèle SWMM (Rossman, 2010). Pour la première fois, la modélisation hydrologique s’appuie aussi sur l’utilisation de sorties de modèles atmosphériques pour les dépôts des particules fines (PM10), hydrocarbures et métaux. Les résultats montrent que l’emploi de modèles totalement distribués peut arriver à reproduire de manière très fine les dynamiques des particules, des hydrocarbures et des métaux. Même si à ce stade la plateforme développée nécessite des améliorations pour adapter aux utilisations dans le champ opérationnel, ceci constitue une avancée pour le domaine de modélisation du transfert hydrologique des polluants routiers en milieu urbain / Nowadays, the increasing use of separate stormwater systems causes a frequent transport of urban pollutants into receiving water bodies (lakes, rivers). However, current studies still lack of the knowledge of urban build-up and wash-off processes. In order to address urban management issues, better understanding of physical mechanism is required not only for the urban surfaces, but also for the sewer systems. In this context, the modelling of hydrological transfer of urban pollutants can be a valuable tool.This thesis aims to develop and assess the physically-based and distributed models to simulate the transport of traffic-related pollutants (suspended solids, hydrocarbons, heavy metals) in urban stormwater runoffs. This work is part of the ANR "Trafipollu" project, and benefit from the experimental results for model calibration and validation. The modelling is performed at two scales of the urban environment: at the local scale and at the city district scale.At the local scale of urban environment, the code FullSWOF (second-order finite volume scheme) coupled with Hairsine and Rose model (1992a; 1992b) and detailed monitoring surveys is used to evaluate urban wash-off process. Simulations over different rainfall events represent promising results in reproducing the various dynamics of water flows and particle transfer on the urban surfaces. Spatial analysis of wash-off process reveals that the rainfall-driven impacts are two orders of magnitude higher than flow-drive effects. These findings contribute to a significant improvement in the field of urban wash-off modelling. The application of soil erosion model to the urban context is also an important innovation.At the city district scale, the second step consists of coupling the TREX model (Velleux, England, et al., 2008) and the CANOE model, named "TRENOE" platform. By altering different options of model configurations, the adequate numerical precision and the detailed information of landuse data are identified as the crucial elements for achieving acceptable simulations. Contrarily, the high-resolution topographic data and the common variations of the water flow parameters are not equally significant at the scale of a small urban catchment. Moreover, this coupling showed fundamental problems of the model structure such as the numerical scheme of the overland flow (only 4 directions), and the empirical USLE equations need to be completed by raindrop detachment process.To address these shortcomings, the LISEM - SWMM platform is developed by coupling the open-source LISEM model (De Roo, Wesseling, et al., 1996), which is initially developed for soil erosion simulations, and the SWMM model (Rossman, 2010). For the first time, the hydrological model is also supported by the simulations of atmospheric dry deposits of fine particles (PM10), hydrocarbons and heavy metals. The performance of water flow and TSS simulations are satisfying with the calibrated parameters. Considering the hydrocarbons and heavy metals contents of different particle size classes, simulated event mean concentration of each pollutant is comparable to local in-situ measurements. Although the platform at current stage still needs improvements in order to adapt to the operational applications, the present modelling approach contributes to an innovative technology in the field of modelling of hydrological transfer of the traffic-related pollutants in urban environment