Water-induced soil erosion and shallow landslides interact with each other and need to be studied in an integrated approach to understand hillslope sediment yields. The principal aim of this thesis was to study and model soil erosion and shallow landslides in an integrated way. The thesis presents results from laboratory and catchment-scale studies and modelling. A laboratory flume under a rainfall simulator was used for shallow landslide and soil erosion experiments using sandy and silty loess soils. In the experiments, landslide initiation, retrogressions and slip surface depths were measured and monitored directly or by using video camera recordings. Sediment and runoff were collected from the flume outlet every minute during landslides and every 10 minutes before and after landslides. Changes in the soil slope, after landslides, were recorded. Initially, six experiments including two repetitions were conducted using sandy soils at a 30º and 10º compound slope configuration, but with different soil profile depths. The experimental results showed that total and landslide-driven sediment yields were affected by the original soil profile depth; the greater the depth, the higher the sediment yield. Later, twelve other experiments were conducted on different slopes using silty loess soils. The experimental observations were used to validate an integrated modelling approach which includes WEPP for runoff and soil erosion modelling, a slope stability model for simulating shallow landslides, and a simple soil redistribution model for runout distance prediction. The model predictions were in good alignment with the observations. In all (sandy and silty loess) experiments, peak sediment discharges were related to the landslide events, proximity to the outlet and landslide volume. The post-failure soil erosion rate decreased as a function of changes in the slope profile. The GeoWEPP-SLIP modelling approach was proposed for catchment-scale modelling. The approach simulates soil erosion using the Hillslope and Flowpath methods in WEPP, predicts shallow landslides using a slope stability model coupled with the WEPP’s hillslope hydrology and finally uses a simple rule-based soil redistribution model to predict runout distance and post-failure topography. A case study application of the model to the Bowenvale research catchment (300 ha) showed that the model predictions were in good agreement with the observed values. However, the Hillslope method over-predicted the outlet sediment yield due to the computational weighting involved in the method. The Hillslope method predicted consistent values of sediment yield and soil erosion regardless to the changes in topography and land-cover in the post-failure scenarios. The Flowpath method, on the other hand, predicted higher values of sediment yield in the post-failure vegetation removal scenario. The effects of DEM resolution on the approach were evaluated using four different resolutions. Statistical analyses for all methods and resolutions were performed by comparing the predicted versus measured runoff and sediment yield from the catchment outlet and the spatial distribution of shallow landslides. Results showed that changes in resolution did not significantly alter the sediment yield and runoff between the pre- and post-failure scenarios at the catchment outlet using the Hillslope method. However, the Flowpath method predicted higher hillslope sediment yields at a coarser resolution level. Similarly, larger landslide areas and volumes were predicted for coarser resolutions whereas deposition volume decreased with the increase in grid-cell size due to changes in slope and flowpath distributions. The research conducted in the laboratory and catchment presented in this thesis helped understand the interactions between shallow landslides and soil erosion in an integrated approach.

L’étude présentée a pour objectif de mieux comprendre l’influence de l’hydrodynamique sur le transfert de produits phytosanitaires lors d’écoulement dans les fossés agricoles. Quatre paramètres ont plus particulièrement été étudiés : i) la vitesse de l’écoulement de surface ; ii) la submergence, définie comme le rapport moyen entre la hauteur des formes du lit et la hauteur d’eau ; iii) la forme du lit et iv) le degré de saturation en eau du lit avant la contamination par les produits phytosanitaires. L’influence de ces paramètres a été étudiée en conditions semicontrôlées à travers la mise en oeuvre d’expérimentations en canal expérimental, au fond duquel a été fixé pour chaque expérimentation un substrat type constitué d’un assemblage de fibres de chanvre utilisé pour approcher de manière simplifiée un système naturel relativement complexe et hétérogène (en considération des nombreux substrats rencontrés dans le lit des fossés naturels : herbes, feuilles mortes, végétation en décomposition, paille, sédiments, …). Les fibres de chanvre ont été choisies après une étude préliminaire en laboratoire visant à comparer les possibilités d’adsorption et de désorption des produits phytosanitaires sélectionnés sur diverses fibres couramment utilisées dans l’industrie géotextile (jute, lin, chanvre, polyamide, polypropylène et polyester). Les expérimentations en canal expérimental ont été menées de manière à tester l’influence spécifique des paramètres étudiés. La comparaison des différentes expérimentations indique qu’une augmentation de la vitesse de l’écoulement de surface augmente la quantité de produits phytosanitaires transférés de l’écoulement de surface vers le lit, ainsi que la cinétique de ce transfert. Le constat est le même lorsqu’on augmente la submergence, ou bien lorsqu’on passe d’une géométrie de petites formes en « dunes » à une géométrie de grandes formes en « créneau ». Enfin, le fait que le lit soit saturé en eau en début d’expérimentation réduit fortement le transfert de la lame d’eau vers le lit. On observe également que l’influence de chaque paramètre semble être interdépendante. Devant le coût en temps et en argent de ces expérimentations, un modèle a été sélectionné dans la littérature, codé et utilisé pour conforter et compléter l’interprétation des données expérimentales. / The aim of this work was to better understand the influence of hydrodynamics on the transfer of pesticides in water flows in agricultural ditches. Special attention was given to four parameters: i) the speed of the surface water flow, ii) the submergence, defined as the mean ratio between the height of the bedforms and the water depth, iii) the shape of the bedforms and iv) the water content of the bed substrate before contamination by pesticides. The influence of these various parameters was investigated with an experimental flume, using a standard of substrate made of hemp fibres. Hemp fibres were chosen as a simplified model of natural substrates such as grass, dead leaves, decaying vegetation, straw and sediments, which are rather complex and heterogeneous. Hemp fibres were selected from preliminary laboratory studies which allowed to compare the adsorption and desorption characteristics of various pesticides for different fibres often used in the geotextile industry (jute, linen, hemp, polyamide, polypropylene and polyester). The flume experiments were designed in order to assess the influence of the four parameters mentioned above. The comparison between the different experiments shows that an increase in surface water speed results in an increase in the amount of pesticides transferred from the water to the bed substrate and the transfer kinetics. A similar result is observed when increasing the submergence, or when going from small sinusoidal bedforms to larger rectangular bedforms. In addition, the data show that the transfer of pesticides from surface water to the bed substrate is strongly decreased when the substrate was saturated with water at the beginning of the experiments. Also, it appears that there is a link between the influence of the differents parameters. When considering the long time duration and the cost of the experiments, it was decided to use a model from the literature which was adapted and used to confirm and extend the interpretation of the experimental data.

La Société le Nickel ’SLN’ est l’un des plus grands exploitants de nickel en Nouvelle Calédonie. Le Stockage des déchets de l’exploitation minière sur le site d’exploitation se fait sous forme de tas manufacturés appelés versants. Pour assurer la stabilité de ces versants, il est nécessaire d’assurer un stockage hors d’eau de ces stériles miniers. Pour se faire, une mise en place des ouvrages en enrochement enterrés se fait au fur et à mesure de l’avancement des travaux d’installation des versants. Le rôle de ces ouvrages enterrés est de maintenir l’écoulement d’eau à travers le milieu poreux. Le premier objectif de la thèse est donc de mieux comprendre les interactions hydrodynamiques dans les milieux très perméables dans le cas d’écoulements pleinement turbulents. La première partie de ce travail concerne une étude expérimentale menée à l’Institut National Agronomique de Tunisie (INAT) mais aussi à l’Institut de mécanique de fluide de Toulouse (IMFT) dont le but est d’étudier à la fois l’évolution de la ligne d’eau dans les milieux macroporeux et la variation de la perméabilité de ces derniers en fonction du nombre de Reynolds. Des simulations avec des modèles de perte de charge les plus utilisés dans la littérature ont été effectuées, pour tester leur capacité à reproduire la ligne d’eau dans les conditions de nos expériences. Les deux relations de Forchheimer (1901) et d’Ergun (1952) donnent des résultats proches de nos expériences avec des valeurs de coefficients adaptés pour notre cas, alors que la relation de Barree et Conway (2004) ne semble pas reproduire les résultats attendus. L’eau qui circule dans les ouvrages enterrés est une eau chargée d’où l’intérêt d’étudier le transport sédimentaire dans ces milieux macroporeux. Pour se faire, une étude expérimentale a été réalisée sur le transport et le dépôt des particules dans des milieux synthétiques constitués de particules de diamètre moyen de 15 cm / The 'SLN' Company is one of the largest nickel operators in New Caledonia. The mining wastestorage at the operation site is as a slope. To ensure the stability of these slopes, it is necessary toensure an out of water storage of the waste rock. To do that, an underground rockfillimplementation is installed. The purpose of these buried structures is to maintain the flow of waterthrough its pores. The first objective of the thesis is to better understand the hydrodynamicinteractions in highly permeable porous media and in case of fully turbulent flows. The first part ofthis work relates to an experimental study carried out at the National Agronomic Institute of Tunisia(INAT) and the Institute of Fluid Mechanics of Toulouse (IMFT) whose aim is to study the waterprofile evolution in macro porous media and the permeability variation as a function of theReynolds number. Simulations with the most widely used pressure drop relationships in theliterature were performed to check their ability to reproduce our experimental water profile. Thetwo relations of Forchheimer (1901) and Ergun (1952) give the most satisfactory results withadapted coefficients values for our case, whereas the Barree and Conway (2004) formula doesn’tseems to give expected results. Given that the circulating water in these buried structures is acharged water, we studied the sedimentary transport in macro porous media. To do this, anexperimental study was carried out on the transport and deposition of particles in the studiedporous media which mean diameter is d=15 cm.