Suspended sediment production and transfer in mesoscale catchments : a new approach combining flux monitoring, fingerprinting and distributed numerical modeling

The study of soil erosion by water and the transfer of suspended solids from watersheds to rivers is crucial given the environmental and socio-economic issues with regards to growing human influence and the expected intensification of these processes under climate change. The objective of this thesis is to understand how rainfall variability controls the activation of different sediment source zones and the dynamics of hydro-sedimentary flows in two mesoscale Mediterranean catchments, i.e. the Claduegne (42 km², subcatchment of the Ardeche) and the Galabre (20 km2 , subcatchment of the Durance) which are members of the OZCAR critical zone research infrastructure.In the first part, the contributions of the erosion zones to sediment fluxes at the outlet of the Claduegne catchment were quantified at high temporal resolution with a low-cost sediment fingerprinting approach. Two sets of tracers (Color and X-ray fluorescence tracers) and three mixing models were compared to assess the sensitivity of estimated source contributions to these methodological choices. Marly-calcareous badlands were identified as the main sediment source. A similar approach carried out on the Galabre catchment area showed that badlands on molasses were the main source. The comparison of tracer sets and mixing models, showed that the methodological choices generated important differences. Thus, we suggest a multi-tracer-multi-model ensemble approach to obtain more robust results. The application of this approach to a large number of sediment samples highlighted the important within and between event variability in the contributions of different sediment sources, raising questions about the hydro-sedimentary processes that cause this variability.We hypothesized that this variability resulted from variable suspended sediment transit time distributions governed by the interplay of (i) catchment characteristics such as the location of different sources and how they are linked to the outlet (referred to as structural sediment connectivity) and (ii) the spatio-temporal characteristics of rain events that activate and impact transfer velocities (i.e. functional connectivity).Thus, in the second part, a distributed numerical model based on the resolution of Saint Venant equations coupled to a multi-source erosion module was used to evaluate the respective roles of structural and functional connectivity. Sensitivity analysis of the discretization and parameterization choices (i.e. threshold of contributing drainage area to identify the river network, values of roughness coefficients on hillslopes and the river) showed that the location of the sediment sources in the watershed was more important than the modeling choices when the parameters were limited to realistic range. A general temporal pattern of source contributions was observed. This was consistent with the results of the fingerprinting approach and the distribution of distances from the sources to the river and the outlet. The same pattern persists for different rainfall durations or intensities but became much more variable when bimodal hyetographs or spatially variable precipitation was applied. In addition, the location of the rainfall with respect to the sources determined the average contributions of the sources and thus differences between rainfall events.The two approaches, sediment fingerprinting and numerical modeling, were found to complement each other and their combined application has a high potential for understanding how interactions between structural and functional connectivity control the dynamics of sediment fluxes in mesoscale catchments.