Distributed modelling of erosion hazard in the Loire River basin and new perspectives of validation

Soils erosion has accelerated after WWII in the cultivated plains of Northwestern Europe. This phenomenon is explained by the modification of landscapes and agricultural practices. To combat erosion in an efficient way, it is crucial to quantify the erosion, transfer and deposition processes within large basins, as well as their spatial redistribution. These estimations are important in order to meet the guidelines required by the European Water Framework Directive, including the ecological good status of water bodies. The objective of this thesis is to quantify erosion processes and sediment redistribution during the last several decades and across multiple spatial scales within the Loire River basin (117,000 km²). First, the erosion hazard assessment model MESALES was improved to include the latest input databases available (reflecting the evolution of vegetation cover, rainfall erosivity and soil surface texture). The results were integrated at the scale of water bodies to identify those sub-catchments that were the most exposed to erosion. The improvement of the model led to a modification of the hazard estimation in half of the water bodies in the Loire River basin. Second, the potential of measuring caesium-137 (¹³⁷Cs) inventories in soils to validate the spatially-distributed erosion model outputs was investigated. A map of the initial fallout of this fission product produced by atmospheric thermonuclear tests in the 1960's was first created. To do so, a data base of undisturbed reference sites was developed by combining several existing samples' banks, and additional samples collected during the thesis. The spatial distribution of initial fallout across the entire basin was achieved through the use of rainfall data corresponding to the 1960-1964 period as covariates, which confirms the lack of significant ¹³⁷Cs fallout following the Chernobyl accident in this region. Then, the ¹³⁷Cs inventories were measured in all the 478 soils profiles collected in the framework of the Soil Quality Monitoring Network (RMQS in French). The mean soil redistribution rates of the 1954-2008 period were calculated using the proportional model. The results showed an over-representation of accumulation sites among the RMQS' sites and the difficulty to identify covariates allowing to map the spatial patterns of soil redistribution at this scale from this type of monitoring network intended to map soil properties rather than their movement across the landscape. Finally, a subcatchment was selected in an intensively cultivated plain (Brosse, Indre-et-Loire, 24 km²). The evolution of land cover and rainfall during the 1959-2010 period was reconstructed to model erosion in this catchment with the event-based Water Sed model. The outputs were confronted to those erosion rates estimated based on the analysis and dating of sediment core samples collected in the pond at the outlet. The results showed a good correlation between the model outputs and the erosion rates estimated from the cores. These results permitted to distinguish between the anthropogenic effect, which governs the general erosion trends across the five decades, from the climatic influence, which explains the seasonal variability. This thesis demonstrated the possibilities and the limitations associated with the use of the latest data bases and sample banks available to model and reconstruct soil erosion rates at multiple spatial and temporal scales within large river basins, which constitute the optimal management units in the European Union. (author)