Qu'est ce que les modèles nous apprennent sur la connectivité de l'eau et des sédiments ?

Connectivity has been embraced by the geosciences community as a useful concept to understand and describe hydrological functioning and sediment movement through catchments. Mathematical modelling has been used for decades to quantify and predict erosion and transport of sediments, e.g. in scenarios of land use change or conservation measures. Being intrigued by both models and the connectivity concept, as a group of modellers we aimed at investigating what different models could tell us about connectivity. Therefore, we evaluated the response of contrasted spatially-distributed models to landscape connectivity features and explained the differences based on different model structures. A total of 53 scenarios were built with varying field sizes and orientations, as well as the implementation of soil conservation measures. These scenarios were simulated, for two rainfall intensities, with five event- and process-based water and soil erosion models – EROSION3D, FullSWOF_2D, LandSoil, OpenLISEM and Watersed. Results showed that rainfall amount plays the most important role in determining relative export and connected area of runoff and sediment in all models, indicating that functional aspects of connectivity were more important than structural connectivity. As for the role of structural landscape elements, there was no overall agreement between models regarding the effects of field sizes, crop allocation pattern, and conservation practices; agreement was also low on the spatial patterns of connectivity. This overall disagreement between models was unexpected. The results of this exercise suggest that the correct parameterization of runoff and sediment production and of routing patterns may be an important issue. Thus, incorporating connectivity functions based on routing would help modelling forward. Our results also suggest that structural connectivity indices may not suffice to represent connectivity in this type of catchment (relatively simple and monotonous land cover), and functional connectivity indices should be applied.