Modélisation des incertitudes de ruissellement de bassins cultivés par algorithme de drainage stochastique

[Departement_IRSTEA]Territoires [TR1_IRSTEA]SYNERGIEInternational audienceAnthropogenic ditch drainage networks have a strong impact on the runoff of small cultivated catchments and are more and more considered in hydrological modelling. However, maps of ditch drainage networks are not usually available which results in uncertainties of water flow-paths. In this context, this study aimed to assess runoff uncertainties entailed by uncertainties of ditch drainage network. We used a coupling approach to propagate uncertainties generated by a random network generation method in a hydrological model. First, we used a stochastic vector drainage algorithm running within the lattice of the field boundaries valued by elevation. It simulated equi-probable networks on a small cultivated catchment, with respect to morphology and uncertainties of elevation data. A thousand simulations represented uncertainty of the spatial organization and density of the network. Next, we propagated uncertainties of the water flow-paths through the hydrological model MHYDAS. Uncertainty of network runoff was high: the coefficient of variation of total volume was equal to 21\% at a subcatchment scale and equal to 18\% at the catchment scale. This uncertainty can be partly related to uncertainty of the network density. In addition to uncertainties of network runoff, uncertainty occurred about diffuse flow-paths too, due to the change of the topology of the fields. This uncertainty of overland flow was higher than for the network (coefficient of variation of overland flow indicator equal to 123\%) and closely related to ditch drainage density. Finally, this study (i) proposed a way to map runoff uncertainty at different scales in the case of an unknown actual network, (ii) allowed to evaluate the relative importance of the ditch drainage network in runoff simulations