MHYDAS-DRAIN: A spatially distributed model for small, artificially drained lowland catchments
Résumé
Although artificial drainage measures such as tile and ditch drainage, which shorten the residence time of water in the soil, do not only enhance diffuse pollution of surface water bodies, but also substantially alter the hydrology of lowland catchments, they are rarely explicitly included into spatially distributed catchment models. This is even more the case if flow anomalies like preferential flow cause a further acceleration of water flux and solute transport. Here, we present the spatially distributed modelling concept MHYDAS-DRAIN to account for and to evaluate these phenomena. As a starting point for model development, we chose the model MHYDAS which takes into account the discontinuities and the spatial variability of farmed catchments. The modelling domain consists of a system of interconnected 'hydrological units' which are derived by the overlay and intersection of geographical information and are linked to a drainage network. For the development of MHYDAS-DRAIN it was hypothesised that the tile drain discharge is composed of two components accounting for preferential flow and matrix flow. The fast flow component is modelled by a transfer function approach while the slow drainage discharge is calculated by the Hooghoudt equation. In open ditches, an additional baseflow component contributes to the total discharge. All flow routing is realised by an analytical diffusive wave approximation. The model was then applied to a small experimental catchment in the pleistocene lowland area of North-Eastern Germany. The model's parameter space was explored by a multi-target sensitivity analysis based on Latin hypercube sampling, Monte Carlo and regression analysis. This allowed the choice of efficient calibration parameters. The comparison and cross-evaluation of different calibration approaches demonstrated that parameter values depend on the calibration criteria as well as on the spatial and temporal resolution of the modelling domain. Modelled flow volumes, discharge rates and groundwater levels agreed reasonably well with measured data both in an hourly and daily temporal resolution. Although the fast flow component contributed-according to the modelling results-only a few percent to the total tile drainage discharge, this may still be of importance for solute transport. Snow events, however, like those of the winter of 2005, could not yet be simulated successfully, and the model proved to be sensitive to input data uncertainty. Nonetheless, the model is useful to account for the spatially variable properties of an artificial drainage system and should be applied to larger scales