Predictions of rainfall-runoff response and soil moisture dynamics in a microscale catchment using the CREW model

Abstract. Predictions of catchment hydrology have been performed generally using either physically based, distributed models or conceptual lumped or semi-distributed models. In recognition of the disadvantages of using either of these modeling approaches, namely, detailed data requirements in the case of distributed modeling, and lack of physical basis of conceptual/lumped model parameters, Reggiani et al. (1998, 1999) derived, from first principles and in a general manner, the balance equations for mass, momentum and energy at what they called the Representative Elementary Watershed (or REW) scale. However, the mass balance equations of the REW approach include mass exchange flux terms which must be defined externally before their application to real catchments. Developing physically reasonable "closure relations" for these mass exchange flux terms is a crucial pre-requisite for the success of the REW approach. As a guidance to the development of closure relations expressing mass exchange fluxes as functions of relevant state variables in a physically reasonable way, and in the process effectively parameterizing the effects of sub-grid or sub-REW heterogeneity of catchment physiographic properties on these mass exchange fluxes, this paper considers four different approaches, namely the field experimental approach, a theoretical/analytical approach, a numerical approach, and a hybrid approach combining one or more of the above. Based on the concept of the scaleway (Vogel and Roth, 2003) and the disaggregation-aggregation approach (Viney and Sivapalan, 2004), and using the data set from Weiherbach catchment in Germany, closure relations for infiltration, exfiltration and groundwater recharge were derived analytically, or on theoretical grounds, while numerical experiments with a detailed fine-scale, distributed model, CATFLOW, were used to obtain the closure relationship for seepage outflow. The detailed model, CATFLOW, was also used to derive REW scale pressure-saturation (i.e., water retention curve) and hydraulic conductivity-saturation relationships for the unsaturated zone. Closure relations for concentrated overland flow and saturated overland flow were derived using both theoretical arguments and simpler process models. In addition to these, to complete the specification of the REW scale balance equations, a relationship for the saturated area fraction as a function of saturated zone depth was derived for an assumed topography on the basis of TOPMODEL assumptions. These relationships were used to complete the specification of all of the REW-scale governing equations (mass and momentum balance equations, closure and geometric relations) for the Weiherbach catchment, which are then employed for constructing a numerical watershed model, named the Cooperative Community Catchment model based on the Representative Elementary Watershed approach (CREW). CREW is then used to carry out sensitivity analyses with respect to various combinations of climate, soil, vegetation and topographies, in order to test the reasonableness of the derived closure relations in the context of the complete catchment response, including interacting processes. These sensitivity analyses demonstrated that the adopted closure relations do indeed produce mostly reasonable results, and can therefore be a good basis for more careful and rigorous search for appropriate closure relations in the future. Three tests are designed to assess CREW as a large scale model for Weiherbach catchment. The first test compares CREW with distributed model CATFLOW by looking at predicted soil moisture dynamics for artificially designed initial and boundary conditions. The second test is designed to see the applicabilities of the parameter values extracted from the upscaling procedures in terms of their ability to reproduce observed hydrographs within the CREW modeling framework. The final test compares simulated soil moisture time series predicted by CREW with observed ones as a way of validating the predictions of CREW. The results of these three tests, together, demonstrate that CREW could indeed be an alternative modelling framework, producing results that are consistent with those of the distributed model CATFLOW, and capable of ultimately representing processes actually occurring at the larger scale in a physically sound manner.