No abstract

Abstract. Despite the extensive research into hillslope and channel interactions in headwater catchments, surprisingly little attention has been paid to such processes in lowland rivers. In particular, previous studies have not addressed the influence of hillslope contributions and have concentrated solely on in-bank floods rather than more complex out-of-bank cases. Accordingly, we combine field monitoring and numerical modeling to study hillslope, floodplain, and channel interactions for a lowland river. Piezometric, precipitation, and river stage data were used to parameterize and test a new twodimensional finite element model of saturated-unsaturated flow applied to two vertically aligned cross sections through a lowland floodplain. Data for two major out-of-bank flood events were simulated which appeared to show the presence of a significant unsaturated zone extending up to 5 rn below the surface. The model simulated reasonably well the pressure head field that was recorded at a number of piezometers located internal to the computational domain on each transect, and we conclude that floodplain hydrology is predominately a two-dimensional (lateral) process. Three-dimensional (down reach) flow effects would seem to become more significant at the beginning and end of each event.The simulations also showed that the unsaturated zone remained close to saturation at all times and that it was not significant in terms of the floodplain hydrology. Examination of velocity vector patterns showed the formation of a strong groundwater ridge within the floodplain. This led to the development of strong velocities directed toward hillslope areas as the inundation front approached the hillslope/floodplain junction. This suggests that surface water may move into hillslope areas adjoining the floodplain during major floods. Thus the extent of the hyporheic zone may be larger than previously thought. Numerical Model DevelopmentA finite element model, ESTEL, was developed to simulate hillslope and channel fluxes to the floodplain alluvium during overbank flood events. Like WaTab2D [Whiting and Pomeranets, 1997] this model is also based on an exact formulation for time-dependent unconfined groundwater flow and hence provides an advance on models such as MODFLOW, which are based on the Dupuit-Forcheimer assumption. However, unlike WaTab2D, ESTEL applies to both saturated and unsaturated flow conditions, as this may be important in particular floodplain situations. The model was programmed using an object-oriented numerical library previously developed to solve the shallow water equations. For a complete description of the model and its development the reader is referred to Desitter et al. [1998, 2000], and here we provide only a brief summary of its major attributes.We solve the so-called "mixed" where t is the time (dimension T), 0 is the volumetric moisture content (L 3 L-3), h is the pressure head (L), K is the hy-

Abstract:When studying hydrological processes with a numerical model, global sensitivity analysis (GSA) is essential if one is to understand the impact of model parameters and model formulation on results. However, different definitions of sensitivity can lead to a difference in the ranking of importance of the different model factors. Here we combine a fuzzy performance function with different methods of calculating global sensitivity to perform a multi-method global sensitivity analysis (MMGSA). We use an application of a finite element subsurface flow model (ESTEL-2D) on a flood inundation event on a floodplain of the River Severn to illustrate this new methodology. We demonstrate the utility of the method for model understanding and show how the prediction of state variables, such as Darcian velocity vectors, can be affected by such a MMGSA. This paper is a first attempt to use GSA with a numerically intensive hydrological model.

The ongoing discussion on the value of hydrological modelling and uncertainty analysis initiated by Beven (2006b) has lately seen an increasing public interest. For instance, during the interactive discussion on the uncertainty of environmental models held at the 2008 EGU joint assembly (Beven, 2008), Murugesu Sivapalan argued that despite the very significant scientific advances made in the topic of uncertainty analysis, there has only been limited progress in the development of better models. As Topmodel, one of the most popular hydrological models, is nearing its 30th birthday (Beven and Kirkby, 1979), we think that this is an issue worth exploring in more detail.Modelling is an essential aspect of hydrological science and uncertainty analysis is certainly a powerful means of identifying sources of errors and uncertainty within a modelling exercise. However, uncertainty analysis should not be an aim but a tool. We believe that two steps should follow the uncertainty analysis. The first step is to use the model uncertainties (which represent those aspects of the hydrological systems that are least understood) to collect new or better data, improve data assimilation techniques or study specific processes. In the second step, the model should be improved using these data, methods and concepts.Those two steps are equally important. However, it seems that the first step has been the main topic of the ongoing discussion (e.g., Todini and Mantovan, 2007) and that the second step, which transfers new or improved concepts to actual model code, has received far less attention. Nevertheless, it is an essential part of the widely supported prediction in ungauged basins (PUB) strategy to improve hydrological models (Sivapalan et al., 2003). One of the aims of PUB is indeed the harmonization of modelling efforts, such that community efforts can interact and new developments can be shared, scrutinized and improved.We agree that new paradigms are important in the search for new approaches to hydrological modelling. But in our opinion, the implementation of new ideas and concepts is also seriously hindered by current modelling practice, to the extent that basic requirements for the advancement of science, i.e. verification of published studies, comparison of models and building further on available knowledge, are not fulfilled. This commentary aims to contribute to the debate by investigating these obstacles in more detail.Over the last decades, substantial developments in computer simulation have boosted the development of the numerical models. New algorithms have been developed and the analytical or numerical solution of theoretical process representations have been improved (e.g., Lane, 1998; Troch et al., 2003). Access to faster and larger computing facilities has also led to typical problems being solved at much higher spatial and temporal resolution. Innovations in monitoring and exploration techniques have led to an increased data availability and new exploration techniques have improved our understanding of hydrologica...

No abstract