Hydrological process representation at the meso-scale: the potential of a distributed, conceptual catchment model

Abstract

In order to achieve a process-oriented simulation of hydrological processes in a meso-scale basin (10¹–10³ km²), the spatially and temporally variable basin inputs (precipitation and energy) and runoff generation processes need to be adequately addressed by the model. The catchment model TAC^D (tracer aided catchment model, distributed) is based on experimental results including tracer studies at the mountainous Brugga basin (40 km²). This raster-based model (50×50 m²) works on an hourly basis, thus capturing the spatially and temporally variable inputs and processes. The model contains a process-realistic description of the runoff generation mechanism, which is based on a spatial delineation of eight units with the same dominating runoff generation processes. This defines the model structure and enables efficient model parameterisation. The model uses linear and non-linear reservoir routines to conceptualise runoff generation processes, and includes a routing routine (kinematic wave approach) to simulate surface runoff.

The model is successfully applied to a 1-year period following minimal calibration (model efficiency 0.94). In addition, the runoff from both an independent 3-year period for the Brugga basin and a sub-basin (15.2 km²) is modelled well (model efficiencies 0.80 and 0.85, respectively) without re-calibration. The use of tracer data (i.e. dissolved silica) measured in outlet discharge demonstrates that the temporal mixing pattern of different runoff components is modelled correctly (multiple-response validation). The results show that a validated process-based model that correctly simulates the origin of runoff components and flow pathways must be the basis for integrating solute transport modelling of non-conservative species. Such a model can serve as tool to make predictions and test hypotheses about the first-order controls on hydrological responses.

Introduction

The satisfactory modelling of hydrological processes in meso-scale basins (approximately 10¹–10³ km²; Blöschl, 1996) is essential for optimal protection and management of water resources at this scale. If the aim of a modelling study is to reproduce only the daily runoff dynamics, a simple lumped model that can be calibrated to observed data is sufficient (e.g. Jakeman and Hornberger, 1993, Beven, 2001). However, if distributed calculations are required as part of an environmental modelling approach, a more complex and distributed model needs to be applied. In particular, if estimations (extrapolations) for the future under changed circumstances (i.e. altered land use, climate change) are sought, purely statistical or lumped approaches reach their limit and a model that contains process-realistic descriptions for all hydrological processes is required.

The processes dominating hydrological response differ at various spatial scales (e.g. Blöschl and Sivapalan, 1995, Blöschl, 1996). Even if many questions concerning scale issues remain unanswered, the following generalized scheme can be assumed for temperate zone catchments without large proportions of urban land use. (i) In micro-scale catchments (i.e. headwater catchments less than about 1 km²), response to rainfall is dominated mainly by the runoff generation processes at the hillslopes and near stream areas (e.g. Anderson and Burt, 1990, McDonnell, 1990, Montgomery et al., 1997). All processes that define the lateral movement of water on top of the soil, or within the unsaturated and saturated zones are the first-order controls. Thus, soil properties and land use play key roles. Finally, the spatial distribution of rainfall can be assumed to be much more uniform than in larger basins. (ii) In meso-scale basins, processes from smaller scales combine in a complex way to produce an integrated response. Identifying basin wide areas with the same dominating runoff generation processes is currently a major challenge in runoff generation process research (e.g. Peschke et al., 1999, Scherrer and Naef, 2003, Uhlenbrook, 2004). Channel processes (i.e. runoff routing, groundwater–surface water interactions) gain increasing importance with increasing catchment area at this scale. (iii) At the macro-scale (basins larger than 1000(0) km²), the spatial and temporal distribution of rainfall or snow melt and the routing of runoff are dominant. Recently, Bárdossy et al. (2002) showed the marginal influence of different soil properties or land use covers (i.e. increasing urbanization) on flood runoff of larger events in the river Rhine.

In summary, the interesting challenge of working at the meso-scale as an intermediate scale is three-fold. First, the distributed runoff generation processes need to be understood and captured by the chosen model. Second, the spatial and temporal variability of atmospheric forces and runoff concentration is significant and needs to be included in the model. Third, as this scale is crucial for water management issues, addressing several societal demands, it requires particular attention.

To capture the hydrological processes at the meso-scale, different models with various ranges of complexity were suggested (Singh, 1995). On one hand, increased computer capabilities have made possible the development and application of distributed and largely physically based models that work at a highly detailed spatial and temporal resolution (e.g. MIKE-SHE, Refsgaard and Storm, 1996; KINEROS, Smith et al., 1995). However, the enormous data requirements often prevent the extensive use of these models, particularly in basins larger than experimental headwaters. In addition, further limitations of these models are obvious, e.g. the application of small-scale physical equations derived in laboratories to larger heterogeneous areas, or the sub-grid variability (for detailed discussion see Beven, 1996, Beven, 2001). On the other hand, conceptual rainfall runoff models (e.g. TOPMODEL, Beven and Kirkby, 1979; HBV, Bergström, 1992) are less complex and the required input data are available for most applications. Nevertheless, the model parameters are often not physically based or clearly related to catchment properties and an equifinality problem (different models or parameter sets reach equally good simulation results) exists (e.g. Beven and Binley, 1992). This makes model application in ungauged basins difficult. In addition, a high degree of model structure uncertainty exists (e.g. Grayson et al., 1992, Seibert, 1999), as the structure should be consistent with the perceptual (conceptual) model of the investigated basin. The perceptual model describes the ‘basin functioning’ and is based on experimental investigations. Of course, this could be different for every basin, thus the application of a standard model and fitting by parameter calibration might lead to models that ‘… are working right, but for the wrong reasons’ (Klemes, 1986). Other sources of model uncertainty, such as the error of input data and its spatial regionalization to basin scale as well as the lack of understanding of the dominant processes and their mathematical descriptions are discussed in further detail by Beven (2001).

Because of these uncertainties it is clear that calibrating model parameters to produce a close fit between measured and simulated runoff is not a rigorous enough test of the model, as it does not guarantee accurate representation of internal hydrological processes. This highlights the need for additional data to evaluate model performance (e.g. Kuczera and Mroczkowski, 1998). This additional data can be, for instance, hydrochemical data (Mroczkowski et al., 1997), groundwater levels (Lamb et al., 1998, Seibert, 1999), environmental isotopes sampled during events (Seibert and McDonnell, 2004), or sampled continuously (Uhlenbrook and Leibundgut, 2002), and the distribution of saturated areas (Ambroise et al., 1995, Franks et al., 1998, Güntner et al., 1999a).

The objectives of this paper are, first, to introduce a modelling approach able to translate runoff generation process understanding into a catchment model for the meso-scale. The second is to model continuously all dominating hydrological processes (not only event based), distributed (50×50 m²) on an hourly resolution using only minimal calibration. Third, this paper seeks to use information in addition to total basin discharge for analysing the model performance (multiple-response validation). Therefore, runoff data from a sub-basin is applied to test the model's ability to predict space–time variability of stream flow. In addition, the concentration of a natural tracer (i.e. dissolved silica) is used to check the temporally variable composition of runoff components. Finally, although this model is site-specific in its current form, the applicability of the modelling approach to other catchments based on knowledge of local processes is discussed.