Description

Robert Schwarze, Werner Dröge, Knut Opherden (1999):

in:    Hydrologie und Regionalisierung: Ergebnisse eines Schwerpunktprogramms (1992 – 1998); Deutsche Forschungsgemeinschaft; Hrsg. von Hans-B. Kleeberg; Weinheim, New York, Chichester, Brisbane, Singapore: Wiley-VCH, 1999

Problem formulation

The regional depicition of water and matter fluxes in catchments is dependent on information on the origin of water. Reliable hydrological information about discharge components, flow paths and residence times are especially important for the assessment of potential impacts of changes to the environment on water and matter budgets. There are many open research questions in hydrology about the definition of structural parameters and residence times of water in the different compartments of the catchment. This is especially true for meso-scale mountaineous bedrock catchments, which are often only examined in regards to flood emergence. Even in such areas, subsurface discharge is about 60-80% of total yearly discharge. This makes the regional analysis and modelling of subsurface discharge components of meso-scaled bedrock catchments an important task for hydrologists.

Problem scale

This project aims at catchments with areal scale between 1 to a few 100 km² (mesoscale). Bigger scales are pointless, as such catchments are non-homogeneous regionally (i.e. changing from mountaineous to plains). The process of groundwater flow is seperated into two phases: groundwater recharge (GWR) and groundwater discharge. The question of scale for the estimation of GWR is defined by meteorological inputs, land use and morphology and soil characteristics of the catchment.
Underlying hypothesis for the transition of scales for the calculation of GWR is the independence of time and spatial scales of soil-plant-atmosphere models. This hypothesis is only valid if GWR is only dependent on meteorological boundary conditions and the previously mentioned catchment characteristics. Interacting processes such as lateral fluxes or the reciprocation of local evaporation on the meteorological inputs must be negligible. The work groups dealing with evaporation in the SPP 'Regionalization in hydrology' show that these assumptions are feasible for small to medium-sized catchments. On this basis, the GWR can be modelled for catchments of any size, as long as the distribution of input data and regional parameters is known and the catchment is dividible into homogeneous subareas.
Regarding groundwater discharge in mesoscaled bedrock catchments, relatively sparse information about the geohydraulic parameters has to be assumed generally. This often makes it impossible to parameterize numerical physically-based geohydraulic models on the mesoscale within justifiable effort. The problem of scale is solved through the aggregation of existing geological information to hydrologically homogeneous subareas in this project. The groundwater discharge in these subareas is solely dependent on input (GWR) and the hydrogeological properties of the subarea. The models used are thus independent of scale under the assumption that bigger catchments can be divided into subareas and large-scale interdependencies between these subareas are negligible. Following these assumptions leads to further statemens regarding the problem of scale. The models can only be applied for catchments where the groundwater discharge in the hydrogeologically homogeneous areas reaches the surface water bodies. Areas where groundwater recharge and discharge are not identical, due to lateral processes involved, can only be studied with reservations. That is the case for large-scale riparian areas at rivers, as well as areas of lateral groundwater flow due to geological stratification or fault zones, among others.

Regionalization aim

The aim of the study was the development of suitable modelling approaches for the calculation of subsurface water flows in catchments between 1-500km². Due to the problems of parameterizing numerical groundwater models for bedrock, as described earlier, conceptional models are deployed for the estimation of flux components and residence times. A vital prerequisite is the ability to physically interpret the models and their parameters. This neccesitates the development of a physically based, regionally applicable parameter model for the objective derivation of conceptual parameters of spatial characteristics. In this study, these tasks had to be solved to create the basis for the regional modelling of groundwater fluxes in hydrologically unobserved catchments.

State of the art

Multiple studies showed that for both unconsolidated and carstic aquifers, subsurface flow is dominant compared to direct or surface discharge (e.g. SCHWARZE et al. 1991, SKLASH & FARVOLDEN 1979, STAUFFER & JOB 1982, HERRMANN & SCHÖNIGER 1992). Hydroisotopic and geochemic analyes proved that, contrary to current methods for determining groundwater flow (classic seperation methods e.g. in DYCK 1978, FH-DGG 1977, HGN 1993), indirect (usually equatable to subsurface) flow components make up between 60 to 80% of the annual total discharge, while often exhibiting residence times of several years to decades in the catchments. This is further confirmed through the investigations within the SPP (SCHWARZE et al. 1995).
The following assertions can be made about the mechanism of discharge development:

• precipitation or snow melt, respectively, infiltrate into the soil and seep through the unsatured zone to the groundwater table. Reduced hydraulic conductivity in the unsatured zone, e.g. from stratification, can lead to interflow.
• increase of groundwater table in dependency of the hydraulic conductivity of the aquifer, leading to a gradient of increased hydraulic potential which results in an increasing groundwater flow.
• exfiltration of the groundwater through pressure transmission into surface waters.

This mechanism is also present during flooding, where direct discharge is often around 10%, even for big flood events (SCHWARZE et al. 1995, UHLENBROOK & LEIBUNDGUT 1997).

The following procedures are applicable for determining subsurface flow:

• environmental isotopes and artificial tracer experiments allow predictions regarding flow paths and residence times of flow components for catchments smaller than around 50km² using physically interpretable, conceptual flow models. Such models are based on the analytical solution of the inverse problem via mathematical coupling of isotope input and output functions (i.a. SCHWARZE et al. 1991, 1995; LIEBSCHER et al. 1995; HERRMANN 1997), which primarily enhances process understanding. Application on the meso scale is usually not possible due to the high experimental, analytical and financial expenditures involved.
• another inverse method is DIFGA (SCHWARZE 1985, SCHWARZE et al. 1989, 1991), a flow component analysis. Bedrock catchments usually exhibit hydrologically different subareas due to their subsurface heteorgeneity, which leads to differentiable subsurface flow components. DIFGA identifies flow components, storage compartements and a full water balance from daily precipitation and discharges utilizing linear storages. These outputs are important model parameters of rainfall-runoff models and can thus be regionally determined using spatially distributed data. Due to the low data requirement, DIFGA can be used in a high number of investigated catchments.
• hydrograph seperation with a non-linear storage model (WITTENBERG 1994, 1997) assumes that storage compartments of all slow flux proportions are hydraulically connected and can thus be simulated with a single fictional storage. This is especially fitting for the hydrologic properties of unconsolidated rockbed. The simulated base flow can be used to estimate and regionalize groundwater recharge and retention.

Both DIFGA and the hydrological tracer and isotope methods are efficient means for the regional process analysis in SPP, which were further developed in this project and used as the basis for the process description and modelling. No comparable methods could be found in international literature.

The following methods are available for the subsurface flow modelling:

• the usual conceptual rainfall-runoff or water budget models do only have a formal, inappropriate estimation of subsurface flow to aid their optimal simulation of total flow hydrographs. Furthermore, most model parameters can not be measured directly but have to be inferred via model calibration. Nonetheless, these model have undeniable advantages (simplicity, robustness, few parameters, etc.) and are thus used in regionalization.
• physically-based process models such as TOPMODEL (BEVEN et al. 1995), SHE (ABBOTT et al. 1986) and ABBOTT & REFSGAARD 1996 incorporate approaches to simulate water flow in saturated and unsaturated zone faithful to the processes involved. Problematic for the regional application of such models is the fact that the neccessary model parameters might be measurable in principle, but can not be obtained for larger areas with reasonable expenditure. Furthermore, transition of scale is not fully solved, as most such model approaches have only be developed and are valid for small areas (such as hillslopes).
• numerical groundwater flow models are also physically-based. They can be used regionally if the problems of parameter identification and scale transitions are solved through a suitable parameter model. Somewhat comprehensive data (like groundwater levels and bore data, among others) for the application of numerical groundwater models are available for the unconsolidated aquifers of the NORDDEUSCHTE QUARTÄR, at least. The practicality of this approach is documented by D'AGNESE et al. 1996, DUVAL et al. 1996, LEIJNSE & PASTOORS 1996, SCHÖNIGER 1997 and SOMMERHÄUSER et al. 1997, among others. Unfortunately, the regional application of numerical groundwater models is not feasible currently for the areas of bedrock aquifers due to substantial problems in parameterization.

In conclusion: Conceptual approaches can be applied for the regionally transmissible modelling of subsurface flow on the meso scale, as long as the models and their parameters can be interpreted physically and the model parameters can be derived from regional characteristics with objective methods. Numerical groundwater models can be applied for unconsolidated aquifers, provided that the representative model parameters are determined in preprocessing. This approach generates physically plausible model results on basis of comprehensive data under usage of a regionally valid parameter model. Both model types depend on a realistic, regionally differentiable estimation of the GWR.

Methodology of the solution

The hydrological model should satisfy the following requirements:

• Description of different groundwater-relevant flux components in catchments between 10 - 500 km² size,
• Exclusive use of comprehensive data,
• Applicability to hydrologically unmonitored catchments.

A solution was developed to satisfy these requirements by simulating groundwater-relevant components with a conceptual hydrologic method. For this method, model parameters are inferred solely from areal characteristics via a new, physically-based algorithm. The flowchart of the solution is as follows:

• Determination of flow components in about 140 catchments from a hydrological catchment analysis:
  •     - analysis of the slow flow components and their water budget with DIFGA (SCHWARZE et al. 1994, 1997; KÖNIG et al. 1994) based on long-term daily discharges,
  •     - analysis of residence times, storage compartments and ratio of fast flow components based on the investigation of environmental isotopes (SCHWARZE et al. 1995).
• Analysis of geological, morphological and pedological areal characteristics and land use for a study region in Germany between Görlitz and Göttingen (40.000 km² bedrock area). The result of this second step is a classification of the study area primarily along hydrogeological aspects (concept of lithofacies). This is characterized by:
  • classification regulations for the determination of hydrogeological units based on the geological maps under utilization of further areal characteristics,
  •  application of the results from the hydrological analysis from step 1 to these units,
  •  development of pyhsically-based rules for the inference of model parameters for these hydrogeological units utilizing comprehensive data,
  • definition of a possible model parameter values (expected value and span) for each hydrogeological unit.
• Development of a parameter model (physically-based inference of parameters for process-describing models) as a case- and rule-based expert system:
  • application of the lithofacies concept to subdivide the study area into subareas of presumed hydrogeological homogeneity,
  • determination of the possible span of model parameter values dependent on the hydrogeological unit of the subarea,
  • application of a set of physically-based rules for the reduction of the possible parameter value span. These rules incorporate further hydrogeological, morphological, etc. information that is available for the subareas.
• Determination of the slow flow components per catchment using the hydrological model SLOWCOMP. SLOWCOMP describes the groundwater flow process with parallel linear reservoirs and is parameterized with the method described above. The model is integrated into a water budget model following the integration approach from Green & Ampt and physically-based approaches for evaporation and unsaturated flow (an ongoing development based on AKWA-M of MÜNCH 1994).