General conclusions

Due to the duality of flow and generally large catchment areas of karstic spring, karst aquifer systems are efficient fresh water sources. The duality is a consequence of hydraulic property differences between different karst compartments and adds a high degree of anisotropy. Several conceptual models and approaches are frequently applied to characterize flow in karst aquifers, or to be more precise, the phreatic part of the system. Integral characterization techniques are scarce because of the low understanding and high complexity of different processes and the high data demand to describe those processes sufficiently. Large-scale abstraction tests are suitable to cover the scale of dominant aquifer heterogeneities in the phreatic zone. On the one hand, measured drawdown curves contain information about the conduit system as well as the fissured matrix and therefore about the two interacting components of the characteristic dual flow behaviour. On the other hand, drawdown curves are a spatial and temporal superposition of different heterogeneities, which means that the ambiguity is normally high.

This thesis presents the application of discrete conduit-continuum models to assess karst specific heterogeneities and their spatial distribution. It is focused on the interpretation of the drawdown curve measured inside the triggered conduit. The different chapters examine a variety of heterogeneities which influence the drawdown at (1) early time, (2) intermediate time and (3) late time as well as the transition periods in between.

The early time response of a large-scale pumping test is influenced by direct storage. The consideration of direct storage is essential for karst related analyses.

Without a fast responding storage, for example the conduit-associated drainable

corresponds to the conceptual model of DROGUE (1992), which is only one potential conceptual karst model. For the Cent Fonts catchment (Languedoc, France), as an example of a mixed flow karst system, this conceptual model is non-applicable. The concept of the fast responding storage is well known from pumping test interpretations. The dimensionless wellbore storage can be applied to compare drawdown curves independent from hydraulic properties or the pumping rate. Another dimensionless parameter, which can be defined to affect the drawdown behavior during early time, is the skin damage factor. The skin damage factor accounts for pressure changes in the interface between a highly conductive feature and the matrix.

By using the dimensionless wellbore storage and the skin damage factor for the interpretation of water abstraction directly from idealized karst conduits, a characterization scheme is developed. The characterization scheme considers different degrees of karstification on two different scales and represents the interaction of the conduit and the matrix during the first transition period. The first scale is the local scale, ranging from a centimeter up to a few meters along the conduit. The second scale is the regional scale, defined by the hydraulic matrix parameter. The characterization schema excludes the representation of diffuse flow karst systems. Two flow systems can be distinguished depending on the skin damage factor: (1) matrix restrained flow regimes and (2) interface restricted flow regimes. The interface restricted flow is defined by low exchange permeability. In mature karst systems the low exchange permeability can be explained by additional flow restrictions, for example consequences of turbulent exchange flow, but also restricted inflow form other karstic features in the vicinity of the conduit. The possibility of the extension of the characterization schema on conduit-influenced flow regimes exists. Due to the restricted applicability of the skin damage factor, as well as the dimensionless wellbore storage, an adaption of the parameter or a definition of new parameters is demanded.

Conduit-influenced flow regimes are characterized by restricted conduit conductivity (finite conductivity) influencing karst spring hydrographs (KOVACS ET AL., 2005) as well as the drawdown behavior of large-scale pumping tests.

The infinite conduit conductivity is the consequence of increased friction factor and therefore accounts for head gradients along the flow direction. The head losses depend on the mean roughness height and the conduit diameter and to a

lesser extent on the tortuosity. Finite conduit conductivity influences the flow pattern during linear flow. Due to the increased head losses, the conduit flow is restricted. Consequently, the exchange flow with the matrix increases near the wellbore and is therefore not uniform along the conduit. For finite conduit conductivity, the linear conduit flow is superposed by the radial matrix flow in the vicinity of the wellbore resulting into bilinear flow. The characteristics of bilinear flow are higher conduit drawdown at the beginning of pumping and a smoother slope of the drawdown and derivative curve (quarter-unit slope) as well as increased matrix drawdown in the vicinity of the wellbore. The non-uniform exchange flow along the conduit is the major difference between conduit-influenced flow regimes and matrix restrained flow regimes. For conduit-influenced flow regimes, the application of turbulent flow equations is demanded. Due to the infinite conduit conductivity of matrix-restrained flow regimes, laminar flow equations can be sufficient even for high Reynolds numbers. For those karst systems, the ratios between mean roughness height and conduit diameter are most likely not covered to by traditional Moody diagram (MOODY, 1944) used to predict the friction factor depending on the ratio and the Reynolds number and separate between laminar and turbulent flow.

Traditional tools for pumping test interpretation normally assume an infinite aquifer extent. This assumption is used as a defined flow pattern for the type curve analysis and the estimation of the matrix transmissivity. For the interpretation of large-scale pumping tests, heterogeneities or boundary conditions need to be considered as influencing the drawdown behaviour during reservoir flow. Those boundary conditions influence the shape of the drawdown curve resulting in a flow pattern different to radial flow. Therefore, the extension of the cone of depression is not a function of Darcian flow. The calculation of the transmissivity based on Darcian flow is not valid for flow patterns different to radial flow.

The interpretation of diagnostic plots can be enhanced by the application of the flow dimension concept. The flow dimension is generally applicable to all periods of pumping and adds additional information without further research effort. The application of the flow dimension on the simulation drawdown curve of a DCC model extends the traditional approaches, which are focused

represents the flow pattern on large scale as a consequence of superposed heterogeneities and can therefore be used to exclude certain conceptual models or fractions of conceptual model, e.g. boundary conditions. In addition to the forward simulation of different conceptual models (pattern matching), DCC models in combination with calibration tools are also able to estimate hydraulic parameters.

The influences of the hydraulic parameters in respect to the three main periods during pumping tests are summarized in Table 7.1.

Table 7.1: Dominating parameters of the different periods during pumping tests

Period Hydraulic parameters Dimensionless