Development and application of a modelling approach for distributed karst aquifer characterization and groundwater residence time derivation

Abstract

Spring protection in karst aquifers is particularly challenging since their high complexity thwarts their characterization by traditional field investigated methods. Especially the properties of the highly conductive conduit system are often poorly known. Therefore, most studies in karst aquifers are limited to spring responses and do not give any information on spatial distributions. Spatial information is required for the implementation of spring protection methods, however. Above all, the delineation of spring catchment areas and the distributions of groundwater residence times are essential for defining protection areas and estimating the effects of contamination events. The aim of this thesis is developing a modelling approach for the spatially distributed characterization of karst aquifers and the simulation of their groundwater residence time distributions. The main objectives during model development are determining the necessary model complexity, the kind and amount of required field data and the new information on aquifer structure and hydraulic parameters provided by the model, i.e. the contribution of the model to aquifer characterization. The simulations are divided into three modelling steps each of which focusing on a concrete simulation aim. The first aim is the delineation of spring catchment areas, the second the simulation of the velocity distribution within the conduit system and the third the spatial residence time distribution within the aquifer. The simulations are applied to the area of the Gallusquelle spring, a well-investigated field site in south-western Germany, where the results can be checked with field data. The models increase step-by-step in their complexity and parameter requirements so that the required minimum complexity for each simulation aim can be deduced. For spring catchment delineation, the average annual spring discharge of the Gallusquelle and the hydraulic head distribution derived from 20 observation wells are successfully employed for calibration. The spring discharge of five other springs within the model area is used for checking the plausibility of the results. Regarding the modelling approach, a hybrid model is recommended explicitly representing the karst conduits. The approximate location of the conduits is required as input data, while the large-scale lateral changes in conduit cross-section can be deduced from the model. The flow velocities in the conduit system are calibrated adding the breakthrough curves of two artificial tracer tests as objective functions. This greatly reduces the ambiguity of the model, so that not only the lateral change in conduit cross-section but also the total conduit volume can be deduced. Further, the roughness of the conduit system can be estimated with this approach. The simulation shows that the conduit roughness varies systematically throughout the conduit system of the Gallusquelle, which is necessary to take into account for reproducing the velocity distribution. For simulating the residence time distribution, a new modelling approach is developed combining a hybrid and a double-continuum approach. This new approach is successfully applied for two process studies. It is able to simulate the groundwater ages, life expectancies and residence times in the conduit network, the fissured system and the porous matrix of karst aquifers. The approach is applied for the Gallusquelle area and shows reasonable results. However, a lack of spatially distributed field data for calibration prohibits the verification of the residence time simulation at this stage. For this, groundwater age measurements at the surrounding springs would be required. However, sensitivity studies show that groundwater residence times have the potential to assist with the derivation of aquifer thicknesses, if such measurements are available.