Preferential flow has growing importance in several research fields such as groundwater supply, contaminant transport and geotechnical engineering. Modeling of preferential flow allows for the estimation of its influence on the hydrogeological settings of a study site since preferential flow can alter the velocity and rate of groundwater flow by several orders of magnitude. Geophysical methods such as resistivity tomography and self-potential mapping have been frequently employed for the identification of preferential flow paths in subsurface.
Self-potential mapping and resistivity tomography were applied on a real-world study site at Deisermillen, in the Upper Mosel River Valley, Luxembourg. In 1964 a dangerous landslide occurred at the study site which destroyed several residential buildings and a regional motor road. Thereafter, three investigation campaigns were conducted which delivered a detailed insight into the lithological and hydrogeological settings of the study site. On the study site conductive colluvium sediments overlie less conductive clayey marls. Observations of groundwater heads in piezometer wells and pumping tests indicated an unexpected high rate of groundwater flow in the less conductive marls caused by preferential flow. Geophysical surveys and analysis of drill core records were employed to conceptualise and parametrise preferential flow in a numerical model.
The geophysical investigation showed possible preferential flow paths in the subsurface, however discrimination between different groundwater flow systems in the uppermost colluvium and preferential flow conduits in the deep-seated marls was not possible. Therefore, a detailed analysis of drill core records and an observation of groundwater heads in piezometers were employed in conjunction with pumping tests. A lateral preferential flow conduit was identified in less conductive sediments of marl which probably resulted from the dissolution of a thin gypsum interlayer. A 3D fully coupled physically based hydrogeological model was compiled for the study site using the HydroGeoSphere code. The lateral preferential flow conduit was integrated into the model’s domain by adapting the model’s mesh and sublayers which allowed for the attainment of an exact correspondence between observed and modelled groundwater heads.
This study shows that the newest numerical models allow for modeling of the preferential flow with a high degree of precision, however conceptualization and parametrisation of preferential flow paths in a numerical model based on geophysical surveys is not straightforward. Therefore, additional sources of data such as from the analysis of drill core records and hydrogeological field tests are necessary.