Analysis of groundwater flow through earth dams and embankments is commonly conducted by developing and running 2D-vertical profile numerical models along selected transects (Gikas and Sakellariou, 2008 ), assuming they are aligned with the major direction of groundwater flow, and negligible flow components at different directions.
This above approach seems to work sufficiently well for the majority of the problems, representing standard geometrical structures and simplified flow patterns. In problems involving complex geometries and geological structures, the use of the 2D profile-modelling methodology fail to realistically represent the pattern and the dynamics of groundwater flow in the system.
The application of numerical models based on 3D-structured mesh, is burdened by several limitations and functional constraints that reduce their suitability in simplified cases. The requirement of large number of elements/cells in many different layers with significant number of redundant elements at pinch out parts, discretization difficulties in approaching key structural details, significant model run times, and convergence difficulties are few examples of the above limitations.
Recent developments of the industry-standard numerical code FEFLOW (Diersch ,2104, DHI, 2018), in the field of 3D-unstructured finite element mesh, proves to work sufficiently well in overcoming the above limitations, providing improved solutions in simulating groundwater flow systems with significant 3D flow components in complex geometrical structures.
This paper presents the application of 3D unstructured - mesh finite element technique in simulating transient seepage of groundwater flow through the body and the vicinity of an earth water dam, in Mornos river catchment, Greece. This dam is 135m high and constitutes the major source of potable water supply of the metropolitan area of Athens, Greece.
The model was designed to incorporate all the structural components of the dam and the surrounding bedrock. It was calibrated at transient state conditions, against monthly water level data observed at various piezometers installed at multiple locations in the body of the dam. The model was run for a simulation period of 30 years, replicating with sufficient accuracy the field observations and the anticipated patterns of groundwater flow. Potential impacts of preferential flow patterns and seepage through underlying bedrock fractures are discussed. Alternative model designs comprising traditional structured mesh as well as 2D profile modelling and the unstructured mesh technique are explored.