Simplified Groundwater Modeling Approach for Quaternary Deposits in Laxemar-Simpevarp, Sweden

Abstract

Groundwater is an important source of drinking water in Sweden. Groundwater level, level fluctuations, and flow direction need to be considered when a risk exists for water contamination. This paper presents the results from 3D groundwater modeling in quaternary deposits in a 71 km2 in Laxemar-Simpevarp, 320 km south of Stockholm close to the nuclear power plant Simpevarp. The site is of significant importance due to the risk for groundwater contamination. For this purpose, groundwater modeling system (GMS) was used. The model domain is characterized by a complex geology and large topographic variations. The main objectives were to predict the groundwater heads and flow directions. The modeling was done in two stages. In the first stage, a model was created using five heterogynous layers. However, the model calibration results showed large differences between the simulated and observed values. The complex stratigraphy and existence of thin layers that intersect at some points negatively affected the model performance. In the second stage, the model was simplified by reducing the existing five layers to two layers including a bedrock layer. The optimum values for hydraulic conductivity were derived using the stochastic modeling module. Results from the multilayer model showed flow toward the sea and in the quaternary deposits but not in high elevated rocks. The same results were found for the simplified two layer model in the first layer. Water flow in the second layer can be representative of the water flowing in the bedrock fractures. Results from the simplified model after applying the stochastic modeling method show better agreement between the simulated and observed values in the monitoring wells. The average residual was shown to decrease by 66% in the two-layered model. The 3D groundwater model GMS was successfully applied to the large Laxemar-Simpevarp region. Model simplification by reducing number of layers and including the bedrock beneath the soil proved to be essential for a better model performance. The same idea could be extended to similar groundwater modeling problems.