Quantifying Groundwater-Surface Water Interactions with Transfer Function Models

Abstract

This project aims to develop and evaluate a data driven method for analysing groundwater-surface water interactions. These interactions are analysed by using time-series analysis and transfer function modelling to create models for ground water head, stream flow and stream level. The method is applied to a case study: a tunnelling project in Kolmården, Sweden, where a recipient stream in a nature reserve risks being affected by tunnel-induced groundwater drawdown due to leakage. The method consists of two modelling stages. In the first stage groundwater head models (GWMs) are developed. In the second stage, the GWMs are used to extend existing groundwater head time series and develop three types of models: models that estimate baseflow (BFM), stream total flow (TFM), and stream level (SLM). Additionally, the results are simulated using 18 different methods for estimation of potential evapotranspiration, to see how different methods impact results. Results from the first modelling stage indicate higher model performance for models representing groundwater level in soil, compared to rock, and for models using longer calibration periods across all four seasons. Results from the second modelling stage present very poor results for the stream baseflow models (BFMs), which is proba bly related to the baseflow separation procedure’s inability to account for specific local conditions. Total stream flow models (TFMs) and stream level models (SLMs) perform much better, with R2 -values around 0.6 in validation. However, all models show signs of overfitting, with R2 -values above 0.8 in calibration. The contribution from groundwater head to the variations seems to be higher for the SLMs. The developed method constitutes a simplified and data driven approach for analysing groundwater-surface water interactions, making it possible to study, e.g., how tunnel induced groundwater drawdown might affect a watercourse. Further development is recommended, e.g., using additional model evaluation metrics and simulating different leakage rate scenarios with pump test data. The method shows promise for allowing effective modelling of the dynamic response of groundwater-fed surface water to hydroclimatic variables, and can be seen as an additional tool for representing hydrogeological systems.