Evaluating the Performance of Transfer Function Models for Identifying Groundwater Disturbances in Infrastructure Projects

Sammanfattning

Hydrogeology is inherently challenged by the difficulty of making direct observations. To address this, numerical groundwater models are commonly developed tosimulate hydrogeological conditions at specific study sites. However, such modelsrequire extensive input data and are often time- and resource-intensive. As a result,alternative, data-driven modeling techniques are of growing interest.This study evaluates the performance of a data-driven Transfer Function model inidentifying and analyzing groundwater head disturbances caused by undergroundexcavation activities. The transfer function model used for this project is theopen-source Python package Pastas. Focusing on the Haga Station site within theVästlänken infrastructure project in Gothenburg, Sweden, the research comparestransfer function models with a numerical benchmark model developed using MODFLOW. Synthetic and constructed disturbance scenarios in the form of excavationshafts were modeled in the benchmark model to generate time series of groundwaterhead, leakage and infiltration. These time series were then used to test the transferfunction model’s ability to detect controlled disturbances. To validate the results, aqualitative assessment was also performed using observed field data.The findings demonstrate that transfer function models developed using Pastas canreplicate general groundwater trends and detect some disturbance signals. However,the models showed significant variation between different well simulations, with inconsistencies in how disturbances were interpreted. Additional limitations includethe overestimation of recharge and the underestimation of leakage into shafts following transfer function model calibration. Despite these challenges, the transferfunction approach shows promise due to its low data requirements and its ability tomodel complex hydrogeologic study sites. This work contributes to a better understanding of transfer function model capabilities and their potential application inhydrogeological assessments of infrastructure projects.