Evaluating an Automated Approach for Groundwater Model Setup in Sweden

Abstract

Groundwater modelling is a complex and time-consuming process, especially in re gions with detailed topography and heterogeneous geology. Therefore, this thesis evaluates and compares two modelling approaches - a traditional MODFLOW-based workflow using ModelMuse, and an automated framework using HydroModPy. The comparison is applied to a real-world case study of the Kolmården Tunnel, a section of the Ostlänken high-speed railway project in Sweden. The aim was to evaluate whether HydroModPy can replicate key behaviours of an expert-built MODFLOW model and assess its adaptability to Swedish input formats and modelling practices. Both models were implemented from the same input data, and their development, calibration, and outputs were systematically compared based on hydraulic head, boundary flows, and model performance at selected observation points. HydroModPy offered valuable automation for pre-processing and grid gen eration but lacked flexibility for custom geological layering, user-defined grids, and site-specific boundary conditions unless custom code modifications were made. In contrast, ModelMuse-based workflow provided full control over model architecture and produced more stable simulation results. The study concludes that HydroModPy is a promising tool for automated model development and is capable of handling complex groundwater systems - provided the user has sufficient programming expertise. Without this, its application in de tailed studies remains limited. The project further highlights the importance of early expert involvement in data interpretation and model design. These findings are rel evant to model developers aiming to streamline model construction and researchers evaluating automation tools for hydrogeological impact assessments in complex in