Underwater road tunnels have become critical infrastructure in Norway and, more recently, in the Faroe Islands. Initially excavated to bypass fjords, they later developed to replace ferry services across longer marine straits. The 24 km-long Suðuroy Tunnel will be the longest underwater fixed link in the archipelago and among the longest worldwide. Due to its exceptional length and prolonged poorly lit driving conditions, special safety measures are planned, including two large equal-design caverns functioning as U-turn roundabouts for emergency vehicle evacuation. This study examined the feasibility of excavating these underground widenings within the geological and geotechnical context of the site, primarily composed of sub-horizontal basaltic lava flow cores interbedded with lava breccias and volcaniclastic sandstones. To address knowledge gaps in local lithostratigraphy, rock strength, deformability, and in-situ stress conditions, geological and geomechanical characterisation was carried out using desk studies and field mapping. Preliminary cavern geometries were defined for numerical modelling, with two design solutions emerging: one with a central pillar and one full-span cavern without internal supports.

Continuum-based simulations explored various cavern sizes. Within basaltic flow core units, the stress regime was mostly moderate, with maximum principal stress less than half the uniaxial compressive strength. Tensile stresses reached 80–85% of the rock’s tensile strength, and displacements remained under 4.5 mm, suggesting cavern excavation feasibility. The ratio of horizontal to vertical in-situ stress had a greater influence on stability than rock mass parameters such as Young’s modulus, Poisson’s ratio, or compressive strength. Caverns in both planned positions along the alignment showed similar stress and displacement responses and comparable geomechanical behaviour across the two sites. Simulations involving lava breccia and volcaniclastic sandstone indicated higher instability risks: breccia exhibited potential for spalling and damage accumulation, while sandstone approached the onset of failure. Based on modelling results, two optimal cavern configurations were proposed: (i) a 20-m outer radius with a 4-m central pillar, and (ii) an 18-m radius full-span excavation without support.

This study advances understanding of large-scale underground widenings designed for road roundabouts. It also highlights the need for more site-specific data on stress fields and rock mass strength for the current project, as the models primarily relied on desk-based estimates. Complementary approaches, such as combining continuum modelling with discrete element methods and incorporating structural reinforcement, would provide further insight into brittle failure mechanisms and help evaluate the full feasibility and improvement of such cavern excavations.

The presence of wells in reservoir simulations can be represented by utilizing a well model within the finite volume framework of the simulator. A commonly applied approach is the Peaceman well inflow model, which places wells at the center of a specified grid cell of the reservoir model. Nonetheless, several studies have investigated the ability of placing wells off-center. Capturing the exact well position can significantly impact the reservoir response of the simulation model, particularly for heterogeneous reservoirs, as spatial heterogeneity plays a more pronounced role when wells are no longer located at the center of the grid cell. Furthermore, exact well position modeling would allow for more realistic well presentation within the reservoir simulator. Although the performed studies have successfully modeled
wells off-center, they are course-grid approaches. Therefore, this study investigates the application of an unstructured grid-based well model to accurately account for both the exact well location within a grid block and the surrounding reservoir heterogeneity in the calculation of a steady-state well index.

To achieve this, an improved well inflow model is established using Gmsh (three-dimensional finite element mesh generator) and the open Delft Advanced Research Terra Simulator (open-DARTS). The model implements an unstructured grid on a locally structured reservoir geometry, where typically five structured blocks are integrated. The model is verified against the Peaceman well inflow model for a center position and values match the Peaceman well model with a 5% accuracy. The model is validated for different well placements, permeability profiles and a considered number of structured reservoir blocks. Additionally, the validation study includes an accuracy assessment of the model against a full unstructured approach. The improved well index model is finally applied to the wells of the Geothermie Delft (GTD) project, enabling well index estimations based on true well trajectories.

Validation of the model shows that both well placement and local reservoir permeability impact the computed well index. A comparison of the proposed well index evaluation with the fully unstructured model indicates that incorporating more surrounding reservoir blocks within the well index calculation leads to closer correspondence with the fully unstructured approach. Additionally, the study reveals that the spatial discretization method within the finite volume framework used in open-DARTS influences well index values and can result in errors, depending on the choice of discretization approach. Finally, the results of the application of the proposed approach to the real trajectory of wells in the Geothermie TU Delft (GTD) project demonstrate that the improved well index model generally yields lower well index estimations than the Peaceman well model. The effect of this difference is further tested by simulating the full reservoir with the newly obtained well index values. These results depict that the well pressure profile for both the injector and producer varies depending on the well model, especially for later simulation times. The improved well model captures more variations in the pressure profile, emphasizing its capability for more accurate well modeling.