This report is the final KTH working package (research activity) report within the SiL2.0 (Stomnät i Luften 2.0) project. One of the main issues in the SiL2.0 project is to ensure a common and robust infrastructure for positioning based on GNSS, providing that all measurements are in the same reference system (horizontal and vertical) regardless of the chosen measurement method. The Swedish Transport Administration (Trafikverket) uses machine guidance and Global Satellite Positioning Systems (GNSS) technology when constructing different infrastructures, e.g. roads and railways. This working package aims to determine an improved geoid model over construction areas using a dense network of GNSS- leveling points. This improved model is called Project Adapted Geoid Model (PaGM). A precise geoid model is needed to correctly convert the measured altitudes obtained from GNSS i.e. h (with respect to the reference ellipsoid) to altitudes above the geoid surface (H). In other words, a geoid model is a mathematical representation of the Earth's gravity field, which is used to convert between ellipsoidal heights and orthometric/normal heights. Lantmäteriet released a geoid model called SWEN17_RH2000 (Ågren et al. 2018), with an uncertainty of about 8-10 mm (1 sigma) on the mainland in Sweden. Its uncertainty is higher (2-3 cm) in some areas, e.g. lakes and mountains. This study aims to assess the quality of the geoid model in different locations and investigate how to improve its quality with the help of dense GNSS-leveling points, i.e. points with precisely known height over ellipsoid and over geoid. Finally, a project-adopted geoid model is determined for future precise positioning using network real-time kinematic (NRTK). Within this project, we have studied PaGM using dense GNSS-leveling points along the project Norrbotniabanan. Norrbotniabanan includes planning and construction of a new railway between Umeå and Luleå, a distance of approximately 250 km. Similar studies were also conducted for the high-speed railway projects Ostlänken and Göteborg-Borås, but this report covers the results for only the Norrbotniabanan PaGM. Different scenarios are evaluated to determine the PaGM e.g. employing different GNSS-leveling data obtained by static (4 and 48 hours static data) and NRTK methods. This project (working package) is the continuation of previous preliminary studies in Sweden, such as Ågren and Ohlsson (2016) and Vium Andersson (2017).