Assessing the Impact of Methodological Differences on Geoid Model Performance

The geoid serves as a critical reference surface for precise mapping applications, particularly in the context of satellite-based positioning systems like the Global Navigation Satellite System (GNSS).While GNSS offers efficient positioning solutions, it relies on an ellipsoidal surface that lacks physical meaning for vertical reference, highlighting the need for accurate geoid height models.A precise geoid model is essential for converting geodetic heights into orthometric heights, which are crucial for practical applications.This study investigates potential discrepancies among geoid models derived from different methods, focusing on the Stokes-Helmert (SH), remove-compute-restore (RCR), and Kungl Tekniska Högskolan (KTH) methods.The primary differences among these methods lie in their approaches to modifying the Stokes formula and their reduction schemes.Conducted in the central part of Java Island, Indonesia, this study uses terrestrial gravity observations to model the geoid and GNSS/leveling data for validating the geoid models.The RCR method demonstrated the highest accuracy, with an RMS error of 8.4 cm, outperforming the KTH method (9.2 cm) and the SH method (10.7 cm).Discrepancies between SH and RCR models were less pronounced, with differences around 30 cm, compared to over 1 meter between KTH and the other methods.The comparison with the global EGM2008 model showed that the gravimetric geoid models were more accurate, with RMS differences reaching up to 10 cm, primarily due to systematic differences with the EGM2008 model.Statistical analysis using t-tests with 95% confidence intervals indicated that the differences among SH, RCR, and KTH methodologies were not statistically significant.Despite the RCR method's apparent superior performance, these differences did not achieve statistical significance.The study notes the limitations of using a relatively limited terrestrial gravity dataset and emphasizes the need for incorporating additional gravity data, such as recent airborne gravity datasets, to improve geoid model performance.Future research should also aim for denser GNSS/leveling observations with stricter measurement requirements to provide a more robust absolute assessment of gravimetric geoid models.