Investigations on different spectral techniques to improve the gravimetric geoid model for the central part of Java, Indonesia, using terrestrial, airborne, and altimetric-based gravity observations

Abstract In the modern era, deriving accurate geoid models is crucial for various engineering and geoscience applications. The geoid facilitates the transformation of ellipsoidal heights from Global Navigation Satellite System (GNSS) measurements to orthometric heights and aids in determining geostrophic ocean surface currents. This study assesses different spectral methods, including two-dimensional Fast Fourier Transform (FFT), multiband spherical FFT, and one-dimensional FFT, for evaluating the Stokes’ function in gedsaata and Stokes’ kernel. The geoid modeling process is divided into three major stages: gravity data assessment, regularization of scattered gravity datasets onto the topographical surface, and geoid computation under the Remove-Compute-Restore (RCR) mode with Residual Terrain Model (RTM) reduction. Results indicate that the multiband spherical FFT method outperforms others, leading to the adoption of the Institut Teknologi Bandung-Central Java 2024 (ITBCJ24) geoid model, which achieves a root mean squared error (RMSE) of 0.068 m relative to GNSS-leveling observations and a relative accuracy of 4.409 ppm. The study also highlights the significance of gravity data coverage and distribution by comparing geoid models derived from terrestrial, airborne, and altimetric-derived gravity observations against the model relying solely on terrestrial gravity observations. Although the latter performs similarly at validation points, prominent discrepancies in geoidal height outside the validation region highlight the importance of uniform and dense gravity data coverage. Finally, the ITBCJ24 geoid model is compared to existing geoid models in Indonesia, showcasing its superiority with RMSE values of 0.117 m and 0.166 m for Indonesia’s regional geoid model of INAGEOID2020 and Earth Geopotential Model 2008 (EGM2008) model, respectively.