AbstractThis study investigated the accuracy improvement of the gravimetric geoid model in Turkey using densified grids of the gravity anomalies with the least-squares modification of the Stokes integral with additive corrections (LSMSA) method. In LSMSA computations of the geoid models, the surface gravity anomalies, which were derived from the complete Bouguer anomaly grid data set by forward modeling the topographic masses, were used. In the computations, in addition to the original 5′-resolution (∼9 km) Bouguer anomaly data set, its densified versions to 1′-resolution (∼2 km) also were used. The densification of the gravity grids was carried out using both planar- and spherical-type complete Bouguer anomalies. Thus, the effects of both gravity grid densification and the applied approximation for the Bouguer reduction process in gravity gridding on the geoid model accuracies were examined in order. The Bouguer gravity reductions were carried out using classical formulas for the planar approximation, whereas the gravimetric terrain corrections from the high-resolution SRTM2gravity global model were employed for the spherical approximation in numerical tests. The calculated geoid models were validated at 100 homogeneously distributed GPS/leveling (GPS means Global Positioning System) benchmarks over the country. In the tests, the geoid undulations derived from the models also were compared with the observations at six tide gauge stations along the coastlines. The latter provided an additional check for the assessment of the models’ compatibility at the coastal boundaries. In conclusion, the accuracy of the best geoid model solution (8.6 cm by means of the standard deviation of geoid undulation differences at GPS/leveling benchmarks) was achieved using 1′-resolution gravity anomalies restored from the spherical complete Bouguer anomalies in Turkey.