The Earth’s gravity field dedicated missions provide homogeneous and uniformly accurate information on the long wavelengths of the Earth’s gravity field. Many different global geopotential models have introduced by Earth and space research centers during the last years.
The geopotential models derived from the GRACE satellite measurements are compared with the other models at different frequency levels. As expected, the achieved results show that the GRACE-derived models are more accurate in low frequency. Although the GRACE-derived models have lower accuracy in medium frequencies they outperform the other alternatives.
Moreover, accuracy of the geopotential models are usually evaluated by comparing the geoidal heights derived from the models with those of the local GPS-leveling stations. Due to the different signal content of two observation types, it may lead to incorrect conclusion. For compatibility, we propose the idea of filtering the high-frequency signals of the GPS-leveling observations. We have employed spatial filters with the uniform and Gaussian kernels to filter the high-frequency components of the terrestrial (GPS-leveling) data.
Herein, we introduce an innovative approach for evaluation of the geopotential models. In the new method, the geodial height differences over baselines with different length are used for the evaluation of the geopotential models. The models behave differently for the baselines of different lengths and orientations. Sub-meter accuracy can be obtained using different models for baselines with lengths up to 50 km. However, for the longer baselines the geoidal height accuracy behaves differently. Furthermore, the lowest accuracy is observed for the baselines in the south-north direction. It might correspond to the accumulative error of the leveling network of Iran which spans the whole country from the Persian Gulf northward.
It is also recommended to perform data screening and outlier detection for terrestrial data. Among 490 GPS-leveling stations used in this study, 40 stations have been removed because of their significant differences (more than 2 meters) with the global models.
Compared to espite the classical point-wise method, the most recently released model, EIGEN-CG04 is the most accurate one according to our analysis over the GPS-leveling network of Iran.