GECO: a global gravity model by locally combining GOCE data and EGM2008

• Published in: Studia Geophysica et Geodaetica
• Main Authors: GILARDONI, MADDALENA, REGUZZONI, MIRKO, D. Sampietro
• Other Authors: Gilardoni, Maddalena, Reguzzoni, Mirko, D., Sampietro
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2016
• Subjects: Earth’s gravity models; satellite missions; geoid combination; Antarctica; Antarc*
• Online Access: http://hdl.handle.net/11311/1018542; https://doi.org/10.1007/s11200-015-1114-4; http://link.springer.com/article/10.1007/s11200-015-1114-4

The EGM2008 model is nowadays one of the description of the global gravitational field at the highest resolution. It is delivered with two, not fully consistent, sources of information on its error: spherical harmonic coefficient variances and a geographical map of error variances, e.g. in terms of geoid undulation. In the present work, the gravity field information derived from a GOCE satellite-only global model is used to improve the accuracy of EGM2008 model in the low to medium frequencies, especially in areas where no data were available at the time of EGM2008 computation. The key issue is to set up the error covariance matrices of the two models for an optimal least-squares combination: the full error covariance matrix of GOCE spherical harmonic coefficients is approximated by an order-wise block-diagonal matrix, while for EGM2008, the pointwise error variances are taken from the provided geoid error map and the error spatial correlations from the coefficient variances. Due to computational reasons the combination is directly performed in terms of geoid values over a regular grid on local areas. Repeating the combination for overlapping areas all over the world and then performing a harmonic analysis, a new combined model is obtained. It is called GECO and extends up to the EGM2008 maximum degree. Comparisons with other recent combined models, such as EIGEN-6C4, and a local geoid based on new gravity datasets in Antarctica are performed to evaluate its quality. The main conclusion is that the proposed combination, weighting the different input contributions not only on a global basis but also according to some local error information, can perform even better than other more sophisticated combinations in areas where the input global error description is not reliable enough.