Abstract:In this paper, a global gravity field model entitled TJGOCE01 up to the degree and order 250 is recovered from about 6 months of GOCE gravity gradient data and 12 months of GOCE orbit data, using a direct approach. The gravity gradient data are filtered by an infinite impulse bandpass digital filter, and gross errors are detected and removed using the threshold and a moving window threshold methods. The gravity gradient observational equations are established in the gradiometer coordinate system directly and the kinematic orbit observation equations are established by using a modified short-arc approach. The weights of the two kinds of observations are determined based on their prior accuracies, and the regularization approach based on Kaula’s rule is applied to solve the normal equation. The geoid error and cumulative geoid error up to degree and order 250 of TJGOCE01 model with respect to EIGEN6C2 model are 19.4 mm and 177.9 mm, respectively. The test results with the GPS leveling data in North America show that the root mean square error of TJGOCE01 model is 0.544 m, which is superior to the second generation models developed by the European Space Agency using time-wise and space-wise approaches.