GOCE level-2 Tzz and Txx gravity gradients at satellite altitude are used in combination as input data to predict surface free air gravity anomalies over the Nordic region using Least Square Collocation. We test the performance of using covariance functions created
separately from Tzz gradients and surface free air gravity anomalies to predict terrestrial gravity anomalies with a direct comparison of 19 mgal level and much better (7-8 mgal) when surface data are filtered at the GOCE resolution. We found that the results are slightly
dependent of covariance function. Combined with DTU10 Mean Sea Surface model, both the NKG-2004 quasi-geoid model of the Nordic and Baltic Area and the one obtained using second generation GOCE spherical harmonic coefficients based on time-wise method can successfully reproduce the higher level of the Baltic Sea relative to the Atlantic Ocean.