Models are routinely used to remove the effects of the global ocean tides from GRACE data during processing to reduce temporal aliasing into monthly GRACE solutions. These models have typically been derived using data from satellite altimeter missions such as TOPEX/Poseidon. Therefore the Arctic ocean components of the tide models are not constrained by altimetry data, potentially resulting in errors that are likely to alias into monthly GRACE gravity fields at all latitudes.

Seven years of GRACE inter-satellite accelerations are inverted to solve for corrections to the amplitude and phase of major solar and lunar ocean tides at latitudes north of 50°N using a mascon approach. The tide model originally applied to our data was FES2004, truncated to maximum degree l_max = 90. Simulations are performed to verify that our inversion algorithm works as designed. Uncertainty estimates are derived from tidal solutions on land, and by subtracting two independent solutions that each use 3.5 years of data. Features in the M₂ and K₁ solutions that rise above the noise floor likely represent errors in the FES2004 model. Errors due to truncating the spherical harmonic expansion of FES2004 are too small, and errors in the land mask model (needed to transform sea surface heights into mass) only affect coastal areas and do not produce similar relative amplitudes in both tidal constituents. In the oceans north of 50°N, these residual estimates tend to reduce the FES2004 amplitudes for M₂, K₁, O₁, and P₁.

The power spectra of accelerations are analyzed, and reductions in the variance of accelerations not used in our inversion suggest that our results can be used to improve GRACE processing.