The Gravity Recovery and Climate Experiment (GRACE) provides estimates of the earths static and time-variant gravity field. Solutions from various processing centres (GFZ, CSR, GRGS, JPL etc.) enable us to determine water mass redistributions on the globe. Given that land signals are generally large compared to anomalies over the ocean, an assessment of the latter requires a particularly careful filtering of the data. We utilized the Finite Element Sea-Ice Ocean Model (FESOM) to develop a filtering algorithm which relies on the spatial coherency of ocean bottom pressure (OBP) anomalies. For every position in the ocean, OBP anomalies are coherent over a considerable area. Thus, this pattern can be used to determine time series of GRACE-derived OBP anomalies at any position by weighting the time series in the vicinity with the correlation to the original time series. First results show that the highly variable currents like the Gulf Stream, the Kuroshio and the Zapiola Eddy can well be detected in the RMS monthly variability of GRACE data filtered with the pattern filtering method whereas these currents were only vaguely perceptible in Gauss filtered solutions. A global validation with in situ OBP data from a global database which has been compiled at the AWI shows that the correlation between in situ and GRACE data is improved by using the new filter in comparison to Gauss filtering. In this study we discuss the different results obtained from filtering data from different processing centres. Global, basin and local scales are considered as well as monthly and weekly time scales. Comparison to in situ data indicates that the method of data processing has an impact on the results. On a global scale and for example for the representation of the currents, these differences are less prominent.