This Thesis focuses on the evaluation of the CUBE-algorithm. This was mainly developed for the application on measured data from multibeam echosounders. The measured values are generally unevenly distributed and partly inaccurate. The aim of the algorithm is to calculate a regular grid with estimates for the water depth in fixed positions out of these values. The special thing about that is that, in addition to the actual measured values, there is extra information about the data included the calculation. These are in particular information about the precision and measuring properties of the used measuring instruments. With the help of this knowledge it is tried to determine the true water depth out of the measured values to exclude inaccurate data from the regular grid. In this thesis the algorithm was applied to deep sea multibeam echosounder data, which has been measured by research vessel polarstern and assessed regarding the reliability with which incorrect values are excluded from the grid. The polarstern is operated by the Alfred Wegener Institute and used to explore the polar seas. At the institute the programs HIPS and SIPS of the company CARIS and Fledermaus by QPS are used to clean the data manually. The programs each contain an implementation of the CUBE-algorithm. After the application of the algorithm to the measured values with both programs, the visible differences in the results were analyzed. The analysis was carried out using the program ArcMap by Esri. Using a color-coded visualization of the grid and through difference grids the optical and metric deviations of the results of the algorithm to the results of a manual data cleanup were examined and compared with each other. The analysis of the comparisons has shown, that the deviations from the grid created by the CUBE-algorithm to manually cleaned data are dropped compared to uncleaned data. That means, referred to the used deep sea multibeam echosounder data, that the algorithm succeeds in excluding a part of the inaccurate data. The reliability with which this is achieved increases with the density of accurate values. In relation to Fledermaus the resuls from HIPS and SIPS showed the lover deviations compared to the manually cleaned data.