One of the most promising new applications of remote observation satellite systems (RO) is the near-shore bathymetry estimation based on spaceborn multispectral imageries. In recent years, many experiments aiming to estimate bathymetry in optically shallow water with the use of remote optical observations have been presented. In this paper, optimal models of satellite derived bathymetry (SDB) for relatively turbid waters of the South Baltic Sea were presented. The obtained results were analysed in terms of depth error estimation, spatial distribution, and overall quality. The models were calibrated based on sounding (in-situ) data obtained by a single-beam echo sounder, which was retrieved from the Maritime Office in Gdynia, Poland. The remote observations for this study were delivered by the recently deployed European Space Agency Sentinel-2 satellite observation system. A detailed analysis of the obtained results has shown that the tested methods can be successfully applied for the South Baltic region at depths of 12-18 meters. However, significant limitations were observed. The performed experiments have revealed that the error of model calibration, expressed in meters (RMSE), equals up to 10-20% of the real depth and is, generally, case dependent. To overcome this drawback, a novel indicator of determining the maximal SDB depth was proposed. What is important, the proposed SDB quality indicator is derived only on the basis of remotely registered data and therefore can be applied operationally.