Аннотация:The probability distribution of several characteristics, in particular, sea surface and subsurface temperature and sea salinity simulated by the ocean circulation model of Nucleus for European Modeling of the Ocean in conjunction with data assimilation are determined. The data assimilation method, called as the Generalized Kalman filter method developed by the authors earlier and published in a number of studies is used. For assimilation the Agro drifter data have been applied for different model levels from sea surface until 2000 m. In order to define the probability distribution of sought model characteristics the Karhunen–Loeve decomposition of the covariance function has been utilized. The results of numerical experiments have been presented and analyzed.