Assimilation of observations is a powerful tool to improve the predictive capabilities of models for radionuclide transport and fate. We describe results of numerical experiments on assimilation of the data on radionuclide contamination of the Black Sea in the result of Chernobyl accident using the three-dimensional model of circulation and radionuclide transport THREETOX. Data assimilation can be formulated as a procedure that contains two steps: update and forecast. On the update step THREETOX is run from the time of release or deposition to the time of the forecast using the updated input from all the measured data available at this period to be assimilated. On the forecast step THREETOX is run from the forecast time to the end of the modeling period using the results of the update step to produce the forecast of radionuclide transport. The data assimilation method called “method of iterations to optimal solution" (IOS) was used at the update step. The approach has been applied to the assimilation of the observational data of 137Cs concentration measured in the “Typhoon" surveys for the period June 1986 to September 1990. From the results of numerical experiments we conclude that usage of a data assimilation procedure can essentially improve predictive capability of the models for radionuclide transport. The computational time for the data assimilation method chosen is small in comparison with the general time of calculations. Results from the study will provide a better understanding of the processes of radionuclide transport in the seas. This novel approach was implemented in the EU DSS RODOS for the real-time simulation of radioactivity transport in the marine environment.