Some problems associated with forecasting water cooling in the entrance to the Baltic Sea are examined, using a diagnostic study from the severe ice winter of 1986/87 as a starting point. In the study, measured and numerically simulated data demonstrate that surface water cooling can be well simulated with the mathematical model presented by Omstedt (1987). Forecast tests during the same winter also indicate that the model may provide useful forecast information. However, because of the limited numbers of temperature and salinity profile measurements, the forecasts had to be accomplished with diagnostic calculations. These calculations will then produce the necessary initial data for the forecasts.