This work uses empirical data from the HELCOM database and a new empirically-based model to predict the concentration of cyanobacteria in the Baltic Proper. The aim has been to estimate nitrogen fixation. The inherent variabilities/ patchiness in the variables regulating nitrogen fixation are great. This means that different approaches may provide complementary information so that several relatively uncertain estimates may together provide less uncertainty in the estimate for nitrogen fixation in a given system. We show that there is marked variability in nitrogen fixation among different years (a factor of 20 between the year 2001 with the smallest value and 2005 with the highest value of about 900 kt/yr of N-fixation). The mean value for the period from 1997 to 2005 was 190 kt/yr. TN/TP based on median monthly data has been higher than the Redfield ratio of 7.2 since 1994. 6.5% of all individual data (n = 3001) from the surface-water layer (44 m) in the Baltic Proper for samples with temperatures higher than 15°C (when risks of getting cyanobacteria blooms are favoured) have TN/TP lower than 7.2. The mean TN/TP is 20 for surface-water sites with temperatures higher than 15°C, indicating that the average trophic conditions in the Baltic Proper are likely more limited by phosphorus than nitrogen. Nitrogen fixation is an important contributor to the nitrogen concentration and we give overall budgets for nitrogen and phosphorus in the Baltic Proper, including nutrient data from land uplift, which is the most important contributor for nutrients and often neglected in discussions about sources of nutrients to the Baltic Sea.