In nanofiltration (NF) and reverse osmosis (RO) membrane systems, 75-90% of thefeed water is converted to product, and as a consequence, dissolved compounds areconcentrated in the membrane concentrate. Subsequently, sparingly solubleinorganic compounds such as calcium carbonate and barium sulphate becomesupersaturated and may precipitate in a membrane element (scaling). Scaling leadsto an increase in energy consumption and chemical cleaning frequency. This isundesirable and, therefore, an accurate prediction of the risk of scaling isimportant. Traditionally, the risk of scaling in a membrane system is estimatedby calculation of the supersaturation ratio of sparingly soluble compounds in thebulk of the membrane concentrate. However, due to concentration polarization, thesupersaturation ratio near the membrane surface and, therefore, the risk ofscaling, is higher than in the bulk of the membrane concentrate. That is why foraccurate prediction of scaling, the supersaturation ratio at the membrane surfaceshould be calculated by incorporating concentration polarization. In this study,a model is presented, called the Supersaturation Prediction Model, which is ableto calculate the supersaturation ratios of sparingly soluble compounds at themembrane surface. The model is a combination of a concentration polarizationmodel from the literature and a model for the calculation of the supersaturationratio of sparingly soluble compounds. The concentration polarization model fromthe literature was fitted to measured average concentration polarization factorsof magnesium sulphate in a spiral wound reverse osmosis membrane element operatedwith different permeate fluxes and linear flow velocities of the membraneconcentrate. The fitted concentration polarization model was also able to predictthe average concentration polarization factors of sodium chloride and magnesiumchloride in a 2,5"*40" spiral wound nanofiltration membrane element.Nanofiltration pilot research showed that the maximum achievable conversion(without the risk of scaling) was about 83.5%, when the design consisted of a 2-1staging with 6 membrane elements per pressure vessel. The maximum achievableconversion was 3.5% lower when only 3 membrane elements were applied per pressurevessel (while maintaining the same staging and average flux). The SupersaturationPrediction Model showed that the ion specific concentration polarization factorswere much higher when 3 membrane elements per pressure vessel were applied. Themodel also predicted that, among others, calcium carbonate, barium sulphate andiron carbonate were supersaturated at the membrane surface. The supersaturationratios were on the same level at a conversion of 80 % and at a conversion of83.5%, with the design containing 3 membrane elements per pressure vessel and thedesign containing 6 membrane elements per pressure vessel respectively. Includes 17 references, tables, figures.