In predicting the nanofiltration (NF) process, concentration polarization (CP) should be seriously considered. Since many empirical expressions for the mass transport in the CP layer were reported, it seems difficult to choose a proper one. This paper discusses the CP development along the bulk flow direction in crossflow NF using a theoretical model, and then compares the differences of the concerned empirical expressions for the mass transport across the CP layer. The differences that the expressions caused were also presented with the NF experiment of the dilute PEG600 solution. The empirical expressions and S-K model were coupled to fit the experiment result. In the regressions of S-K model parameters, the pressure loss along the bulk flow channel and the bulk concentration variation were both considered. It was found that the concentration polarization along the bulk flow direction increased greatly before the boundary layer was fully developed. However, the polarization degree within a mesh length of 0.003 m was not serious. On the other hand, the empirical correlations gave sharply different CP degrees. Schock's and Hickey's expressions gave a comparatively close mass transport coefficient only when the bulk flow rate are around 0.15 m.s(-1). Eriksson's expression showed that the CP in the spiral-wound element was not serious only if the bulk flow rate was not much less than 0.1 m.s(-1). The results suggested that Eriksson's expression was still the best choice for the polarization correction while the other candidates seemed to overestimate the polarization.