The performance of nanofiltration (NF) and reverse osmosis (RO) membranes involves complex interactions between water chemistries and properties of the membrane as previously observed by other researchers: however, little has been done on comparing the effects of different parameters at once. In this study, a factorial design was employed to integrate four independent factors, i.e., membrane type, dissolved organic carbon (DOC) composition, divalent cation and monovalent cation, to investigate their influences on specific flux decline and fouling. Quantitative linear models were also built to predict the specific flux decline and fouling formation. The model developed to evaluate the specific flux decline displayed a high prediction capability with an adjusted R-2 of 0.93, and with membrane properties and DOC composition exhibiting the highest influences. It was found that mechanical sieving was identified as the most important rejection mechanism and biofouling contributed more to the flux decline than abiotic fouling did. Membrane biofouling was found to be strongly affected by monovalent cation concentration and membrane properties, such as roughness and zeta potential. High membrane roughness and zeta potential (i.e. more negative membranes) along with high monovalent cation concentrations in the feed water were observed to contribute to biofouling control. Finally, the humic acid composition was identified as the main abiotic foulant, which was enhanced by high initial specific flux values. (C) 2004 Elsevier B.V. All rights reserved.