Relationships between natural organic matter characteristics, reverse osmosis pretreatment and membrane performance

Natural organic matter (NOM) is one of many causes of reverse osmosis (RO) membrane fouling, but its fate in potential pretreatment processes is not well defined. The main objective of this research was to develop empirical models of organic fouling using NOM characteristics. The polarity rapid assessment method (PRAM) and ultrafiltration (UF) fractionation evaluated changes to NOM polarity, molecular weight (MW) distribution and non-polar and polar charge/size ratios during conventional filtration (CF) and CF with pre-ozonation and biofiltration (CF-O3/BF). CF and CF-O3/BF reduced NOM hydrophobicity, average MW and charge. Both treatment trains reduced larger NOM (>3 kDa) but had variable effects on smaller NOM (<3 kDa). CF-O3/BF also reduced the non-polar and polar charge/size ratios. These data showed PRAM and UF fractionation detected changes to NOM polarity, MW distribution and charge density during treatment.;This research also compared the efficacy of CF, microfiltration (MF), CF-O3/BF and MF with pre-ozonation (MF-O3) as pretreatment for RO desalination of Colorado River water. MF provided the best RO pretreatment because it had superior particulate removal. CF did not perform as well as MF because poor particulate removal caused colloidal fouling. The transformation of NOM by the ozone unit process in MF-O3 and CF-O3/BF increased organic fouling and cause severe declines in specific flux and salt rejection.;NOM characteristics also were correlated with specific flux and salt rejection changes. These data showed hydrophobicity and non-polar charge/size ratio had strong correlations with specific flux. Salt rejection correlations were strongest for weak anion exchange capacity and one of the polarity resins (CN). OF fractionation data were weakly correlated with specific flux and salt rejection. Linear regression was used to model the effect of organic fouling on RO performance using NOM characteristics. The non-polar charge/size ratio produced the most efficient specific flux model, although the one using hydrophobicity was nearly as good. The salt rejection model, which was not as accurate as the one for specific flux, included the weak anion exchange capacity of NOM. Modeling organic fouling and tracking NOM changes during treatment demonstrated two applications of the NOM characterization methods used in this research.