An improved procedure was developed and applied to fractionate natural organic matter (NOM) prior to characterization and subsequent ozonation under conditions representing typical drinking water treatment practice. Two surface waters, each a source of potable water, were investigated. One was a eutrophic lakewater with characteristic late summer non-volatile dissolved organic carbon (NVDOC) concentrations of 10 to 15 mg C/L. The other was a river water with a much lower NVDOC (2 to 5 mg C/L during most seasons).

NOM was fractionated using Amberlite® XAD-8 resin followed by XAD-4 and cation exchange (Bio Rad® AG MP-50). At pH 2, the XAD resins isolate acidic and hydrophobic NOM (humic and fulvic acids), whereas the cation exchange concentrates hydrophilic bases. Improving the procedure by incorporating XAD-4 and AG MP-50 resins allowed a significantly higher amount of NOM to be recovered and fractionated. Using this additional material a more thorough investigation of disinfection by-product (DBP) precursors could be attempted. When compared to previously reported fractionation procedures utilizing XAD-8 resin, the yield of XAD-isolable material approximately doubled with the addition of XAD-4 resin, and increased a further 25 % when ion exchange was included.

NOM isolated on XAD-4 resin was similar in character to that isolated on XAD-8, but displayed a lower molecular weight and higher aliphatic content (as determined by infrared spectroscopy) than that isolated on XAD-8. The characteristics of the hydrophilic bases isolated on AG MP-50 resin were different from those of the XAD-isolates, most notably in their greater aliphatic content when compared to the XAD-4 fractions, and in their substantially higher nitrogen content.

Bench scale ozonation experiments were conducted under various conditions of pH, alkalinity (total carbonate species concentration) and ozone dosage to assess by-product formation for a range of possible process conditions. Classes of DBPs which were measured included aldehydes, oxoacids and low molecular weight carboxylic acids. Several samples were also post-chlorinated or post-choriiminated to simulate typical drinking water treatment conditions. These were examined for the production of DBPs including trihalomethanes, haloacetic acids, chloral hydrate and cyanogen chloride.

Of the ozonation parameters studied, pH was shown to exert the largest effect on both ozonation and post-chlor(amfiliation DBP formation. Greater amounts of ozonation DBP precursors weie generally isolated using the XAD-4 resin when compared with XAD-8. This was true for both water sources and for all DBP classes. Lakewater NOM generally produced more DBPs per mass than did river water NOM. In contrast, higher amounts of precursor material for the halogenated DBPs, especially following ozonation, were isolated on the XAD-8 resin and from river water. These results illustrate the dependence of DBP formation on source water characteristics as well as the precursor isolation method employed. DBP yields were not directly related to the concentrations of precursor material utilized, emphasizing some of the problems of extrapolating data obtained at bench scale to conditions encountered in typical water treatment.

In summary, NOM was isolated from a eutrophic lakewater and from river water, both sources of potable water. To improve the recovery of NOM, Amberlite® XAD-4 and Bio Rad® AG MP-50 resins, in addition to XAD-8 were used for NOM fractionation. Solutions of fractionated material were ozonated, and in some cases subsequently chlorinated or chloraminated to investigate the origins of various classes of DBPs of interest in drinking water treatment.