Natural organic matter (NOM) serve as precursors for disinfection byproducts (DBPs) in drinking water production making NOM removal essential in predisinfection treatment processes. We identified molecular formulas of chlorinated DBPs after chlorination and chloramination in four Swedish surface water treatment plants (WTPs) using ultrahigh resolution Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS). Chlorine-containing formulas were detected before and after disinfection and were therefore classified to identify DBPs. In total, 499 DBPs were detected, of which 230 have not been reported earlier. The byproducts had, as a group, significantly lower ratio of hydrogen to carbon (H/C) and significantly higher average carbon oxidation state (COS), double bond equivalents per carbon (DBE/C) and ratio of oxygen to carbon (O/C) compared to Cl-containing components present before disinfection and CHO formulas in samples taken both before and after disinfection. Electrophilic substitution, the proposed most significant reaction pathway for chlorination of NOM, results in carbon oxidation and decreased H/C while O/C and DBE/C is left unchanged. Because the identified DBPs had significantly higher DBE/C and O/C than the CHO formulas we concluded that chlorination of NOM during disinfection is selective toward components with relatively high double bond equivalency and number of oxygen atoms per carbon. Furthermore, choice of disinfectant, dose, and predisinfection treatment at the different WTPs resulted in distinct patterns in the occurrence of DBP formulas.

Absorbance, 3D fluorescence and ultrahigh resolution electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry (ESI-FT-ICR-MS) were used to explain patterns in the removal of chromophoric and fluorescent dissolved organic matter (CDOM and FDOM) at the molecular level during drinking water production at four large drinking water treatment plants in Sweden. When dissolved organic carbon (DOC) removal was low, shifts in the dissolved organic matter (DOM) composition could not be detected with commonly used DOC-normalized parameters (e.g. specific UV254 absorbance - SUVA), but was clearly observed by using differential absorbance and fluorescence or ESI-FT-ICR-MS. In addition, we took a novel approach by identifying how optical parameters were correlated to the elemental composition of DOM by using rank correlation to connect optical properties to chemical formulas assigned to mass peaks from FT-ICR-MS analyses. Coagulation treatment selectively removed FDOM at longer emission wavelengths (450-600 nm), which significantly correlated with chemical formulas containing oxidized carbon (average carbon oxidation state ≥ 0), low hydrogen to carbon ratios (H/C: average ± SD = 0.83 ± 0.13), and abundant oxygen-containing functional groups (O/C = 0.62 ± 0.10). Slow sand filtration was less efficient in removing DOM, yet selectively targeted FDOM at shorter emission wavelengths (between 300 and 450 nm), which commonly represents algal rather than terrestrial sources. This shorter wavelength FDOM correlated with chemical formulas containing reduced carbon (average carbon oxidation state ≤ 0), with relatively few carbon-carbon double bonds (H/C = 1.32 ± 0.16) and less oxygen per carbon (O/C = 0.43 ± 0.10) than those removed during coagulation. By coupling optical approaches with FT-ICR-MS to characterize DOM, we were for the first time able to confirm the molecular composition of absorbing and fluorescing DOM selectively targeted during drinking water treatment.

Non-target analysis of potentially toxic disinfection by-products at four waterworks in Sweden.

Rising organic matter concentrations in surface waters in many Nordic countries require current drinking water treatment processes to be adapted. Accordingly, the use of a novel nanofiltration (NF) membrane was studied during a nine month period in pilot scale at a large drinking water treatment plant in Stockholm, Sweden. A chemically resistant hollow-fibre NF membrane was fed with full scale process water from a rapid sand filter after aluminum sulfate coagulation. The combined coagulation and NF process removed more than 90% of the incoming lake water dissolved organic carbon (DOC) (8.7 mg C L(-1)), and 96% of the absorbance at 254 nm (A254) (0.28 cm(-1) incoming absorbance). Including granulated active carbon GAC) filter, the complete pilot plant treatment process we observed decreases in DOC concentration (8.7-0.5 mg C L(-1)), SUVA (3.1-1.7 mg(-1) L m(-1)), and the average nominal molecular mass (670-440 Da). Meanwhile, water hardness was practically unaffected (<20% reduction). Humic substances (HS) and biopolymers were almost completely eliminated (6510-140 and 260 to 10 μg C L(-1) respectively) and low molecular weight (LMW) neutrals decreased substantially (880-190 μg C L(-1)). Differential excitation emission matrices (EEMs), which illustrate the removal of fluorescing organic matter (FDOM) over a range of excitation and emission wavelengths, demonstrate that coagulation removed 35 ± 2% of protein-like material and 65 ± 2% of longer emission wavelength, humic-like FDOM. The subsequent NF treatment was somewhat less selective but still preferentially targeted humic-like FDOM (83 ± 1%) to a larger extent than protein-like material (66 ± 3%). The high selectivity of organic matter during coagulation compared to NF separation was confirmed from analyses with Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS), and liquid chromatography with organic carbon detection (LC-OCD), as coagulation exclusively targeted oxidized organic matter components while NF removed both chemically reduced and oxidized components. DOC removal and change in DOC character in the GAC filters showed marked differences with slower saturation and more pronounced shifts in DOC character using NF as pre-treatment. Fluorescence derived parameters showed a similar decrease over time of GAC performance for the first 150 days but also indicated ongoing change of DOM character in the post NF GAC filtrate over time even after LC-OCD indicated steady state with respect to outgoing carbon. During our trial iron concentrations were low (<30 ppb) and thus A254 could be directly related to the concentration of HS (R(2) = 0.9). The fluorescence derived freshness index (β:α) proved to be an excellent variable for estimating the fraction of HS present in all samples. Given the recommended limit of 4 mg L(-1) for chemical oxygen demand (COD) for Swedish drinking water, coagulation will need to be supplemented with one or more treatment steps irrespective whether climate change will lead to drier or wetter conditions in order to mai...

No abstract