Natural Organic Matter (NOM) in South African water sources and its removal using ceramic membranes in water treatment plants

Abstract

South African water treatment plants employ the conventional water treatment processes which include processes such as coagulation/flocculation, sedimentation, filtration and disinfection. However, these traditional unit processes do not effectively remove natural organic matter (NOM) (around 35% at conventional pH). Detrimental to water treatment and distribution, NOM is the major contributor to the fouling of all membrane types (MF, UF, NF and RO), is a precursor to the formation of disinfection by-products (DBPs), impacts on the organoleptic properties, accelerates the clogging of the pores of activated carbon and thus decreases their effectiveness to remove pollutants, and certain fractions of NOM promote biological growth in the distribution. Previous research showed good correlation of NOM reactivity and treatability by employing comprehensive and sensitive analytical approaches such as biodegradable dissolved organic carbon (BDOC), polarity rapid assessment matrix (PRAM), fluorescent excitation emission matrix (FEEM), and synchronous fluorescence scan (SFS). Coupling the aforementioned techniques with chemometric methods such as parallel factor component analysis (PARAFAC), fluorescent regional integration (FRI), two dimensional correlation spectroscopy (2D-COR), spectroscopic indices such as humification index (HIX), freshness index (β:α) and fluorescence index (FI) would provide even deeper information on otherwise latent features of NOM fractions. There has not been a comprehensive study in South Africa pooling these analytical methods and chemometric techniques for characterization of NOM found in drinking water sources of South Africa. Additionally, there are no studies in South Africa that seek to introduce advanced treatment process such as ceramic membrane filtration to complement the conventional treatment process. To achieve the principal aim of assessing the applicability of ceramic membranes for the removal of NOM, it required that the NOM character and the extent of NOM removal by conventional methods be established. Data from the eleven drinking water plants spatially located in the five water quality regions of South Africa were pooled and a PARAFAC model fitting four components was established and validated based on the slit half criteria, and the distribution of the components at each water source was quantified using their maximum fluorescence intensities xi (Fmax). The value of Fmax was higher for terrestrial humic-like (C1) and fulvic like (C2) components in comparison to humic-like (C3) and protein-like (C4) components. It was established that coagulation was more effective for the removal of the humic-like fractions when compared with the other fractions. In the rapid sand filtration stage (RSF), bulk NOM removal (in terms of UV254) was found to be higher than that of fluorescent natural organic matter (FNOM) fraction, regardless of location of the plant. This suggests that non-FNOM fractions were removed much more effectively than FNOM fraction during the RSF stage. The selectivity of ceramic membranes was further explored by using ceramic membranes of different molecular weight cut off (MWCO) within the nanofiltration range so as to identify latent removal efficacies of ceramic membranes within a narrow range. Ceramic NF membranes with a disc configuration of 90 mm diameter, 2.5 mm thickness, and an effective filtration area of 0.00563 m2, with a porosity of 30% were used. It was established that higher MWCO (750 Da) membranes rejected FNOM fractions more compared to lower MWCO (450 Da) membranes. This was attributed to the effect of permeation drag. The results indicated that FNOM fractions could effectively be removed by higher MWCO membranes than lower MWCO membranes. In addition, higher MWCO membranes were particularly selective towards the removal of tryptophan (TPL) and humic like (HL) fractions of FNOM. Further, by employing 2D-Shige modelling technique it was deduced that the HL and TPL like fractions were more susceptible to removal than the fulvic-like (FL) fraction. These findings corroborated FNOM removal studies, which suggest that the rate of removal of FNOM fractions (TPL and HL, in particular) was higher than that of other fractions. The propensity to fouling, (i.e the deposition of NOM on the membrane surface and inside the membrane pores consequently leading to the reduction of flux) by model NOM fractions on ceramic membranes was investigated. The effect of membrane surface modification on fouling resistance was also studied by comparing the performance of both TiO2 atomic layer deposited (ALD)-coated and pristine membranes. Using model NOM foulants, the results showed that sodium alginate (SAL) caused the most extensive fouling on pristine membranes, and the ALD coating reduced the fouling potential of SAL by 35%. Cake filtration was xii found to be the least significant fouling mechanism occurring in feed solutions composed of bovine serum albumin (BSA) and SAL, and the most significant fouling mechanism of feed solution made up of humic acid (HA) and SAL. The fouling mechanisms were almost similar for both the coated and the pristine membranes. For the coated membrane the propensity to fouling correlated with complete blocking, standard blocking and intermediate blocking (R2 = 0.74; 0.74 and 0.40, respectively). For the pristine membrane, R2 values of 0.76; 0.75 and 0.60 were established for the respective fouling mechanisms of complete blocking, standard blocking, and intermediate blocking. However, cake filtration was found to be the most significant fouling mechanism for the coated membrane (R2 = 0.99), and the least significant fouling mechanism for the pristine membrane (R2 = 0.37). Coating was found to increase the hydrophilicity of the ceramic membranes; this was evidenced by the contact angle measurements, which showed a 23% decline in hydrophobicity of the coated membrane relative to the uncoated membrane.