Disinfection by-products in high quality recycled water production with high pressure membranes

Abstract

Reuse of high quality recycled water (HQRW) is a new strategy for meeting current and future demand for potable water, whereby highly treated secondary effluent is used to augment existing water supplies. Reverse osmosis (RO) membranes are commonly used to generate HQRW from secondary effluent but require the application of disinfectants during the pre-treatment stages to limit biofouling on their surface and to avoid loss of performance. As an unintentional consequence of this treatment, disinfection by-products (DBPs) are formed. Due to their potential adverse effects on human health, the understanding and control of DBP formation and fate during HQRW production at advanced water treatment plants (AWTPs) is vital. The overall objective of this thesis was to fundamentally understand the individual DBP formation in secondary effluent and their subsequent fate across high-pressure membranes at bench, pilot and full-scale. Individual DBPs were studied within the groups of trihalomethanes, iodinated trihalomethanes, haloacetonitriles, halonitromethanes, haloketones, haloacetamides and haloacetaldehydes. At full-scale AWTPs DBPs were consistently measured from the point of disinfection, across the various barriers until the final treated water. Operational parameters that are of relevance at full-scale AWTPs and potentially can affect DBP formation were investigated under controlled conditions at bench-scale. To this aim a factorial experimental design (32) in conjunction with response surface modelling was employed to study the effect of reaction time (0.5 - 24 hours), pH (5.5 - 8.5), temperature (23˚C - 35˚C), and disinfection strategy (chlorination, pre- and in line-formed chloramines) and the interaction between those parameters. For the first time it was shown that DBP formation from secondary effluent was similarly affected by the various operational parameters than in drinking water matrices, following the order chlorination >> in line-formed monochloramine > pre-formed monochloramine during the first 24 hours of reaction. During chloramination time was identified to be the major influencing factor for nitrogenous DBP formation while pH is affecting the formation of only chlorine containing DBPs to a greater extent. While temperature only had a minor positive impact on chloral hydrate formation during chloramination, temperature was of equivalent importance than the pH for trichloronitromethane formation. In the case of brominated-DBPs, time was revealed to be the dominant parameter during chlorination, where increasing reaction time lead to decreased bromine incorporation. On the other hand, pH and time were almost equally important during chloramination, but inversely correlated. In conclusion, halogenated DBP formation can be minimized by applying pre-formed chloramination instead of forming chloramines in line, especially at contact times less than 24 hours. DBP rejection by RO membranes at the AWTPs ranged from almost complete rejection for the trichlorinated species to as low as 10% rejection for the dihaloacetonitriles. In order to fundamentally understand the DBP rejection mechanisms by RO and nanofiltration (NF) membranes, an investigation at both bench- and pilot-scale was performed under controlled conditions. In agreement with the observations at full-scale, DBP rejection by RO and NF membranes varied to a great degree between negative and almost full rejection. Multivariate linear regression revealed that the intrinsic physicochemical properties of the uncharged DBPs could substantially affect their removal by the membranes. Results showed that besides their size, polarity was an important descriptor for DBP rejection. Furthermore, the effect of feed solution properties and operational parameters on the removal of DBPs was investigated. Using the large dataset (n=500) obtained from the bench-scale investigations, a model for rejection prediction at various operational conditions was developed and validated. The developed model can successfully simulate the rejection by simply using a linear relationship of three common molecular properties, the transmembrane flux and temperature. The modelled results revealed temperature to be the major influencing operational parameter. Predictive models were used to elucidate factors and transport mechanisms that determine the rejection of DBPs by the membranes. Model predictions applying the phenomenological (Spiegler Kedem) approach were in good agreement with the experimental data. Experiments performed with a pilot system (4” spiral wound modules in series) allowed for development and validation of a mechanistic predictive model to simulate DBP rejection in pilot- and full-scale RO processes. Pressures, flows as well as DBP rejections could be modelled across a pressure vessel containing six membrane modules and a three stage plant design. It was shown that the recovery has a substantial impact on DBP rejection.