Microbial drinking water quality is of great importance to human health. Drinking water distribution systems (DWDSs) are designed as the final barriers for delivering and maintaining the biosafety of drinking water. Though the drinking water produced is usually safe and clean, it is common that the water quality deteriorates during the distribution. Such deteriorations can be linked to the establishment of biofilm in DWDSs, where the majority of the biomass is residing (> 95%). The formed biofilms are reportedly leading causes of the undesired taste, odor, and color of the drinking water, corrosion of the pipes, decay of the disinfectants, and proliferation of pathogenic microbes, giving rise to public health concerns.
In the Netherlands, the control of the biofilm growth in DWDSs is achieved by producing bio-stable drinking water with extremely low nutrients (e.g., AOC < 10 µg C/l). On the other hand, water utilities in many countries usually apply chemical disinfectants (e.g., free chlorine, monochloramine) to control the biofilm growth in DWDSs. Nevertheless, biofilm formation is inevitable, regardless of the strategy. Additionally, there is no standard method to monitor the biofilm growth in DWDSs, which makes the understanding and management of DWDS biofilms more challenging. Efforts have been made to explore the biofilm formation and structure through pilot studies. However, most of these investigations have been conducted in a short time frame (e.g., within weeks to a max of 84 days), where the developed biofilms were far from mature and significantly different from those in the real DWDSs. To uncover how biofilm develops and what roles disinfectants play during the biofilm development, a newly-built pilot system was followed for a 64-weeks period under different disinfection regimes: no disinfectants (NC), free chlorine (FC), and monochloramine (MC) (Chapter 2). The results showed that residual disinfectants presented intensive suppression of the biofilm growth and shaped the biofilm communities. Specifically, MC exhibited stronger suppression of the biofilm activity (i.e., ATP), whereas FC expressed intense selection pressure on the microbes and established more homogenous and less complex biofilm community, with Proteobacteria comprising on average 82% of the relative abundance. The temporal trends highlighted the essential developmental stages in biofilm formation from initial colonization to accumulation and selection and stabilization, which occurred at different rates under each of the conditions, and were associated with significant dynamic changes in biofilm bacterial communities. Reaching stabilization took longest in the MC condition (> 64 weeks), followed by the NC (~ 36 weeks) and FC (~ 19 weeks) conditions. Holistically, the early stages in the biofilm formation in the NC condition were primarily dominated by stochastic processes where colonizers originating from treated water randomly attached to and settled on the pipes, while deterministic processes progressively increased in their relative contributions at the end of the accumulation stage and became predominant at the later stages. In the MC condition, the biofilm succession was governed by stochastic processes during the entire test, even though some deterministic processes occurred during the accumulation stage. Conversely, in the FC condition the biofilm succession was driven by deterministic processes already from the initial development stage.
DWDSs are highly dynamic ecosystems, where the liquid (i.e., bulk water, suspended particles) and solid (i.e., biofilm, loose deposits) phases interact intensively during transport of the water from treatment to consumer. The cells and/or particles that were introduced with the treated water may attach to and/or settle on the pipes, forming biofilm/loose deposits when the hydraulic forces are weak. Conversely, the biofilm/loose deposits might release cells/particles to the bulk water during hydraulic disturbances, affecting the drinking water quality negatively. The hydraulic conditions in DWDSs are very complex and dynamic. They exhibit daily patterns, with high flow rates at high water demand periods (e.g., morning and/or evening hours) and long stagnancy or low flows during the night. However, most monitoring occurs using grab samples at one point in time. Thus, continuous online sampling is required to obtain a representative image of the particles and microbes in drinking water. In Chapter 3, a novel online monitoring and sampling system (OMSS) was developed to investigate the spatiotemporal variations of the planktonic and particle-associated bacteria in an unchlorinated DWDSs. The 16S rRNA gene sequencing combined with SourceTracker2 was used to trace and reveal the origin of the changes in the planktonic and particle-associated bacteria, assigning sampled biofilm and loose deposits as sources. The results showed that, spatially, the particle loads significantly increased from treatment plant within distribution networks, while the trend in the quantity of the particle-associated bacteria was the opposite. Similar to the trend of particle loads, the number of the observed OTUs in both planktonic and particle-associated bacteria increased from the treatment plant within the distribution network. The spatial results implied a dominant role of sedimentation of particles entering the DWDS from the treatment plant, while the observed increases in particles and the associated bacteria primarily originated from the distribution network, which were confirmed by the increased contributions from loose deposits and biofilm determined by SourceTracker2. Temporally, daily peaks in the water quality, including particle-associated bacterial quantity, observed operational taxonomic unit (OTU) number, and contributions of biofilm and loose deposits, were sensitively captured during the high water demand (morning/evening peaks). The temporal results revealed clear dynamic interactions between the liquid (i.e., bulk water, suspended particles) and solid (i.e., biofilm, loose deposits) phases in DWDSs.
Driven by increasingly stringent drinking water regulations and challenges to drinking water quality, efforts are underway to further improve water quality. These initiatives include source water switching, upgrading treatment processes, and implementing changes to disinfectant strategies. Such actions change the quality and composition of the treated water that enters the DWDS. This may have transition effects, which in this thesis refers to the water quality deteriorations contributed by the release of cells and particles from biofilm and/or loose deposits due to the irregular changes in supply-water quality. It is largely unknown whether, where and when the transition effects will happen. In Chapter 4, transition effects were investigated through characterizing the particles before (T₀), during (T_3-weeks) and after (T_6-months) introducing additional treatment steps (softening, second rapid sand filtration and adding carbon dioxide) to the existing treatment. The results showed that the upgraded treatment significantly improved the water quality after 6 months’ time. However, significant water quality deterioration was observed at the initial stage (T_3-weeks) when the quality-improved treated water entered into the network. This manifested as a significant increase in total suspended solids (TSS) by 50-260%, active biomass (ATP) by 95-230%, and Mn by 130-250%. Furthermore, pyrosequencing results revealed sharp differences in microbial community composition and structure of the bacteria associated with particles between T₀ and T_3-weeks, implying the potential contributions from biofilm or loose deposits in the DWDS. Interestingly, the domination of Nitrospira spp. and Polaromonas spp. in the distribution system at T_3-weeks, which were detected at rather low relative abundance at treatment plant, further confirmed the potential contributions from biofilm or loose deposits.
Though the study in Chapter 4 confirmed the occurrence of the transition effects, the question how fast/how long the transition effects will occur/last, where the deteriorations originate from, and what actions can be carried out to minimize the transition effects is not clear. The sampling was conducted in a relatively short time frame (i.e., 6 months), with only a few time points (i.e., T₀, T_3-weeks, T_6-months) and without the collection of biofilm and loose deposit samples. Additionally, as what we can see from the results from Chapter 3, it could be imagined that the transition effects might be enhanced during high water demand when shear forces are high. In order to fill the knowledge gaps, the OMSS was applied, accompanied with SourceTracker2, in an unchlorinated DWDS where partial RO was introduced (Chapter 5). The study was conducted before (T_B), immediately after (T₀), one month (T_1M), two month (T_2M), one year (T_1Y) and two years (T_2Y) after the partial RO introduction. Noticeably, significant transition effects in DWDS were captured right after the RO introduction, with increases in the particle loads, bacterial quantity, community diversity, and significant differences between bacterial communities in particles at treatment plant and distribution network. The disturbances lasted one month until T_1M, after which they ceased to be observable around T_2M. The captured deteriorations were confirmed by the increased contributions of loose deposits and biofilm (both the number of the immigrants and their abundance) at T₀ and T_1M determined by SourceTracker2 and neutral community model. While the peak transition window spanned about one month, it took considerably longer, until one year (T_1Y) and two years (T_2Y) later, for the microbial ecology to re-stabilize and for improvements in water quality to become noticeable. In addition, the peaks in the water quality deteriorations were enlarged during the high water demand (morning/evening peaks), which implies that current monitoring could potentially underestimate the extent of the quality deterioration. Remarkably, the observation that loose deposits contributed more to the transition effects than biofilm challenges the traditional standpoint, and provided new insights into the management of the transition effects, where the risks of the transition effects can be largely reduced by conducting flushing before the introduction of treatment changes to remove the loose deposits. In light of the destabilization caused by the changed water quality, flushing with new-quality water might be more rewarding.
To conclude, through conducting studies at both field and pilot scales, the effects of the (changes in) operational conditions on the microbial drinking water quality in DWDSs were comprehensively explored. The findings in the thesis offer novel insights into the drinking water quality management. The knowledge gained from investigating the biofilm succession dynamics under different disinfectant regimes has significantly deepened our understanding of managing drinking water biofilms. These insights serve as valuable information when making informed decisions about the appropriate strategies to employ. The implementation of the developed OMSS is capable of capturing both periodic and aperiodic changes in drinking water quality, making it an essential tool in minimizing assessment deviations and ensuring accurate evaluations of drinking water quality. Moreover, the established methodology holds promise for application in various systems, including those that utilize chlorination. By identifying and characterizing the transition effects resulting from changes in supply water quality, such as treatment upgrades or the introduction of reverse osmosis, the study highlights the significance of considering these effects in water management practices. These observations underscore the importance of addressing the impact of transition effects on drinking water quality and provide practical implications for minimizing their negative consequences.