Review Potential impacts of changing supply-water quality on drinking water distribution: A review

the switching of treated water quality by destabilization and the release of pipe material and contaminants into the water and the subsequent risks. At the end of this article, a framework is proposed for the evaluation of potential transition effects.

Drinking water treatment makes water potable by removing contaminants present in the source water. Depending on the contaminants present, different technologies and their combinations can be used for drinking water production (Moel et al., 2006). In both developing and industrialized countries, a growing number of contaminants are entering water supplies due to human activity: heavy metals, pharmaceuticals, endocrine disruptors, perfluorinated compounds, flame retardants or biocides (Shannon et al., 2008;Schriks et al., 2010;Ternes et al., 2015). Public health and environmental concerns drive efforts to tighten water quality regulations (WHO, 1996;WHO, 2004;WHO, 2011;Qu et al., 2012) and further treat waters previously considered clean. These efforts have greatly promoted the development of water treatment science and technology and the upgrading of treatment plants over several decades (Pinheiro and Wagner, 2001;Miner, 2002;Shannon et al., 2008).
The treated drinking water is delivered to individual dwellings, communal buildings and other customers' through pressurized distribution networks (DNs), including drinking water distribution systems (DWDSs) and premises plumbing (Mays, 1999;Ainsworth, 2013). Those DNs -consisting of pipes, pumps, valves, storage tanks, reservoirs, meters, fittings, and other hydraulic appurtenances transport the drinking water from a centralized treatment plant or well to customers' taps (Moel et al., 2006;NRC, 2006). In general, DWDSs are buried underground, range in length of tens to several hundreds of kilometers (totaling 0.4 million kilometers in the Netherlands and1.6 million kilometers in the U.S.) with varying diameters and materials (metals, plastic, and cement pipes) in multi-loops (Verberk et al., 2007;Vreeburg, 2007;AWWA, 2012). Premise plumbing refers to the parts of DN that connect a building to the water main pipe and the pipes inside buildings, which are characterized by high surface area to volume ratio, long retention times, warmer temperature, and close contact with water (NRC, 2006;Nguyen et al., 2012;Wang et al., 2013;Proctor and Hammes, 2015). The DNs is the final barrier that protects drinking water from contaminants during distribution and ensures the quality of the supplied product (Mays, 1999). For the drinking water provision, distribution is as important as production, because the quality of tap water can only be as good as the condition of the pipes it flows through (Moel et al., 2006;Misko et al., 2010;Ainsworth, 2013).

Water quality stability and deterioration during distribution
There is a broad consensus that the ultimate goal of drinking water supply should be seen as providing good quality at the customer's tap rather than only at the point it leaves the treatment plant. The treated drinking water enters the distribution system containing physical loads (particles), microbial loads (cells) and nutrient loads (organic and inorganic nutrients) (Liu et al., 2013c;Prest et al., 2016). Given the occurrence the long retention times (also referred as "water age") especially at the dead-end nodes (Ainsworth, 2013) and in premises plumbing the simultaneously impact of physiochemical and microbiological processes can result in the deterioration of the quality of the water that reaches customer's tap compared to the original water produced at the treatment plant (Vreeburg and Boxall, 2007;Liu et al., 2013c;Proctor and Hammes, 2015). Such water quality deterioration has been observed and reported worldwide: for example in higher turbidity and particle counts (Vreeburg et al., 2004;Verberk et al., 2007;Liu et al., 2016), larger cell numbers (Van der Kooij, 1992;Hammes et al., 2010;Liu et al., 2016), and greater presence of selected indicator micro-organisms (van der Wielen et al., 2016) at taps compared to treatment plants. In extreme cases, discolored water (also reported as dirty water, red water) (Sly et al., 1990;Vreeburg and Boxall, 2007;Li et al., 2010) was observed at the taps.
Those observed quality deteriorations are related to the accumulation of distribution network harbored material (DNHM) and ongoing processes such as pipe corrosion, biofilm matrix formation and detachment, loose deposit accumulation and resuspension, which occur in the DN pipes over decades (Lee et al., 1980;LeChevallier et al., 1987;Van Der Wende et al., 1989, Smith et al., 1997Gauthier et al., 1999;Chaves Simões andSimões, 2013, Liu et al., 2014) (Fig. 2A). Consequently, to guarantee the delivery of high quality water to customers, the concept of water quality stability was proposed and set. The standards included: re-suspension potential measurements (RPM, turbidity < 0.8NTU) for physical stability (Vreeburg et al., 2008); a saturation index (SI, -0.2 to 0.3) for chemical stability (Verberk et al., 2009); and assimilable organic carbon limit (AOC, < 10 mg C/l for unchlorinated water) for biological stability (Van der Kooij, 1992). Achieving these stability standards has become another important driving force behind the upgrading of treatment processes (Pinheiro and Wagner, 2001;Miner, 2002;Qu et al., 2012).

Irregular supply-water quality changes and transition effects
At the operational level, the tightening of water quality regulations, the development of water purification technologies, and the awareness of the deterioration of water quality have led to the upgrading of treatments at water utilities and thus to a betterquality treated water at treatment plants worldwide (Pinheiro and Wagner, 2001;Miner, 2002). This upgraded treatment constitutes a cause of irregular changes in supply-water quality; other causes are the use of alternative source water ( Zhang, 2009;Li et al., 2010;Utecht and McCoy, 2016) and variations in disinfectant strategies . Irregular changes therefore refer to situations that are qualitatively different from the regular operation of a DWDS. Accordingly, the accompanying transition effects refer to the often dramatic impact of these changes in the networks, and particularly on distribution network-harbored material (DNHM).
Under regular operations, DNs can function as sinks (contaminant accumulation, as in particle sedimentation) or sources (contaminant release, as in loose deposit resuspension) of trace contaminants in the bulk water (Vreeburg and Boxall, 2007;Liu et al., 2014;Makris et al., 2014). The material retained in DNs (pipe scales, biofilm matrix and loose deposits) develops and stabilizes under the historical environmental conditions when the supply water is continuously flushed out. If destabilized and mobilized, this DNHM will become a major source of contaminants, which can impact supply-water quality. The transition effects are defined as the physiochemical and microbiological water quality problems caused by the destabilization and mobilization of DNHM and its release into the bulk water. The transition effects are in turn the consequence of irregular changes in supply-water quality. Examples of transition effects include the water discoloration observed in 80% of the distribution area of Beijing, China, when the quality of supply water was improved in 2008 (Li et al., 2010); and the recent Flint (Michigan, U.S.) drinking water crisis, in which elevated lead levels were measured in children following changes in source water and treatment (Hanna-Attisha et al., 2016).
Transition effects have so far been poorly documented and their quantitative definition, based on measured parameters, requires further research on integral evaluation and a better understanding of the composition and characteristics of the corresponding DNHM. Supply-water quality is of course subject to regular fluctuations related to daily variations in source-water quality and treatment performance. In addition, there may also be periodic changes, such as operationally related changes (filter backwash), seasonal changes of surface-water quality, annual switching of water sources, and changes in the mixing ratio of multiple water sources. All of these periodic changes in supply-water quality fall under the rubric of regular changes, and are part of the environment in which the DNHM has developed and stabilized in the distribution system (biofilm, pipe scales and biofilm matrix). Such regular changes are not considered in this review paper because they do not lead to DNHM destabilization and mobilization.

Objective of this review
Water utilities pay close attention to the safety and quality of drinking water during distribution, focusing on various aspects and situations, such as pressure transients, cross-connection, deterioration of buried infrastructure, permeation and leaching, nitrification, microbial growth and biofilm, water storage facilities, water age, and pipe repair. Even the accumulation of chemical and biological material in DWDSs have become a widely accepted target of attention. However, the transition effects which can lead to esthetic and health problems have been neglected. The available literature on the characteristics of the material developed and contaminants harbored in DNs and the potential transition effects following the switching of supply-water quality were reviewed for this article, which also incorporates the authors' first-hand knowledge and research. The particular object of this critical review is development and destabilization of DNHM. In the conclusion, we propose a framework for the evaluation of potential transition effects before switching supply-water quality. The discussion emphasizes water quality improvements, but the review, knowledge, and proposed framework could also be applied to other irregular water quality changes produced by a variety of causes.

Material accumulation in DNs and its characteristics
Four elements have been defined and studied in DNs: 1) the bulk water that flows through the pipe networks; 2) the suspended solids which are particulate matter that is suspended in the water and transported through the network; 3) the pipe surface with the associated material, e.g., biofilm, extracellular polymeric substance (EPS), scaling; and 4) the loose deposits which are particulate matter that has accumulated and is retained in the pipes (Liu et al., 2014;Proctor and Hammes, 2015;Liu et al., 2016). The materials in the first two elements only present in the system for a relatively short period of time, i.e., the period it takes for the water to flow through. Additional forces, such as bio-adhesion, weight, and chemical bonds, effectively counter the regular hydraulic forces, resulting in the retention of the material over a longer period of time (i.e., years), under the last two elements.
Physically, the force can be the weight of suspended solids, if they become large and heavy enough due to (bio-)aggregation and precipitation (Vreeburg and Boxall, 2007) to the point that, once settled, they cannot be resuspended by the hydraulic turbulence in a DN; one example would be the sand and activated carbon particles collected from field networks as loose deposits (Camper et al., 1986;Brazos and O'Connor, 1996;Liu et al., 2014), as shown in Fig. 1B. Chemically, the forces can be chemical bonds of corrosion products, pipe scales (passivation layers), and association of ironbonded oxidation products (e.g., ionic attractive force, hydrogen bonding, van der Waals interactions, electrostatic attractive forces, repulsive forces) (Mceneill and Edwards, 2001;Lytle et al., 2004;Flemming and Wingender, 2010;Peng et al., 2010) (Figs. 1C and 2). Microbiologically, the viscous forces between the adhering bacteria and substratum surfaces can be attributed to the selfproduced extracellular polymeric substances which can help to harbor more material to form a biofilm matrix (Beech and Sunner, 2004;Flemming and Wingender, 2010;Chen et al., 2014;Fish et al., 2016), as shown in Fig

. Pipe material
The pipe material is important because of its influence on water quality and its relationship to energy efficiency (Broo et al., 2001). For example, the release of lead will pose health risks for consumers (Rabin, 2008); the release of nutrients from rubber material will promote microbial growth in the distribution system (Broo et al., 2001); the presence of tuberculation will make the pipes rough inside, resulting in increased pumping costs and, in some extreme cases, leakages (Mceneill and Edwards, 2001). Worldwide, the traditional inorganic pipe materials are being replaced by organic pipe materials, even if the use of stainless steel is increasing. Iron pipes with interior cement-mortar lining and exterior plastic protection, as well as other cement-based materials, are still being installed (Broo et al., 2001). Although lead and asbestos cement-based material are being phased out and replaced, there are still substantial lengths being used in operational distribution systems (Rabin, 2008). An overview of the pipe material used in distribution networks in the Netherlands from 1955 to 2014 is shown in Fig. 3 (Vreeburg, 2007;Geudens, 2015).

Aging pipes
The design lifespan of distribution pipes, regardless of the material used, is 65e90 years (AWWA, 2012). The majority of the distribution pipes in developed countries (e.g., the U.S.) will reach the end of their lifespan in the coming 20 years (Moe and Rheingans, 2006). Therefore, a new era of water industry for pipe repairing and replacement is beginning. For the U.S., the estimated replacement cost will be at least $1 trillion in the coming 20 years (AWWA, 2012;Qureshi and Shah, 2014).
The challenges of aging pipes: 1) Aging pipes are more vulnerable to contaminants (Moe and Rheingans, 2006;AWWA, 2012;Qureshi and Shah, 2014). As recorded by a Midwestern water utility in the U.S., pipe breaks sharply increased from 250 breaks/year in 1970 to 2200 breaks/ year in 1989 (Moe and Rheingans, 2006). The challenge is especially serious in developing countries where maintenance of the distribution system infrastructures is lacking due to inadequate resources. Water loss level (as a percentage of water supplied) for developed countries, middle-income countries, and developing countries, are 5e24%, 15e24% and 25e45% respectively (Moe and Rheingans, 2006). In cases of low or negative pressures in the pipes, the contaminated water, wastewater, and other contaminants surrounding the pipes can be drawn into the pipes through cracks. In such cases, even if measures are taken to improve water quality at the treatment plant, customers will not fully benefit from them. 2) The material accumulated in a distribution system (biofilm, scaling and loose deposits) develops over time. Therefore, the older the pipes the greater the accumulation of material in the form of pipe scales, biofilm matrix and loose deposits (Makris et al., 2014). 3) Because the long design lifespan and limited accessibility, the DN consists of pipes installed at different times. As a result, it is a variegated reflection of both past and present knowledge, strategies, technologies, and operations related to drinking water distribution.

Loose deposits
Despite the application of different treatment processes, certain (varying) amounts of particles remain present in treated and distributed water. These particles may be particles present in the raw water that were not removed by the treatment (Clark et al., 1992;Tobiason et al., 1993), or may have been introduced during the treatment for example, powdered activated carbon particles (Camper et al., 1986;Brazos and O'Connor, 1996) or aluminum micro-flocs (Gauthier et al., 2001). Particles can also be generated during distribution as a result of precipitation and flocculation processes, biofilm detachment and corrosion phenomena (Sly et al., 1990). While travelling in distribution systems, the particles can be transported throughout the network by the water flow or be deposited as loose deposits which can be re-suspended by daily hydraulic turbulence, depending on their characteristics and the flow conditions (Vreeburg and Boxall, 2007;Liu et al., 2013c;Liu et al., 2014;Liu et al., 2016).
Sedimentation and resuspension are two-way continuous processes until the loose deposits have accumulated to a degree that they can no longer be resuspended by daily hydraulic variations. Then, niches are created characterized by a high surface area and micro-environments that may lead to the formation of biofilm and the accumulation of different elements (Ca, Fe, Mn, and As) (Liu et al., 2014). Even anoxic and anaerobic conditions can occur in such circumstances. Gradually, the accumulated material may develop into pipe scales (Makris et al., 2014).
Therefore, the presence and development of loose deposits involves a combination of physical and bio-chemical processes. Amounts of loose deposits ranging from 30 to 24,500 mg/m have been found in distribution systems worldwide Carri ere et al., 2005;Vreeburg et al., 2008;Liu et al., 2014) containing high amount of biomass (Zacheus et al., 2001;Liu et al., 2013c;Liu et al., 2014), organic matter (Gauthier et al., 1999;Lehtola Fig. 3. Material used in distribution network in the Netherlands over time (1995e2014) (Geudens, 2015). et al., 2004) and inorganic matter (Echeverría et al., 2009). Previous studies have mainly focused on discoloration problems (Vreeburg and Boxall, 2007). Recently, however, loose deposits have attracted renewed attention due to the high volume and diverse of the microbes they harbor (Liu et al., 2013c;Liu et al., 2014;Proctor and Hammes, 2015;Prest et al., 2016;van der Wielen and Lut, 2016). The reported composition of loose deposits is summarized in Table 1.

Biofilm matrix
Biofilm refers to the structures formed by microorganisms via adhesion, nucleation, and growth on surfaces. Specifically, biofilm is defined as matrix-enclosed bacterial populations adhering to each other and/or to surfaces or interfaces, including microbial aggregates and floccules and populations within the pore spaces of porous media (Costerton et al., 1995). In most biofilms, the microorganisms account for less than 10% of the dry mass, whereas the matrix can account for over 90%. The matrix is extracellular material, mostly produced by the organisms themselves, which consists of a conglomeration of different types of biopolymers and is known as EPS (Flemming and Wingender, 2010). In addition to EPS, the mineral scaffolds (bio-mineralization) also play a crucial and conserving role in the biofilm, providing resistance to environmental stresses and increasing the overall fitness of the microbial community (Oppenheimer-Shaanan et al., 2016). The research has found that, the biomineralization media promotes the resilience of biofilm and limits the penetration of antibacterial agents into biofilms.

Pipe scales
The presence of scales (passivation layers, corrosion products, and associated contaminants) in water pipes has attracted both scientific and societal interest (Renner, 2008;Schock et al., 2008;Makris et al., 2014). Scales form on drinking-water pipes as a result of a combination of in situ processes or upstream corrosion, post-treatment deposition of solid materials, and deposits formed through the reaction of the treated water and the pipe material with natural substances that pass through the treatment process (Schock et al., 2008). The pipe scales formed in DNs consist mostly of tubercles formed via precipitation and re-precipitation mechanisms of pipe-originating nucleating elements, e.g., aluminum (Al), copper (Cu), iron (Fe), and lead (Pb) (Makris et al., 2014). Their formation has been primarily attributed to electrochemical surface corrosive phenomena, and to the dissolution and precipitation of metal salts in the pipe networks (Mceneill and Edwards, 2001;Makris et al., 2014).
Various types of pipe scales have been found, with metal pipes being especially susceptible (Prest et al., 2016). The nature and composition of scales in depend on both the water quality and the pipe material. Thin manganese oxide was found in PVC pipes, and was much easier to detach compared to metal pipe scales (Cerrato et al., 2006). In addition, the water source (Yang et al., 2012) and disinfectant residuals (Wang et al., 2012a) present in distributed water are also important factors in the reported variations in formed pipe scales. Lead (Pb) and copper (Cu) pipes rely upon the chemistry of supply water to develop passivating or immobilizing solid phase on the pipe surfaces (Schock et al., 2008). Other trace elements in water, such as aluminum (Al), radium (Ra) and arsenic (As), also accumulate in scales on DN pipes. The most commonly observed and studied pipe scales are alkaline/calcareous scales, iron scales, aluminum scales, corrosion-induced asbestos scales, and lead scales: these are reviewed and summarized in Table 3.

Destabilization of DNHM
Transition effects can occur when the balance of forces are destabilized (Fig. 2) as a result of the irregular changes in supplywater quality. This destabilization leads to the mobilization of DNHM and the release of contaminants into the water column.

Physical destabilization
Under normal water distribution operations, the phenomenon of discoloration has been attributed to the disturbance of equilibrium between weight and the hydraulic resuspension force (weight ¼ hydraulic force) of loose deposits as a result of sudden hydraulic peaks (increasing hydraulic resuspension force, hydraulic force > weight) (Vreeburg and Boxall, 2007). This principle is also applied, for instance, in resuspension potential measurements (Vreeburg et al., 2008), self-cleaning networks , hydraulic flushing and loose deposits sampling (Liu et al., 2014). In the event of irregular change in supply-water quality, a similar destabilization can occur: weight decreases through particle dissolution which makes the particles smaller and transferable to the water column. This weight decrease is caused by the destabilization of chemical forces, for example the process in Equation (1) (Moel et al., 2006) and applies to other compounds associated with loose deposits, biofilm matrix and pipe scales.

Chemical destabilization
Chemical destabilization is mostly caused by the differences of pH, redox potential, and ion composition in the supply water. This is especially relevant in the remobilization of contaminants bound by pipe scales on metal pipes via desorption and/or dissolution (Peng et al., 2010). For example, large amounts of Fe rich particulates (>300 mg/l) were released after a change was made from nondisinfection to chlorination (Equation (2)) (Reiber and Dostal, 2000); and in another study, the switch in source water caused Fe concentration peaks (>10 mg/l) in the first month because of the presence of high SO 4 2À concentration (Equations (3) and (4)) (Wang et al., 2009). Moreover, the replacement of chlorine for monochloramine caused the release of up to 4800 mg/l Pb (Equation (5)) (Edwards and Dudi, 2004). There have been several reports of serious red-water complaints following switches in source water in Southern California, Tucson (Arizona), Tampa (Florida), in the United States; and, in a northern city in China (Brodeur et al., 2006;Tang et al., 2006;Yang et al., 2012). It is believed that the pipe material and supply-water quality determine the characteristics of the pipe scales, and that different pipe scales adapt differently to supply-water quality changes (Yang et al., 2012).

Microbiological destabilization
Microbial community composition and function can be potentially impacted by the changes in environmental conditions (Allison and Martiny, 2008;Shade et al., 2012). The drinking water microbial communities are sensitive to physiochemical and microbiological water quality changes (i.e., disinfectants, nutrients concentration and composition, inorganic elements), which could    (Ewing, 1935;Makris et al., 2014) aluminum scales Al 2 SiO 5 and AlPO 4 (Kvech and Edwards, 2001;Kvech and Edwards, 2002) corrosion-induced asbestos scales release Ca and asbestos fibers into water (Axten and Foster, 2008) lead scales PbO,PbO 2 , Pb 3 (CO 3 ) 2 (OH) 2 , Pb 5 (VO 4 ) 3 Cl (Schock, 1980;Schock et al., 2008;Gerke et al., 2009) induce specific selection pressure on microbial population and drive community diversification (Gomez-Alvarez et al., 2016). For example, the temporary switch between chlorination and chloramination revealed reversible shifts in microbial communities (Hwang et al., 2012;Wang et al., 2014). The response of microbial community composition and function to water quality changes can lead to the microbiological destabilization of DNHM. In extreme case, discoloration (and associated health concerns) may arise; for example, the increased sulfate levels in feed water have caused significant increase of sulfur-oxidizing bacteria, sulfate-reducing bacteria and iron-oxidizing bacteria, which are associated with severe iron corrosion scale release and red water (Yang et al., 2014). In cases of water quality improvements achieved through the input of better source water or upgraded treatment, the available nutrients (e.g., AOC) are reduced, which leads to a decrease in the biological activity in the water and the biofilm (Van der Kooij, 1992; Van der Wielen and Van der Kooij, 2010;Liu et al., 2013b). Furthermore, the lower nutrient concentration and resulting decrease biomass production will also lead to a lower production of EPS and therefore a reduction of the bio-adhesion to the pipe surface (Hsieh et al., 1994;Laspidou and Rittmann, 2002). In practice, the physiochemical and microbiological destabilization will occur simultaneously. The mineral dissolution as mentioned above in accordance with Equations (1)e(5), will gradually result in the loss of biofilm structure stability (biofilm mineral scaffolds) and a diminished ability of biofilm to resist environmental condition changes, such as those associated with the changes in physiochemical characteristics of the contacting water (Oppenheimer-Shaanan et al., 2016). On the other hand, the loss of biofilm matrix protection will increase the contact of the water with pipe scales, which will further accelerate desorption and dissolution of inorganic contaminants. Meanwhile, the changes of water characteristics will influence the microbial community composition and function that result in bio-destabilization of DNHM. This has been illustrated by the red water events in Beijing, where increased sulfate in supply-water caused microbial community composition changes revealed by increase in sulfur oxidizing bacteria, sulfate reducing bacteria and iron oxidizing bacteria and red water events associated with high iron concentrations (Yang et al., 2014).

Timeline of destabilization processes
Although the issue is obviously important, no study explicitly focuses on the timeline of transition effects. Once destabilization occur, the chemical bonds and forces may be weakened faster than the microbiological forces, owing to the stability provided by the entanglement of biopolymers within the biofilm matrix (Flemming and Wingender, 2010) and to the viscous bond between the attached surface and bacteria (Chen et al., 2014). From the microbial community perspective, changes of community composition were not observed in the studies over short periods (0.2 years on average), but the studies covering significant longer periods (4.9 years on average), indicate that there is a lag in the responses of microbial communities to water quality changes (carbon amendments in this case) (Allison and Martiny, 2008). This means that there will be a lag in the case of bio-destabilization caused by microbial community composition and function changes. Therefore, it is hypothesized that if the transition effects are dominated by chemical destabilization, they will happen and be detected over short periods (i.e., weeks), and evidenced by reported pipe scales dissolution and release (Reiber and Dostal, 2000;Edwards and Dudi, 2004;Wang et al., 2009). In contrast, if the transition effects are dominated by microbiological destabilization, they will only be detected and observed after a longer period of time following the switching (i.e., months to years). This is illustrated by the reported water meter clogging that occurred two years following switching in the supply area of the Oasen drinking water company in the Netherlands, where the distribution system consisted mainly of PVC pipes without corrosion problems (Fig. 4, results not published).

Problems associated with transition effects
3.2.1. Aged pipes As discussed above, aging pipes are vulnerable to the intrusion of contaminants because of the increased number of pipe breaks and negative pressure events (Moe and Rheingans, 2006). In the case of transition effects caused by physiochemical and microbiological destabilization, the corrosion, pipe material leaching, and pipe scale release caused by the irregular changes in supply-water composition, may accelerate the pipe damaging process. Paradoxically, the irregular changes of supply-water quality (e.g. improved quality due to switching source water and to upgraded treatment) may raise the risk of contaminant intrusion by increasing the chances of pipe damages.

Tap water quality and safety
Physically, the release of DNHM can cause problems, such as blue water (high copper concentration) (Edwards et al., 2000) and red water (high iron concentration) (Wang et al., 2009;Li et al., 2010;Yang et al., 2014), and in some cases water meter clogging which leads to pressure loss at the taps (Fig. 4). Chemically, the release of heavy metals (Pb, As, Cu, Fe, Mn) may pose health risks to customers (Lytle et al., 2004;Liu et al., 2010;Lytle et al., 2010;Hanna-Attisha et al., 2016;Utecht and McCoy, 2016). Microbiologically, the detachment of biofilm and the release of cells may lead to high cell densities in the water column (Makris et al., 2014). As (opportunistic) pathogenic microorganisms have been frequently detected in biofilms and loose deposits in DNs (Pryor et al., 2004;Torvinen et al., 2004;Feazel et al., 2009;Chaves Simões andSimões, 2013, Falkinham et al., 2015), their release into water column can, in principle, constitute a pathogen mobilization process and pose health risks to consumers (Beuken et al., 2008;Schwake et al., 2016). An overview of reported water quality problems associated with water quality changes is summarized Fig. 4. Picture of particulate matter that clogged the water meter in Oasen's Hendrik Ido Ambacht supply area.
in Table 4; most problems relate to discoloration and Pb. However, knowledge about the influence of such a transition on microbial ecology and bacteria release is lacking.
It should be realized that material release may occur without being apparent in visible/noticeable changes in color and pressure for the end-users; this could include the leaching of certain compounds or the presence of (opportunistic) pathogens, phenomena that involve even higher risk to customers than do discoloration and pressure loss. For example, the Pb problem in the Flint water triggered a serious public health crisis, as evidenced by high Pb levels in children's blood; these levels were detected before the contamination was recognized (Hanna-Attisha et al., 2016;Schwake et al., 2016). Even if the precise mechanisms are unknown, the field result collected at Flint also suggested high Legionella numbers in premise plumbing: a consequence of the interruption of distribution system corrosion control (Schwake et al., 2016).
Once the transition effects occur, the DN no longer functions as a sink but rather as a source of material leached into water. It is important to quantify the level to which this can impact water quality, in order to decide whether actions should be taken to guarantee the quality and safety of drinking water at customers' taps. This maximal quality deterioration potential (QDP) can be determined by assuming the DNHM is immediately released into the contacting water column, as presented in Equations (6) and (7) below: T DSHM : total material harbored by certain length of DN pipe (e.g. 1 m). Niches: different niches available in DN pipes, e.g. biofilm, loose deposits, pipe scales, etc. QDP: quality deterioration potential in the selected pipe. V water : volume of water in contact with DNHM in the selected length pipe.
Taking the reviewed data sets from reported distribution system study as an example (Liu et al., 2013c), the biofilm formed on 110 cm water main pipe can lead to a QDP for microbiological quality of 0.4 Â 10 7 cells/ml and 200e1600 ng ATP/l. These numbers increase further when the contribution of niches of loose deposits are also taken into account (by 60e600 ng ATP/l). Such assessment can be done on different parameters and multiple niche environments within DN pipes.

Distribution process
The destabilized and mobilized DNHM may be retained in the DNs as loose deposits instead of flowing out at customers' taps if the generated particulate matter has a weight that can balance the hydraulic force. In this situation, the organic components, which were previously protected by pipe scales and biofilm matrix from contacting disinfectant residuals in the water column will, when release, cause faster decay of disinfectant residuals and form disinfection by-products (DBP) in the system where the disinfectant residuals maintained during distribution. The mobilized but retained particulate matter also offers more surface area and available nutrients for bacterial growth.

Control strategies for preventing potential transition effects
The potential transition effects and the associated problems will seriously discourage customers from directly using tap water and can easily counteract the numerous efforts devoted to source-water switching and treatment upgrading. In extreme cases, there can be esthetic and health risks. Therefore, attention and effort should be focused on developing strategies to prevent potential transition effects.

Reconditioning the treated water
The potential transition effects are caused by irregular changes in the physiochemical and microbiological characteristics of the produced water. Therefore, properly reconditioning the treated water can be helpful to prevent transition effects during distribution. The critical parameters to be adjusted should be determined according to the properties of aged pipes, pipe scales, loose deposits and biofilm matrix in the downstream DNs (e.g. PH, redox, SI, corrosion inhibitors). However, the reconditioning option should be carefully examined to make sure treated water will not be recontaminated. One example of such a reconditioning is the remineralization of reverse osmosis permeate by adding the required minerals for health reasons and to meet regulatory requirements (Withers, 2005;Liang et al., 2013;Lupt akov a and Derco, 2015).  (Schwake et al., 2016;Utecht and McCoy, 2016) Similarly, in the Flint case, the addition of corrosion inhibitors to the treated water following switching might have prevented the crisis; while for the Beijing case, strategies to deal with the potential influences of high sulfate concentration could have prevented the red water events during the switching.

Cleaning the pipe network beforehand
As the distributed drinking water quality can only be as good as the condition of the distribution network, one of the proposed control strategies is to clean the pipe network before switching supply-water quality. The commonly used method for distribution pipe cleaning is known as 'unidirectional flushing', which employs a large water volume at sufficiently high velocity (1.5 m/s) water to mobilize the sediments and carry away the loose deposits through fire hydrants (Ellison, 2003). The problem with the method is that it can be wasteful of water and ineffective in removing adherent contaminants (e.g., tuberculation, scale, biofilm), especially when there are difficulties in achieving required velocity (Ellison, 2003;Quarini et al., 2010). Air scouring is similar to water flushing, except that the air is introduced into the fluid column to create a high-velocity, turbulent flow. It is generally more efficient in removing biofilm and soft scales than flushing, but it is at least twice as costly as flushing. An alternative technique is pigging, which propels or pulls (pigs, a bullet-shaped object) through the pipelines using water pressure. Hard pigs have a higher cleaning efficiency than soft pigs. Recently, ice pigging has been introduced and become established, providing the advantages of high cleaning efficiency, no sticking, water saving and reduced costs (Bellas and Tassou, 2005;Quarini et al., 2010;Dang et al., 2014). In extreme cases, when the above-mentioned cleaning techniques are not sufficient, the pipes should be rehabilitated or replaced.
From a practical perspective, considering the complexity and costs of each method, the water flushing and air scouring options are more appropriate compared to pigging and pipe replacement. This especially true for the mega-cities, such as New York and Beijing. In instances where regulations allow for quality variations and exemptions, and the transition effects pose no health risks to customers, the temporary exceeding of water quality requirements, and subsequent compliance, maybe acceptable (EPA 1998a,b).

Monitoring
Another strategy involves monitoring the transition effects after switching supply-water quality, so that proper action can be taken in a timely manner by the water utilities if any destabilization occurs. Generally, there are three challenges to monitoring the release of contaminants during transition effects. Namely, 1) the uncertainty concerning whether and when the release is going to happen; because the contaminant-release events often go undetected since the release is episodic and non-periodic (Makris et al., 2014); 2) the released contaminants might be under the detection limits, because any released contaminants are diluted in a large volume of water which flow through the DN and cannot, therefore, be detected; 3) the sampling artifacts and analytical limitations in the currently used protocols exclude suspended colloidal or particulate material, such as for Pb, As, and Cr, from detection (Parks et al., 2004;Triantafyllidou et al., 2007).
To overcome the first challenge mentioned above, one suggestion would be to monitor water quality parameters in real time, for instance by online particle counting (Verberk et al., 2007) and online cell counts using flow cytometry (Hammes et al., 2012;Besmer et al., 2014). Measurements taken at multiple locations in the distribution system before the water quality changes can provide a background database; new measurements would then be taken at the same locations during and after the supply-water quality changes. A data comparison obtained at these different instances will offer valuable information on the transition effects. However, online detection is limited. There are no available online devices for heavy metals (such as As, Pb) and pathogens. In addition, if the contaminant-releasing events are too low to be noticed by comparing the background and transition period measurements, online monitoring will give no significant results.
As a solution to both the second and third challenges, it is proposed that suspended particles be studied, as we have described previously (Liu et al., 2013a;Liu et al., 2016) be studied. In short, this method consists of a pre-concentration step for the suspended particles by filtrating about 200 L of water through 1.2 mm filters with a multiple particle filtration device. The collected particles can then be analyzed on physiochemical (e.g., total suspended solids and elemental composition) and microbiological parameters (biomass quantification and bacterial identification). In this way, the particulate contaminants can be pre-concentrated and the contaminant-releasing events can be detected, captured and characterized (Liu et al., 2016). As a combined solution to all the three challenges, an online particle sampling, concentrating, and monitoring system would have to be developed.
For the purposes of summary, an overview of the aging pipes and the harbored material retained in the pipe systems, types of potential destabilization, associated problems, and timeline for the releasing events to occur, and various cleaning methods is given in Table 5.

Conclusions and outlooks: transition effects evaluation framework
Until now, no information has been available on the evaluation of the potential transition effects of the planned switching of supply-water quality, nor are there any proper guidelines for the avoidance of potential esthetic and health risks. Based on this review, and bearing in the mind the WHO's outlined Water Safety Plans (Davison et al., 2005), the following framework comprising a system assessment, management plans and operational monitoring is proposed as a means of evaluating the potential transition effects before the switching of supply-water quality (Fig. 5).
System assessment: Step 1 involves the parallel evaluation of: 1.1) water quality changes and 1.2) the current situation of the distribution system. If there is no change in the produced water quality nor fouling in the distribution system, the supply water can be switched directly. Otherwise, the evaluation should be continued with Step 2. The research questions in Step 1 can be summarized as follows: Q1) Are there quality improvements/changes as a result of the new treatments/sources? Q2) Is there DNHM in the DN? Q3) What is the composition of the DNHM material? Q4) Where is the DNHM present?
Step 2 involves an assessment of whether the transition effects will occur or not. Q5) Will the transition effects occur? Q6) What will be released into water from DNHM during the transition period? Q7) What risks does the released material present?
If there are no transition effects observed, the supply water can be switched safely. If transition effects are observed, the evaluation should continue to the next step.

Management plans
Step 3 involves an evaluation of the control strategies to avoid transition effects. The possible strategies include flushing the distribution pipes (water, water and air), ice pigging, and the addition of required agents to stabilize the quality improved water (e.g., corrosion inhibitors). Q8) Can the transition effects be prevented by pre-cleaning the DN? Q9) How should the distribution pipes be cleaned? Q10) Are there any essential compounds that should be added to the quality-improved water, for instance for reconditioning, to maintain stabilization of the DN and DNHM?
If the transition effects can be prevented by pre-cleaning the pipes or by reconditioning the quality-improved water, the supply water can be switched safely. The monitoring program should be carried out as described in Step 4 after the switching of the supply water. In case transition effects are observed, and no strategies succeed in controlling them, the supply water should not be switched until effective control strategies are developed.

Operational monitoring
After the decision is made to switch the supply water, the distribution of quality-improved drinking water should be continuously monitored for material releases from the DWDS into the water. If material releases are detected, the evaluation should return to Step 3, with a search for cleaning and control methods, unless the material releases are minor and have no harmful effects on consumers. In no material releases are detected, the distribution  of quality-improved drinking water may continue. This monitoring program/method can be used as an early-warning system, so that appropriate actions can be taken in time to prevent any undesired transition effects.
In the event of the application of the framework to the situation of Beijing, the system assessment will be able to diagnose: 1) the composition of pipe scales and biofilm community; 2) differences of supply-water quality regarding the sulfate concentrations; 3) potential transition effects of community-composition changes regarding sulfur-oxidizing bacteria, sulfate-reducing bacteria and iron-oxidizing bacteria, and 4) possible red-water events associated with high iron concentrations. The management plans should be able to prevent discoloration, either through the effective cleaning of DNs or by proper adjustment of treated-water quality, for example, through efficient sulfate removal. The monitoring program should cover bacterial community composition, suspended particle load, and sulfate and iron concentrations, all of which can offer valuable information on what is happening in the DNs after switching supply-water quality.