Iron and copper release in drinking-water distribution systems.

Introduction

Unlined metal pipes are widely used in drinking-water distribution systems. In the United States, more than 60 percent of distribution pipes are composed of cast iron, ductile iron, and steel (McNeill & Edwards, 2001). Copper pipe is a primary material for home plumbing systems. Corrosion of all unlined metal pipes is ubiquitous and can result in a contaminant release that adversely affects health and the aesthetic quality of the water.

Corrosion in drinking-water distribution systems results from chemical reactions between pipe metals (or plumbing fixtures) and finished water; it results in formation of a solid corrosion product: scale. Disruption of the solid scale is undesirable, as more contaminants are typically released until new scale is formed. If the composition of the water frequently varies, the equilibrium between the corrosion scale and the water is disrupted, and contaminant release occurs. Many studies have been done on the formation and properties of corrosion scales to gain insight into corrosion mechanisms as affected by water quality in distribution systems (Blengino, Keddam, Labbe, & Robbiola, 1995; Palit & Pehkonen, 2000; Refait, Abdelmoula, & Genin, 1998; Sarin, Snoeyink, Bebee, Kriven, & Clement, 2001). These scales can be porous or impervious. Impervious scales can stop corrosion, but porous scales can accelerate corrosion and increase iron and copper release.

Many water quality parameters, including pH, alkalinity, sulfate, chloride, phosphate, silicates, natural organic matter, dissolved oxygen, disinfectantresiduals, and temperature, can affect iron and copper release under different specific conditions (Boulay & Edwards, 2001; Broo, Berghult, & Hedberg, 1997; Edwards, Schock, & Meyer, 1996; Larson & Skold, 1958; Sander, Berghult, Broo, Johansson, & Hedberg, 1996). The mechanisms involving metal pipe corrosion and corrosion product release, however, are still unclear, and contradictory results have been reported by different researchers.

Because of a lack of well-accepted guidelines for corrosion and metal release control, a pilot study was needed to reveal water quality problems that might occur in a distribution system because of changes in water source and treatment process. Findings from such a study could help utility operators work out effective measures to ensure high-quality drinking water for consumers.

With the joint support of the American Water Works Association Research Foundation (AWWARF) and Tampa Bay Water (TBW) in Florida, a large-scale pilot study was carried out at the University of Central Florida. The purpose was to determine what effect blending finished waters from ground, surface, and saline sources would have on distribution system water quality Treatment and distribution systems were built for this pilot study.

This paper focuses on seven finished waters produced from different treatment systems. The quality of these waters is characterized, and their influences on iron and copper release are interpreted in terms of primary water quality parameters and changing operation conditions.

Methods

Water Treatment Systems

Finished waters were produced by seven pilot treatment systems (Table 1). The selection of the treatment processes was based on research interests and the needs of TBW. The integrated membrane system (S2) was selected to provide a point of high water quality. The nanofiltration system (G4) was selected to provide data for a TBW member government that was considering nanofiltration of all treated waters. The softening system (G3) was selected to provide data for a second member government that was considering softening all waters in an existing softening facility. The remaining four systems (G1, G2, S1, and RO) were existing water treatment systems. Source groundwater was drawn from the project site, the Cypress Creek well field in Pasco County, Florida. Surface water was transported weekly in a 6,500-gallon stainless steel trailer from the Hillsborough River to the site for treatment. All finished waters were disinfected, and their pH values were adjusted before distribution.

Some of the finished-water parameters were adjusted to meet TBW standards through the addition of chemicals. Calcium and alkalinity (as sodium bicarbonate) were added to S2 and RO waters, alkalinity was added to S1, and sea salts were added to RO-treated water to simulate the quality of finished water from the TBW Regional Desalination Facility.

All requirements for disinfection contact time (the time that disinfectant and water were required to be in contact) were met for free chlorine or ozone, and the water had a total chlorine residual (monochloramine and free chlorine, measured as [Cl.sub.2]) of 4 mg/L before it entered the pipe distribution systems. All finished waters were stabilized with respect to calcium carbonate (CaC[O.sub.3]) by pH adjustment: The waters would neither dissolve preformed solid CaC[O.sub.3] nor form new CaC[O.sub.3].

Construction and Operation of Pipe Distribution Systems

The pipe distribution systems (PDSs) consisted of 18 lines made of PVC, lined cast iron, unlined cast iron, and galvanized pipes. All the pipes used to make the PDSs were excavated from existing distribution systems. Fourteen PDSs were hybrid lines that consisted of PVC, lined cast iron, unlined cast iron, and galvanized pipes. Four of the 18 PDSs were made of a single material. The PVC, lined-cast-iron, and unlined-cast-iron pipes were all 6 inches (15.2 cm) in diameter, and the galvanized pipe was 2 inches (5.1 cm) in diameter. All 14 hybrid lines had identical components: 20 feet (6.1 m) of PVC, 20 feet (6.1 m) of lined cast iron, 12 feet (3.7 m) of unlined cast iron, and 40 feet (12.2 m) of galvanized pipe.

All PDSs were equilibrated for 144 days, from July 16, 2001, to December 6, 2001, with finished groundwater, which was the historical source. At the end of the equilibration period, the background water quality was stable, with 0.1-0.15 mg/L effluent total iron, 0.8 NTUs turbidity, and 10 CPUs (Co-Pt units) of apparent color.

Distribution of finished waters to the PDSs began on December 6, 2001, and continued for one year. The water flowed in a single pass; there was no re-circulation. During the yearlong operation period, the finished-water pipeline assignments were switched every three months to minimize experimental error. Therefore, the study could be divided into four operational phases, Phase I through Phase IV, which corresponded well with the four seasons, winter to autumn. In addition, for G3 and G4, the water source blending ratios varied from phase to phase. Moreover, the division into phases made it possible to evaluate the effect of temperature on water quality changes in distribution systems. The PDS hydraulic residence time (HRT) was five days in Phase I, Phase II, and Phase III, except for the last two weeks of Phase III. PDS residual disinfectant dissipated too rapidly during the summer because of high temperature. Consequently, the HRT was reduced to two days for the last two weeks of Phase III and until the end of the study. All the PDS lines were flushed once a week during five-day HRT period and once every two weeks during the two-day HRT period. Because of the limited length of PDS lines, the flow velocities were much lower than those of full-scale operation. Thus, flush operations were performed to remove sloughed material if present. The water used to flush was the same treated water as that feeding each individual line, and the volume of flush water was equivalent to five pipe volumes.

Copper Loop Corrosion Monitoring

Each of the PDSs was succeeded by a copper loop (new copper pipe) 30 feet (9.1 m) long and 5/8 inch (1.6 cm) in diameter, which contained a lead coupon and simulated home plumbing systems. We monitored these loops for lead and copper to provide information for compliance with the Lead and Copper Rule (LCR). The water in the copper loops was typically stagnant but was periodically flushed in simulation of home use. Samples were collected after a standing period of six hours, as required by the LCR. The corrosion of lead has been reported elsewhere (Tang, Hong, Xiao, & Taylor, 2006) and is not a subject of this paper.

Sampling and Analysis

We conducted sampling and analysis throughout the whole investigation period in order to establish a detailed database. Some relatively unstable water quality parameters were monitored daily or weekly at the site, including temperature, turbidity, apparent color, dissolved oxygen (DO), free chlorine, total chlorine, and pH. The relatively stable parameters were analyzed less frequently (once every two weeks) in a school laboratory; these parameters included calcium, total iron, alkalinity, sulfate, chloride, conductivity, total dissolved solids (TDS). Copper from the corrosion loops was analyzed once every two weeks in Phase I, Phase II, and Phase IV, and once a week in Phase III.

It needs to be pointed out that the iron release data used in this paper were collected from hybrid pipelines that carried seven different finished waters. The iron release was the sum from both unlined cast iron and galvanized pipes (PVC and lined cast iron pipes did not release iron, as had been proved by preliminary monitoring). Also, the copper release data were associated only with seven finished waters. No blended scenarios are presented in this paper.

Results and Discussion

Characteristics of Finished Waters

Finished water quality depended on source and treatment. Lime softening of the Cypress Creek groundwater produced finished water with low-to-moderate calcium, alkalinity, and TDS. Passing the groundwater through an RO membrane and adding sea salts produced a water with high chloride and low hardness, alkalinity, and total dissolved solids. Coagulation of the surface water produced a low-alkalinity and high-sulfate water. Phase-average water quality parameters for the seven finished waters are shown in Table 2.

Although all these waters met regulatory limits, there was significant variation in water quality. As noted, G1 water had extremely high alkalinity and calcium, S1 water had extremely high sulfate, and RO water had the highest chloride content. Compared with G1, lime-softened groundwater (G2) had nearly 60 percent less alkalinity and calcium. G3 had the second highest level of sulfate because its source contained treated surface water. S2 had the second highest level of chloride, which came from chemical adjustment. Comparing G3 with G4 and S1 with S2 shows that membrane nanofiltration significantly reduced sulfate levels, from 79 to 6 mg/L and 183 to 12 mg/L, respectively. Since finished waters were stabilized on the basis of CaC[O.sub.3] equilibrium, the average pH values of all waters were kept within a small range, from 7.9 to 8.5.

Table 2 also indicates the water quality variation by season. G3 alkalinity greatly increased during Phase III and Phase IV. G4 alkalinity was much higher in Phase II and Phase IV than in Phase I and Phase III. G3 and S1 sulfate concentrations gradually decreased, from Phase I to Phase IV. These variations were due to the fluctuation of water sources with the season and the difference in treatments required.

Effect of Water Quality on Total Iron Release

Conductivity is related to the types and concentrations of total ions in water and significantly affects corrosion. Within the pH range of drinking water, conductivity is primarily controlled by the major anions (HC[O.sub.3.sup.-], S[O.sub.4.sup.2-], [Cl.sup.-]) and cations ([Ca.sup.2+], [Mg.sup.2+], [Na.sup.+]). As shown in Figure la, conductivity (denoted as [kappa]) increased linearly with the sum of the major anions (denoted as Ai and including bicarbonate, sulfate, and chloride) over the yearlong operation of the field facility. The unit of Ai, meq/L, means milli-equivalent per liter.

Larson and Skold (1958) studied the corrosion of iron materials and found that sulfate and chloride promoted corrosion, that alkalinity inhibited corrosion, and that the corrosion rate increased as the following ratio (called Larson Ratio, denoted as LR) increased:

LR = ([[Cl.sup.-]] + [S[O.sub.4.sup.2-]])/[[HCO.sub.3.sup.-]] (1)

In this formula, the unit of measure for all anions is expressed as meq/L. The Larson ratio (LR) was used to correlate with the iron release in this study. Figure 1b shows the correlation of the LR with total iron release for seven finished waters under five-day HRT. Clearly, a significant linear relationship exists between total iron release and the LR. In addition, on the basis of the correlation of conductivity ([kappa]) with anions (Ai) and the correlation of total iron release (Fe) with LR, as shown in Figure 1, the total iron release can also be related to conductivity and bicarbonate concentration:

Fe = 0.0018([kappa]/C) - 0.137/C - 0.07 (2)

where Fe is the total iron release in mg/L, [kappa] is the conductivity in [micro]S/cm, and C is the bicarbonate concentration in meq/L.

Conductivity and alkalinity (within the pH range of drinking water, bicarbonate is the dominant contributor to total alkalinity) can be more conveniently monitored; therefore, Equation 2 can be useful for water quality management.

The average total iron release from the hybrid PDSs receiving one of the seven different finished waters during Phase I, Phase II, and Phase III (a five-day HRT) is shown in Figure 2. It can be seen that the total iron release of these waters differs significantly; the amount of iron release associated with the different waters is, in descending order, as follows: S1 > G3 > RO > S2 > G4 > G2 > G1. The average LR values of these waters over the yearlong operation compared as follows: G1(0.33) < G2(0.66) < G4(0.75) < S2(1.56) < RO(1.99) < G3(2.46) < S1(4.16). This result implies that the LR could be used to proactively evaluate the total iron release in distribution systems with varying water quality.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

Effect of Changing Operational Conditions on Iron Release

With the different temperatures of the different seasons, the iron release exhibited different behaviors. In Phase III (the summer), the temperature was much higher than in Phase I and Phase II. Linear regression was applied to describe the relationships between the LR and total iron release in low-temperature period (Phase I and Phase II) and high-temperature periods (Phase III) (Figure 3a). The slope of the regression line corresponding to Phase III was significantly greater than that corresponding to Phase I and Phase II. This result indicates that iron release was more sensitive to the LR in Phase III than in Phase I and Phase II.

[FIGURE 3 OMITTED]

In order to maintain disinfectant residuals, the HRT was reduced to two days for the last two weeks of Phase III and Phase IV. Figure 3b shows the iron release trends and the LR under five-day and two-day HRTs. The flow velocity increased with the reduction in HRT. It was observed, however, that iron release decreased with the reduction in HRT. This result indicates that hydraulic scouring was not a dominant contributor to total iron concentration in the PDS effluent waters.

[FIGURE 4 OMITTED]

Copper Release of Different Finished Waters

The water entering each copper loop came directly from the effluent of the corresponding PDS line. Some relatively stable water quality parameters, such as alkalinity, sulfate, chloride, and calcium, did not exhibit significant changes after flowing through PDS lines. Disinfectant residuals and dissolved oxygen, however, decreased markedly after passing through PDS lines (data not shown). The pH values also changed slightly. Statistical methods were used to identify the water quality parameters significantly influencing copper release. The p-values for the correlation between each parameter and copper release are given in Table 3. Obviously, alkalinity and pH are the parameters most significantly related to copper release during all four phases. Factor analysis (data not shown here) demonstrated that alkalinity had a positive linear relationship with calcium (with an average [R.sup.2] value of 0.66). Thus, the effect of calcium on copper release cannot be distinguished from that of alkalinity. Although the copper release is associated with the process of copper corrosion (oxidation of metal copper), the total chlorine residual and dissolved oxygen were not detected as significant factors for copper release. The result indicates that copper release behaviors were controlled by the dissolution of the corrosion scales, not by the corrosion rate of metal copper.

Figure 4 shows the correlation of copper release with alkalinity and pH. Copper release increases with increased alkalinity and decreased pH, a result that is in agreement with the findings of Edwards and co-authors (1996). It is also interesting to find that pH variation within the small range of 7.6-8.4 (for PDS effluent waters), which had no significant effect on iron release, had a strong influence on copper release. The sensitivity of copper release to pH suggests that pH adjustment should be taken as a crucial measure for the control of copper levels in copper plumbing systems.

The copper release levels of seven finished waters during the yearlong period are shown in Figure 5a. Obviously, G1 had the highest copper release, while the differences in copper release among other waters were relatively small. The extremely high alkalinity of G1 was responsible for the severity of copper release in that system.

[FIGURE 5 OMITTED]

The roles of sulfate and chloride in copper release were complicated, and the mechanisms involved are still unclear. In our study, sulfate did not under most circumstances significantly affect copper release (Table 3). Chloride correlated significantly with copper release during Phase I, Phase II, and Phase IV, but this effect was mainly due to the negative correlation of chloride and alkalinity. Therefore, the specific effect of chloride on copper release could not be identified. Another aspect of Figure 5a is that during the yearlong experimental period, copper fluctuated much more in the G1 system than in the other systems, although G1 water quality was quite stable. It can be deduced that temperature might affect copper release behavior. The average copper release for G1 and the average temperature changes in different phases are illustrated in Figure 5b. It can be inferred that temperature had a positive correlation with the copper release for G1. The highest copper release for G1 occurred in Phase III (the summer season).

Conclusions

A high positive correlation between total iron release and the Larson ratio was observed, and this result could be applied to predict the potential for iron release in finished waters with different mineral characteristics. Total iron release was more sensitive to the Larson ratio under higher temperatures. Reducing the HRT was an effective method of cutting down total iron release levels in a distribution system. Alkalinity and pH were the two water quality parameters that most significantly affected copper release. Copper release correlated positively with alkalinity and negatively with pH.

With its low alkalinity and high sulfates levels, finished surface water treated by ferric sulfate coagulation caused the highest iron release of all the waters. Finished groundwater treated by the conventional method produced the lowest iron release but the highest copper release. The iron release of desalinated water was mainly due to its high chloride concentration and low alkalinity. High temperature could enhance the copper release of finished water with high alkalinity.

The findings of our investigation could be of great significance for utility operation and drinking-water management. When water sources are switched or new treatment processes are introduced, it is necessary to evaluate the corrosivity of the new finished waters with respect to the existing distribution system, with special attention paid to changes in alkalinity, conductivity, and pH.

The indexes traditionally used to assess the chemical stability of finished waters are almost all based on CaC[O.sub.3] equilibrium, which is not effective in predicting iron and copper release in distribution systems. The conductivity and alkalinity ratios may be used as a supplementary index for determination of the chemical stability of the water with respect to metal release from iron-based pipes. If copper pipes are included in the distribution systems, finished water of extremely high alkalinity should be avoided; meanwhile, the pH should also be stringently adjusted and maintained.

The transportation capacity for which the distribution system was designed should match well with the water demand so that water does not stagnate for long periods of time. Utility operators and water management personnel can take proactive measures by adopting alternative water treatment technologies and effective water quality monitoring devices to produce finished waters with high chemical stability.

Acknowledgements: The authors acknowledge the project sponsors AWWARF and Tampa Bay Water. Special thanks go to all the personnel who provided maintenance of the project site and data analysis, both in the field and at the University of Central Florida laboratory.

Corresponding Author: Baoyou Shi, Associate Professor of Environmental Science and Engineering, Chinese Academy of Sciences, State Key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, 18 Shuangqing Road, Haidian, Beijing 100085, China. E-mail: [e-mail].

REFERENCES

Blengino, J.M., Keddam, M., Labbe, J.P., & Robbiola, L. (1995). Physico-chemical characterization of corrosion layers formed on iron in a sodium carbonate-bicarbonate containing environment. Corrosion Science, 37(4), 621-643.

Boulay, N. & Edwards, M. (2001). Role of temperature, chlorine, and organic matter in copper corrosion by-product release in soft water. Water Research, 35(3), 683-690.

Broo, A.E., Berghult, B., & Hedberg, T. (1997). Copper corrosion in drinking water distribution systems--The influence of water quality. Corrosion Science, 39(6), 1119-1132.

Edwards, M., Schock, M.R., & Meyer, T.E. (1996). Alkalinity, pH, and copper corrosion by-product release. Journal of American Water Works Association, 88(3), 81-94.

Larson, T.E., & Skold, R.V. (1958). Laboratory studies relating mineral quality of water to corrosion of steel and cast iron. Journal of American Water Works Association, 14(6), 43-46.

McNeill, L.S., & Edwards, M. (2001). Iron pipe corrosion in distribution systems. Journal of American Water Works Association, 93(7), 88-100.

Palit, A., & Pehkonen S. (2000). Copper corrosion in distribution systems: Evaluation of a homogeneous [Cu.sub.2]O film and a natural corrosion scale as corrosion inhibitors. Corrosion Science, 42, 1801-1822.

Refait, P., Abdelmoula, M., & Genin, M.R. (1998). Mechanisms of formation and structure of green rust one in aqueous corrosion of iron in the presence of chloride ions. Corrosion Science, 40(9), 1547-1560.

Sander, A., Berghult, B., Broo, A.E., Johansson, E.L., & Hedberg, T. (1996). Iron corrosion in drinking water distribution systems--The effect of pH, calcium and hydrogen carbonate. Corrosion Science, 38(3), 443-455.

Sarin, P., Snoeyink, VL., Bebee, J., Kriven, W.M., & Clement, J.A. (2001). Physico-chemical characteristics of corrosion scales in old iron pipes. Water Research, 35(2), 2961-2969.

Tang, Z., Hong, S.K., Xiao, W. & Taylor, J.S. (2006). Impacts of blending ground, surface, and saline waters on lead release in drinking water distribution systems. Water Research, 40(5), 943-950.

Baoyou Shi, Ph.D.

James S. Taylor, Ph.D., P.E.

TABLE 1 Finished Waters--Sources and Major Treatment Processes

Finished Water  Original Source Water  Major Treatment Process

G1              Groundwater            Aeration
G2              Groundwater            Lime softening, sedimentation,
                                       filtration
G3              Blended source*        Lime softening, sedimentation,
                                       filtration
G4              Blended source         Membrane nanofiltration
S1              Surface water          Ferric sulfate coagulation,
                                       ozonation, biologically activated
                                       carbon filtration
S2              CSF water**            Membrane nanofiltration
R0              Groundwater            Membrane reverse osmosis (R0)

* Blend of finished G1, R0, and surface water after coagulation-
sedimentation-filtration.
** Surface water after coagulation-sedimentation-filtration.

TABLE 2 Average Water Quality Parameters of Finished Waters During Each
Phase

                       Alkalinity*       S[O.sub.4.sup.2-]  [Cl.sup.-]
Phase  Finished Water  (mg/L)       pH   (mg/L)             (mg/L)

I      G1              206          8.2   25                30
       G2               91          8.1   26                23
       G3               51          8.2  108                62
       G4               66          8.2    7                55
       S1               65          7.9  237                44
       S2               67          8.2   11                61
       RO               71          8.2    6                99
II     G1              210          7.9   25                29
       G2               95          8.0   25                27
       G3               52          8.1   82                42
       G4              105          8.3    6                40
       S1               56          7.9  217                41
       S2               67          8.1   13                60
       RO               69          8.2    5                93
III    G1              204          7.9   26                33
       G2               87          8.0   25                24
       G3               64          8.0   77                54
       G4               80          8.1    7                47
       S1               61          8.0  160                32
       S2               65          8.0   14                69
       RO               65          8.0    6                86
IV     G1              208          8.0   23                26
       G2               89          8.1   25                19
       G3               84          8.0   55                42
       G4              104          8.3    4                33
       S1               60          8.2  124                29
       S2               72          8.5    9                76
       RO               66          8.3    4                91

Phase  [Ca.sup.2+] (mg/L)  Conductivity ([micro]S/cm)

I      85                  508
       34                  285
       46                  457
       25                  285
       83                  606
       25                  361
       25                  389
II     85                  545
       33                  331
       31                  434
       45                  355
       69                  673
       24                  370
       27                  445
III    83                  582
       31                  332
       31                  497
       27                  345
       43                  599
       27                  422
       30                  449
IV     81                  582
       32                  331
       34                  479
       33                  352
       32                  570
       26                  462
       25                  474

* Alkalinity: mg/L of CaC[O.sub.3].

TABLE 3 Statistical-Analysis p-values: Correlations of Copper Release
with Water Quality Parameters

Phase  pH    Alkalinity  [Cl.sup.-]  S[O.sub.4.sup.2-]  [Ca.sup.2+]

I      .000  .000         .003        .023              .000
II     .000  .000         .000       0.476              .000
III    .000  .000        0.486       0.375              .000
IV     .000  .000         .009       0.234              .000

Phase  Si[O.sub.2]  Total Chlorine  Dissolved Oxygen

I       .011        0.431           0.209
II     0.421        0.246            .048
III     .032         .053            .067
IV      .002         .032            .044