• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Red Water Release in Drinking Water Distribution Systems

Posted on: Sunday, 18 September 2005, 03:01 CDT

The implementation of groundwater conservation measures has forced utilities with a historical reliance on groundwater sources to consider alternative sources to augment their supplies or to eliminate their groundwater dependence. Switching from traditional source water, however, can cause unacceptable changes in water quality that result from destabilization and the release of chemical and biological films from the interior surfaces of the existing distribution systems. Data from a two-year study were used to identify significant water quality parameters and to develop a predictive nonlinear model to estimate the corrosivity of blends based on water quality. The results of the statistical analysis indicate that alkalinity, chlorides, sulfates, sodium, and dissolved oxygen of the source water or blend of source waters have a significant effect on release of corrosion by-products in the form of red water. Temperature and hydraulic retention time were the significant physical and operational parameters identified.

Tampa Bay Water (TBW) is a regional water supplier in Florida that serves nearly two million consumers distributed over six member governments. TBW has historically relied on groundwater sources to meet its 250 mgd (946 ML/d) average daily demand. To reduce dependency on a single source and to protect the aquifer and the surrounding environment, the Southwest Florida Water Management District and TBW agreed to a series of groundwater withdrawal reductions. This plan required TBW to develop other sources of drinking water. As part of the first phase of the plan, a surface- water treatment plant and a desalination plant were constructed. To evaluate whether the introduction of new source waters, blends of source waters, or both could cause unacceptable changes in water quality in the existing distribution systems, the University of Central Florida (UCF) conducted a pilot-scale research study under the aegis of the AWWA Research Foundation (AwwaRF).

Water quality changes in drinking water distribution systems occur as a result of complex and often interrelated physicochemical and biological processes (Hedberg & Johansson, 1987). The mineral composition of the source water usually determines whether adverse conditions (conditions that will lead to corrosion and subsequent deterioration of quality) exist within the distribution systems (Pisigan & Singley, 1987). Biological conditions are also inextricably linked to the mineral contents of the source water (Carter et al, 2000; Fransolet et al, 1988). As a result, the effect of the chemical composition of the source water on the quality deterioration within the distribution system must be evaluated.

Historically, the main focus has been on the calcium carbonate (CaCO^sub 3^) solubility of the source water. Water slightly oversaturated with CaCO^sub 3^ will precipitate a calcite layer on the pipe surface. This precipitate would protect the surface from subsequent contact with the water, thereby reducing corrosion. Based on this principle, a number of corrosion indexes have been developed, including the Langelier index (LI). The LI is popular among drinking water utilities and professionals as an indicator of the stability of source water. A negative LI indicates undersaturated water as well as possible removal of the passivating layer of CaCO^sub 3^ formed on the pipe surface. The general practice, then, is to maintain a positive LI. This can be achieved by increasing alkalinity, calcium, temperature, or pH (Singley, 1981). Usually pH is increased by adding sodium hydroxide (NaOH). Blending water from different sources changes the water quality and characteristics in a complex way. Studies have shown that waters considered individually stable with respect to CaCO^sub 3^ precipitation may become unstable when blended (Trussell & Thomas, 1971).

Research has also shown, however, that CaCO^sub 3^ solubility is not the only water quality aspect that corrodes distribution systems (Pisigan & Singley, 1985). A number of additional corrosion indexes have been suggested to characterize corrosion potential that are not specifically related to CaCO^sub 3^ solubility. The Larson ratio implicates chlorides and sulfates as contributors of corrosion. It has been observed that an increase in alkalinity could overcome the adverse corrosive effects of chlorides and sulfates (Larson & Skold, 1957, 1958).

Unlined cast-iron and galvanized-iron pipes are the major source of iron in the distribution system and thus a major cause of red water problems. Corrosion in cast-iron distribution systems is a function of a variety of water quality and flow characteristics. These parameters are usually interlinked by a complex relationship. For example, an increase in calcium hardness, pH, and alkalinity may be a mitigating factor in corrosion, whereas chlorides, dissolved oxygen, sulfates, and residual chlorine may increase the iron uptake in the distribution systems (Singley et al, 1985). Pisigan and Singley (1984) conducted a series of corrosion tests on metal specimens in jars and in pipe loops. They used the data obtained from these tests to develop a nonlinear model in which the corrosion rate was modeled as a function of different water quality parameters. Corrosion-rate increases were correlated to increases in chlorides, sulfates, alkalinity, and dissolved oxygen. Decreases in corrosion rates were correlated to increases in calcium, saturation index, and time.

It is necessary to make a distinction between red water problems in distribution systems and the actual corrosion rate of the pipe material. Iron pipes will corrode to varying degrees under any water quality. During a period of usage, the chemical precipitates and other corrosion by-products that form on the internal surface of the distribution system reach a steady state. Changes in the source water quality can lead to disruption of this equilibrium and a significant deterioration in water quality. The water quality deterioration in iron pipes is related to the destabilization of the surface films resulting either from hydraulic effects or changes in source water characteristics. Sander et al (1997) note that the detailed mechanisms for pipe corrosion coupled to different water compositions are not well understood. A number of water quality parameters have been implicated for causing iron release in the distribution systems. These parameters include pH, dissolved oxygen content, organics and nutrient content, chlorine dose and dissipation, chlorides, sulfates, microorganisms, calcium, and metals concentration. It would be difficult to incorporate all of these mechanisms into a unified theory of corrosion in distribution systems. It is feasible, however, to analyze the corrosion by- products or by-product indicators and identify which parameter is statistically and mechanistically significant. This analysis would help in evaluating blends based on water quality considerations for red water control.

MATERIALS AND METHODS

Pilot distribution system design. A pilot distribution system (PDS) was constructed at the TBW Cypress Creek well field, near Tampa, Fla., using pipes provided by the participating member governments. These pipes were transferred to the project site in a manner that would cause minimum disruption of the internal films and scales. The idea was to study the effect of the proposed changes on a PDS that could mimic the actual member government distribution system as closely as possible.

The PDS is composed of 18 different distribution lines. Lines 1- 14 are hybrid lines made up of four different materials: polyvinyl chloride (PVC), unlined-iron, lined-iron, and galvanized-iron pipes. The average length of each hybrid line is 91 ft (27.7 m), of which 19 ft (5.8 m) is PVC, 19 ft (5.8 m) is lined iron, 13 ft (4 m) is unlined iron, and 40 ft (12.2 m) is galvanized-iron pipe, in that order. The PVC, lined-iron, and unlined-iron pipes are 6 in. (0.15 m) in diameter and the galvanized-iron pipe is 2 in. (0.05 m) in diameter. Line 15 is composed entirely of unlined-iron pipe. Similarly, lines 16, 17, and 18 are composed of lined-iron, PVC and galvanized-iron pipes, respectively.

TABLE 1 Mode of production of simulated source waters*

TABLE 2 Blends used for different PDS lines over the period of the study

Once on site, the pipes were assembled and allowed to equilibrate with TBW groundwater over a period of five months (Cullen, 2002). After equilibrium was established, different water blends were introduced into the PDS. The project was divided into six 3-month phases. Similar blends were used in alternate phases to evaluate the effect of seasonal conditions on the PDS and related water quality.

Pilot source waters. The pilot-plant facility was constructed so that different source waters with different chemical characteristics could easily be produced. Table 1 shows the different pilot processes and their mode of production. The finished water qualities from these pilot processes were set to match TBWs existing or proposed future sources of water. Thedifferent blends evaluated in this study, which were chosen by TBW to model anticipated operations, are shown in Table 2. Table 3 enumerates the water qualities for the different source waters. All of the source waters were chloraminated and stabilized to maintain a positive LI.

Pilot distribution system operation. The PDS was operated initially (in phases 1-3) at a five-day hydraulic retention time (HRT) to simulate dead-end conditions. The HRT was changed to two days during phases 4 and 5 to permit chlorine residual maintenance. Different source waters and their blends were introduced into the PDS by dosing pumps that feed individual influent standpipes for each line. The lines were flushed once a week during the five-day HRT period and once every two weeks for the two-day HRT period. The flush velocity was 1.0 f/s (0.3 m/s) for at least three pipe volumes. Sampling for a number of water quality parameters was conducted once a week at the influent and effluent standpipes.

Water quality measurements. Samples were collected and analyzed in the field and at the UCF laboratory. Blind duplicates and spikes were used to determine the accuracy of measurements, and dynamic control charts were used to determine whether the results were acceptable.

STATISTICAL MODEL DEVELOPMENT

Red water releases in distribution systems result from a complex interrelation of physical, chemical, and biological effects. Empirical models offer a robust tool for predicting the effect of influent water quality on the changes within the distribution systems. To evaluate the effect of water quality on red water release in the distribution system, a nonlinear regression model was developed. Analysis of variance (ANOVA) was used to identify the parameters that have significant effect as predictors of change in apparent color (measured in cpu). The change in apparent color, rather than total iron, was used as the dependent variable because investigators observed that the total iron was in particulate form and had a strong correlation to the apparent color. A high correlation between apparent color and total iron (Figure 1) indicates that apparent color is an appropriate surrogate measurement for total iron under the conditions of this study. Apparent color was more convenient to analyze on a regular basis on the site using a spectrophotometer.

TABLE 3 Average finished water quality values for simulated source waters

A balanced experimental design is necessary to avoid confounding effects among parameters. This can be achieved by evaluating different levels of chlorides, sulfates, alkalinity, and other water quality parameters in the source water for the PDS. Confounding in the database, however, complicated the resolution of the singular effect of each variable. Specific confounding effects of potential concern are noted between alkalinity and chlorides, alkalinity and sodium, alkalinity and sulfate, sodium and chlorides, and temperature and dissolved oxygen. These difficulties arise, for example, from the different characteristics of the sources (groundwater is high in alkalinity but low in chlorides), the requirement for electroneutrality (sodium is coupled with chlorides because sea salt is added to simulate desalination), and any seasonal effects (elevated temperatures are associated with low- dissolved-oxygen content). Any mitigation of corrosion-product release should also recognize the complexities and electroneutrality constraint associated with altering finished water chemistry.

FIGURE 1 Apparent color as a substitute for total iron measurement

FIGURE 2 Model performance for average pilot distribution system response from phases 1-4

For the nonlinear modeling, water quality parameters analyzed during phases 1-4 of the study were used. Water quality for each area of the PDS was averaged over the respective phase, giving a total number of 70 data sets that were used in the modeling. Predictor water quality variables used were alkalinity, calcium, conductivity, pH, sulfate, chloride, sodium, silica, UV^sub 254^, dissolved oxygen, temperature, and HRT. The dependent variable used was change in apparent color.

The F statistic (ANOVA) was used to determine whether dropping a term in the full model caused significant deterioration in model fit. The conventional process of eliminating nonsignificant variables could not be used because correlated parameters caused a confounding effect. Consequently, the investigators decided to evaluate all the combinations of the parameters to identify the best possible model. In all, 4,098 model combinations were evaluated, and the final model was selected based on the ANOVA.

TABLE 4 Variation in water quality parameters used in development of statistical model

The empirical model selected identifies alkalinity, chlorides, sulfates, sodium, dissolved oxygen, HRT, and temperature as the significant variables. Figure 2 shows the excellent correlation between the actual change in color and the one predicted by the model. The pseudo-R^sup 2^ is 0.83, which indicates that 83% of the variation in the data can be explained by the model. Contrary to conventional wisdom, calcium and pH were not identified as significant in the statistical model. The main reason for this omission is that all the waters were stabilized to achieve a positive LI. Stabilization of the source waters resulted in a very narrow range of pH. The simulated source waters and the blends used in this study mimic either existing or planned changes in operation in the next few years. Thus a balanced design for determining the individual effects of the water quality parameters was incidental rather than planned. Table 4 gives the range of water quality parameters that were used in the nonlinear model development. The variation in the different water quality parameters during the two- year study encompasses a wider range than would be expected during blending of the different source waters. Application of the model, then, would not require extrapolation.

TABLE 5 Analysis of variance table for verification of empirical model by independent data set from phases of the study

Model verification. The empirical model was verified by conducting a fifth independent phase of study. Phase 5 considered a wider range of alkalinity, chlorides, and sulfates than was available from phases 1-4 to support model development (Eq 2). The source waters were prepared by adding alkalinity in the form of sodium bicarbonate, chlorides from sea salt, and sulfates from calcium sulfate. The purpose of these additions was to simulate a particular water quality rather than a specific process. An additional set of verification experiments was pursued using PDS lines 8-10, with blend changes every two weeks to adjust alkalinity within a range of 60 to 250 mg/L as CaCO^sub 3^. As Figure 3 shows, the performance of the model was consistent when verified on the independent data set from phase 5. To verify the hypothesis that the empirical model developed from phases 1-4 averaged PDS data (Eq 2) was applicable to independent data from phase 5, ANOVA was used (Table 5). The results indicate that there is no significant difference between the performances of the model when it is used on an independent data set from phase 5.

MITIGATION

From the empirical model, the only water quality parameter that appears to be controllable with some success by chemical addition is alkalinity. The other parameters are inherent in the processes that produce the source waters. For instance, desalinated water will have an elevated chloride level that can be reduced only by modifying the desalination process. Surface water will have elevated sulfates that result from ferric sulfate coagulation. Temperature is based on seasonal (and sometimes even diurnal) variations. HRT is governed by the dynamics of water demand. Figure 4 shows the response of the PDS to changes in alkalinity from phases 1-4 of the study. Alkalinity, however, shows an inverse relationship to the release of color. A clear line of demarcation can be seen at an alkalinity of 80 mg/L as CaCO^sub 3^; there is a drastic reduction in the release of color above an alkalinity of this level. With the exception of a few source waters (incidentally the ones high in sulfate and chloride; circled in Figure 4), waters with alkalinity greater than 80 mg/L as CaCO^sub 3^ tended to be less corrosive.

The empirical model developed from phases 1-4 as well as the verification data from PDS lines 1-6 and PDS lines 8-10 from phase 5 suggest that an increase in alkalinity is desirable to reduce the release of color. But chemical measures of alkalinity supplementation should be practiced with care. For instance, an increase in alkalinity resulting from the addition of sodium bicarbonate would also increase the sodium content of the water. On the basis of the empirical model, this corresponding increase in sodium may offset the intended benefit of increased alkalinity. Whenever feasible, alkalinity supplementation by blending with the high-alkalinity groundwater (>200 mg/L as CaCO^sub 3^) or the addition of lime plus carbon dioxide is recommended.

FIGURE 3 Verification of model by independent data from phase 5

FIGURE 4 Dependence of color release on alkalinity of source water

CONCLUSIONS

Red water release in distribution systems, as measured by increase in apparent color, is caused by the release of corrosion products from unlined and galvanized-iron pipes.

An empirical statistical model identified alkalinity, chlorides, sulfates, sodium, dissolved oxygen, temperature, and HRT as being significant to change in apparent color in the hybrid PDS. Alkalinity has a strong negative correlation to increase in color, but chlorides, sulfates, sodium, dissolved oxygen, temperature, and HRT showed a positive correlation to increase in color.

Calcium and pH were not identified as significant variables during the statistical modeling, perh\aps because all the waters were stabilized for CaCO^sub 3^ solubility. The empirical model was verified with independent data from a subsequent set of experiments (phase 5).

Alkalinity is the only significant variable identified in this article that can be effectively controlled by chemical addition. On the basis of the two-year PDS data, alkalinity greater than 80 mg/L as CaCO^sub 3^ seems to have a beneficial effect on reducing release of color. Therefore, it is recommended that in addition to minimizing chlorides and sulfates (which may not always be possible because of production characteristics), alkalinity above 80 mg/L as CaCO^sub 3^ should be targeted.

ACKNOWLEDGMENT

The authors would like to thank Chris Owen (TBW project coordinator), Roy Martinez (AwwaRF project officer), and the following member governments: Pinellas County, Hillsborough County, Pasco County, Tampa, St. Petersburg, and New Port Richey. Pick Talley, Robert Powell, Dennis Marshall, and Oz Wisener from Pinellas County and Luke Mulford from Hillsborough County are specifically recognized for their contributions. We also thank the UCF environmental engineering students and faculty who contributed to this project.

With the implementation of groundwater conservation measures, many utilities with a historical reliance on groundwater sources have been obligated to consider alternative sources to augment their supplies or eliminate their groundwater dependence. Switching from traditional source water, however, can bring about unacceptable changes in water quality resulting from destabilization and release of chemical and biological films from the interior surfaces of the existing distribution systems. For example, red water release in distribution systems is caused by the release of corrosion products from unlined- and galvanized-iron pipes.

In this two-year study, the effect of changing water quality on red water release in distribution systems was evaluated under a wide range of conditions to simulate blending of different source waters (groundwater, surface water, and desalinated water). First, investigators identified alkalinity, chlorides, sulfates, sodium, dissolved oxygen, temperature, and hydraulic retention time (HRT) as significant water quality parameters. Next, a predictive nonlinear model was developed to estimate the corrosivity of blends based on water quality.

The results of the statistical analysis indicate that alkalinity, chlorides, sulfates, sodium, and dissolved oxygen of the source water or blend of source waters have a significant effect on release of corrosion byproducts in the form of red water. Alkalinity has a strong negative correlation to increase in color, but chlorides, sulfates, sodium, dissolved oxygen, temperature, and HRT showed a positive correlation to increase in color. Temperature and HRT were the significant physical and operational parameters identified.

This work resulted in a valuable tool for evaluating release problems resulting from different blends of water.-RSH

A full report of this project, Effects of Blending on Distribution System Water Quality (91065F), will be available in early 2006 as a downloadable PDF file from the AWWA Bookstore (1- 800-926-7337) or from awwa.org/bookstore. Reports are free and currently available to AwwaRF subscribers by calling 303-347-6121 or from www.awwarf.org.

The implementation of groundwater conservation measures has forced utilities with a historical reliance on groundwater sources to consider alternative sources to augment their supplies or eliminate their groundwater dependence.

Red water release in distribution systems, as measured by increase in apparent color, is caused by the release of corrosion products from unlined and galvanized iron pipes.

Any mitigation of corrosion-product release should also recognize the complexities and electroneutrality constraint associated with altering finished water chemistry.

Alkalinity has a strong negative correlation to increase in color, but chlorides, sulfates, sodium, dissolved oxygen, temperature, and hydraulic retention time showed a positive correlation to increase in color. On the basis of the two-year pilot distribution system data, alkalinity greater than 80 mg/L as CaCO^sub 3^ seems to have a beneficial effect on reducing release of color.

An empirical statistical model identified alkalinity, chlorides, sulfates, sodium, dissolved oxygen, temperature, and hydraulic retention time as being significant to change in apparent color in the hybrid pilot distribution system.

REFERENCES

Carter, T.J., et al, 2000. Relationship Between Levels of Heterotrophic Bacteria and Water Quality Parameters in Drinking Water Distribution Systems. Water Res., 34:5:1495.

Cullen, C.J., 2002. Equilibration of Pilot-scale Distribution Systems. Master's thesis. University of Central Florida, Orlando.

Fransolet, G., et al, 1988. The Role of Bicarbonate in Bacterial Growth in Oligotrophic Waters. Jour. AWWA, 80:11:57.

Hedberg, T. & Johansson, E., 1987. Protection of Pipes Against Corrosion. Water Supply, 5:13-4:SS20.

Larson, T.E. & Skold, R.V., 1957. Corrosion and Tuberculation of Cast Iron. Jour. AWWA, 49:10:1294.

Larson, T.E. & Skold, R.V., 1958. Laboratory Studies Relating Mineral Water Quality of Water to Corrosion of Steel and Cast Iron. Corrosion-NACE, 15:285t.

Pisigan, R.A. & Singley, J.E., 1984. Evaluation of Waters' Corrosivity Using Langelier Index and Relation to Corrosion Rate Models. Corrosion, 84:149.

Pisigan, R.A. & Singley, J.E., 1985. Effects of Water Quality Parameters on the Corrosion of Galvanized Steel. Jour. AWWA, 77:11:76.

Pisigan, R.A. & Singley, J.E., 1987. Influence of Buffer Capacity, Chlorine Residual, and Flow Rate on Corrosion of Mild Steel and Copper. Jour, AWWA, 79:2:62.

Sander, A., et al, 1997. Iron Corrosion in Drinking Water Distribution Systems-Surface Complexation Aspects. Corros. Sci., 39:1:77.

Singley, J.E., 1981. The Search for a Corrosion Index. Jour, AWWA, 73:11:578.

Singley, J.E., et al, 1985. Corrosion Prevention and Control in Water Treatment and Supply Systems. Pollution Technol. Rev., 122:12.

Trussell, R.R. & Thomas, J.F., 1971. A Discussion of the Chemical Character of Water Mixtures. Jour. AWWA, 63:1:49.

ABOUT THE AUTHORS

Syed A. Imran is a visiting fellow at the National Research Council of Canada, Ottawa, Canada. This article is based on his dissertation research at the University of Central Florida, Orlando. John D. Dietz (to whom correspondence should be addressed) is an associate professor with the Department of Civil and Environmental Engineering, University of Central Florida, P.O. Box 162450, Orlando, FL 32816-2450; e-mail jdietz@mail.ucf.edu. Ginasiyo Mutoti has a PhD from the University of Central Florida and is now a process engineer at Timmons Group, Richmond, Va. James S. Taylor is the Alex Alexander Professor and director of the Environmental Systems Engineering Institute at the University of Central Florida. A.A Randall is an associate professor in the University of Central Florida's Department of Civil and Environmental Engineering, and C. D. Cooper is a professor in that department.

If you have a comment about this article, please contact us at journal@awwa.org.

Copyright American Water Works Association Sep 2005

Source: American Water Works Association. Journal

More News in this Category