Impacts of blending ground, surface, and saline waters on lead release in drinking water distribution systems

Impacts of blending ground, surface, and saline waters on lead release in drinking water distribution systems

ARTICLE IN PRESS WAT E R R E S E A R C H 40 (2006) 943 – 950 Available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/watres I...

337KB Sizes 6 Downloads 98 Views

ARTICLE IN PRESS WAT E R R E S E A R C H

40 (2006) 943 – 950

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Impacts of blending ground, surface, and saline waters on lead release in drinking water distribution systems Zhijian Tanga, Seungkwan Hongb,, Weizhong Xiaoa, James Taylora a

University of Central Florida, Civil and Environmental Engineering Department, P.O. Box 162450, Orlando, FL 32816, USA Korea University, Civil and Environmental Engineering, Anam-dong, Seungbuk-ku, Seoul, 136– 713, Korea

b

art i cle info

A B S T R A C T

Article history:

The impacts of distribution water quality changes caused by blending different source waters

Received 25 April 2005

on lead release from corrosion loops containing small lead coupons were investigated in a pilot

Received in revised form

distribution study. The 1-year pilot study demonstrated that lead release to drinking water

28 November 2005

increased as chlorides increased and sulfates decreased. Silica and calcium inhibited lead

Accepted 20 December 2005

release to a lesser degree than sulfates. An additional 3-month field study isolated and verified

Available online 3 February 2006

the effects of chlorides and sulfates on lead release. Lead release decreased with increasing pH

Keywords:

and increasing alkalinity during the 1-year pilot study; however, the effects of pH and

Lead corrosion

alkalinity on lead release, were not clearly elucidated due to confounding effects. A statistical

Water blending

model was developed using nonlinear regression, which showed that lead release increased

Water quality

with increasing chlorides, alkalinity and temperature, and decreased with increasing pH and

Statistical regression

sulfates. The model indicated that primary treatment processes such as enhanced coagulation

Pipe distribution systems

and RO (reverse osmosis membrane) were related to lead release by water quality. Chlorides are high in RO-finished water and increase lead release, while sulfates are high following enhanced coagulation and decrease lead release. & 2006 Elsevier Ltd. All rights reserved.

1.

Introduction

In recent years, the use of alternative source waters such as seawater is common practice in the water community since traditional resources are limited and regulation constraints and demand are increasing. For instance, Tampa Bay Water (TBW) has decreased groundwater utilization and increased utilization of surface and saline sources for drinking water supply because of groundwater depletion and regulatory constraints. Allowable surface water use is limited by availability and varies by season, but is maximized to reduce groundwater demand. Sea water utilization by reverse osmosis membrane (RO) is intended to be constant. Hence, a constant distribution system water quality cannot be maintained. Although each source-specific finished water met all regulatory constraints, variations in blended finished water quality can adversely impact lead release to the finished water in the distribution system.

As stated by AWWA (1996) and Birden et al. (1985), generally source waters are free of lead, but significant amounts of lead may be present in the tap water due to dissolution of lead corrosion products, which are formed in domestic plumbing systems. AWWA (1996) reported that pH, alkalinity, sulfate, chloride, dissolved oxygen (DO) and temperature have significant impacts on lead corrosion in drinking water. Boffardi (1988, 1990, 1995), Johnson et al. (1993), Schock (1980, 1989), Schock and Gardels (1983), and Walker and Oliphant (1982) stated that increasing pH can significantly decrease lead release when pH is less than 8.0. Alkalinity addition has been reported to reduce lead, particularly in poorly buffered waters. The optimum alkalinity for lead corrosion control is associated with pH. The adjustment of pH and alkalinity has been widely practiced to mitigate lead release (Risser, 1997; Juadge, 1994; Chen et al., 1994; Vinci, 1991). Chloride and sulfate can also affect lead corrosion by

Corresponding author. Tel.: 82 2 3290 3322; fax: 82 2 928 7656.

E-mail addresses: [email protected] (S. Hong), [email protected] (J. Taylor). 0043-1354/$ - see front matter & 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2005.12.028

ARTICLE IN PRESS 944

WA T E R R E S E A R C H

complexation and/or other mechanisms (AWWA, 1996). Oliphant (1983) reported that lead corrosion was decreased by sulfate. Johnson et al. (1993), Lee et al. (1989) and Boffardi (1988) showed that high temperature common to finished drinking water enhanced lead corrosion and lead release. The main objective of this work was to investigate the effect of blending on lead release in drinking water distribution system. This work was part of a pilot-scale pipe distribution study in which the impacts of blending ground, surface and saline waters on water quality (including general water qualities, biological activities, and the corrosion and release of iron, copper and lead) were investigated systematically. The study was conducted over a 3-year period, which included 18 months of field operation and data collection. To accomplish this objective, a 790 m2 roofed research facility was constructed, which included seven different water treatment processes, 18 pilot distribution systems (PDS) and 18 Cu–Pb corrosion loops that immediately followed the PDSs. Finished waters were produced from ground, surface, and saline sources using the pilot systems, which simulated TBW’s water treatment facilities. The finished waters were blended at various ratios and fed to pilot pipe distribution systems. The copper corrosion loops contained small lead coupons that were sized to simulate the lead solder surface area in homes, and received PDS effluent, which represented several combinations of TBW-finished water quality. A variety of water quality parameters were monitored and correlated to lead release from the corrosion loops. Based on pilot study data, the effects of blending ratios on lead release were investigated, and statistically significant nonlinear regression models were developed to predict lead release to finished water as a function of water quality.

2.

Experimental

2.1.

Source water production and blending

The seven processes that were used to produce finished water are shown in Table 1. The first four waters (G1, G2, S1 and RO) represent existing facilities; however, the primary TBW water treatment facilities are represented by G1, S1 and RO as only one utility utilizes softening (G2) for drinking water

40 (2006) 943– 950

treatment. The last three waters (G3, G4, and S2) were designed to simulate potential water treatment facilities in the future. All of the processes that had disinfection utilized chloramines for residual maintenance and were stabilized with respect to calcium carbonate before PDS discharge. All finished waters were disinfected with free chlorine to CT criteria except S1, which utilized ozone. NH4Cl was added immediately after disinfection to form chloramines for residual maintenance except in S1, where free chlorine and NH4Cl were added simultaneously. These processes represent TBW’s and MG’s current and future methods of producing finished drinking water. Residual maintenance with chloramines was necessary to control trihalomethane and haloacetic acid formation to less than the maximum contamination limit; consequently, chloramines were used in this work.

2.2.

Pipe distribution systems (PDSs)

The PDSs were constructed with aged pipes approximately 10–50-years old that were taken from actual distribution systems and were provided by the utilities that were supplied by TBW. These utilities had utilized only groundwater since inception. PVC, lined cast iron (LCI), unlined cast iron (UCI), and galvanized steel (G) pipes were used to make the PDSs. The PDSs included 14 hybrid PDSs that were made from all four materials, and four different PDSs that were made of only a single pipe material. The nominal length of each line is approximately 30 m. Pipe diameters are 15 cm with the exception of galvanized that is 5 cm. The PDSs equilibration was declared when the PDS effluent apparent color, turbidity, alkalinity and pH did not change for four consecutive observations, which occurred in 4 months following construction. The pilot systems were operated at a 5-day hydraulic residence time (HRT), were flushed with five pipe volumes of water at 0.3 ft/sec every week. This velocity was selected to limit accumulation of films on the pipe interior. Feed and effluent samples were monitored for a large number of parameters including pH, alkalinity, temperature, chloride, sulfate, calcium, magnesium, sodium, silica, DO, HPC and residual.

Table 1 – Source water and related processes Code

Source water

Processes

G1 G2 S1

Conventional groundwater Lime-softened groundwater Enhanced surface water treatment

RO

Desalination of reverse osmosis

G3 G4

Lime-softened blend of G1, S1 and RO Membrane-treated blend of G1, S1 and RO

S2

Integrated membrane system surface water

Aeration, disinfection stabilization Lime softening, disinfection, stabilization Ferric sulfate coagulation, flocculation, settling, filtration, disinfection, biologically activated carbon filtration, chlorination, stabilization Reverse osmosis membrane, inorganics addition for stabilization, aeration, disinfection, and stabilization Lime softening, disinfection, stabilization Nanofiltration treatment, inorganics addition for stabilization, aeration, chlorination, stabilization Ferric sulfate coagulation, flocculation, settling, filtration, nanofiltration treatment, inorganics addition for stabilization, aeration, chlorination, stabilization

ARTICLE IN PRESS

The annual averages of the effluent PDS water quality is summarized in Table 2, which clearly shows that source water quality and treatment process determined finished water quality. Conventionally treated ground water (G1) had high alkalinity, which was source specific and unchanged by conventional treatment. Lime-softened groundwater (G2) had much less alkalinity and was impacted by treatment. S1 finished water had higher sulfates than all other finished waters and was influenced by treatment as the sulfates were primarily introduced by enhanced coagulation. S2 finished water had moderate alkalinity and low sulfates, and was

24.2 24.2 23.9 24.0 24.0 23.9 24.1 24.1 24.0 23.8 24.0 24.2 23.9 23.8 6.5 6.9 6.2 6.3 6.3 5.6 5.2 6.2 6.0 5.6 6.5 5.5 6.5 6.1 0.060 0.057 0.043 0.028 0.024 0.029 0.028 0.041 0.048 0.048 0.034 0.046 0.037 0.040 562 338 466 379 600 412 441 580 501 526 565 557 481 444 13.7 13.7 10.4 10.0 10.0 8.8 3.5 11.9 10.9 9.8 8.8 10.8 11.2 10.9 84.8 34.1 35.4 35.0 56.5 28.0 28.7 70.3 66.1 69.2 56.1 66.4 45 40.8 18.0 13.8 37.3 29.8 48.7 45.5 52.3 34.1 29.5 29.8 41.5 32.7 32.2 31.4 26.1 26.0 79.0 6.8 190.2 12.7 5.8 105.9 19.9 19.4 93.4 77.4 112.6 73.4 28.9 23.0 49.7 43.9 37.1 68.5 91.7 35.0 45.9 49.4 53.5 40.1 32.7 39.7 207 89 61 88 60 66 69 139 151 158 104 147 73 77

UV-254 (cm1) Cond (US/cm) SiO2 Ca+2 (mg/L) SO2 4 (mg/L) Cl (mg/L)

100 100 100 100 100 100 100 55/45 60/30/10 68/32 23/45/32 62/27/11 50/50 62/24/14

3.1. Effects of blending source waters on distribution system water quality

G1 G2 G3 G4 S1 S2 RO G1/S1 G1/S2/RO G1/RO G1/S1/RO G1/S1/RO G2/S1 G2/S1/RO

Results and discussion

Alk (mg/L)

3.

pH

‘‘Sigma Plot’’ software was used to develop lead release models as a function of water quality using linear and nonlinear regression. Initially all measured water quality parameters were included in the models; however, only statistically significant (less than 0.05 confidence level) terms remained in the final model. The models were developed by iteration. Following regression, the term with the highest pvalue over 0.05 was discarded and the regression was repeated. This iteration was repeated until only statistically significant terms remained.

Na+

Statistical regression analysis

Table 2 – Average water quality of different blended waters

2.4.

DO (mg/L)

A copper and lead corrosion loop was installed at the end of each PDS to simulate kitchen tap monitoring for LC (lead and copper) Rule compliance. The corrosion loop consisted of 9 m copper tubing with a diameter of 1.6 cm, which contained 1.8 L of water. One pure lead coupon was placed within the copper tubing between two standard copper tube fittings (brass) for each loop system. All other fittings and materials were PVC or other plastic polymers. The coupons had a surface area of 9.88 cm2 (dimension of 7.6  1.3 cm). Assuming a 0.32 cm bead on the inside of each 1.3 cm diameter joint as the surface area exposed to water using solder, the coupon surface area should be equivalent to 17 joint-ends or 7–8 fittings which could be reasonable for plumbing associated with a kitchen sink. To simulate water delivered to consumers, the corrosion loops received water after it passed through the associated pipe distribution system. The corrosion loops were flushed with 7.6 L of water every morning. One liter sample was collected after 6 h stagnation in the afternoon. The 1 L sample was shaken to mix completely and then 125 mL was used for total lead analysis. Using nitric acid, water sample was acidified to pH less than 2, then total lead concentration was measured using an atomic absorption spectrophotometer. Statistical analysis of PDSs and corrosion loop effluent water quality during the first quarter of the field investigation found there was no difference; hence, corrosion loop effluent was only monitored for copper and lead.

7.87 7.92 8.13 8.17 7.92 8.13 8.06 7.94 7.99 7.94 8.01 7.91 8.02 8.03

Temp 1C

Corrosion loops

Ratio (%)

2.3.

945

40 (20 06) 94 3 – 950

Blending

WAT E R R E S E A R C H

ARTICLE IN PRESS

impacted by treatment as nanofiltration reduced alkalinity and easily removed sulfates. RO water was the simulated ROtreated seawater in which sea salt was added to match the finished water from the TBW Regional Desalination Facility, (i.e., 100 mg/L chloride) and was impacted by source and treatment as that although RO significantly reduces chlorides, the chlorides are by far the highest in the saline source and remain highest in the RO finished water. The quality of blended waters varied according to blending ratios of three principal source waters (G1, S1, and RO). The higher blending ratio of G1 water resulted in higher alkalinity. The blended waters with higher S1 and RO sources showed higher sulfates and chlorides, respectively. Lime softened (G3) and/or nanofiltered (G4) finished waters were produced from G1, S1, and RO blends, and had low calcium and sulfate concentrations. Although temperature and DO did vary by season (time) among of the finished waters, temperature and DO were always equal in all sources at any point in time. Due to the nature of source waters and processes, G1 and S1 showed high conductivity (i.e. total dissolved ions), while the nanofiltered or lime softened waters exhibited lower ion concentrations.

3.2.

Alka.>140

50

80< Medium Alka.<140 Alka.<80

40 30 20 10

(a)

0 7.5

7.6

7.7

7.8

7.9

8.0 pH

8.1

8.2

8.3

8.4

8.5

High pH:>8.1 Medium pH:7.9 - 8.1 Low pH:<7.9

50 40 30 20 10 0 0

(b)

30

60 90 120 150 180 Alkalinity (mg/L as CaCO3)

210

240

Fig. 1 – Effect of pH and alkalinity on lead release: (a) pH; (b) Alkalinity.

Effects of blended water quality on lead release

As discussed in the previous section, blending of different source waters caused significant changes in distribution system water quality. The alteration of water quality in distribution systems imposes potential for lead release in drinking water by creating chemical conditions favorable for lead corrosion, and/or by disrupting existing chemical films that were formed in the presence of ground water. In the following subsections, the impacts of variations in distribution water qualities by source water blending on lead release are systematically delineated in terms of primary water qualities including pH, alkalinity, chloride, sulfate, calcium and silicate. Since there was no significant variation in DO, its effect on lead release was not discussed. The lead coupons in the corrosion loops were exposed to finished water for 3 months. Total lead was measured every 2 weeks during 3month incubation.

3.2.1.

40 (2006) 943– 950

Lead (µg/L)

WA T E R R E S E A R C H

Lead (µg/L)

946

Effect of pH and alkalinity

According to the literature, pH and alkalinity are considered to be the most important parameters affecting lead corrosion since they control the solubility of lead corrosion scales such as hydrocerussite (Pb3(OH)2(CO3)2) and cerussite (PbCO3). Fig. 1(a) shows average lead concentrations observed in our pilot study as a function of average pH of blended waters. The experimental data are presented as three groups of alkalinity as shown in Fig. 1(a) with an overall trend line based on all the data. Similarly the effect of alkalinity on lead release is presented in Fig. 1(b) as three pH groups with an overall trend line based on all the data. As shown, there was no clear trend of pH or alkalinity versus lead release, even in the subgroup divided by pH or alkalinity. The average lead concentrations were 10.6 mg/L lead for high pH (48:1), 9.3 mg/L for medium pH (7.9–8.1) and 9.4 mg/L for low pH (o7:9). The alkalinity also did not show any correlation with lead concentration. Specifi-

cally, the average lead concentrations for high (4140 mg=L as CaCO3), medium (80–140 mg/L as CaCO3) and low (o80 mg=L as CaCO3) alkalinities were 9.9, 8.5, and 10.5 mg/L, respectively. No clear effect of pH and alkalinity was observed in the pilot study, probably due to this confounding effect between these two parameters due to stabilization, which caused finished water pH to decrease as alkalinity increased and vice versa in these studies; hence, pH and alkalinity were confounded. In addition, their correlations with lead corrosion may be masked by the dominant effect of chloride and sulfate described in the following subsection.

3.2.2.

Effect of chloride and sulfate

In addition to alkalinity (i.e. HCO 3 at typical drinking water conditions), chlorides and sulfates may affect lead release. Fig. 2 (a) shows the effect of chloride on lead release. The experimental data are divided into three pH groups to minimize the pH effect. As shown, increasing chlorides corresponded to increasing lead levels. This trend holds for all three pH groups. Increasing chlorides could complex lead ions to form more soluble compounds or alter scale formation to enhance particulate lead release, and thus increase total lead release to drinking water. This observation indicates that lead concentration in drinking water will increase as RO blending ratio increases since chlorides are typically higher in RO finished waters. The effect of sulfate on lead release is presented in Fig. 2(b). As shown, high sulfate reduced lead release in all three pH ranges. This trend suggests that lead release will be reduced when the blending ratio of S1 finished water is increased. Previous work supports our finding. Oliphant (1983) stated that sulfate might control lead corrosion by affecting the physical structure of corrosion films, which prevents the release of lead corrosion products.

ARTICLE IN PRESS WAT E R R E S E A R C H

High pH:>8.1

50

Medium pH:7.9 - 8.1

40 (20 06) 94 3 – 950

947

Low pH:<7.9

Lead (µg/L)

40 30 20 10 0 0

30

60 Chloride (mg/L)

(a) 50

90

120

High pH:>8.1

Medium pH:7.9 - 8.1

Low pH:<7.9

30

120 150 180 Sulfate (mg/L)

210

Lead (µg/L)

40 30 20 10 0 0

(b)

60

90

240

270

Fig. 2 – Effect of major anions on lead release: (a) Chloride; (b) Sulfate.

SEM pictures of lead coupon surfaces shown in Fig. 3(a) and (b) indicate that chloride and sulfate concentrations in finished water affected the release of total lead by affecting lead scale mineralogy. The lead surface exposed to the relatively high sulfate concentrations appears significantly smoother and compacter as shown in Fig. 3(a), whereas the lead surface exposed to relatively higher chloride concentrations as shown in Fig. 3(b) appears to have flakes which may be released as particulates much more readily than from the smoother surface. Hence, the appearance of the surfaces indicates particle release in finished waters that contain high sulfates as would be common to finished waters produced by coagulation using a sulfate-based salt would be less than for finished waters that contain relative high chlorides as would be common to finished waters produced by desalination of a highly saline source.

3.2.3.

Fig. 3 – Representative SEM pictures—physical surface structure of corrosion scales formed under various source waters: (a) S1 water; (b) RO water.

Effect of calcium and silica

In Fig. 4(a) and (b), lead concentrations are correlated to calcium and SiO2 of blended waters. In pHo7:9, calcium did not show any effect however, a negative trend was observed for pH47:9, suggesting that calcium precipitation may reduce lead release by forming or enhancing the solid film on the lead surface that can mitigate lead release. SiO2, on the other hand, decreased lead release for all pH ranges. It has been reported in the literature that silicate compounds are often used as an inhibitor to lessen metal corrosion including lead (AWWA, 1996; Anon, 1994). Lastly it should be noted that some of water quality parameters were related to each other due to characteristics of raw source waters and associated treatment processes, which caused confounding effects on lead corrosion. To further clarify the effects of blended water quality, linear correlations (R2) among major water quality parameters were estimated and summarized in Table 3. Alkalinity was

positively correlated with calcium (R2 ¼ þ0:60) due to the relatively high calcium and alkalinity in G1. Chloride was negatively related to silica (R2 ¼ 0:56) again because of the relatively high silica and low chlorides in groundwater. Thus, the effect of calcium as described above may be attributable to alkalinity. Similarly, silica’s inhibitory effect on lead corrosion may be attributable to chloride.

3.3.

Correlation between water quality and lead release

The effect of blended water quality on lead release was not clearly understood from 1 year of pilot blending study because of confounding correlations among primary distribution water quality parameters. In order to resolve this problem, a series of experiments were conducted 3-month using six PDSs. In these experiments, only one water quality

ARTICLE IN PRESS 948

WA T E R R E S E A R C H

40 (2006) 943– 950

parameter was varied systematically while others were kept constant. The changes of pH, alkalinity, and sulfate were accomplished by varying blends of G1, S1, RO and DI (distilled water), and by adding sulfuric acid and sodium bicarbonate. The variation of chloride levels was achieved by adjusting the sea salt added to RO water. In the first month, pH/alkalinity levels varied while chloride and sulfate concentrations were maintained at 33 and 15 mg/L, respectively. Sulfate concentrations were changed in the second month at pH ¼ 7:8, alkalinity ¼ 80 mg=L as CaCO3, and chloride ¼ 33 mg=L. In the last month, chloride levels varied while other parameters were kept constant (i.e. pH ¼ 7:7, alkalinity ¼ 72 mg=L as CaCO3, and sulfate ¼ 19 mg=L).

The effects of pH and alkalinity on lead release in this controlled study are presented in Fig. 5(a) and (b). As shown, lead concentration was reduced as pH and alkalinity increased. The results also clearly demonstrated that lead release increased linearly with increasing chloride and decreasing sulfate, similar to 1-year pilot study described previously. Note that other dissolved compounds such as calcium and silicate were kept relatively constant, thus their confounding effects with alkalinity and chloride were minimized. This observation clearly demonstrated the sole impact of chloride and sulfate on lead corrosion and clarified confusion made from 1-year pilot blending study. But since the pH and alkalinity varied simultaneously by adding

13 12

High pH:>8.1

Medium pH:7.9 - 8.1

Low pH:<7.9

11 Lead (µg/L)

Lead (µg/L)

50 40 30 20

10 9 8 7

10

6

0 20

40

60

(a)

80

100

7.6

7.7

7.8

0

30

60

7.9

8.3

8.4

8.5

90 120 150 180 210 Alkalinity (mg/L as CaCO3)

240

270

(a)

Calcium (mg/L)

8.0 8.1 pH

8.2

13 12

High pH:>8.1

Medium pH:7.9 - 8.1

Low pH:<7.9

11 Lead (µg/L)

Lead (µg/L)

50 40 30 20

10 9 8 7 6

10

5

0 0

2

4

(b)

6

8

10

12

14

16

18

(b)

20

Silica (mg/L)

Fig. 4 – Effect of calcium and silica on lead release: (a) Calcium; (b) Silica.

Fig. 5 – Effect of major water quality parameters on lead release in controlled experiments: (a) pH; (b) Alkalinity (Each data point is the average value of 1 month observations.)

Table 3 – Linear correlations among major water quality parameters Parameter pH Alk Cl SO2 4 Na+ Ca+2 SiO2 Cond UV-254 DO Temp

pH

Alk.

Cl

SO2 4

Na+

Ca+2

SiO2

Cond

UV-254

DO

Temp

+1.00 0.15 +0.11 0.00

0.16 +100 0.13 0.06

+0.11 0.13 +1.00 0.04

0.00 0.06 0.04 +1.00

0.00 +0.23 +0.34 +0.02

0.06 +0.60 0.10 +0.10

0.00 +0.16 0.56 +0.03

0.24 +0.21 0.01 +0.26

0.09 +0.31 0.21 0.08

+0.19 0.00 0.01 +0.02

0.15 0.00 0.00 0.01

0.00 0.06 0.00 0.24 0.09 +0.19 0.15

+0.23 +0.60 +0.16 +0.21 +0.31 0.00 0.00

+0.34 0.10 0.56 0.01 0.21 0.01 0.00

+0.02 +0.10 +0.03 +0.26 0.08 0.02 0.01

+1.00 0.22 0.65 +0.06 0.27 0.12 +0.04

0.22 +1.00 +0.24 +0.38 +0.09 +0.05 0.04

0.65 +0.24 +1.00 0.00 +0.16 +0.08 0.00

+0.06 +0.38 0.00 +1.00 0.00 0.04 +0.02

0.27 +0.09 +0.16 0.00 +1.00 +0.05 0.00

0.12 +0.05 +0.08 0.04 +0.05 +1.00 0.40

+0.04 0.04 0.00 +0.02 0.00 0.40 +1.00

ARTICLE IN PRESS WAT E R R E S E A R C H

sulfuric acid and sodium bicarbonate, the specific effect of alkalinity or pH was still not clearly elucidated.

50 R2 = 0.42 40

Statistical multiple regression

In addition to controlled experiments, statistical multiple regressions was performed on experimental data collected during 1 year of pilot study to further investigate the effects of blended water quality on lead corrosion. This effort was made also to develop empirical models that will be utilized to simulate and predict lead release under various blending scenarios. Regressions were conducted based on 90% of the data (90 data points) collected in 1-year pilot study, and the other 10% data collected in 1-year pilot study and data collected from controlled experiment were used to verify the statistical model. In the model development, quarterly average data were used for data collected in 1-year pilot study and monthly average data were used for data collected in controlled experiment. According to our pilot observations, seven major parameters were chosen for multiple regressions. They included alkalinity, pH, chloride, sulfate, conductivity, DO, and temperature. Other dissolved ions such as sodium, calcium, and silicate were excluded since they were correlated with main parameters mentioned above (refer to Table 1). Regressions were systematically performed according to the procedures introduced previously. A variety of regression functions were examined and evaluated. Among them, the following function gave the best results: Lead ¼ 1:027ðTemperature25Þ  Alkalinity0:677  pH2:726  Chloride1:462  Sulphate0:228 ,

ð1Þ

where lead concentration is given in mg/L; alkalinity is given in mg/L as CaCO3; chloride and sulfate are in given in mg/L; temperature is given in 1C. The statistical coefficients of the statistical model were summarized in Table 4. All of the parameters have a p-value less than 0.05. The statistical model indicates that increasing pH and sulfate will decrease lead release, while high alkalinity and chloride content and conductivity will increase lead release. The models indicate a similar trend for main parameters as the previous study, except for alkalinity. It should be noted that there were correlations between pH and alkalinity in controlled study since both of them varied with respect to addition of acid or base. As a result, increasing pH came with increasing alkalinity, indicating that the effect of alkalinity on lead release shown in controlled study might be caused by pH

Table 4 – Summary of statistical coefficient of regression model Parameters

Alkalinity PH Chloride Sulfate Temperature (Y)

Coefficient

Std. error

t

P

0.6774 2.7262 1.4624 0.2276 1.0267

0.2783 1.2303 0.3586 0.0872 0.0215

2.4339 2.2159 4.0785 2.6101 47.7813

0.0173 0.0297 0.0001 0.0109 o0:0001

Predicted Lead (µg/L)

3.4.

949

40 (20 06) 94 3 – 950

30

20

10

0 0

10

20 30 Actual Lead (µg/L)

40

50

Fig. 6 – Comparison of regression model predictions and actual lead releases.

and the statistical model might reveal the actual trends. The model also demonstrated the importance of temperature; the higher the temperature, the higher the lead corrosion potential. The trends observed from multiple regression regarding primary anions agreed fairly well with previous experimental work as well as corrosion theory. The model developed was used to predict lead release and their predictions were compared with actual experimental data obtained in pilot study. The results are presented in Fig. 6. As shown, the regression model was able to predict the values close to actual lead concentrations, although R2 value for statistical models was 0.42. The model seemed to give slightly lower values when high lead release was observed in the pilot study. However, considering the high variability of lead levels reported in the literature and the difficulty associated with developing mechanistic model for lead corrosion, it can be said that our regression model provides an adequate and practical tool that can be utilized to simulate the effect of source water blending on lead corrosion.

4.

Conclusions

Primary inferences from this study are summarized as follows:

 Due to the nature of raw water quality and associated



treatment processes, G1 water had high alkalinity, while S1 and RO sources were characterized as high sulfate and high chloride waters, respectively. The blending ratio of different treated waters determined the quality of finished waters, which determined the release of total lead in these studies. One year of pilot study showed that increasing chloride enhanced lead levels while high sulfate lessened lead release. The effect of pH and alkalinity, however, was not clearly observed probably due to confounding correlation

ARTICLE IN PRESS 950





WA T E R R E S E A R C H

among major water quality parameters. Such observations are not uncommon when graphical interpretation is used to describe phenomena that may be affected by several independent variables. Additional field study clearly demonstrated that lead release increased with increasing chloride and decreasing sulfate. It was also shown that lead concentration was reduced as pH and alkalinity increased simultaneously. A nonlinear statistical model was developed and verified that accurately predicted total lead release increased as chlorides, alkalinity and temperature increased and decreased as sulfates and pH increased for the conditions of this study. Although such models are empirical, they may be successfully used to assess the potential for total lead release in similar conditions, and indicate that RO finished waters or any finished water high in chlorides will adversely impact total lead release.

Acknowledgments Support for this research was provided by Tampa Bay Water (TBW), and AWWA Research Foundation (AwwaRF). The authors specially acknowledge Roy Martinez, AwwaRF Senior Account Officer, who was the Project Officer, and Chris Owen, TBW Quality Assurance Officer. The TBW Member Governments: Pinellas County, Hillsborough County, Pasco County, Tampa, St. Petersburg, and New Port Richey; and the AwwaRF Project Advisory Committee are recognized for their review and recommendations. Pick Talley, Robert Powell, Dennis Marshall and Oz Wiesner from Pinellas County, and Dr. Luke Mulford from Hillsborough County are also specifically recognized for their contributions. UCF Environmental Engineering graduate students, especially Jorge Arevalo and faculty who worked on this project are recognized for their efforts.

Appendix A.

Supplementary Material

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.watres.2005.12.028.

40 (2006) 943– 950

R E F E R E N C E S

Anon, 1994. Using silicate to lower lead levels in drinking water. Public Works 125 (10), 83–84. AWWA (American Water Works Association), 1996. Internal Corrosion of Water Distribution Systems, second ed. American Water Works Association, Denver, Colo. Birden, H.H., Calabrese, E.J., Stoddard, A., 1985. Lead dissolution from soldered joints. J. Am. Water Works Assoc. 77 (11), 66–70. Boffardi, B.P., 1988. Lead in drinking water—causes and cures. Public Works 119 (12), 67–70. Boffardi, B.P., 1990. Minimization of lead corrosion in drinking water. Mater. Performance 29 (8), 45–49. Boffardi, B.P., 1995. Lead corrosion. J. New Engl. Water Works Assoc. 109 (2), 121–131. Chen, C-A., McAnally, A.S., Kumaraswamy, S., 1994. Lead and copper corrosion control. Environmental Science and Health. Part A. Environ. Sci. Eng. 29 (8), 87–1606. Johnson, B., Yorton, R.T.T., Kim, J., 1993. Evaluation of corrosion control alternatives to meet the lead and copper rule for Eastern Massachusetts. J. New Engl. Water Works Assoc. 107 (1), 24–30. Juadge, J.A., 1994. Use of the bicarbonate/pH process for corrosion control in Fitchburg, Massachusetts. J. New Engl. Water Works Assoc. 108 (2), 126–132. Lee, R.G., Becker, W.C., Collins, D.W., 1989. Lead at the tap: sources and control. J. Am. Water Works Assoc. 81 (7), 52–62. Oliphant, R.J., 1983. Lead contamination of potable water arising from soldered joint. Water Supply 1 (2–3), SS18-5–SS18-12. Risser, T.M., 1997. The use of potassium carbonate for corrosion control in small systems. J. New Engl. Water Works Assoc. 111 (3), 253–257. Schock, M.R., 1980. Response of lead solubility to dissolved carbonate in drinking water. J. Am. Water Works Assoc. 72 (12), 695–704. Schock, M.R., 1989. Understanding corrosion control strategies for lead. J. Am. Water Works Assoc. 81 (7), 88–100. Schock, M.R., Gardels, M.C., 1983. Plumbosolvency reduction by high pH and low carbonate-solubility relationship. J. Am. Water Works Assoc. 75 (2), 87–91. Vinci, A., 1991. Reducing lead contamination in potable water using sodium bicarbonate and sodium hydroxide. In: Proceedings of the AWWA Annual Conference. American Water Works Association, Denver, CO. Walker, R., Oliphant, R., 1982. Corrosion of lead in drinking water. Anti-Corrosion Methods Mater. 29 (4), 8–11.