Evaluating the Effects of Full and Partial Lead Service Line

Jun 23, 2016 - Lead service line replacement (LSLR) is an important strategy for reducing lead ... Environmental Science & Technology 2017 51 (17), 95...
0 downloads 3 Views 755KB Size
Subscriber access provided by Georgetown University | Lauinger and Blommer Libraries

Article

Evaluating the effects of full and partial lead service line replacement on drinking water lead levels Benjamin F Trueman, Eliman Camara, and Graham A. Gagnon Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.6b01912 • Publication Date (Web): 23 Jun 2016 Downloaded from http://pubs.acs.org on June 24, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 44

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Environmental Science & Technology

EVALUATING THE EFFECTS OF FULL AND PARTIAL LEAD SERVICE LINE REPLACEMENT ON DRINKING WATER LEAD LEVELS Benjamin F. Trueman1, Eliman Camara1, and Graham A. Gagnon1,* 1

Civil & Resource Engineering Dalhousie University 1360 Barrington St. Halifax, NS, CAN B3H 4R2 *Corresponding Author: G.A. Gagnon, e-mail [email protected], Tel: 902.494.3268, Fax 902.494.3108 Submitted to Environmental Science & Technology in revised form for review and possible publication, June 2016 KEYWORDS: lead service line replacement; drinking water; distribution system; lead and copper rule

ACS Paragon Plus Environment

Environmental Science & Technology

41 42 43

ABSTRACT

44

lead exposure via drinking water, but jurisdictional issues can sometimes

45

interfere with full replacement of the lead line. The effects of full and partial

46

LSLR on lead levels were assessed using 5 × 1L sample profiles collected at

47

more than 100 single-unit residences. Profiles comprised four sequential

48

standing samples (L1-4) and a free-flowing sample (L5) drawn after a 5-

49

minute flush of the outlet. At 45 sites with full LSLs, 90th percentile lead

50

levels in standing samples ranged from 16.4 to 44.5 µg L-1 (L1 and L4,

51

respectively). In the free-flowing sample (L5), 90th percentile lead was 9.8

52

µg L-1. Within 3 days, full LSLR had reduced L3-5 lead levels by more than

53

50%, and within 1 month, lead levels were significantly lower in every litre

54

of the sample profile. Conversely, partial LSLR more than doubled premise

55

plumbing (L1-2) lead release in the short term and did not reduce L1-2 lead

56

release in the long term. Even 6 months after partial LSLR, 27% of first-

57

draw lead levels were greater than 15 µg L-1 (the U.S. EPA action level),

58

compared with 13% pre-replacement.

Lead service line replacement (LSLR) is an important strategy for reducing

59 60 61 62 63 64

ACS Paragon Plus Environment

Page 2 of 44

Page 3 of 44

Environmental Science & Technology

65

INTRODUCTION

66

Lead service lines (LSLs)—pipes connecting distribution mains to premise

67

plumbing—were installed widely throughout the first half of the twentieth

68

century in North America and even occasionally up until the U.S.

69

congressional ban in 1986.1 In Canada, the National Plumbing Code

70

permitted installation of LSLs until 1975.2 At sites where they were

71

installed, 50-75% of drinking water lead may be attributable to the LSL.3

72

73

Elevated lead in drinking water is a significant public health concern

74

because water lead levels have been shown to correlate positively with blood

75

lead levels.4-7 Childhood blood lead levels below 10 µg dL-1—and early

76

childhood levels as low as 2 µg dL-1—are linked with deficits in cognitive

77

and academic skills.8-10 Chronic lead exposure in adults is associated with

78

renal dysfunction11 and hypertension12 and lead is a known abortifacient.13

79

Lead in U.S. drinking water is regulated under the Lead and Copper Rule,14

80

which specifies an action level of 15 µg L-1 for the 90th percentile first-draw

81

lead level. In Canada, the maximum acceptable concentration in a free-

82

flowing sample is 10 µg L-1, a benchmark that serves as the current Nova

83

Scotia regulation.15 Health Canada also recommends corrective action when

ACS Paragon Plus Environment

Environmental Science & Technology

84

the 90th percentile first-draw lead level exceeds 15 µg L-1.2,16

85

86

Lead service line replacement (LSLR) is an important strategy for reducing

87

lead exposure via drinking water, but joint (public-private) ownership can

88

interfere with full replacement of the LSL. Typically, partial LSLR occurs

89

when the public LSL is replaced with copper and joined to the private LSL

90

left in place.1 Partial LSLR can cause elevated lead in drinking water—

91

disturbance of LSL corrosion scale during replacement may release high

92

levels of lead for an extended period post-replacement. The lead-copper

93

junction is a specific concern due to the potential for galvanic corrosion.3,17

94

Elevated lead release owing to a galvanic lead-copper couple has been

95

demonstrated in laboratory and pilot studies and is a likely factor in

96

persistent high lead levels following partial LSLR.18-21 Mineralogical

97

evidence of galvanic corrosion in lead-copper and lead-brass joints

98

excavated from several distribution systems has also been reported.22

99

100

Limited residential data suggest that partial LSLR may be associated—at

101

best—with insignificant decreases in lead exposure risk. Partial LSLR did

102

not reduce the risk of elevated blood lead among children in Washington,

ACS Paragon Plus Environment

Page 4 of 44

Page 5 of 44

Environmental Science & Technology

103

DC and children living in homes with partial LSLs were more than three

104

times as likely to have elevated blood lead (≥ 10 µg dL-1) compared to those

105

in homes constructed without LSLs.23 Other residential studies—on 5 or

106

fewer sites—have reported lead spikes following partial LSLR and modest

107

or minimal subsequent reductions in lead release.3,24,25 Camara et al.26

108

reported greater lead release following partial LSLR relative to full LSLR

109

but did not provide detailed pre- and post-replacement comparisons.

110

111

In 2011, an advisory board to the U.S. EPA concluded that existing data

112

were inadequate to fully assess the effect of partial LSLR on drinking water

113

lead levels.27 The board found that the few studies available had

114

limitations—including small sample sizes and limited follow-up sampling—

115

but did conclude that partial LSLR could not be relied upon to reduce lead

116

levels in the short term.27

117

118

The objective of this work was to estimate changes in lead exposure due to

119

full and partial LSLR via a standardized profile sampling protocol. This

120

protocol was implemented within a water system where Pb(II) compounds—

121

in contrast to highly insoluble Pb(IV) oxides28—were presumed to dominate

ACS Paragon Plus Environment

Environmental Science & Technology

122

based on distributed water quality and consistent observation of peak lead

123

levels in samples representative of LSLs. Water distribution infrastructure

124

was considered typical of older North American municipalities. Water

125

system characteristics featured several risk factors for elevated lead release,

126

including 1) a significant number of unlined cast iron distribution mains,26 2)

127

distributed water with low alkalinity (20 mg L-1 as CaCO3) and pH (7.3),29 3)

128

a low orthophosphate residual (0.5 mg L-1 as PO43-)29 and 4) a chloride-to-

129

sulfate mass ratio above the critical threshold (0.5 – 0.77) identified in

130

previous work as a driver of galvanic corrosion.30,31 This study contributes to

131

literature by helping to address limitations identified in the EPA advisory

132

board report—small sample sizes and limited follow-up sampling.27 This

133

paper also expands on previous work26 with analysis of a much greater

134

volume of data: 74 and 61 full and partial LSL replacements—including

135

paired before-and-after comparisons of 18 partial replacements—and 13

136

additional sites with LSLs.

137

138

MATERIALS AND METHODS

139

Study area. The study area comprised single-unit residences in Halifax, NS

140

(CAN) that underwent LSLR between 2011 and 2015. In the event of partial

ACS Paragon Plus Environment

Page 6 of 44

Page 7 of 44

Environmental Science & Technology

141

LSLR, electrical continuity between lead and copper was expected but not

142

verified—following open-trench replacement, lead and copper were

143

typically joined via a brass union as described in Clark et al.32 Participating

144

residences were predominantly older homes—in areas of widespread pre-

145

1950 construction, thousands of LSLs are still in use.26

146

147

Sample sites received distributed water from a treatment facility employing

148

free chlorine disinfection—this facility is described in detail elsewhere.33

149

Table 1 lists 2013 – 14 typical values for key treated water quality

150

parameters—no relevant changes in treatment were made over the study

151

period.34 Beginning in 2002, a blended zinc ortho/polyphosphate corrosion

152

inhibitor (75% orthophosphate, 25% polyphosphate) was added at a treated

153

water residual of 0.5 mg L-1 (as PO43-).

154

155

Sample collection. Residents, with direction from utility staff, collected

156

profiles of four sequential 1L standing samples from kitchen cold-water taps,

157

beginning with the first draw following a minimum 6-hour standing period.

158

The 4 × 1L sample profile (L1-4) was followed by a 5-minute flush of the

159

outlet and subsequent collection of a fifth 1L sample (L5). Profile sampling

ACS Paragon Plus Environment

Environmental Science & Technology

160

was carried out before and at four follow-up rounds (3 days, 1 month, 3

161

months, and 6 months) after LSLR. Residents were instructed to record

162

exact stagnation times (median 7 hrs. 40 min., min. 6 hrs., max. 23 hrs.), and

163

to sample at a constant flow rate, but were not instructed to remove faucet

164

aerators prior to sample collection. Sampling instructions given to residents

165

are provided as Supplementary Information—data were excluded from

166

analysis when these instructions were not followed.

167

168

A complete series of pre- and post-replacement sample profiles was not

169

available for every residential site, owing to incomplete resident

170

participation. For this reason, sample sizes for before-and-after comparisons

171

differ by follow-up round. However, lead levels at a given follow-up round

172

were not significantly different (two-tailed rank-sum tests, α = 0.05) at sites

173

where residents participated at the next round compared to sites where they

174

did not—reporting lead levels to residents did not appear to influence

175

subsequent participation.

176

177

In order to assess the effect of water temperature on lead release, 13 × 1L

178

sample profiles were collected from kitchen cold-water taps at two other

ACS Paragon Plus Environment

Page 8 of 44

Page 9 of 44

Environmental Science & Technology

179

sites with LSLs (denoted sites A and B), following a 5-minute flush of the

180

outlet and subsequent 30-minute stagnation. The final litre of the 13 × 1L

181

profile was a free-flowing sample collected after a second 5-minute flush of

182

the outlet. Sample collection by the authors at these 2 sites, as opposed to

183

residents, necessitated use of an alternate sampling protocol, however

184

differences in collection methods limit the comparisons that can be made

185

between these two sites and the rest of the data. Samples were collected

186

weekly over a period of 7 or 8 weeks (sites A and B, respectively). Water

187

temperature was measured, using a glass thermometer, in the first and last

188

litre of each profile from the second week on.

189

190

Analytical methods. Total lead, copper, iron, and aluminum were measured

191

by ICP-MS (Thermofisher X Series II) according to Standard Methods 3125

192

and 3030.35 Reporting limits for Pb, Cu, Fe, and Al were 0.4, 0.7, 6.0, and

193

4.0 µg L-1, respectively. Lead was also quantified in 0.45 µm filtrate for a

194

subset of 386 sample profiles. Filtration via cellulose nitrate membrane

195

filters was generally performed within 2 days of sample collection (but prior

196

to acid-preservation), so results should be interpreted with care—changes in

197

speciation between collection and filtration cannot be ruled out. Loss of lead

ACS Paragon Plus Environment

Environmental Science & Technology

198

to sample bottles was estimated at 1.2% (std. dev. 17.3%) by comparing

199

total lead in 10 mL aliquots drawn from well-mixed 1L samples before and

200

24 hours after acid preservation of the entire sample (N = 82 5 × 1L sample

201

profiles).36 Losses due to filtration were estimated at 21.1% (std. dev. 0.4%)

202

by comparing lead in 10 mL aliquots filtered once with lead in 10 mL

203

aliquots filtered twice, using a new filter each time.36 Polyethylene (HDPE)

204

bottles and caps were immersed in ~ 2M reagent grade HNO3 for a

205

minimum of 24 hours and rinsed three times with ultrapure water prior to

206

use, and method blanks were prepared by holding ultrapure water preserved

207

with trace metal grade HNO3 in acid-washed bottles for 24 hours at 4˚C.

208

209

Data analysis. Lead levels at each of the four follow-up rounds after LSLR

210

were compared with pre-replacement lead levels. Sample profiles collected

211

after full LSLR (Table 2) were compared, using two-tailed rank-sum tests,37

212

with an independent set of 45 profiles collected at sites with full LSLs.

213

Differences between groups were multiplicative (i.e., best described as

214

ratios) but a natural log transformation yielded additive differences (i.e., best

215

described as constants). Changes in lead level were quantified using a

216

Hodges-Lehmann estimator, c, where c = median (yi / xj) for all i=1,...,n and

ACS Paragon Plus Environment

Page 10 of 44

Page 11 of 44

Environmental Science & Technology

217

j=1,...,m. Variables x and y denote lead levels observed at m and n sites with

218

full and fully replaced LSLs, respectively. The quantity c estimates the ratio

219

of lead levels between the two groups, where y = cx.38

220

221

Sample profiles collected before and after partial LSLR were paired by

222

address and compared using two-tailed signed rank tests by profile litre and

223

follow-up round.37 Natural log transformations were applied to the paired

224

data to achieve symmetry in the distribution of differences. Multiplicative

225

differences in before-and-after replacement lead levels were also quantified

226

using a Hodges-Lehmann estimator—details, though similar to those

227

provided above, may be found elsewhere.38 No control of the family-wise

228

error rate for multiple comparisons was employed.

229

230

RESULTS AND DISCUSSION

231

Pre-replacement lead levels. Lead and copper levels representing 5 × 1L

232

sample profiles collected at 45 sites with full LSLs are provided in Figure 1.

233

Peak copper levels occurred in L1 (90th percentile: 151 µg L-1) and peak lead

234

levels in L4 (90th percentile: 44 µg L-1). For single-unit residences, peak

ACS Paragon Plus Environment

Environmental Science & Technology

235

lead levels are often observed by L4 of the sample profile.39 The higher

236

median (11 µg L-1) and increased variability in L3 lead levels (90th

237

percentile: 29 µg L-1) suggest that L3 stagnated at least partially within the

238

LSL at some of the 45 sites. This is consistent with previous work—Cartier

239

et al.40 estimated median (mean) premise plumbing volumes in 88 pre-1970

240

homes at 2.0 L (2.3 L). At sites with full LSLs, L1 significantly

241

underestimated peak lead levels—in systems where lead (II) compounds

242

form preferentially, lead in water that contacted the LSL during stagnation

243

may be considerably higher than lead in the first-draw sample.26,28,41

244

245

Longer sample profiles (13 × 1L) collected at two residential sites with LSLs

246

(sites A and B, Figure 2) provide insight into the ability of the 5 × 1L profile

247

to estimate lead exposure. Plumbing configuration can be inferred by

248

comparing lead and copper levels over each profile, although mixing among

249

profile litres and the 5-minute flush prior to stagnation (sites A and B only)

250

may have influenced apparent plumbing volumes. Site A is a typical

251

example of the single-unit residences that underwent LSLR (the large

252

apparent premise plumbing volume of 8-9 L at site B is explained by sample

253

collection from a second-level kitchen). The apparent premise plumbing

ACS Paragon Plus Environment

Page 12 of 44

Page 13 of 44

Environmental Science & Technology

254

volume at site A was 2-3 L—copper declined sharply after the first 2 L and

255

quickly approached the level of the flushed sample (L13, 10 µg L-1). Peak

256

lead levels were observed by L4, although L3-9, or parts thereof, appear to

257

have stagnated within the LSL. In light of LSL lead release observed at sites

258

A and B following 30 minutes of stagnation (max. 28 and 14 µg L-1,

259

respectively), apparent LSL lead release at the 45 sites represented in Figure

260

1 was lower than expected (min. 6-hour stagnation, 90th percentile of 44 µg

261

L-1 in L4). This suggests that L4 may sometimes have fallen short of the

262

LSL. Furthermore, while L4 does appear to have reached the LSL in at least

263

some cases, it may not have reached the lead-copper junction at sites with

264

partial LSLs, as sample profiles reported in previous work have shown.42

265

Thus, 4 × 1L standing sample profiles may only provide an indirect

266

assessment of the effect of galvanic corrosion following partial LSLR.

267

268

Influence of water temperature. Pre-replacement profiles were collected in

269

summer—collection dates at the initial round had a July median. This

270

introduces water temperature as a potential confounding variable. One-

271

month follow-up collection dates had a September median, and on this

272

interval the variation in water temperature is likely to have been low—

ACS Paragon Plus Environment

Environmental Science & Technology

273

typical 5-minute flushed sample temperature was 17.9 ˚C in July (this study)

274

and 17.8 ˚C in October.43 Comparisons between full and partial LSLR at the

275

same follow-up round are also unlikely to have been strongly influenced by

276

temperature. However, partial LSL profile collection dates at the 3 and 6-

277

month follow-up rounds had December and April medians, respectively.

278

Pre- and post-replacement comparisons on these intervals are subject to

279

decreases in water temperature of approximately 10 ˚C—typical 5-minute

280

flushed sample temperature in February was 7.2 ˚C.43

281

282

The effect of water temperature on lead release is complex—the positive

283

effect of temperature on the rate of electrochemical reactions may be

284

counterbalanced by the reduced solubility of lead minerals at higher

285

temperatures, depending on the composition of corrosion scale.44

286

Temperature effects also depend on whether the source of lead is premise

287

plumbing—where seasonal variation is expected to be minimal—or LSLs—

288

where variation is greater.17 Previous work has shown that lead release can

289

be temperature-dependent,40,45 and data from 13 × 1L profiles collected at

290

sites A and B show that lead release was moderately-to-highly correlated

291

with water temperature (R2 = 0.46 – 0.98, average of 0.79). Between 13 and

ACS Paragon Plus Environment

Page 14 of 44

Page 15 of 44

Environmental Science & Technology

292

19 ˚C, a 1 ˚C increase in 5-min. flushed sample temperature accompanied an

293

average 1.1 µg L-1 increase in lead release from LSLs (Figure S1), although

294

other seasonally varying water quality parameters, such as free chlorine

295

residual, could have influenced observed lead levels.

296

297

Water temperature was correlated with lead release from premise plumbing

298

at sites A and B as well—likely a consequence of the short (30-min.)

299

stagnation time. More generally, lead release from premise plumbing does

300

not often exhibit strong temperature dependence, provided that stagnation

301

time is sufficient for standing water to reach building temperatures (e.g., 6

302

hours).17 In a survey of 365 U.S. drinking water utilities, 90th percentile first-

303

draw lead levels—collected following a minimum 6-hour stagnation—were

304

not a function of season.46 Within the present study area, first-draw lead

305

levels (min. 6-hr. stagnation, 34 residential sites) were no higher in October

306

than in February.43

307

308

Owing to the effect of water temperature on lead release from LSLs, long-

309

term (3 and 6 month) pre- and post-replacement comparisons were only

310

interpreted for the portion of the sample profile least likely to have been

ACS Paragon Plus Environment

Environmental Science & Technology

311

influenced by temperature—L1 and L2. It is possible that, for sites with very

312

small premise plumbing volumes, L2 lead levels were affected by

313

temperature variation, but no significant drop in L2 lead levels from summer

314

to winter was observed following partial LSLR (Table 2). For L3 and L4,

315

however, decreasing water temperature likely contributed to— and may have

316

been entirely responsible for—observed reductions in lead release.

317

318

Effect of full LSL replacement. For each profile litre and follow-up round,

319

the fraction of pre-replacement lead remaining after LSLR was estimated

320

(Table 2)—full LSLR reduced lead levels in every litre of the sample profile

321

within 1 month. Before-and-after differences were multiplicative (not

322

additive), meaning that sites with high lead levels pre-replacement tended to

323

see greater reductions in lead post-replacement. For a given profile litre, a

324

ratio less than one signifies a reduction in lead over pre-replacement levels

325

and a ratio greater than one signifies an increase.

326

327

At sites that underwent full LSLR, public and private LSL sections were not

328

often replaced on the same day, and a delay of several months was not

329

uncommon. In many cases, the pre-replacement profile represented the

ACS Paragon Plus Environment

Page 16 of 44

Page 17 of 44

Environmental Science & Technology

330

partial LSL configuration. In order to properly evaluate the effect of full

331

LSLR, lead levels after full replacement were compared, without pairing,

332

against 45 sites with full LSLs—41 of these 45 underwent partial LSLR

333

only. Unpaired comparisons have the advantage of larger sample sizes, but

334

they are expected to be less accurate when other sources of lead are present

335

(e.g., in L1 and L2). A key benefit of pairwise comparisons is that they tend

336

to account for variation due to uncontrolled factors (e.g., other sources of

337

lead, variations in flow rate).47

338

339

Reductions in lead following full LSLR were immediate—lead levels in L3-

340

5 were less than half of their pre-replacement counterparts at the 3-day

341

follow-up (Table 2). One month after full replacement, lead release from

342

premise plumbing (L1-2) had dropped significantly as well. Distribution

343

quantiles, representing before-and-after full LSLR comparisons at the 1-

344

month follow-up, are provided in Figure 3. Quantiles adhered well to the

345

ratio estimates listed in Table 2, except in the upper extremes, where outliers

346

occasionally deviated. One month-post replacement, 90th percentile lead

347

levels ranged from 2 to 12 µg L-1 (L5 and L1, respectively), while pre-

348

replacement 90th percentiles ranged from 10 to 44 µg L-1 (L5 and L4,

ACS Paragon Plus Environment

Environmental Science & Technology

349

respectively). Reductions in lead release from premise plumbing (L1-2) may

350

be attributed to gradual flushing of lead that had accumulated pre-

351

replacement, as previous work has suggested.3 Since premise plumbing

352

upgrades were not performed in conjunction with LSLR, leaded solder and

353

brass were not expected to have contributed to changes in lead release post-

354

replacement. Accumulation of lead in premise plumbing may have been

355

driven in part by adsorption to iron deposits or surfaces—as described

356

elsewhere, colloidal lead (< 0.45 µm) was strongly associated with colloidal

357

iron at residential sites within the same water system.36 Previous work has

358

accounted for a correlation between iron and lead at the point of use with

359

reference to the strong tendency for lead to adsorb to iron oxide deposits or

360

galvanized iron plumbing.42,48,49 Manganese deposits in premise plumbing

361

have also been implicated as a sink for—and subsequent source of—lead in

362

drinking water.50

363

364

Effect of partial LSL replacement. Partial LSLR more than doubled

365

premise plumbing (L1-2) lead levels at the 3-day follow-up (Table 2). One

366

month post-replacement, L1 was still elevated by more than 60%, while

367

subsequent standing sample lead levels (L2-4) had not changed significantly

ACS Paragon Plus Environment

Page 18 of 44

Page 19 of 44

Environmental Science & Technology

368

relative to their pre-replacement counterparts. Even 6 months after partial

369

LSLR, no significant reductions in L1-2 lead levels were observed.

370

Reductions in L3-4 lead release at the 3 and 6-month follow-up rounds were

371

expected to have been enhanced by—and could have been entirely due to—

372

decreasing water temperature. (Increases, relative to pre- replacement, in

373

LSL lead release at 3 and 6 months were sometimes observed despite

374

temperature differences—see Figure S2.) Applying a 1.1 µg L-1 ˚C-1

375

correction (described in the Supplementary Information), based on expected

376

water temperature differences, eliminated statistically significant reductions

377

in L3-4 lead release that would otherwise have been attributed to partial

378

LSLR.

379

380

In contrast to standing samples, L5 lead levels were not significantly

381

different from their pre-replacement counterparts at the 3-day follow-up

382

(90th percentile of 14 µg L-1) and were significantly lower at the 1-month

383

follow-up (90th percentile of 6 µg L-1). However, data from this study do not

384

support 5 minutes of flushing as a strategy for protecting against the short-

385

term effects of partial LSLR— 9% of L5 samples were greater than 15 µg L-

386

1

at the 3-day follow-up compared with just 4% pre-replacement. Moreover,

ACS Paragon Plus Environment

Environmental Science & Technology

387

an L5 sample collected 3 days post-replacement measured 230 µg L-1.

388

389

Changes in lead release due to partial LSLR are illustrated in Figure 4

390

(bottom), which displays the distribution of pairwise differences in lead

391

release (after – before), grouped by profile litre for the first two follow-up

392

rounds (3 days and 1 month). Positive differences correspond to an increase

393

in lead release post-replacement, and negative differences correspond to a

394

decrease (sample sizes are provided in Table 2). Increased lead release

395

following partial LSLR is evident, especially at the 3-day follow-up—on this

396

interval, more than a quarter of sites saw increases of 20 µg L-1 in at least

397

one standing sample (L1-4). On the 1-month interval, L3 and/or L4 lead

398

increased by 10 µg L-1 at more than a quarter of sites.

399

400

Increased lead release to L1 and L2 can likely be attributed to accumulation

401

of particulate lead in premise plumbing following replacement-induced

402

destabilization of LSL corrosion scale.3 Galvanic corrosion at the lead-

403

copper junction has been linked with elevated particulate lead release as

404

well.18-21 Occurrence of particulate lead (> 0.45 µm) was more frequent

405

following partial (relative to full) LSLR (Figure 5). At the 3-day follow-up,

ACS Paragon Plus Environment

Page 20 of 44

Page 21 of 44

Environmental Science & Technology

406

11 and 26% of samples collected following full and partial LSLR,

407

respectively, had more than 10 µg L-1 of particulate lead (compared to 3%

408

pre-replacement). Elevated lead release in general was dominated by

409

particles—at higher total lead levels the particulate fraction was larger,

410

approaching unity at lead levels higher than approximately 100 µg L-1

411

(Figure 5). Available data suggest that lead in 0.45 µm filtrate was

412

dominated by colloidal particles (0.05 – 0.45 µm).36

413

414

Post-replacement lead exposure. Serious spikes in lead sometimes

415

followed LSLR and were much more frequent following partial LSLR—

416

elevated lead levels after partial replacement have been reported in previous

417

work as well.3,24,25,51 Trends in post-replacement lead release diverged

418

immediately according to replacement type. At sites with full LSLs (pre-

419

replacement), 29% of standing sample (L1-4) lead levels were greater than

420

15 µg L-1 (45 sites). Three days after partial LSLR, 45% of L1-4 lead levels

421

were greater than 15 µg L-1 (34 sites), while 3 days after full LSLR, just 14%

422

were greater (48 sites). Lead levels exceeding 15 µg L-1, by follow-up round

423

and LSL configuration, are displayed in Figure 4 (top). This threshold

424

represents a concentration above which the Centers for Disease Control and

ACS Paragon Plus Environment

Environmental Science & Technology

425

Prevention considers drinking water unsuitable for consumption by children

426

and pregnant women.52 At the 3-day follow-up, 3 observations also exceeded

427

the U.S. Consumer Product Safety Commission’s acute exposure level for

428

children (700 µg L-1, based on a 250 mL intake)53—all at sites with partial

429

LSLs. These extreme lead levels, including a sample with 10,340 µg L-1,

430

were associated with premise plumbing (L1 or L2). High-velocity, multiple

431

outlet flushing post-replacement is a possible strategy for protecting against

432

these short-term spikes in lead.51

433 434

Unusually high lead levels were also observed at several sites following full

435

LSLR (Figure 4, top). The highest observations at 1 and 6 months were both

436

first-draw samples, and the highest four observations at 3-months represent

437

standing samples (L1-4) collected at a single site. Lead in these 4 samples

438

was more than 90% particulate (> 0.45 µm). These samples were also

439

unusually rich in particulate iron, copper, and aluminum, suggesting that the

440

source was corrosion scale within premise plumbing that had accumulated

441

multiple contaminants over time. In the case of full LSLR, public and

442

private LSL replacements were not often performed simultaneously, and

443

disturbances associated with two replacements—as well as possible galvanic

444

corrosion in the interim—may have contributed to elevated lead release and

ACS Paragon Plus Environment

Page 22 of 44

Page 23 of 44

Environmental Science & Technology

445

accumulation of particulate lead within premise plumbing. While full LSLR

446

was associated with substantial reductions in lead levels, staggered

447

replacements may have caused the true benefit of full replacement to be

448

underestimated.

449 450

In the long term, elevated lead was observed more often at sites with partial

451

LSLs than at sites with either full LSLs or with full copper service lines

452

(post-LSLR). At the 6-month follow-up after partial LSLR, 22% of premise

453

plumbing (L1-2) lead levels—and 30% of service line lead levels (L3 and

454

L4)—were greater than 15 µg L-1 (30 sites). At sites with copper service

455

lines, 7% of L1-2 samples—and none of the L3 or L4 samples—were

456

greater than 15 µg L-1 (45 sites). The frequency of high (> 15 µg L-1) L1-2

457

lead levels was substantially greater— even 6 months post-replacement—at

458

sites with partial LSLs relative to sites with full LSLs (22 vs. 16%

459

respectively). Moreover, the fraction of first-draw samples greater than 15

460

µg L-1 at 6 months was double the pre- replacement fraction (27% vs. 13%).

461

Despite the possibility that the sample profile did not reach the lead-copper

462

junction—and that the flow regime did not represent a worst-case scenario—

463

these data captured the greater tendency, identified in previous work,19,20 for

464

elevated lead at sites with partial LSLs compared to sites with full LSLs.

ACS Paragon Plus Environment

Environmental Science & Technology

465

466

Implications for controlling drinking water lead. This study used a

467

standardized profile sampling protocol to assess the effect of full and partial

468

LSLR on lead release to drinking water. The strength of this work was the

469

comparatively large volume of data collected pre- and post-replacement, and

470

the principal limitations were the inevitable uncertainties associated with

471

sample collection by residents and the lack of information on plumbing

472

volumes and configurations that limits a mechanistic understanding of the

473

results.

474

475

Full LSLR reduced service line (L3-4) and 5-minute flushed sample (L5)

476

lead levels within 3 days. At 1 month, full replacement had caused lead level

477

reductions in every litre of the sample profile. Partial LSLR, on the other

478

hand, caused substantial short-term increases in premise plumbing (L1-2)

479

lead levels and did not significantly reduce L1-2 lead levels within 6 months.

480

Furthermore, first-draw lead levels were greater than 15 µg L-1 at a

481

considerably higher frequency than at sites with full LSLs, even 6 months

482

post-replacement. This finding could have important implications in

483

jurisdictions where drinking water lead is regulated based on the first-draw

ACS Paragon Plus Environment

Page 24 of 44

Page 25 of 44

484

Environmental Science & Technology

sample.

485

486

This study generated a considerable volume of data corroborating previous

487

work that showed 1) full LSLR—in addition to removing the primary source

488

of lead—is effective for reducing lead release from premise plumbing, 2)

489

partial LSLR dramatically increases lead at the point of use in the short term,

490

3) partial LSLR may be worse than leaving the LSL intact due to the

491

potential for elevated lead release in the long term, 4) in Pb(II)-dominated

492

water systems, first-draw lead levels are likely to underestimate lead

493

exposure at residences with LSLs, and 5) lead release from LSLs is

494

sometimes strongly influenced by water temperature.

495

496

The short-term elevated lead levels that sometimes followed partial LSLR

497

are a serious concern—some were high enough to pose acute health risks.

498

High-velocity flushing of outlets,51 use of pipe cutting methods that

499

minimize disturbances to LSLs,3 and point-of-use lead removal54 are

500

potential strategies that could be implemented to reduce health risks

501

associated with this mode of exposure. In assessing the effects of full and

502

partial LSLR, the potential for dramatically elevated lead levels post-

ACS Paragon Plus Environment

Environmental Science & Technology

503

replacement is an important consideration. While the rapid reductions in lead

504

that typically follow full LSLR outweigh the risk of short-term disturbance-

505

induced spikes, the modest long-term benefits from partial LSLR described

506

in some previous work24,25 may be overshadowed by the greater risk of

507

elevated lead in both the short and long term.

508

509

SUPPORTING INFORMATION

510

Supplementary figures, a description of temperature correction methods, a

511

transcription of the sampling instructions given to residents. This material is

512

available free of charge via the Internet at http://pubs.acs.org.

513

514

ACKNOWLEDGEMENTS

515

This work was funded by the Natural Sciences and Engineering Research

516

Council (NSERC)/Halifax Water Industrial Research Chair in Water Quality

517

and Treatment (Grant No. IRCPJ 349838-11). The authors also acknowledge

518

the technical support of Heather Daurie, Elliott Wright, Krysta Montreuil,

519

Alisha Knowles, Everett Sneider, and Halifax Water staff.

520

ACS Paragon Plus Environment

Page 26 of 44

Page 27 of 44

Environmental Science & Technology

521 522

TABLES

523 524

Table 1. Typical values for treated water quality parameters, pre-distribution. Parameter Zinc ortho/polyphosphate (as PO43-) Alkalinity (as CaCO3) Free chlorine Hardness (as CaCO3) Total organic carbon pH Chloride Sulfate Turbidity Iron Lead

525 526 527 528 529

Table 2. Estimated fraction of pre-replacement lead levels remaining after full and partial LSLR.

Partial LSLR (paired) Full LSLR (unpaired)c 530 531 532 533

Typical value 0.5 mg L-1 20.0 mg L-1 1.2 mg L-1 12.0 mg L-1 1.5 mg L-1 7.3 9 mg L-1 8.5 mg L-1 0.06 NTU 0.45 µm (µg L−1 )

(10340,10288) (902,876) (749,723) (475,464) (441,426) (324,283) (217,211) (147,140) (143,74) (473,459) ● (340,328) (162,150)

● Pre−replacement ● Partial LSLR

100

Full LSLR



75 ● ●

y=x

50 ●

25

0

Page 38 of 44

●●

● ● ●● ● ● ● ●● ●● ● ● ● ●● ● ● ● ● ●●●● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●●● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●●● ●● ● ● ● ● ● ●● ● ●● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

0

25

50



75



● ● ●

100 −1

Total lead (µg L

y = 0.21 x (~100% of Pb < 0.45 µm)

125

150

)

Figure 5. Particulate (> 0.45 µm) lead release as a function of total lead, pre-replacement and at the first two follow-up rounds (72 hrs. and 1 mo. post-replacement). Points beyond the plot limits are represented as (x,y) coordinates and the line y = 0.21 x represents the estimated loss due to 0.45 µm filtration (~100% of lead < 0.45 µm).

ACS Paragon Plus Environment

6 mo. post−LSLR (µg L−1 )

● & Technology 200 Environmental Page 39 of 44 Science ●

y=x

150 100 50 0

● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ACS Paragon ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0

50



Plus 100

Quantiles

● Full LSLR ● Partial LSLR Environment 150

Pre−LSLR (µg L−1 )

200

180

Environmental Science & Technology Page 40 of 44

Total lead (µg L−1 )

40

160

Copper Lead

35

140

30

120

25

100

20

80

15

60

10

40

5

20

ACS Paragon Plus Environment 0

0 L1

L2

L3

Profile litre

L4

L5

Total copper (µg L−1 )

45

25

100

20

80

15

60

10

40

5

20

0

0

Copper Lead

25

100

20

80

15

60

10

40

5

20

0

ACS Paragon Plus Environment 1

2

3

4

5

6

7

8

9

Profile litre

Total copper (µg L−1 )

Total lead (µg L−1 )

PageEnvironmental 41 of 44 Science & Technology

10

11

12

13

0

1 mo. post−replacement (µg L−1 )

80Environmental Profile litre ● ● ●

60

● ●

Science & Technology Page 42 of 44

L1 L2 L3 L4 L5

y=x ●

0.65 ●

40

● ●

0.43 ●

20

(103, 32)

● ●





● ● ●

0

●● ●● ● ●●● ● ● ●● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ●● ●● ● ● ●● ● ●● ● ●●● ● ● ● ●● ● ● ● ● ● ● ● ●●● ● ● ● ●●● ● ● ● ● ●● ● ● ●●● ●●●● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●

0.20 0.13 0.13

ACS Paragon Plus Environment

0

20

40

60

−1

Pre−replacement (µg L

80

)

−1

lead (µg L ) PageEnvironmental 43 of 44 Total Science & Technology 10

100

1000

10000

Before 3 d. 1 mo.

Observations > 15 µg L−1

3 mo.

Before Full LSLR Partial LSLR

6 mo. ● ●

Increase in total lead (µg L−1 )

80

60

85 246 312 10,330

88

676

238

227





Sample round ●

40



3 d. 1 mo.



20



0 ● ●

−20



ACS Paragon Plus Environment L1

L2

L3

L4

Profile litre

L5

Environmental Science & Technology Page (10340,10288) 44(902,876) of 44

● Pre−replacement ● Partial LSLR

Lead > 0.45 µm (µg L−1 )

100

(749,723) (475,464) (441,426) (324,283) (217,211) (147,140) (143,74) (473,459) ● (340,328) (162,150)

Full LSLR



75 ● ●

y=x

50 ●

25

0



●●

● ● ●● ● ● ● ●● ●● ● ● ● ●● ● ● ● ● ●●●● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●●● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●●● ●● ● ● ● ● ● ●● ● ●● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●



● ● ●

y = 0.21 x (~100% of Pb < 0.45 µm)

ACS Paragon Plus Environment

0

25

50

75

100 −1

Total lead (µg L

125

)

150