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