United States
Environmental Protection
Agency
Water Engineering
Research Laboratory
Cincinnati OH 45268
Research and Development
EPA/600/S2-87/042  Nov. 1987
 Project Summary
 Corrosion  in Water  Distribution
 Systems  of  the  Pacific
 Northwest
Steve H. Reiber, John F. Ferguson, Mark M. Benjamin, and
Dimitris Spyridakis
  Continuous linear polarization mea-
surements (10-min  increments) of
corrosion rates  on  copper surfaces
demonstrated an immediate response
to changing water quality conditions
found in Pacific Northwest distribution
systems. Significant relationships were
found between copper corrosion rates,
pH, and free chlorine residual in these
low-mineral waters. Regression analy-
sis of copper corrosion rates versus four
common chemical quality parameters
was used to develop a statistical model
of the relationship of copper corrosion
to changing water quality conditions.
  This study evaluated the impact of
changing water quality conditions on
copper  surfaces  coated with varying
depths of oxide films. The presence of
a well aged film (greater than 50 /urn
depth) generally  decreased  the corro-
sion rate by more than 50 percent when
compared with corrosion on surfaces
coated with a relatively fresh film layer
(less than 20 /urn depth).
  The Seattle Water Department's
program of corrosion control through
lime and soda ash addition was eval-
uated during its 12-month implemen-
tation period. A gradual pH increase of
approximately 2.0 units during this
period decreased the corrosion rates of
aged copper surfaces by approximately
9 /urn/year.
  Selected chemical water quality
parameters were measured in the
distribution  networks  of Seattle,
Tacoma, and  Anacortes. Significant
differences in quality variation were
found as a function of distribution level.
The higher the distribution level (i.e.,
the closer  to the consumer tap), the
greater the degree of variation. More
elaborate treatment processes (rapid
sand filtration preceded by coagulation
and  flocculation) induced  a greater
degree of variation in the distribution
system than systems that did not filter
their water.
  Particle counts and turbidity levels in
the distribution  system were only
partially  correlated. The  strongest
turbidity  correlation  was  found for
particles of less than 10-//m diameter,
which also constituted the great major-
ity  of particles  in all  the waters
measured.
  This Project Summary was devel-
oped by  EPA's Water Engineering
Research Laboratory, Cincinnati, OH,
to announce  key findings of the
research project  that  is fully docu-
mented in a separate report of the same
title (see Project Report ordering
information at back).

Introduction
  The Pacific Northwest has some excep-
tionally pristine surface water sources of
unusually low mineral content and high
clarity. These sources are extremely low
in alkalinity, pH, and buffer intensity, and
they are undersaturated with respect to
CaCOs- The potential for deposition of
protective carbonate films is minimal,
and excessive metal corrosion on plumb-
ing appurtenances is a common problem.
  The general objective of this study was
to  demonstrate  the effectiveness and
capabilities of the U.S. Environmental
Protection Agency mobile water quality
monitoring laboratory (MWQML) in

-------
monitoring the variation of water quality
parameters and the impact of chemical
dosing  on  corrosion rates in  water
systems of the Pacific Northwest. Other
activities included  evalution  of  the
Seattle  Water Department's corrosion
control program, determination of how
instantaneous copper corrosion rates are
affected by  short-term variation in
selected water quality parameters,  and
an  assessment  of  the relationship
between distribution system water qual-
ity  and distribution  level,  raw  water
source, and treatment level.
  The communities of Seattle, Tacoma,
and Anacortes participated in this Study.
Selection was based on raw water source
(which for all cases was surface water
originating on the western slopes of the
Cascade Mountains), the type  of treat-
ment provided,  and  the history of
corrosion-related  problems within the
system. Table 1 summarizes some basic
information about the sampling program
distribution systems


Procedures
  The  MWQML  is  a self-contained,
mobile,  microprocessor-controlled labo-
ratory capable of continuously sampling,
analyzing, and recording 20 water quality
and corrosion parameters. Specific
capabilities of the MWQML are directed
at the automated monitoring of common
chemical quality parameters in the
distribution  system,  including  pH,
temperature, conductivity, chlorine
residual, turbidity, alkalinity, hardness,
chloride, and others.  Specialized instru-
mentation includes  a  three-channel
Petrolite* linear-polarization corrosion-
rate monitor, an  HIAC/Royko particle
counter, and an  loinics digital titrator.
Control and  data  acquisition functions
are provided  by an HP 9845B micropro-
cessor. Data storage is on floppy disk.
  Standard procedure was to locate the
MWQML at selected  sampling sites
within  the  distribution  systems  for
periods of 2  to 4  weeks. This time was
necessary  to develop a sufficient  data
base of diurnal   chemical  variations.
Sampling sites  included  endline  user
taps, raw water sources,  reservoir
outlets, water blending points, and other
points of interest within the  systems.
Data collected and stored by the MWQML
were periodically uploaded by means of
modem connection to the Cyber 1500
Table 1.    Information Summary on Sampling Program Distribution Systems
Service
Community
Seattle
(North)

Tacoma
Anacortes
Raw Water
Pop.
500,000+

70,000+
20,000+
Source
So. Fork Toll River

Green River
Skagit River
Treatment
Screening and Cli ana
F-corrosion control
initiated 6/82
C/2 only
C/z, coagulation, floe-
                                                           culation,  and rapid
                                                           sand filtration
 "Mention of trade names or commercial products
  does not constitute endorsement or recommenda-
  tion for use
system at the University of Washington
Academic Computing  Center.  Basic
statistical analysis and  data  reduction
were performed using the 1984 edition
of "Minitab" from Pennsylvania  State
University.
  Corrosion rate  measurements  were
made with  a  three-channel Petrolite
model  1000 linear polarization instru-
ment.  Two different sets  of  copper
electrodes were  used to approximate
copper plumbing with different degrees
of oxide  coating  and  age.  The  fresh
electrode set was relatively clean and
covered with a cuprous oxide layer of less
than 20 /urn depth. The aged electrode
set was coated with a heavy oxide layer
exceeding 50 /urn.
Results and Discussion

The Seattle Corrosion Control
Program
  Seattle bega n its program of corrosion
control  in  June  1982. The  program
called for the  gradual  increase  of
distribution water  pH  and alkalinity
through  the addition of lime and soda
ash. Target dosages for the two chem-
icals were 2.0 and 9.0 mg/L, respec-
tively,  which resulted in an  average
overall alkalinity increase of 5.0 to 17.0
mg/L (as CaCOs). The adjustment period
extended over 12 months. During this
period,  copper  corrosion rates were
measured on aged surfaces at selected
sites throughout the  northern Seattle
distribution area. Figure 1 summarizes
weekly  median pH  values and aged
copper  surface corrosion rates during
the adjustment period. Overall, the  pH
was shifted from a median distribution
 level of 5.8 at the beginning of the
 program to a median level of  7.7 after
 12 months of  operation, producing a-
 reciprocal  decrease  in corrosion rates
 from a weekly average of 20 to 12/um/'
year, a decrease of approximately 40
percent.
  Figure  2 presents a graph of weekly
median rates for aged copper corrosion
as a function of median pH. The inverse
relationship  between pH and the distri-
bution copper corrosion rate is clear.

Water Quality Variation and
Copper Corrosion
  Monitoring for short-term  water
quality variations and their  impact on
copper corrosion  rates was  conducted
at several points in each of the sampled
distribution systems. Figure 3 presents
a 5-day summary of water quality and
copper corrosion rate  variations at one
end-line  tap in the  northern Seattle
distribution  area. The strong  positive
correlation of both the aged and fresh
electrode corrosion rates with the free
chlorine residual is apparent. Also clear
is a pronounced inverse relationship of
the corrosion rates with the pH changes,
including the larger pH spikes. Influence
of temperature and conductivity are not
readily apparent.
   Multiple linear regression  on lumped
data sets was performed to evaluate the
relative importance of the four  water
quality parameters on both the fresh and
aged copper surfaces. A forward,  step-
wise regression  was  used with four
independent parameters regressed in
the  following order:  (1) pH,  (2) free
chlorine, (3) temperature, (4) conductiv-
ity. Table 2  presents the linear regres-
sion model developed for corrosion rates
on the aged and fresh surfaces. Table
3 presents  the  analysis of variance
(ANOVA) for the two regression models.
From this table, it is apparent that the
regression models account for 95 and
76 percent, respectively, of the overall
corrosion rate  variance on the two
surfaces, and that in both cases, the free
chlorine residual is the most significant
predictor.

-------
   w

    9


    8

   ' 7

    6

    5
        ,00
            Ooooo
°  °
                                                  °°00000
                                                                                     pH (Weekly Median)

                                                                                          oo0oo0oo
   30
            0  0000
                                               n
                                 °000000    00
                                                                                  Copper Corrosion Rates
                                                                       °°0o°°o  o  0°°°
                                                                       40
                                                                                        SO
      0               10               20              30

       June 82                               Weekly Progression

Figure 1.   A verage weekly pH and copper corrosion rates in the Tolt distribution area during the first year of the control program.
                                                                                    60
   10 T
    9-
    8-
    7-
                                 8°8g
                                            8
                                                  Joo
                                                              0  0
                                                                          Corrosion rates measured at
                                                                          selected points throughout
                                                                           the distribution system.
      5                      W                       15

                                             Copper Corrosion Rates um/yr

Figure 2.   Copper corrosion rates in the Seattle System as a function of weekly median pH.
                                                                              20
                                                                                   25

-------
  30
  25 <
   20 <
                                                                                                Fresh Cu Electrode
                                                                                                Aged Cu Electrode
    8 *
    7 ,
   18


   17

   16

   15
Temperature
   031
 -J
 ^02
   0.1'
    0
                                                                                                 Total Chlorine
                              24
                                                      48
       0
    9/6/83                                 Experimental Progression (hours)

Figure 3.     Variation of water quality and corrosion rate with time (data set S906).

                                      4
                                                                              72
    96
                                                                                                                              120

-------
 Table 2.
Parameter
Linear Regression Models of Copper Corrosion on Aged and Fresh Surfaces

                                              Coefficient
                             Units
                                   AgedSurface
Fresh Surf ace
Corrosion rate
Y intercept
pH
Free chlorine residual
Temperature
Conductivity
lim/year
—
pH units
mg/L
°C
umhos/cm

0.553
-0.078
0.810
0.026
0.005

1.820
-0.091
1.013
0.013
0.0009
 Table 3.
Analysis of Variance for Aged and Fresh Copper Surface Corrosion Regression
Models
Item
Total
Residual
Regression
pH
Free Chlorine
Temperature
Conductivity
Degrees of
Freedom
350
346
4
1
J
r
1
Sequential Sum
of Squares
Aged
2.357
0.118
2.238
0.354
1.491
0.385
0.008
Fresh
1.543
0.373
1.170
0.001
1.055
0.076
0.036
Coefficient of
Determination R2
Aged


0.950
0.15
0632
0.163
0.003
Fresh


0.758
0.001
0.684
O.050
0.023
Quality Variation in the
Distribution Network

  In each of the sampled communities,
comparisons were made  of  specific
chemical parameters at different levels
of the distribution system to evaluate the
impact of residence time and system
contact on parameter variability. One
example appears in Figure 4 where free
chlorine  residual  at the point of dis-
charge from  the Anacortes treatment
plant is  compared with the residual
found at  an end-line tap  at the extreme
end of the distribution system. Water
discharged from  the plant carries  a
residual that varies erratically from 0.4
to 0.6 mg/L. The unevenness of the
residual is apparently due to minor flow
and dosage variations and incomplete
mixing  in the clear well. Chlorine
residuals at the consumer  tap  are not
erratic, but they show a smooth and
pronounced diurnal variation. The high-
est  concentration corresponds with the
period of peak demand, indicating that
retention  time  in  the system affects
residual  levels. Presumably,  longer
contact times provide greater opportun-
ity for chlorine depletion through reac-
tion with  plumbing materials and dis-
solved chlorine-consuming substances.
Similar variation and  periodicity were
observed in other chemical and corro-
sion parameters in all of the distribution
systems examined.
                              Turbidity Versus Particle
                              Concentration
                               For the northern Seattle distribution
                              system,  a comparison  was  made of
                              turbidity  values  measured   by  a
                              continuous-flow-ratio turbidimeter with
                              particle concentrations measured by an
                              HIAC/Royko counter. Figure 5 presents
                              4  days  of  concurrent  turbidity  and
                              particle  data,  with the  latter  broken
                              down into three different size ranges.
                              A  moderately strong correlation exists
                              between the  smallest  particle size
                              concentration and the turbidity data (R
                              = .687). However, the correlation dimin-
                              ishes dramatically for the larger particle
                              concentrations. These and other results
                              indicate that the smaller particle sizes
                              (less than 2 fim diameter), which numer-
                              ically make up the  great  majority of
                              particles in all the distribution systems
                              examined,  also contribute the  most to
                              turbidity variation.
                               Regression analysis of turbidity data
                              using the three particle-size-range
                              concentrations  as the  independent
                              predictors  showed that the particle
                              concentrations accounted for only 50
                              percent  of  the observed turbidity vari-
                              ation. Filtered particle  samples were
                              examined by scanning electron micros-
                              copy, which revealed that the bulk of
                             the smaller particles consisted  of sili-
                             ceous diatom  skeletons and mineral
                             agglomerates.  Evidence  indicates that
                             the mineral agglomerates may be cor-
 rosion  products generated on exposed
 steel surfaces in the distribution system.
 Most of the observed diatom forms were
 common to clean surface water sources
 and  may result from  algae growth in
 impoundments  and on open  reservoir
 embankments.

 Conclusions
   This  study has explored the relation-
 ship  between  water quality  variation
 distribution system location and copper
 corrosion rates. Some of the  principal
 conclusions drawn from the study  are
 as follows:

  1.   Seattle's  program of corrosion
      control through  moderate  pH
      adjustment and alkalinity addition
      has effectively reduced corrosion of
      copper plumbing  materials  by
      approximately 40 percent.

  2.   Free chlorine residual is the most
      important of the  common water
      quality parameters  for predicting
      copper corrosion  rates.  In  some
      cases over half of the variation in
     corrosion rate can be attributed to
     variation in  the  free  chlorine
      residual.

  3.  Variable distribution system resi-
     dence time as a result of demand
      cycling can  induce  substantial
     variation  in  pH, turbidity, and
      chlorine residual  of the delivered
     water.  Some treatment processes
     (filtration,  coagulation/floccula-
     tion) induce  further variation  by
     chemically destabilizing the water,
     which  is followed by  reequilibra-
     tion within the distribution system.

  4.  Turbidity is only partially useful as
      a  predictor of particle concentra-
     tion. Small particles of less than 5-
     /jm diameter correlate most closely
      with turbidity  levels, however,  the
      concentration variation of all size
      ranges studied accounted for only
      60 percent of the  overall variation
      in  observed turbidity levels.

  The full report  was  submitted  in
fulfillment of Cooperative Agreement  No.
810508 by the  Department  of  Civil
Engineering, University of Washington,
under the sponsorship of  the  U.S.
Environmental Protection Agency.

-------
         7 *
        0.5 '
   o
    tb
       0.5 i
                      •*•
                                  24
                                                                                                       Finished Water
                                                                                                       Endline Tap
                                                         48                    72


                                                Experimental Progression (hours)


Figure 4.    Comparison of free chlorine residuals in finished and delivered water in the Anacortes System.
                                                                                                      96
                                                                                                                             120
         20



         10



          0


       200
        700
      2000
        0.75'



          0.5'
       0.25
                                 Particle Cone. (20-30 urn)
                                        /?,„* = .304
                                Particle Cone. (10-20im)

                                       flturb = .534
                                Particle Cone. (2-10 fjm)

                                          = • 687
                                 Turbidity
                                                                    •#•
                                  24
                                                          48                    72


                                                    Experimental Progression (hours)
                                                                                                      96
                                                                                                                             120
Figure 5.    Comparison of turbidity and particle  concentrations  in  the  Seattle  Water Distribution System (U. of Wa. Campus).

                                      6

-------
     John F. Ferguson, Mark M. Benjamin, Dimitris Spyridakis, and Steve H. Reiber
       are with the University of Washington, Seattle, WA 98195.
     Marvin Gardels is the EPA Project Officer (see below).
     The complete report entitled "Corrosion in Water Distribution Systems of the
       Pacific Northwest." (Order No. PB 87-197 521/AS; Cost: $13.95, subject
       to change)
     will be available only from:
             National Technical Information Service
             5285 Port Royal Road
             Springfield, VA22161
             Telephone: 703-487-4650
     The EPA Project Officer can be contacted at:
             Water Engineering Research Laboratory
             U.S. Environmental Protection Agency
             Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
     BULK RATE
POSTAGE & FEES
        EPA
   PERMIT No. G-
Official Business
Penalty for Private Use $300

EPA/600/S2-87/042
                                                                                                             I II
                                                                                                             CO
                                                                                                             ro
                                                                               
-------