ARTICLE IN PRESS

WAT E R R E S E A R C H 41 (2007) 519 – 529

Available at www.sciencedirect.com

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

REVIEW

Discolouration in potable water distribution systems:

A review

Ir J.H.G. Vreeburga,, Dr. J.B. Boxallb

a
Technical University Delft, Kiwa Water Research, The Netherlands
b
University of Sheffield, UK

art i cle info ab st rac t

Article history: A large proportion of the customer contacts that drinking water supply companies receive
Received 15 June 2006 stem from the occurrence of discoloured water. Currently, such complaints are dealt with
Received in revised form in a reactive manner. However, water companies are being driven to implement planned
1 September 2006 activities to control discolouration prior to contacts occurring. Hence improved under-
Accepted 12 September 2006 standing of the dominant processes and predictive and management tools are needed.
Available online 14 December 2006 The material responsible for discolouration has a variety of origins and a range of
Keywords: processes and mechanisms may be associated with its accumulation within distribution
Potable water systems. Irrespective of material origins, accumulation processes and mechanisms,
Discolouration discolouration events occur as a result of systems changes leading to mobilisation of the
Proactive management accumulations from within the network. Despite this conceptual understanding, there are
Particles very few published practicable tools and techniques available to aid water companies in the
Cleaning planned management and control of discolouration problems. Two recently developed and
published, but different approaches to address this are reviewed here: the PODDS model
which was developed to predict levels of turbidity as a result of change in hydraulic
conditions, but which is semi-empirical and requires calibration; and the resuspension
potential method which was developed to directly measure discolouration resulting from a
controlled change in hydraulic conditions, providing a direct assessment of discolouration
risk, although intrinsically requiring the limited generation of discoloured water within a
live network. Both these methods support decision making on the need for maintenance
operations.
While risk evaluation and implementation of appropriate maintenance can be
implemented to control discolouration risk, new material will continue to accumulate
and hence an ongoing programme of maintenance is required. One sustainable measure to
prevent such re-accumulation of material is the adoption of a self-cleaning threshold, an
hydraulic force which a pipe experiences on a regular basis that effectively prevents the
accumulation of material. This concept has been effectively employed for the design of
new networks in the Netherlands. Alternatively, measures could be implemented to limit
or prevent particles from entering or being generated within the network, such as by
improving treatment or preventing the formation of corrosion by-products through lining

Corresponding author. Tel.: +31 30 606 95 76; fax: +31 30 606 11 65.
E-mail address: j.h.g.vreeburg@tudelft.nl (J.H.G. Vreeburg).
0043-1354/$ - see front matter & 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2006.09.028
 ARTICLE IN PRESS
520 WA T E R R E S E A R C H 41 (2007) 519– 529

or replacing ferrous pipes. The cost benefit of such capex investment or ongoing opex is
uncertain as the quantification and relative significance of factors possibly leading to
material accumulation are poorly understood. Hence, this is an area in need of significant
further practical research and development.
& 2006 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520
1.1. Particles in the distribution system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
2. Measurement and modelling techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
2.1. Turbidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
2.2. Resuspension potential method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
2.3. Cohesive transport model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
2.4. Other risk estimation tools and techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
3. Cleaning of networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
3.1. Velocity criteria for flushing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
3.2. Shear stress criteria for flushing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
3.3. Self-cleaning threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
3.4. Longer term implications for cleaning strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
5. Summary/conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528

1. Introduction a reactive manner. Within modern customer focussed water

companies such reactive maintenance is no longer accepta-
A large proportion of the customer contacts that drinking ble, particularly within the regulatory framework of the UK.
water supply companies across the world receive stem from Water companies urgently need a practicable understanding
the occurrence of discoloured water. Fig. 1 shows a typical of the processes and mechanisms leading to discolouration
break down of customer contacts for a UK water company, incidents and to develop management tools and techniques.
while Fig. 2 shows an example of discoloured water supplied This paper reviews the techniques to assess the discoloura-
to a customer in Holland. tion risk and the strategies available to control this.
Discoloured water incidents as shown in Fig. 2 greatly affect
customer’s confidence in tap water quality and the quality of
service provided by water companies. Although good custo-
mer perception is a major driver for water companies (van
Dijk and Van der Kooij, 2005), thorough understanding of the
mechanisms and processes that lead to discolouration are
currently lacking or at least not applied widely. Hence, water
companies can only respond to discolouration complaints in

7% 0%

40%
34%

19%
No Water Pressure Problems
Discoloured Water Other Aesthetic Problems
Illness Complaint

Fig. 1 – Typical break down of reasons for customer contacts Fig. 2 – Example of discoloured water leading to customer
for a 5 year period for a UK water company. complaints.

Although referred to as discolouration the visual effect
observed by customers is rarely colour in a strict water quality
sense, defined as contaminants absorbed into the chemical
composition of the water. Typically, if a ‘discoloured’ water
sample is left to stand for a prolonged period (over night) it
will clear, and material will deposit. Hence, it can be
concluded that it is particulate matter that the customer
experiences as ‘discolouration’. The measurable parameter
requiring investigation is therefore turbidity. However, differ-
ent particles have significantly different effects on perceived
turbidity, or discolouration. A combination of factors includ-
ing obscuration, reflection, refraction, diffraction and scatter
contribute, although scattering usually dominates. Russell
(1993) states that peak scattering occurs for particles at
around half a micron diameter with a rapid fall off for
suspensions of larger or smaller sizes.
Particulate accumulations are also known to have a relation
with biological activity (Gauthier et al., 1999). One to 12% of
the organic matter in the particulate accumulations may
consist of bacterial biomass, making the deposits an im-
portant factor in hygienic safety of drinking water.
Discoloured water events are difficult to study in real
systems because they often occur over short durations for
unpredictable reason. Fig. 3a and b show typical short
duration events captured by turbidity instruments installed
in systems in the Netherlands and UK, respectively. The
figures show that discolouration events have the same
characteristics: a sharp rise in turbidity that reduces within
a few hours, despite considerable differences between the
(a)
30
Turbidity (NTU)
20
10
0 originating from biofilters. The distribution system itself can
10/6/04 10/7/04 9/8/04 8/9/04
(b) Date
Fig. 3 – (a) Typical discolouration event measured in a water
distribution system in The Netherlands. (b) Typical
discolouration event measured in a UK water distribution
system.
ARTICLE IN PRESS
WAT E R R E S E A R C H 41 (20 07) 51 9 – 529 521
Dutch and UK networks. Prince et al. (2003) shows similar
results in monitoring turbidity and velocity at different
locations in the Melbourne drinking water distribution
system. The largest growth period of the network in the
Netherlands was in the period 1945–1980, hence the average
age of the network is 45 years and the predominant pipe
materials are PVC and AC. Conversely, in the UK the networks
have not experienced such intensive investment leading to
systems that are still dominated by cast iron pipes dating
back over the last 100 years and beyond. The Australian
network researched by Prince et al. (2003) is of more recent
date than the Dutch network and has seal coated concrete
and PVC as dominant pipe material. The treatment histories
of the systems are also different with systems in the
Netherlands having long adopted a very high standard of
treatment and a policy of no chlorination, while the UK has
seen a variety of levels of service, although now all treatment
is to a high standard. The Australian network is supplied with
unfiltered water, dosed with chlorine, fluoride and lime.
These historic factors are key to understanding the levels of
services and the processes leading to the occurrence of
discolouration events as shown in Figs. 3a and b. This
difference is also manifest in the reactive trigger levels that
companies use to instigate cleaning in response to disco-
louration, typically around 4 contacts per 1000 properties in
UK compared with 1–2 contacts per 1000 in the Netherlands
and 6 contacts per 1000 properties in Australia. This shows
that despite obvious differences in systems, the same
discolouration problems occur. Intuitively, discolouration in
the Dutch systems should be almost non-existent with the
history of good treatment, very low leakage and a network
with a limited amount of cast iron pipes, but the figures show
that discolouration does occur and other processes are
involved besides corrosion. This may also highlight the
inconsistent nature of customers, with propensity to make
contact predominately when the quality of the water changes
from what is perceived as ‘normal’.
1.1. Particles in the distribution system
Discolouration is associated with the mobilisation of accu-
mulated particles from within distribution networks. Such
particles have different sizes and densities and hence
probably have different origins, often characterised as either
external sources or from processes occurring within the
system. Particles can enter the distribution network as
background concentrations of organic and inorganic material
from the source water (Lin and Coller 1997; South East Water,
1998; Kirmeyer et al., 2000; Slaats et al., 2002; Ellison, 2003),
due to incomplete removal of suspended solids at the
treatment plant (Gauthier et al., 2001; Vreeburg et al., 2004b)
or be added to the water by the treatment plant itself, such as
carbon and sand particles, alum or iron flocs and bioparticles
8/10/04 7/11/04
also produce particles such as from pipe and fitting corrosion
and lining erosion (Stephenson, 1989; Ruta, 1999; Gauthier et
al., 2001; Clement et al., 2002; Slaats et al., 2002; Boxall et al.,
2003), biological growth (Le Chevallier et al., 1987; Stephen-
son, 1989; Clark et al., 1993; Meches, 2001) and chemical
reactions (Stephenson, 1989; Sly et al., 1990; Walski, 1991; Lin
 ARTICLE IN PRESS
522 WA T E R R E S E A R C H
and Coller, 1997; Kirmeyer et al., 2000); or external contam-
ination that may occur during operations such as pipe repairs
(Gauthier et al., 1996; Slaats et al., 2002), intrusion (Gauthier
et al., 1999; Kirmeyer et al., 2000; Prince et al., 2001) and
backflow. Possibly the most common and significant biologi-
cal process is biofilm formation which can result from the
presence of assimilable organic carbon in the water or the
pipe wall (van der Kooij, 2002). The effects of these complex
and interacting processes is further complicated by exposure
to various different physical and chemical conditions during
passage through distribution systems including contact with
a range of different pipe materials and ages and different
hydraulic conditions. The formation and growth of particles
is a very complex process which is currently poorly under-
stood. Factors such as contact times, contact surface and
hydraulic condition are likely to play an important role in
controlling these processes. These sources, external and
internal, rarely contribute directly to discolouration events
but facilitate the gradual accumulation of material within the
distribution system.
Next to the sources and growth of particles, it is important
to understand the hydraulic behaviour of the particles to
determine the fate of the particles in the network. Boxall et al.
(2001) presented results for the distribution of particle sizes
found in discoloured water samples, suggesting a repeatable
distribution of particle sizes irrespective of network condi-
tions, source water etc. They suggested that the size range of
the particles was predominately less than 0.050 mm, with an
average size of around 0.010 mm and a significant number of
particles in the sub 0.005 mm range. Boxall et al. (2001) went
on to show that it is unlikely that gravitational settling alone
will be a sufficient force for accumulation of such particles as
turbulent forces generated by even the lowest flows within a
distribution system are likely to be sufficient to over come
gravity settling forces, particularly for the smaller sized
particles found within discolouration samples which will
dominate discolouration due to their light scattering proper-
ties. Fig. 4 shows material accumulation due to corrosion
processes around the complete circumference of pipe sam-
ples and a lack of invert deposit, consistent with these
concepts. Samples such as these have been installed in a
laboratory facility and significant discolouration generated by
exposing them to flushing flow rates, despite the disturbance
of weakly adhered material caused by obtaining the samples.
Whatever processes dominate the accumulation of material
within a distribution system, the processes may be concep-
tualised as the loss of material from the bulk fluid to the pipe
Fig. 4 – Material accumulation around the complete
perimeter of a cast iron pipe samples.
41 (2007) 519– 529
wall. The exception to this is corrosion of ferrous pipes and
fittings, which may contribute directly to layers at the pipes
wall, as well as material to bulk fluid which may then
accumulate in other non-ferrous pipes (Smith et al., 1999;
Boxall et al., 2003; Seth et al., 2004).
To investigate the process of loss of material from the bulk
fluid to the pipe wall a 6 m long pipe test facility was build out
of 100 mm internal diameter Perspex pipe. The facility was
run in a pressurised re-circulating mode, with a downstream
flow control. The source water was initially dosed with iron
chloride to a concentration of 10 mg/l iron corresponding to a
turbidity of 10 FTU. After 4 days of recirculation the turbidity
had dropped below 0.5 FTU, resulting in accumulation on the
pipe walls as shown in Fig. 5 for the section of the pipe where
the flow is stable, unaffected by curvature or entry/exit
conditions. What is particularly remarkable about Fig. 5 is
that at low velocities only the lower half of the pipes
accumulates iron flocs, while with the higher velocity flocs
accumulate over the complete pipe perimeter. However, the
size of the flocs formed will be a function of the flow regime
and the relatively pure iron chloride flocs formed are likely
to be larger than the size of particles typically seen in
discoloured water samples.
This phenomenon could be explained with turbophoresis
(Young and Leeming, 1997). Turbophoresis is the process that
describes the turbulent transportation of particles from more
turbid regions to less turbid regions in a flow pattern. The
turbophoretic force is dependant of the gradient of turbulence
over the flow profile. In pipe flow this means that particles are
transported from the bulk fluid to less turbid regions near the
wall where they can be trapped in cohesive layers. With
higher velocities the gradient is greater as the turbulence at
the pipe wall must always be zero, resulting in a larger force
driving particles from the centre to the wall of the pipe. In
light of this theory it can be suggested that at a flow rate of
0.14 m/s the turbophoresis force exceed the gravitational
force resulting in uniform supply of material at the pipe
surface, while at 0.06 m/s the gravity and turbophoresic forces
where nearer to equilibrium.
Overall it can be concluded that the mechanism leading
to discolouration events are complex, poorly understood
and interactive. However, the processes may be understood
through a relatively easy concept. The cause of discoloura-
tion is particles attached by some means to the pipe wall.
Pipe diameter 100 mm
Flow 0.06 m/s Flow 0.14 m/s
Fig. 5 – Photographs of accumulated material within a
Perspex pipe loop after 4 days re-circulation with high load
ferric chloride solutions (10 mg/l).
 ARTICLE IN PRESS
WAT E R R E S E A R C H
Accumulation during normal flow Mobilization due to above normal flow
Fig. 6 – Conceptual model of the fundamental processes
leading to the occurrence of discolouration within potable
water distribution systems.
In normal flow the particles stay in their place and do not
affect the aesthetic quality of the water. If flows are increased
above normal, scouring forces and shear stress increase
consequently and then the particles may be mobilised,
sometimes leading to customer complaints, Fig. 6.
2. Measurement and modelling techniques
While various research projects have been undertaken to
investigate and improve understanding of the processes and
mechanisms leading to discolouration, as reviewed above,
there is a relatively small body of published material relating
to the derivation of the practicable tools and techniques that
are needed by the international water supply industry.
2.1. Turbidity
Discrete instruments turbidity meters have been available as
proven and reliable instrumentation for some time, while
treatment work control has driven the development of
continuous, low range instruments for processes control.
However, more robust instrumentation, with greater dynamic
range and improved logging and communications technology
are now available suitable for deployment on distribution
systems. Such equipment allow continuous monitoring at
several locations at the same time, making it possible to
record the changes in turbidity and hence to identify causal
factors (Slaats et al., 2002; van der Hoven and Vreeburg, 1992).
Data obtained from such turbidity meters have been used to
develop techniques to aid water companies to identify and
quantify discolouration risks within distribution networks
(Vreeburg, 1996).
2.2. Resuspension potential method
Irrespective of the origin, the presence and mobility of
deposits determines the discolouration risk. The resuspen-
tion potential method (RPM) as developed in the Netherlands
(Vreeburg et al., 2004a,b) is based on measuring the mobility
of the material in a network.
The RPM consists of a controlled and reproducible increase
of the velocity of 0.35 m/s in a pipe on top of the actual
velocity. The hydraulic shear stress as a result of the
increased velocity causes particles to mobilise, affecting the
turbidity of the water. The method is mainly applied in
100–150 mm pipes hence the absolute difference in shear
stress caused by the uniform velocity increase is not very
large. The velocity of 0.35 m/s was empirically determined
(Vreeburg et al., 2004a,b). The turbidity effect is monitored
41 (20 07) 51 9 – 529 523
and translated to a ranking of the discolouration risk. The
method is applied as follows:
 Isolate the pipe for which the discolouration risk is to be
assessed, as for uni-directional flushing (Antoun et al.,
1999). The isolated length should be at least 315 m to be
sure that only this single pipe is affected.
 Open a fire hydrant such that the velocity in the pipe is
increased by the additional 0.35 m/s above the actual
velocity and maintain for 15 min, after this reduce the
flow to normal (total length affected is 315 m).
 Monitor turbidity in the pipe throughout the 15 min of
extra velocity and beyond that until turbidity returns to
the initial level.
The result obtained from an RPM test is the turbidity
response of a pipe. A typical example is shown in Fig. 7
highlighting four regions of the trace that are utilised to rank
discolouration risk.
The RPM elements are:
 Base level turbidity, the level preceding the hydraulic
disturbance, used in estimating the time for the turbidity
to return to initial levels.
 Initial increase in turbidity during the first 5 min at the
start of the hydraulic disturbance quantifies the top layer
that is immediately available and gives the instantaneous
mobility of the particles resulting in peak turbidity.
 Development of turbidity during last 10 min of the
hydraulic disturbance.
 Resettling time and pattern to base (initial) turbidity level
after stopping the disturbance. The duration here is
important for the discolouration risk, the longer the
turbidity levels remain increased the greater the risk of
discolouration complaints.
Discolouration risk ranking is based on five elements: the
maximum and average turbidity in the first 5 min and last
Fig. 7 – Typical turbidity trace resulting from an RPM test,
showing the four regions used to rate the discolouration
risk.
 ARTICLE IN PRESS
524 WA T E R R E S E A R C H
Turbidity with RPM, before cleaning
Disturbance 13:09 - 13:23
50
40
Turbidity [FTU]
30
20
10
0 0
13:09 13:11 13:13 13:15 13:17 13:19 13:25
Time
Fig. 8 – Results of RPM method applied pre- and postcleaning to evaluate the effectiveness of the operations.
10 min of the disturbance and the time to clear. Each of these
can be rated on a scale from 0 to 3 and summed, characteris-
ing the experiment to a single figure on a scale of 0–15. The
ranking tables can be adjusted based on the results obtained
and instrumentation used (i.e. average turbidity levels) to
obtain a spread of risk scores, providing the flexibility to tailor
the method to different networks, Vreeburg et al. (2004a,b).
The RPM was developed within the joint research program
of the Dutch water companies (Bedrijfstakonderzoek BTO)
and has been applied by the Dutch water companies for more
than a decade. The method is used to evaluate both the need
for cleaning and through application following maintenance
to evaluate effectiveness of cleaning regimes. Fig. 8 gives the
result of a RPM before and after cleaning by flushing a pipe.
Regular assessment with RPM in network can provide
information on the necessary frequency of cleaning.
2.3. Cohesive transport model
As stated earlier Boxall et al. (2001) carried out theoretical
analysis of the interaction of particles of the size found to
predominate in discoloured water samples with respect to the
hydraulic forces generated within distribution networks,
concluding that forces and mechanisms above and beyond
gravity settling forces must be in effect to inhibit particle
movement. Rather than trying to identify or quantify specific
contributing processes, they suggested a semi-empirical
model that could be used to account for the effects of any
such processes. The model they proposed was based on
theory developed to describe the erosion of estuarine mud
Parchure and Mehta (1985) and as applied to in-sewer
deposits by Skipworth et al. (1999).
The model is based on the concept that discolouration
material is held in stable cohesive layers attached to the pipe
walls of distribution systems and that these layers are
conditioned by the usual daily hydraulic regime within the
system. Within the model the material layers are described by
a profile of discolouration potential versus layer strength,
with an increase in potential corresponding to a decrease in
strength. This strength, and hence layer state, is dictated by
the shear stresses imposed by hydraulic conditions. Hence
areas with low daily maximum hydraulic forces, such as
dead-end pipes, redundant loops, over sized pipes, zone
boundaries, extremities of loops etc. will have low strength
characteristics and high discolouration potential, as has been
noted in practice. The occurrence of disequilibria hydraulic
41 (2007) 519– 529
Turbidity with RPM, after cleaning
Disturbance 9:59 - 10:15
50
Turbidity [FTU]
40
30
20
10
09:59 10:02 10:05 10:08 10:11 10:14
Time
conditions (burst, re-zoning, increased demand etc.) may
expose the layers to shear stress in excess of their condi-
tioned cohesive strength and lead to a mobilisation of the
cohesive layers, resulting in a discolouration event.
It should be noted that neither the source of material nor
the mechanisms and processes leading to accumulation and
binding of particles are considered explicitly within the
modelling approach. However, through calibration of the
empirical parameters describing layer strength characteris-
tics, mobilisation and accumulation mechanisms a range of
processes and materials may be simulated.
The model is used to predict turbidity as a result of
hydraulic disturbance, and has been termed PODDS (predic-
tion of discolouration events in distribution systems). PODDS
has been coded into EPANET (Rossman, 2000) and runs as a
water quality element that utilises the EPANET hydraulic
solution, substance tracking and transport algorithms. The
incorporation of such a modelling approach into a calibrated
hydraulic model allows the simulation of the discolouration
risk (potential and impact) posed by different network areas
and hydraulic scenarios. Once calibrated the model may be
used to plan pro-active management strategies such as the
flushing of systems to reduce the risk of discolouration
events.
The model has been validated for data collected from
flushing operations in the UK (Boxall and Saul, 2005), as
shown in Fig. 9, and for data collected in Australia (Boxall and
Prince 2006).
2.4. Other risk estimation tools and techniques
The authors are aware of two other projects currently in
progress for the development of practicable tools and
techniques to aid the water supply industry with the
identification and estimation of discolouration risk. The
discolouration risk management (DRM) tool which is being
developed and rolled out to UK industry by Ewan Group plc
and Yorkshire Water Services (Dewis and Randall-Smith,
2005). DRM is a risk based assessment tool incorporating
likelihood of pipe failure, discolouration and consequence
and is based on ‘expert panel’ risk trees. As such DRM is a
pragmatic asset management and investment tool, but is
limited by the discolouration knowledge and understanding
of the required expert panels. The particles sediment model
(PSM) is being developed and trialled in Australia by the
Cooperative Research Centre (CRC). However, as far as the
 ARTICLE IN PRESS
WAT E R R E S E A R C H 41 (20 07) 51 9 – 529 525

10
250
9
measured
simulated
200 8
measured 7

Turbidity, N (NTU)
turbidity, N (NTU)

150
simulated 6

100 4

3
50
2

1
0
0
5000 7500 10000 12500 15000
12000 13000 14000 15000 16000 17000 18000 19000

time, t (seconds) Time, t (seconds)

Flushing of a 76mm cast iron pipe Flushing of a 100mm UPV

70 100

90
60 measured measured
simulated 80 simulated
50 70
Turbidity, N (NTU)

turbidity, N (NTU)
60
40
50
30
40

20 30

20
10
10

0 0
2000 4000 6000 8000 10000 12000 23000 24000 25000 26000 27000 28000
Time, t (seconds) time, t (seconds)

2 stage flushing of a 76mm asbestos cement pipe Flushing of a 171mm asbestos cement pipe

Fig. 9 – Results of PODDS model simulation for four different flushing operations (after Boxall et al., 2005).

7
6
authors are aware, little or no material is currently in the
Turbidity (FTU)

public domain relating to this. 5

4
3
3. Cleaning of networks 2
1
Once an estimate of discolouration risk has been obtained it
0
is necessary to implement some network cleaning technique wed thu fri sat sun mon tue wed
to manage the risk. Three non-structural methods are Days of the week
commonly applied within the drinking water industry
world-wide: water flushing, water/air scouring and swab- Fig. 10 – Turbidity pattern after water/air scouring of a 4’’ CI
bing/pigging. Although all the methods are capable of main.
cleaning a network, there are large differences in costs and
effectiveness.
In general water flushing is the simplest and most cost water/air scouring) on a 100 mm cast iron main. The pattern
effective way of reducing the risk of discolouration. More of rising turbidity during stagnation in the night hours and
complete removal of all material from a pipe is only possible falling turbidity in the morning peak hours is typical for an
with more abrasive methods like swabbing and pigging. active corrosion process, Smith et al. (1999), Slaats et al.
However, these methods are comparatively expensive and (2002).
disruptive to the distribution systems. They also have the
potential to aggravate corrosion processes in ferrous pipes if 3.1. Velocity criteria for flushing
some form of internal lining is not applied following their
application. Established corrosion layers can effectively In the Netherlands a strict velocity criterion of 1.5 m/s has
protect underlying ferrous material, aggressive cleaning can been pursued as a cost effective mains cleaning technique
expose this underlying surface which will then start to since the 1990s. However, the efficiency and effectiveness of
corrode more rapidly generating material at the pipe wall the method is significantly influenced by a number of factors:
and releasing ferrous ions into the bulk fluid. Fig. 10 shows the strict use of a clear water front (through the use of valve
the turbidity after effects of abrasive cleaning (in this case operations to control flow routes and to separate the cleaned
 ARTICLE IN PRESS
526 WA T E R R E S E A R C H
and still fouled network); thorough cleaning of all pipes in the
network; and ensuring 2–3 times turn over of the volume of
water in the pipe effected. An advantage of controlling flow
routes is that the point of flushing can be located near to a
ditch or a storm water run off. The flush point can be
equipped with a dedicated flush point with a higher capacity
than a normal hydrant, allowing high volume flows.
Flushing guidelines, such as the Water Mains Cleaning
handbook (Miller, 1994) in the UK suggest diameter and
specific gravity dependent velocities for effective entrainment
and removal of material from within distribution systems.
Typically, such figures are based on classical sediment
transport theory, such as presented by Stephenson (1989).
UK guidance values range from 0.7 m/s in a 50 mm pipe to
1.3 m/s in a 200 mm pipe.
3.2. Shear stress criteria for flushing
From theoretical analysis based on particle size distribution
measurements of discoloured water samples, Boxall et al.
(2001) suggested that traditional sediment transport theory is
not appropriate for describing the generation of discoloura-
tion within distribution systems. Suggesting that the pro-
cesses are better described through consideration of the
interaction of hydraulic shear stresses and the pipe wall/
water interface with material layers. Similarly, Ackers et al.
(2001) recognised the importance of shear stress for the
mobilisation of material and recommended a value of
2.5 N/m2 that should be achieved by flushing. However, this
value is based on previous research and design principles for
sewer systems and may not be appropriate for distribution
systems.
Simplified modelling may be attempted using the PODDS
modelling assumptions that the discolouration potential of
layers is a function of their strength characteristics, dictated
by the daily hydraulics and that increases in shear stress
above this value, excess shear stress, will produce succes-
sively more turbidity, as demonstrated by Boxall and Dewis
(2005). The greater the excess shear the greater the disco-
louration. It should be noted that this is a very simplified
approach, as the layer characteristics have been found to be
∅ 75 Flat
VP 350 kPa
∅ 110
Flat house
Hospital
VP 350 kPa
Fig. 11 – (a) Conventional distribution network (van Boomen Vreeburg, 1999). (b) Distribution network after reduced fire
fighting demands (van Boomen Vreeburg, 1999).
41 (2007) 519– 529
dependent on several factors, including source water, pipe
material and diameter, Boxall and Saul (2005). Relative
discolouration risk, based on shear stress criterion, could be
judged as the average of the excess shear stresses caused by
different changes in hydraulic conditions, with steady state
shear stress (t) readily evaluated from:
D
t ¼ rg S0 , (1)
4
where r is shear stress, g acceleration due to gravity, D pipe
diameter and S0 hydraulic gradient or head loss estimated by
every hydraulic network model. Hence, unlike velocity
criteria, shear stress criteria are diameter, roughness and
velocity squared independent.
3.3. Self-cleaning threshold
It has been suggested that material will tend to accumulate in
areas with low velocities, such as dead ends, over sized pipes
and redundant loops. Such features are common in most
networks as the systems are deigned to comply with large fire
fighting demands that are typically far greater than consumer
demands, particularly at the extremities of large systems and
in smaller systems. A typical over sized looped network is
shown in Fig. 11a, the white pipe is part of the main transport
system, while the distribution pipes are looped and of smaller
diameter. The velocities in such systems are low and the
loops will probably experience flow reversals and tidal points,
lead to long residence times and risk of discolouration.
On the basis that higher velocities within the systems
would reduce the potential for material accumulation and
hence reduce the risk of discolouration a new approach to
network design has been instigated in the Netherlands since
1999 (van Boomen Vreeburg, 1999). The approach is based on
a fundamental rethinking of the fire fighting demand. In close
cooperation with national fire fighting agencies in the Nether-
lands fire fighting codes have been amended to stipulate just
30 m3/h in residential areas in which buildings meet modern
fire fighting codes. With this reduced requirement pipe sizing
is designed to meet expected customer demands and velocity
criteria. A velocity of at least 0.4 m/s is stipulated as being
sufficient to prevent accumulation of material. The value of
∅ 40
∅ 63 VP 350 kPa
∅ 110 64
116 85
60 m3/h
Hospital
VP 350 kPa
 ARTICLE IN PRESS
WAT E R R E S E A R C H
0.4 m/s is determined as a pragmatic value that is achievable
and weakly substantiated by the shear stresses caused by
flow acceleration up to this value. The new design philosophy
results in a distribution network that is branched with pipes
of relatively small diameter, such as shown in Fig. 11b
showing declining diameters with distance to ensure the
minimum velocity in the pipes whilst maintaining security of
supply. The new design rules are widely applied in the
Netherlands resulting in a reduction in the average pipe
diameter by length in new networks, the rules are not only
accepted on the basis of improved water quality but also
because the new networks are on average 20% cheaper to
construct than conventional ones. The savings are mainly
achieved by the reduction in length of the pipes because the
loops are not closed anymore. Further development of the
design procedure comprises research to the pattern of water
consumption of individual houses, resulting in better esti-
mates of maximum flows (Blokker and Vreeburg, 2005).
The concept of a self-cleaning threshold, defined as a shear
stress that a pipe experiences regularly, due to normal daily
demand, that prohibits the accumulation of sufficient mate-
rial within the pipe and hence pose no discolouration risk was
investigated by Boxall and Prince (2006). They evaluated the
forces necessary to inhibit accumulation of material through
evaluation of imposed shear stress rather than velocity,
arriving at a value of 1.12 N/mm2 for a large diameter asbestos
cement main for clay driven discolouration problems in
Melbourne, Australia.
3.4. Longer term implications for cleaning strategies
Particles in the distribution network are the key to the
aesthetical water quality in the network. There are several
ways to control the amount of particles in the network the
practical impact of which are summarised in Fig. 12.
The solid line represents the (re)charging of the network
with particles, increasing the risk of discolouration. Once a
certain threshold is reached remedial action is taken to clean
the network, reducing the risk of discolouration. If nothing
further is changed, the network will recharge with material
from various sources re-accumulating resulting in the need
for further cleaning. This procedure gives the cleaning
frequency for this network. However, there are various
possibilities for improving the quality or prolonging the
Effect of Effect of investment / management /
cleaning operational strategies
Discolouration risk
Threshold level
Cleaning
Time
frequency 1
Cleaning frequency 2
Fig. 12 – Potential to manage discolouration risk in a
network.
41 (20 07) 51 9 – 529 527
cleaning frequency, such as reducing the rate of material
accumulation through investment (improving treatment
works or replacing/relining iron pipes), management (disin-
fection regime to control biofilm growth) and operational
(change the system hydraulics to impose increased hydraulic
forces). However, the quantification and relative value of such
strategies in different situations is beyond current methods
and understanding.
4. Discussion
Discolouration is a problem that is as old as public drinking
water supply. Until a few years ago this phenomenon has had
relatively little attention, however, with other improvements
in the supply of drinking water discolouration is now the
single most common reason for customer contacts. Research
has been undertaken in the last decade into the underlying
mechanisms of discolouration, going beyond the intuitive and
accepted causes like corrosion, such that new tools and
techniques can be developed to support the implementation
of planned operation and maintenance strategies to control
discolouration risk. Most of the research however concen-
trates on the composition of the removed sediment and the
possible impacts on microbiological stability of the water
(Gauthier et al., 1996, 1999, 2001; Carriere et al., 2005; Torvinen
et al., 2004; Zacheus et al., 2001; Barbeau et al., 2005). The
actual discolouration risk is not assessed or evaluated. This
research though has led to the concept that the gradual build
up of cohesive particulate layers in combination with
hydraulic disturbances are the mechanisms leading to
discolouration events.
Initial emphasis for the application of this research has
been the development of practicable tools to identify dis-
colouration ‘hot spots’, producing a physical method (RPM)
and a modelling approach (PODDS). Managing these hot
spots with monitoring and cleaning can control them,
although the application of science behind such cleaning
has not yet led to optimal operations. Unidirectional flushing
is recognised as a good technique, but guidelines are mainly
driven by ‘‘Good management practices’’ (Friedman et al.,
2002) to optimise the costs a flushing program rather than to
maximise the effect.
Guidelines based on shear stresses as well as general
guidelines on velocity illustrate the need for detailed cleaning
programs involving network simulation to design efficient
and effective cleaning protocols. This changes the image of
cleaning from a low rated routine task with a low efficacy to
an important water quality process that involves pro-active
operation of networks. Practise in the Netherlands shows that
the strict application of three simple rules (velocity41.5 m/s,
clear water front and pipe turnover 2–3 times) is effective. Use
of pre- and postassessment of the discolouration risk either
with the RPM or other sediment sampling methods should be
a routine activity in the cleaning programs (Schaap and
Vreeburg, 1999).
Research challenges are now for developing better under-
standing of cleaning efficacy with particular reference to
improving understanding of the processes controlling the
refouling of systems. Such that the cleaning frequency, as
 ARTICLE IN PRESS
528 WA T E R R E S E A R C H
shown in Fig. 12, can be predicted and managed. There are
indications that these frequencies should be estimated in
terms of months rather than years. Preventing or limiting
material accumulation, to increase these cleaning frequen-
cies, can either be influenced by eliminating or restricting the
particle production processes or by resizing the network to
promote self-cleansing capacities.
5. Summary/conclusions
Particles in water distribution network, either as loose
deposits or trapped in cohesive layers, are the main factor
related to discoloured water. Water companies need methods
to quantify, and manage particle accumulations and hence
reduce customer complaints.
For estimating the discolouration risk a model based
method is available called PODDS. PODDS allows water
companies, after sufficient calibration, to predict the dis-
colouration potential of pipes in a network and hence to
prioritise maintenance. Next to the model approach a
practical measuring method has been developed that facil-
itates assessment of the resuspension potential of particles
determining the discolouration risk. The RPM is also applic-
able as evaluating tool for the efficacy of the applied cleaning
techniques.
Remedial action as cleaning by uni directional flushing are
not uniformly described and rather based on good manage-
ment practises than on optimal results. Clear and simple
rules for flushing (1.5 m/s, 2–3 pipe turnovers and clear water
front) prove to be effective in the Netherlands.
Adoption of a self cleaning threshold, a force experienced
regularly by a pipe to prohibit material accumulation, can be
an effective long term approach for controlling discoloura-
tion, this also results in considerably cheaper networks with
better water quality and minimal maintenance needs.
In the longer term sustainable solutions for controlling
discoloured water problems may be found through preventing
the causes of particles, such as by improving treatment,
stopping or controlling regrowth, coagulation and corrosion
in the network. However, the quantification and evaluation of
such strategies in different situations is beyond current
methods and understanding and these represent the main
challenges to meet in the future.
R E F E R E N C E S
Ackers, J., Brandt, M., Powel, J., 2001. Hydraulic characteristics of
deposits and review of sediment modelling. Project Drinking
Water Quality and Health—Distribution systems—DW/03’ UK
Water Industry Research, London, UK.
Antoun, E.N., Dyksen, J.E., Hiltebrand, D., 1999. Unidirectional
flushing: a powerful tool. J. Am. Water Works Assoc. 91 (7),
62–71.
Barbeau, B., Gauthier, V., Julienne, K., Carriere, A., 2005. Dead-end
flushing of a distribution system: short and long-term effects
on water quality. J. Water Supply Res. T. 54 (6), 371–383.
Blokker, E.J.M., Vreeburg, J.H.G., 2005. Monte Carlo Simulation of
residential water demand: a stochastic end-use model. In:
ASCE Conference Proceedings 173, 34, May 15–19.
41 (2007) 519– 529
van Boomen, M., Vreeburg, J.H.G., 1999. Nieuwe ontwerprichtlij-
nen voor distributienetten (New design rules for distribution
networks. Kiwa report SWE99.011, Nieuwegein, December
1999, ISBN 90-74741-78-9 (in Dutch).
Boxall, J.B., Dewis, N.D., 2005. Identification of discolouration risk
through simplified modelling. ASCE ERWI World Water and
Environmental Water Resources, 15–19th May.
Boxall, J.B., Prince, R., 2006. Modelling discolouration in a
Melbourne (Australia) potable water distribution system. J.
Water Supply Res. T. 55, 207–219.
Boxall, J.B., Saul, A.J., 2005. Modelling discolouration in potable
water distribution systems. J. Environ. Eng. ASCE 131 (5),
716–725.
Boxall, J.B., Skipworth, P.J., Saul, A.J., 2001. A novel approach to
modelling sediment movement in distribution mains based on
particle characteristics. In: Proceedings of the Computing and
Control in the Water Industry Conference, De Monfort
University, UK.
Boxall, J.B., Skipworth, P.J., Saul, A.J., 2003. Aggressive flushing for
discolouration event mitigation in water distribution net-
works. Water Sci. Technol.—Water Supply 3 (1/2), 179–186.
Carriere, A., Gauthier, V., Desjardins, R., Barbeau, B., 2005.
Evaluation of loose deposits in distribution systems through
unidirectional flushing. J. Am. Water Works Assoc. 97 (9),
82–92+12.
Clark, R.M., Grayman, W.M., Males, R.M., Hess, A.F., 1993.
Modeling contamination propagation in drinking water dis-
tribution systems. J. Environ. Eng. ASCE 119 (2), 349–354.
Clement, J., Hayes, M., Sarin, P., Kriven, W.M., Bebee, J., Jim, K.,
Beckett, M., Snoeyink, V.L., Kirmeyer, G.J., Pierson, G., 2002.
Development of red water control strategies. Awwa Research
Foundation and American Water Works Association, USA.
Dewis, N., Randall-Smith, M., 2005. Discolouration risk modelling.
In: Proceedings of the Eighth International Conference on
Computing and Control for the Water Industry, 5–7th Sep-
tember, Exeter, UK.
van Dijk, J.C., van der Kooij, D., 2005. Water Quality 21 research
programme for water supplies in The Netherlands. Water Sci.
Technol.: Water Supply 4 (5–6), 181–188.
Ellison, D., 2003. Investigation of pipe cleaning methods. Report
No. 90938, Awwa Research Foundation and American Water
Works Association, Denver, USA.
Friedman, M., Kirmeyer, G.J., Antoun, E., 2002. Developing and
implementing a distribution system flushing program. J.
AWWA. 94 (7), 48–56, 10.
Gauthier, V., Rosin, C., Mathieu, L., Portal, JM., Block, J.C., Chaix, P.,
Gatel, D., 1996. Characterization of the deposits in drinking
water distribution systems. In: Proceedings of the Water
Quality Technology Conference of American Water Works
Association.
Gauthier, V., Gérard, B., Portal, J.M., Block, J.C., Gatel, D., 1999.
Organic matter as loose deposits in a drinking water dis-
tribution system. Water Res. 33 (4), 1014–1026.
Gauthier, V., Barbeau, B., Millette, R., Prévost, M., 2001. Suspended
particles in the drinking water of two distribution systems.
Water Sci. Technol.: Water Supply 1 (4), 237–245.
van den Hoven, Th.J.J., Vreeburg, J.H.G., 1992. Distribution system
analysis by continuous monitoring and network calculation.
Water Supply 10 (1), 117–124.
Kirmeyer, G.J., Friedman, M., Clement, J., Sandvig, A., Noran, P.F.,
Martel, K.D., Smith, D., LeChevallier, M., Volk, C., Antoun, E.,
Hiltebrand, D., Dyksen, J., Cushing, R., 2000. Guidance manual
for maintaining distribution system water quality. Report No.
90798, AWWA Research Foundation and American Water
Works Association, Denver, USA.
van der Kooij, 2002. Managing regrowth in drinking water
distribution systems. Kiwa Report BTO 2002.139, Kiwa Water
Research, Nieuwegein, The Netherlands.
 ARTICLE IN PRESS
WAT E R R E S E A R C H
Le Chevallier, M.W., Babcock, T.M., Lee, R.J., 1987. Examination
and characterization of distribution system biofilms. Appl.
Environ. Microbiol. 53 (12), 2714–2724.
Lin, J., Coller, B.A., 1997. Aluminum in a water supply, Part 3:
domestic tap waters. Water. J. Aust. Water Assoc., 11–13.
Meches, M., 2001. Chapter 13—Biofilms in drinking water
distribution systems. Controlling disinfection by-products
and microbial contamination in drinking water. Clark, R.M.,
Boutin, B.K. (Eds.), National Risk Management Research
Laboratory, Office of Research and Development, US Environ-
mental Protection Agency, Cincinnati, Ohio, USA, p. 13.1–13.10.
Miller, S., 1994. Water mains cleaning handbook. Report UMI1429.
Common interest Research Programme, Reference U-0509,
Water Research Centre, Swindon, UK.
Parchure, T.M., Mehta, A.J., 1985. Erosion of soft cohesive
sediment deposits. J. Hydraulic Eng. 111 (10), 1308–1326.
Prince, R., Goulter, I., Ryan, G., 2001. Relationship Between
Velocity Profiles and Turbidity Problems in Distribution
Systems. World Water and Environmental Resources Con-
gress, American Water Works Association, Orlando, Florida,
pp. 9.
Prince, R., Goulter, I., Ryan, G., 2003. What causes customer
complaints about discoloured water. Water 30 (2), 62–67 ISSN
0310-0367.
Rossman, L.A., 2000. EPANET 2 Users Manual EPA/600/R-00/057.
United States Environmental Protection Agency. Cincinnati,
USA.
Ruta, G., 1999. Assessing Water Quality at Customers Internal
Taps. In: Australian Water and Wastewater Association, 18th
Federal Convention Proceedings.
Russell, S., 1993. Water Industry Instrument Handbook—Book 4
Turbidity. WRc plc, Wiltshire, UK.
Schaap, P.G., Vreeburg, J.H.G., 1999. Cleaning of networks: manual
for set up, performance and monitoring of cleaning programs.
Kiwa report SWE 99.009, Nieuwegein, The Netherlands (in
Dutch).
Seth, A., Bachmann, R.T., Boxall, J.B., Saul, A.J., Edyvean, R., 2004.
Characterisation of materials causing discolouration in pota-
ble water systems. Water Sci. Technol. 49 (2), 27–32.
Skipworth, P.J., Tait, S.J., Saul, A.J., 1999. Erosion of sediment beds
in sewers: model development. J. Environ. Eng. 120 (6),
566–573.
41 (20 07) 51 9 – 529 529
Slaats, P.G.G., Rosenthal, L.P.M., Sieger, W.G., van den Boomen, M.,
Beuken, R.H.S., Vreeburg, J.H.G., 2002. Processes involved in
generation of discoloured water. Report No. KOA 02.058,
American Water Works Association Research Foundation/
Kiwa, The Netherlands.
Sly, L.I., Hodgkinson, M.C., Arunpairojana, V., 1990. Deposition of
manganese in drinking water distribution system. Appl.
Environ. Microbiol. 56 (3), 628–639.
Smith, S.E., Holt, D.M., Delanoue, A., Colbourne, J.S., Chamberlain,
A.H.L., Lloyd, B.J., 1999. A pipeline testing facility for the
examination of pipe-wall deposits and red water events in
drinking water. J. CIWEM 13, 7–15.
Stephenson, G., 1989. Removing loose deposits from water mains:
Operational guidelines. Water Research Centre, Swindon, UK.
South East Water, 1998. Water Quality Report 97/98, South East
Water, Melbourne, Australia.
Torvinen, E., Suomalainen, S., Lehtola, M.J., Miettinen, I.T.,
Zacheus, O., Paulin, L., Katila, M.L., Martikainen, P.J., 2004.
Mycobacteria in water and loose deposits of drinking water
distribution systems in Finland. Appl. Environ. Microbiol. 70
(4), 1973–1981.
Vreeburg, J.H.G., 1996. Brown water, cause and consequence.
Efficiency of cleaning with flushing, water/air scouring and
pigging, Kiwa report SWE 96.008, Nieuwegein, The Nether-
lands (In Dutch).
Vreeburg J.H.G., Schaap, P.G, van Dijk, J.C., 2004a. Measuring
Discolouration Risk: Resuspention Potential Method: Leading
Edge Technology, Prague, IWA.
Vreeburg, J.H.G., Schaap, P.G., van Dijk, J.C., 2004b. Particles in the
drinking water syste: from source to discolouration. Water Sci.
Technol. 4 (5–6), 431–438.
Walski, T.M., 1991. Understanding solids transport in water
distribution systems. In: Proceedings Water Quality Model-
ling in Distribution Systems, AWWA Research Foundation,
USA, pp. 305–309.
Young, J., Leeming, A., 1997. A theory of particle deposition in
turbulent pipe flow. J. Fluid Mech. 340, 129–159, Cambridge
University Press.
Zacheus, O.M., Lehtola, M.J., Korhonen, L.K., Martikainen, P.J.,
2001. Soft deposits, the key site for microbial growth in
drinking water distribution networks. Water Res. 35 (7),
1757–1765.