War. Res. Vol. 28, No. 6, pp.

1497-1503, 1994
Copyright © 1994ElsevierScienceLtd
Pergamon Printed in Great Britain.All rights reserved
0043-1354/94$7.00+ 0.00

B I O F I L M F O R M A T I O N ON CAST I R O N SUBSTRATA IN
W A T E R D I S T R I B U T I O N SYSTEMS
R. M. DONLAN 1., W. O. PIPES 2 and T. L. YOHE3
tCalgon Corporation, Pittsburgh, PA 15017, 2Drexel University, Philadelphia, PA 19104 and
3Philadelphia Suburban Water Company, Bryn Mawr, PA 19010, U.S.A.

(First received October 1991; accepted in revised form .4ugust 1993)

Al~tract--Cast iron substrata were exposed inside drinking water distribution pipes for selected time
intervals and water temperature conditions. The rate of biofilm formation was related to a seasonal effect
and possibly to flow velocity. The water temperature, bulk water heterotrophic plate count and
disinfectant residual were the measured differences between the seasons. The highest rates of biofilm
formation occurred during warm water conditions (15-25°C).

Key words--biofilm, drinking water, cast iron, water temperature, flow velocity, chlorine

NOMENCLATURE Tuovinen and Hsu, 1982; van der Wende et al., 1989)
Reynolds number, Re = L V ' p / u few have actually studied the rate of biofilm for-
Boundary layer thickness, l I = D ~3/ V" Y (u/p)2/3 mation within a drinking water distribution system.
where A possible reason for the paucity of such studies is
Y =empirical value from Fig. 8 in Williamson and that until recently, there were no reliable methods for
McCarty (1976) exposing substrata in these environments. Some stud-
u = viscosity (g cm- i s)
p = density (gcm -3) ies used solid materials from pipe sections or removed
Dw=diffusion coet~cient through water for NH2C1 through fire hydrants (Lechavallier et al., 1987; Nagy
(cm2s-~) and Olson, 1985; Tuovinen et al., 1980), but this
V' = maximum velocity (cm s-~ ) approach did not allow the investigators to study
L = distance from the leading edge or length of test biofilm formation over a period of time. Van der
substratum (cylinder) (cm)
L~ = stagnant liquid layer depth, or boundary layer Wende et al. (1989) used a Roto Torque system to
study biofilm development while Pedersen (1990)
used a biofilm reactor. Both studies passed a side
stream of drinking water through the test device to
INTRODUCTION
examine biofilm formation. Neden et al. (1992) in-
In aqueous systems, bacteria are attracted to surfaces, stalled and utilized underground pipe biofilm moni-
which they readily colonize (Lapin-Scott and Coster- toring stations in drinking water distribution system
ton, 1989; Marshall, 1992). Growth of these bacteria locations and installed 4 in. diameter by 6 in. long
at surfaces is accompanied by production of extra- pipe sections of different materials, including cast
cellular polymers and the term biofilm is used to iron, in each of these stations. Our approach was to
describe colonies of surface adherent microorganisms install replicate cast iron substrata, in the form of a
and their associated excreted extracellular polymers cylinder, directly into the drinking water main
(Marshall, 1992). Biofilms are important in a wide through a corporation valve in order to examine
spectrum of industrially relevant problems, including: biofilm formation under in situ conditions.
dental decay, fouling of ship hulls, biofouling and According to the Water Industry Data Base (Water
microbially influenced corrosion of industrial water Industry Data Base: Utility Profiles Report No.
systems and regrowth in drinking water distribution 20293, American Water Works Association and
systems (Lapin-Scott and Costerton, 1989). AWWA Research Foundation, Denver, Colo.), of
Although a number of studies have examined 1097 water utilities surveyed in the United States (and
biofilms in drinking water systems (Allen et al., 1980; 998 responding), unlined cast iron water mains make
Lechavallier et al., 1987; Nagy and Olson, 1985; up 22.2% of the total miles of pipe in service. The
Neden et ai., 1992; Nix and Bratton, 1986; Olson Philadelphia Suburban Water Company (PSWC)
et al., 1981; Pedersen, 1990; Tuovinen et al., 1980; has approx. 860 miles of unlined cast iron pipe
in service in its distribution system. Cast iron was
*Author to whom all correspondence should be addressed. chosen as the test substratum in this study because of

wR~/c,-P 1497
1498 R. M. DONLAN et al.

its extensive use by P S W C a n d by the water industry Sampling protocol

in general. Cylinders were cleaned in a detergent solution, rinsed first
The objective of this study was to determine bio- in tap water, then deionized water and dried prior to use.
Immediately prior to connection of the CSD to the corpor-
film f o r m a t i o n on cast iron s u b s t r a t a in order to ation stop, the cylinders and all CSD components which
u n d e r s t a n d what factors m i g h t relate to biofilm for- would be in contact with the water inside the water main
m a t i o n on the walls o f cast iron drinking water pipes were disinfected with a strong sodium hypochlorite solution.
within drinking water distribution systems. Exposed cylinders were aseptically removed from the
CSD and placed inside a "whirl pak" bag (Nasco Inc., Fort
Atkinson, Wis.). In the laboratory each cylinder was held in
MATERIALS AND METHODS place directly over a 22 x 250 mm sterilized glass test tube.
Attached microorganisms were removed from cylinder inner
Description of water systems surfaces using an alcohol sterilized scalpel followed by
The experimental part of this investigation was performed rinsing with 25 ml of sterile site water using a Cornwall
in the Philadelphia Suburban Water Company system which syringe (Becton Dickinson Co., Rutherford, N.J.). The site
provides potable water to 62 municipalities within a 339 water was dechlorinated with 80 mg l- t sodium thiosulfate
square mile territory north and west of the city of Philadel- and filter sterilized by passage through a 0.2/~ porosity
phia, Pa. The principal sources of supply are four rural polycarbonate membrane filter (Nuclepore Corp., Pleasan-
streams which are tributaries to the Schuylkill and Delaware ton, Calif.) prior to use. The 25 ml cell suspension was
Rivers. The company also has a 30 million gallon per day refrigerated for no more than a few hours prior to further
allocation from the Schuylkill River. The company main- analysis. The cell suspension was mixed using a Vortex
tains: five impounding reservoirs with a combined storage Mixer (Model K-550-G, VWR Company, Philadelphia, Pa)
capacity of over 9 billion gallons, a l billion gallon ground at a speed of 4-5 for 3 min. It was then ready for subsequent
water recharged reservoir and 35 active deep wells. For the dilution and analysis. Population densities of attached
purpose of this study the sampling sites were chosen that microorganisms were determined using the heterotrophic
received water from three different surface water treatment plate count procedure (HPC). All analyses were performed
plants. Raw water treatment consisted of disinfection, co- in duplicate using the spread plate method and R2A
agulation, flocculation, powdered activated carbon addition medium (Reasoner and Geldreich, 1979), with incubation at
and dual media rapid sand filtration. Post-filtration treat- 20°C for 7 days. Water samples were collected through a
ment included lime addition (pH adjustment) and the copper or plastic sample line which was connected directly
addition of bimetallic phosphate for corrosion control. The to the main near the corporation sampling device installa-
effluent total chloramine concentration ranged from 1.0 to tion point. Sample lines were flushed for approx. 1 min prior
2.5 mg 1-~ over the study period. to sample collection. All samples were tested for: tempera-
ture, total chlorine concentration, heterotrophic plate count
Description of sampling device (HPC), pH, alkalinity, organic nitrogen, ammonia nitrogen,
nitrite nitrogen, nitrate nitrogen, total and ortho phosphate
Hollow cast iron cylinders (15 mm i.d. × 10mm length)
and total organic carbon. Both water temperature and total
were exposed inside water mains at selected distribution
system locations using a Corporation Sampling Device chlorine concentration (LaMott DPD test kit, LaMott
Chemical Co., Chestertown, Md) were determined on site.
(U.S. Patent 4,631,961). This Corporation Sampling Device
(CSD) has been described previously (Donlan and Pipes, HPC samples were collected in sterile plastic bottles.
1988). Samples for chemical analysis were collected in acid-washed
plastic bottles, iced and returned to the laboratory for
This device allows insertion of four cast iron cylinders
analysis. Heterotrophic plate counts on water samples were
into a 1 in. drinking water main through a 1 in. corporation
determined in the same way as cylinder samples using R2A
valve as shown in Fig. 1. The cylinders are oriented to the
direction of flow and after the designated exposure period medium.
All analyses were performed according to Standard
the device is removed and the microorganisms attached to
cylinders are quantified. Methods (APHA, 1985) with two exceptions. Nitrate was
analyzed using a Dionex Model 2020i ion chromatograph
(Dionex Corporation, Sunnyvale, Calif.) using a U.S. EPA
Description of sampling sites approved method (The determination of inorganic anions in
The water at each of the four installation sites differed water by ion chromatography--Method 300.0, U.S. En-
from the other sites in one or several of the following vironmental Protection Agency, Cincinnati, Ohio). Total
characteristics: flow velocity, chloramine concentration, het- organic carbon was determined using a Dorman Model
erotrophic plate count (HPC) and/or solute composition. DC80 TOC Analyzer (Dorman, Santa Clara, Calif.).
These site characteristics are shown in Tables 1, 2 and 3. The Velocity at each sampling site was measured with a pitot
data in Tables 1 and 2 are based upon water samples which rod device and manometer using the procedure described by
were collected regularly during a designated exposure the manufacturer [F. B. Leopold Co. (1982) Bulletin No.
period. The number of samples collected for both hetero- 1301-5022R., F. B. Leopold Co., Zelionople, Pa]. The pitot
trophic plate count and total chlorine concentration was rod was inserted into the water main through the corpor-
designated as N. The mean values were determined based ation stop valve used for CSD installation.
upon this number of samples. The value for SD, or standard
deviation, was a measure of the variability in HPC or
chlorine values collected over the entire exposure interval, QUALITY CONTROL
rather than a determination of variability for a specific
A control for the cylinder sampling procedure was
sample. In order to investigate temperature effects, cylinders
were exposed under either "warm water" or "cold water" provided by placing cast iron cylinders inside C S D
regimes. The period of time during which water tempera- units, t r a n s p o r t i n g them to the exposure site, disin-
tures were between 15 and 26°C was designated the "warm fecting and then rinsing each with sterile deionized
water" period and it occurred between June and November water before returning them to the laboratory a n d
1985. The period of time during which water temperatures
were between 4 and 15°C was designated the "cold water" analyzing as normal. All microbiological analyses
period and it occurred between January and April 1985 and were performed in duplicate. In addition, media a n d
January and April 1986. m e t h o d sterility controls were run with each set of
Biofilm formation in water systems 1499

O ~

"r,~"

06

ii

o.-q

•~ - f..)

• ×
1500 R. M. DONLANet al.

Table 1. Total chlorine concentrations measured on water collected from CSD

sampling sites during warm water (June-November)and cold water (January-April)
periods
Warm water total chlorine Cold water total chlorine
concentration (mg I- i) concentration (mg 1- i )
Sampling site N Mean SD N Mean SD
Bryn Mawr 17 0.89 0.18 14 1.20 0.12
Berwyn 12 0.22 0.09 13 1.03 0.13
Lansdowne 13 0.10 0.09 13 0.77 0.13
Cheltenham 12 0.09 0.06 12 0.30 0.34

microbiological analyses. All D P D measurements The 30 to approx. 80 day exposure period, the
were the consensus of two individuals. Water chemi- longest exposure periods for the "cold water" study
cal analyses were subject to the Philadelphia Subur- (Fig. 3), occurred during March and April. Figure 4
ban Water C o m p a n y laboratory quality assurance shows that water temperatures were rising substan-
guidelines for precision and accuracy. tially during this period. As water temperatures
began to climb during this period, biofilm continued
RESULTS to increase.
In order to determine the effect of temperature
Figures 2 and 3 show biofilm formation data (log during "cold water" conditions on biofilm formation
cfu cm -2) for cast iron cylinders exposed at the Bryn rate, rate data (No. cfu c m - 2 d - 1) from each of the
Mawr, Berwyn, Lansdowne and Cheltenham four sites was plotted as a function of water tempera-
sampling sites under warm water and cold water ture measured at the C S D sampling site during that
regimes, respectively. Each data point represents the specific exposure period. These data are shown in
mean count obtained for each exposure interval and Fig. 5, which plots rate of biofilm formation (log No.
is based on the analysis of three cylinders. Relative cells c m - 2 d -1) as a function of average water tem-
standard deviations (standard deviation as a percent perature for each of the four sites. It shows that rate
of the mean) for all data points were always less than of biofilm formation increases substantially as water
or equal to 5% except when the mean values were less temperatures climb.
than I0 bacteria cm -2. In these circumstances the N o n e of the other water quality parameters
data were much more variable, due to the low measured (pH, alkalinity, organic nitrogen, ammonia
numbers of cells being quantified. Under the " w a r m nitrogen, nitrite nitrogen, nitrate nitrogen, total phos-
water" exposure regime, the Bryn Mawr, Lansdowne phate, ortho phosphate and total organic carbon)
and Cheltenham sites exhibited a very similar amount appeared to correlate with either rate or extent of
of biofilm which approached steady state after 28 biofilm formation.
days and did not increase significantly after this time.
Biofilm at the Berwyn site also approached a steady DISCUSSION
s t a t e after 28 days though the extent of formation was
approx. 1% of the amount formed at the other three Seasonal effects
sites. This indicates for all four sites, that the The observed decrease in biofilm formation rate
difference between cell accumulation/cell growth and extent which occurred under "cold water" con-
and cell death/cell loss due to sloughing approached ditions compared to similar exposure periods under
zero after approx. 30 days for " w a r m water" " w a r m water" conditions may be explained in part
conditions. from the data in Table 1. At each of the four sites,
The "cold water" and " w a r m water" colonization the mean total chlorine concentrations were higher
curves were significantly different at all four sites. (by a factor of 1.3-7.7) under "cold water" conditions
Rates of biofilm formation were much lower during than under " w a r m water" conditions. It is possible
"cold water" exposure for a comparable exposure that increased monochloramine concentrations could
period (cf. Fig. 3 with Fig. 2). have, at least in part, limited biofilm formation.

Table 2. Bulk water heterotrophic plate counts measured on water collected

from CSD sampling sites during warm water (June-November) and cold water
(January-April) periods
Warm water HPC Cold water HPC
(cfu m l i ) (cfu ml- i )
Sampling site N Mean SD N Mean SD

Bryn Mawr 17 870 933 13 214 7.90

Berwyn 12 794 446 I1 15.1 31.0
Lansdowne 13 2000 1950 12 95.5 79.0
Cheltenham 12 3890 4070 ll 380 398
 Biofilm formation in water systems 1501

Table 3. Comparisonof flowdynamicsat each of the four sampling The higher Reynolds number at Berwyn (Table 3)
sites
also points to more turbulent water flow patterns
Mean center Reynolds Boundarylayer
Samplingsite velocity(era s-I ) number thickness(/~m) over the test surfaces at this site. This would have the
effect of diminishing the mean residence time for the
Bryn Mawr 9.0 898 60
Berwyn 29.I 2904 30 cell at the metal surface. Fletcher (1990) described
Lansdowne 9.3 928 60 mean residence time as the difference between the
Cheltenham 1.2 120 182 attractive forces between the cell and the surface and
the fluid shear. If the time required for cell adhesion
Heterotrophic plate counts were also lower during to the surface is greater than the mean residence time,
"cold water" conditions at each site (Table 2). And the cell will be washed away prior to attachment. This
it would be expected that the number of planktonic has been documented experimentally by others
cells would correlate directly with rate of biofilm (Characklis, 1990a). It is possible then that shorter
formation. It would also be expected that water mean residence times at the Berwyn site could have
temperature alone would exert a direct effect on resulted in lower rates of attachment to the metal
biofilm formation, due to effects on bacterial growth surface. Also, higher shear forces would be expected
and metabolism. to cause greater erosion of cells from the developed
Flow-related effects biofilm (Characklis, 1990b).
A comparison of the biofilm formation data in
In this water system, flow dynamics may influence Fig. 2 and 3 indicates that under "cold water"
biofilm formation or detachment either through conditions, the discrepancy between sites (ex. Bryn
affecting mass transfer of monochloramine to the Mawr and Berwyn) for designated exposure interval
biofilm surface, or by the direct action of shear forces (ex. 60 days) is not nearly so great as for the same
on cell adhesion. Table 3 presents flow related infor- period under "warm water" conditions. This leads to
mation for each of the four sampling sites. Berwyn, the hypothesis that seasonal factors (temperature,
the site with significantly lower biofilm formation monochloramine concentration, etc.) are the primary
rates and extent had the highest Reynolds number limiting factors for net cell accumulation. Mono-
(2904) and the thinnest boundary layer (30 mi- chloramine concentrations were significantly lower in
crometers). It would be expected that thinner bound- warm water periods apparently due to higher chlorine
ary layers would allow more rapid diffusion of demand during this period. When seasonal factors
monochloramine to the surface of the biofilm. are not limiting for a given site (under "warm water"
Though Bryn Mawr had the highest average mono- conditions), then flow related variables appear to be
chloramine concentrations under warm water con- limiting.
ditions, Berwyn had the lowest biofilm values. The
thinner boundary layer at Berwyn might partly ex-
plain this discrepancy. Berwyn did have lower bulk Other studies in drinking water systems
water heterotrophic plate counts than the other three We are not aware of other studies which have
sites (Table 2) and this could be another reason for investigated biofilm formation rate on cast iron sub-
lower biofilm values at Berwyn. strata in drinking water distributions systems.

9F
$ - , I

.7-6
7
.......................... t ....................... t
4 5
t
6 4
z .....
O

,..1 3 :i
ii
2 a Bryn Mawr ..... • ..... Lansdowne
':t
- - - o - - - Berwyn - - . - o . - - - Cheltenham
1

I I I I t I
20 4O 60 8O 100 120
Time of exposure (days)
Fig. 2. Biolfilm formation on cast iron under warm water conditions.
1502 R. M. DONLANet al.

• Bryn Mawr ..... • ..... Lansdowne

--- o.-- Berwyn --.-s..-- Cheltenham

~5

,d 3

I I
20 40 60 80 100 120

Time of exposure (days)

Fig. 3. Biofilm formation on cast iron under cold water conditions.

Pedersen (1990) examined biofilm formation on stain- CONCLUSIONS

less steel and PVC surfaces in drinking water. Biofilm
formation on stainless steel (free chlorine re- (1) Biofilms reached steady state within 30 days on
sidual = 0.1 mg 1- l, velocity = 10 cm s- l) was much cast iron substrata under warm water conditions
slower than at any of the sites which we examined. He (15-25°C) and low monochloramine concentrations
determined 4 months to be a period suitable for (0.09-0.89 mg l - l ) .
development of a biofilm on stainless steel compared (2) Biofilm accumulation rate was greatest during
to approx. 1 month on cast iron in our studies. warm water conditions (15-25°C). Based on the
Possible explanations for this discrepancy are the parameters measured, this effect may be due to
difference in metallurgy or the differences in the type increased temperature, lowered disinfectant concen-
of disinfectant present (free chlorine compared to tration, increased bulk water cell count, some
monochloramine in our studies). Free chlorine would other unmeasured factor or to a combination of all
be expected to be more effective than monochlora- three.
mine for suppressing biofilm development on stain- (3) Flow rates may also affect biofilm accumu-
less steel since it would not be inactivated by iron lation rates. Flow-related effects were most pro-
corrosion products (Lechavallier et al., 1990). nounced under warm water conditions (15-25°C).

3 0 [-- ~ Bryn Mawr Warm --- -- Lansdowne Warm

25 I-
I ...... Berwyn Warm

./.
- - - Cheltenham

-~.i~'.~.__.,~.\
Warm

g 20 ." 7 . 7 . 7 : . . . . . ""-. . ." ~ .

.--~
15 - - Bryn Mawr Cold ----- Lansdown¢ Cold ~"-..

...... Berwyn Cold - - - Cheltenham Cold ~ ~

10

I I I I I I I I I I I I I I I I
1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170

Period of exposure (days)

Fig. 4. Water temperature for each CSD sampling site during the entire study.
 Biofilm formation in water systems 1503
6.0 -

5.5 - Bryn Mawr DO

D
5.0 -
Berwyn
A
4.5
Lansdowne t].

'~ 4.0 O

¢q Cheltenharn
i
3.5

-- 3.0

6 2.5
Z
2.0
A O
[]
1.5 -
A
1.0 - LX
Q [] O
0.5
O
I I @ ~ I*1 I I I I I I I I I L I I L L I I I I I
l 2 3 4 5 6 7 8 9 l0 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Average water temperature (°C)

Fig. 5. Rate of biofilm formation vs average water temperature.

Acknowledgements--This work was performed while R. M. Marshall K. C. (1992) Biofilms: a overview of bacterial
Donlan was employed by the Philadelphia Suburban Water adhesion, activity, and control at surfaces. A S M News
Company (PSWC). The authors wish to acknowledge the 58, 202-207.
generous support of this research by the management of Nagy L. A. and Olson B. H. (1985) Occurrence and
PSWC. We are also indebted to Thomas Gittelman and significance of bacteria, fungi, and yeasts, associated
Edward Crist for providing field and laboratory support and with distribution pipe surfaces. In Proceedings Water
Steven Galante for reviewing the Reynolds number and Quality Technology Conference; Advances in Water
boundary layer calculations. Analysis and Treatment, Houston, Tex., pp. 213-238.
Neden D. G., Jones R. J., Smith J. R., Kirmeyer G. J. and
Foust G. W. (1992) Comparing chlorination and chlo-
REFERENCES ramination for controlling bacterial regrowth. J. Am.
Wat. Wks Ass. 84, 80-88.
Allen M. J., Taylor R. H. and Geldreich E. E. (1980) The Nix P. G. and Bratton G. (1986) A coliform equivalent
occurrence of microorganisms in water main encrusta- technique to assess the microbiological quality of drink-
tions. J. Am. Wat. Wks Ass. 72, 614-625. ing water. In Proceedings American Water Works Associ-
APHA (1985) Standard Methods for the Examination of ation Water Quality Technology Conference, Portland,
Water and Wastewater, 15th edition. American Public Ore., pp. 291-304.
Health Association, Washington, D.C. Olson B. H., Ridgeway H. F. and Means E. G.
Characklis W. G. (1990a) Microbial fouling. In Biofilms (1981) Bacterial colonization of mortar lined and
(Edited by Characklis W. G. and Marshall K. C.), galvanized iron water water distribution mains. In
pp. 523-584. Wiley, New York. Proceedings American Water Works Association
Characklis W. G. (1990b) Biofilm processes. In Biofilms National Conference, Denver, Colo., pp. 1027-1039.
(Edited by Characklis W. G. and Marshall K. C.), Pedersen K. (1990) Biofilm development on stainless steel
pp. 195-231. Wiley, New York. and PVC surfaces in drinking water. Wat. Res. 24,
Donlan R. M. and Pipes W. O. (1988) Selected drinking 239-243.
water characteristics and attached microbial population Reasoner D. J. and Geldreich E. E. (1979) A new medium
density. J. Am. Wat. Wks Ass. 80, 70-76. for enumeration and subculture of bacteria from potable
Fletcher M. (1990) Methods for studying adhesion water. Appl. envir. Microbiol. 49,1 1-7.
and attachment to surfaces. In Methods in Micro- Tuovinen O. H. and Hsu J. C. (1982) Aerobic and anaero-
biology, Vol. 22, Techniques in Microbial Ecology bic microorganisms in tubercules from the Columbus,
(Edited by Grigorova R. and Norris R.), pp. 251-283. Ohio water distribution system. Appl. envir. Microbiol.
Academic Press, London. 44, 761-764.
Lappin-Scott H. M. and Costerton J. W. (1989) Tuovinen O. H., Button K. S., Vuorinen A., Carlson L.,
Bacterial biofilms and surface fouling. Biofouling 1, Mair D. M. and Yut L. A. (1980) Bacterial, chemical,
323-342. and mineralogical characteristics of tubercules in distri-
Lechavallier M. W., Babcock T. M. and Lee R. G. bution pipelines. J. Am. Wat. Wks Ass. 72, 626-635.
(1987) Examination and characterization of distri- van der Wende E., Characklis W. G. and Smith D. B. (1989)
bution system biofilms. Appl. envir. Microbiol. 53, Biofilms and bacterial drinking water quality. Wat. Res.
2714-2724. 23, 1313-1322.
Lechavallier M. W., Lowry C. D. and Lee R. G. (1990) Williamson K. and McCarty P. L. (1976) A model of
Disinfecting biofilms in a model distribution system. substrate utilization by bacterial films. J. War. Pollut.
J. Am. Wat. Wks Ass. 82, 87-99. Control Fed. 48, 9-24.