&EPA
United States
Environmental Protection
Agency
EPA/600/R-10/005 I March 2010 I www.epa.gov/ord
Detection of Biological
Suspensions Using Online
Detectors in a Drinking Water
Distribution System Simulator
Office of Research and Development
National Homeland Security Research Center
-------
Page Intentionally Blank
-------
EPA/600/R-10/005 | March 2010 www.epa.gov/ord
Detection of Biological Suspensions
Using Online Detectors in a Drinking
Water Distribution System Simulator
United States Environmental Protection Agency
National Homeland Security Research Center
Cincinnati, OH 45268
Office of Research and Development
National Homeland Security Research Center, Water Infrastructure Protection
-------
Page Intentionally Blank
-------
Table of Contents
Disclaimer vi
Executive Summary ix
Introduction 1
Experimental Design 3
Single Pass Distribution System Simulator 3
Water Quality Sensors 3
Contaminants 3
Evaluation of Water Quality Sensor Response 4
Results and Discussion 5
Results of Sensor Response Experiments 5
Sievers® 900 On-Line Total Organic Carbon Analyzer and Hach CL17 Free Chlorine Analyzer 5
s::can Spectra: :lyser™ and Hach FilterTrak™ 660 sc Laser Nephelometer 5
Real Tech Real UVT Online 9
Fluid Imaging Technologies FlowCAM® 9
JMAR Technologies BioSentry® 9
Discussion and Significance of Sensor Response Results 10
Sensor Cost Considerations 10
Conclusions and Future Work 13
References 15
-------
Disclaimer
The U.S. Environmental Protection Agency through its Office of Research and Development funded and collaborated in the
research described herein under contract number EP-C-04-034 with Shaw Environmental and Infrastructure, Inc. It has been
subject to an administrative review but does not necessarily reflect the views of the Agency. No official endorsement should be
inferred. EPA does not endorse the purchase or sale of any commercial products or services.
Questions concerning this document or its application should be addressed to:
Jeffrey Szabo
National Homeland Security Research Center (NG-16)
Office of Research and Development
United States Environmental Protection Agency
26 W. Martin Luther King Dr.
Cincinnati, OH 45268
(513)487-2823
szabo.jeff@epa.gov
or
John Hall
National Homeland Security Research Center (NG-16)
Office of Research and Development
United States Environmental Protection Agency
26 W. Martin Luther King Dr.
Cincinnati, OH 45268
(513)487-2814
hall.john@epa.gov
-------
List of Tables and Figures
Table 1: Water quality sensors 4
Table 2: Sensor response to contamination reported as absolute change (top), percent change (middle)
and signal-to-noise ratio (bottom) 6
Table 3: Formulation for terrific broth and sporulation media 9
Table 4: Sensor detection and purchase data 11
Figure 1: Sensor baseline and response to contamination with Bacillus globigii spores at four inch pipes densities ... 7
Figure 2: Sensor baseline and response to contamination with Escherichia coli K-12 at four inch pipes densities 8
Figure 3: Multiple angle light scattering device response to Escherichia coli K-12 and Bacillus globigii 10
-------
List of Acronyms
ATCC American Type Culture Collection
CWS Contamination Warning System
DPD N, N-diethyl-p-phenylenediamine
GAC Granular Activated Carbon
GCWW Greater Cincinnati Water Works
GPM Gallons per Minute
HSPD Homeland Security Presidential Directive
MALS Multiple Angle Light Scattering
MSD Metropolitan Sewer District
NTU Nephelometric Turbidity Unit
TOC Total Organic Carbon
US EPA United States Environmental Protection Agency
UV Ultraviolet
-------
Executive Summary
Detection of relatively low density microbial suspensions
(less than 105 cfu/mL) was evaluated with a suite of online
water quality sensors and instruments. Typically, the drinking
water industry uses online sensors to measure parameters
such as free chlorine, pH, and conductivity; however, in
microbial suspensions below 105 cfu/mL, the sensors have a
weak response or are ineffective. Therefore, sensors designed
to detect either particulates in water or organic compounds
that might accompany microbial suspensions (i.e., culture
media) were evaluated for their ability to detect low density
microbial suspensions.
Evaluation took place in a pilot scale drinking water
distribution system simulator (DSS) with sensors attached
through a slip-stream. Technologies investigated were
the Fluid Imaging Technologies FlowCAM®, Hach
FilterTrak™ 660 sc Laser Nephelometer, JMAR BioSentry®,
Real Tech Inc., Real UVT Online and the S: :CAN
Spectro::lyser™.
Biological suspensions were injected into the DSS and
sensor responses were compared to stable baseline values
before injection. The JMAR BioSentry® detected the least
dense biological suspension (600 cfu/ml) while the S::CAN
Spectro::lyser™ and Hach FilterTrak™ 660 sc Laser
Nephelometer performed as well as the online free chlorine
and TOC analyzers (2.5xl04 cfu/ml). The results were
determined by selecting the biological densities that elicited
an obvious visual response from the sensor. However, it
should be noted that changes not obvious to the naked eye
could be detected with event detection algorithms. Operation
and maintenance costs of the sensors are minimal, but
some have high capital costs that must be considered when
weighing their detection ability.
-------
Page Intentionally Blank
-------
Introduction
Detecting contamination in drinking water distribution
systems has challenged water utilities for many years. Water
utilities have traditionally focused on intrusion from incidents
such as pipe breaks or plant malfunctions as the primary
source of contamination, but in recent years utilities have also
considered issues associated with intentional contamination.
Contaminant-specific sensors show promise for detecting
contamination, but their use is limited by factors such as cost,
ease of use, commercial availability and limited acceptability
by the drinking water community. Furthermore, the large
number of contaminant-specific sensors needed to detect the
universe of potential contaminants makes their use inefficient.
The drinking water community has explored the use of
common water quality sensors to help detect contamination
(Allmann et al., 2005; Byer et al., 2005; Kessler et al.,
1998; King et al., 2005; Kroll et al., 2005; Magnuson et al.,
2005). In response to Homeland Security Presidential
Directives (HSPD) 7 and 9, the United States Environmental
Protection Agency (USEPA) has undertaken research using
a multifaceted contamination warning system (CWS) for
drinking water distribution systems, of which online water
quality monitoring is one component (USEPA, 2005a; 2006;
2007; Hall et al., 2007; Szabo et al., 2006; HSPD 7, 2003;
HSPD 9, 2004). Research efforts have remained focused on
commercially available online water quality sensors, since
they offer the dual benefits of water quality data and potential
detection of contamination.
Detecting biological suspensions at low concentrations (less
than 105 cfu/mL) in drinking water has proven challenging.
Previous research has shown that biological suspensions,
injected with the growth media in which they were cultured,
will affect parameters like free chlorine and total organic
carbon (TOC) in chlorinated water (Hall et al., 2007;
USEPA, 2007). This is due to the nutrient media reacting with
and reducing free chlorine and organic carbon in the broth,
which increases TOC. If biological suspensions are washed
and injected without growth media, or injected into a large
volume where the growth media is highly diluted, standard
water quality parameters show little noticeable response.
Therefore, it is sensible to study less common sensors to
determine whether they can detect biological suspensions at
lower densities than standard sensors. This paper describes
detection studies with a suite of specialized water quality
monitors and their response to biological contamination at
1 x 102-2.5 x 104 cfu/ml. As-tested cost information is also
included, which will hopefully provide perspective to any
water utility or other user looking to employ the tested water
quality monitors.
-------
Page Intentionally Blank
-------
Experimental Design
Single Pass Distribution System Simulator
The drinking water distribution system simulator used in this
study was described in Yang et al. (2008). A drinking water
distribution pipe was represented using a once-through (or
single pass) pipe at USEPA's Test and Evaluation (T&E)
Facility in Cincinnati, Ohio. The pipe consisted of 1,200 feet
of 3-inch diameter fiberglass-lined ductile iron. Experiments
were conducted at 22 gallons per minute (gpm), which
corresponds to an average velocity of 1 foot per second (ft/
sec) in the pipe. This flow rate will produce turbulent flow
(Reynolds number approximately 26,000) in the relatively
smooth pipe. Although the pipe was lined with fiberglass,
sections have chipped away, exposing ductile iron. These
sections were heavily corroded and were more representative
of an iron drinking water pipe than the lined sections. Note
that English standard units, commonly used by the U.S. water
utility personnel, have been used throughout this report. For
example, volume is reported in U.S. gallons and velocity
in feet per second (ft/s). However, in keeping with industry
usage, contaminant concentrations are reported in metric
units, in milligrams per liter (mg/L).
Chlorinated tap water was introduced directly from the
Greater Cincinnati Water Works (GCWW) distribution
system into a 750 gallon storage tank where it was
fed by gravity into the 3-inch pipe system. An air gap
was maintained between the GCWW system and this
experimental setup to ensure that there was no back flushing
of the injected contaminant. Free chorine was generally
1.0 +/- 0.1 mg/L, with temperature ranging from 10° to 30° C
depending upon the season. Turbidity was 0.1 nephelometric
turbity units (NTU) or less throughout the year. The water
fed from the 750 gallon overhead tank provided a 10 to 12
pounds per square inch (psi) inside the pipe. Contaminant
injections were performed for 20 minutes by injecting a
10 L mix of contaminant in chlorine-free granular activated
carbon (GAC) filtered tap water at the rate of 0.5 L/min.
Contaminant concentration in the pipe was varied by altering
the amount of contaminant mixed in the 10 L volume.
Control experiments were performed by injecting 10 L of the
GAC filtered tap water without the contaminant at the same
injection rate.
Water Quality Sensors
A suite of sensors measured water quality at 80 feet from the
injection point through a slipstream. Past experimental results
indicated that of the standard water quality parameters,
free/total chlorine and TOC were the best at indicating
contamination in chlorinated tap water (USEPA, 2006a;
2005b). Total chlorine was measured using a Hach CL17
Total Chlorine Analyzer (Hach Company, Loveland, Colo.),
which uses the N, N-diethyl-p-phenylenediamine (DPD)
colorimetric method (Standard Methods, 1998). TOC was
measured using a Sievers® 900 On-Line Total Organic
Carbon Analyzer (GE Analytical Instruments, Boulder, Colo.)
The operation and maintenance of these sensors is described
in Hall etal. (2007).
The following specialized water quality sensors were
evaluated in this study to determine if their response was
better than the standard TOC and chlorine sensors listed
above:
• A laser turbidimeter, the Hach FilterTrak™ 660 sc
Laser Nephelometer (Hach Company, Loveland,
Colo.), was used as an enhanced turbidity monitor. It
operates similarly to a standard turbidimeter, except that
it employs a laser nephelometer, which yields better
resolution.
• A BioSentry® Water Monitoring system (JMAR
Technologies, San Diego, Calif.) represented a laser
based multiple-angle light scattering (MALS) device. In
addition to detecting changes in the number of particles
in water, it has the ability to classify microorganisms
using their unique bio-optical MALS signatures, but this
information was not specifically analyzed in this study.
• Estimates of TOC and turbidity were obtained using
a 100 mm s::can Spectro::lyser™ (s::canMeBtechnik
GmbH, Vienna, Austria). The spectro::lyser operates
using ultraviolet-visible (UV-Vis) absorption
spectrometry in the 200 to 750 nm range.
• Continuous ultraviolet light transmission at UV 254
nm wavelength (UV254) was made with a Real UVT
Online monitor (Real Tech, Inc., Whitby, Ontario,
Canada). Similarly to the spectro::lyser, changes in
UV absorption by aromatics or other light absorbing
compounds in the microbial suspension could be
detected by this device.
• An online flow-cytometer and microscope called
FlowCAM® (Fluid Imaging Technologies, Yarmouth,
Maine) was used as a digital imaging microscope-based
particle detector. Particles are channeled through a flow
cell where they are digitally imaged and can be counted.
Contaminants
Two microorganisms were used in this study. Escherichia
coli K-12 (ATCC 25204) (E. coli) was used as representative
vegetative bacteria. Bacillus globigii (B. globigii)
(obtained from the US Army's Dugway Proving Ground,
Dugway Proving Ground, Utah) was used in spore
form and was considered to be a representative spore-
forming bacterium. Storage and culturing methods
are described in detail in Szabo et al. (2007) and Hall
et al. (2007) for B. globigii and E. coli, respectively.
Wastewater (secondary effluent) was also injected
and was obtained from the Cincinnati Metropolitan
Sewer District (MSD) Mill Creek treatment plant.
-------
Table 1: Water quality sensors
BioSentry®
FlowCAM®
Hach CL17 Free Chlorine Analyzer
Hach FilterTrak™ 660 sc Laser Nephelometer
Real UVT Online
Sievers® 900 On-Line Total Organic Carbon
Analyzer
Spectro::lyser™
JMAR Technologies, Inc., San Diego, California
Fluid Imaging Technologies, Yarmouth, Maine
Hach Company, Loveland, Colorado
Hach Company, Loveland, Colorado
Real Tech, Inc., Whitby, Ontario, Canada
GE Analytical Instruments, Boulder, Colorado
scan Messtechnik GmbH, Vienna, Austria
http://www.jmar.com
http://www.fluidimaging.com
http://www.hach.com
http://www.hach.com
http://www.realtech.ca
http://www.geinstruments.com
http://www.s-can.at
Evaluation of Water Quality Sensor Response
Sensor response to contamination was evaluated by
calculating the absolute and percent change from a stable
baseline to the peak value recorded as the contaminant passed
the sensor. Baseline values were calculated by averaging
the sensor signal over a one hour period before contaminant
injection, with baseline noise represented by standard
deviation. Sensors were polled every minute during test runs,
so 60 pre-injection data points were used for determining
baseline mean and standard deviation. Contamination
injections were performed in duplicate, and results are
presented as the average of those duplicates (see Table 2 in
the results section).
Evaluating the data by calculating the percent change yields
a good system specific response of water quality parameters
to contaminants. However, percent change may be different
in systems that have different water quality. For example, if
a contaminant injected into water with 1 mg/L free chlorine
decreases the chlorine concentration to 0.9 mg/L, a 10%
reduction has occured. If the same contaminant injected at
the same concentration in water with 2 mg/L free chlorine
consumes the same amount of free chlorine, a 5% reduction
has occured. Therefore, sensor response is also characterized
as a signal-to-noise ratio. The maximum absolute change
(baseline to peak) recorded during injection was normalized
by the baseline standard deviation. Analyzing data with
the signal-to-noise approach illustrates sensor response as
the magnitude of the water quality change relative to the
variation in the baseline before injection.
The time period when the injected contaminant was in
contact with the sensors was determined based on the flow
rate and injection duration. Injections were 20 minutes long
and flow velocity was 1 ft/sec, so the injection reached the
80 ft sensor station 1.3 minutes after injection and continued
passing the sensors for 20 minutes. Sensor responses usually
lasted longer than 20 minutes at this station due to dispersion,
which elongated the contaminant plume in the pipe. Peak
sensor responses were taken from the time periods when the
contaminants were in contact with the sensors.
Although water quality sensors typically respond within
seconds of water quality change, the Hach CL17 and
Sievers® 900 On-Line Total Organic Carbon Analyzer used
for this experiment run on cycles of 2.5 and 8 minutes,
respectively. These instruments were polled every minute,
but only returned new values at the end of their cycles.
Still, new values were returned frequently enough that the
changes in water quality were seen for both devices while the
contaminant was passing the sampling point.
-------
Results and Discussion
Results of Sensor Response Experiments
Table 2 summarizes the response of the standard
and specialized sensors to various cell densities (or
concentrations) of E. coli and B. globigii. Time series plots
of some of the specialized sensor responses are presented
for B. globigii and E. coli in Figures 1 and 2, respectively.
These plots show the response of duplicate experiments
in sequence, as well as the baseline data that precedes the
injection.
Sievers® 900 On-Line Total Organic Carbon Analyzer
and Hach CL17 Free Chlorine Analyzer
Total chlorine measured by the Hach CL17 decreased
by 0.04 and 0.08 mg/L upon addition of B. globigii and
E. coli, respectively, at 2.5xl04 cfu/mL. The response of
the JMAR Biosentry® was much larger at this cell density,
but the chlorine response is comparable to the s::can and
laser turbidimeter at 2.5xl04 cfu/mL for both microbial
suspensions. TOC measured by the Sievers® 900 changed by
0.16 and 0.41 mg/L for B. globigii andE. coli, respectively,
at 2.5xl04 cfu/mL. The signal-to-noise ratio of 13.7 for
E. coli at 103 may indicate a change, but no visual change
was noticeable below this level. Sievers® 900 TOC and Hach
CL17 Total Chlorine results confirm what has been reported
in the past: there is a significant decrease in the response of
both instruments when the cell density (or concentration) of
injected microorganisms decreases below 2.5xl04 cfu/mL.
It is important to note that the Sievers® 900 TOC and Hach
CL17 Total Chlorine response occurs only when B. globigii
and E. coli are injected with their respective sporulation and
growth media. If these growth media were washed away
or sufficiently diluted, the Hach CL17 Total Chlorine and
Sievers® 900 TOC sensors show little or no response. Signal-
to-noise values for the Sievers® 900 TOC analyzer are higher
than the other instruments below 2.5xl04 cfu/mL because
TOC baseline was stable during these tests.
s::can Spectro::lyser™ and Hach FilterTrak™ 660 sc
Laser Nephelometer
The response for the s::can spectro::lyser's™ turbidity
measurement channel is the same for each injected density
and provides little indication that a contaminant is present.
The percent change values are misleading since the average
value of turbidity is low, but the data are noisy. When the
peak during injection is compared to the mean, it appears
that a large change has taken place. However, when signal-
to-noise is calculated, the response is one to two, indicating
that the peak is close to the standard deviation in the baseline.
Figures 1 and 2 confirm that visually distinguishing a
turbidity response is difficult.
-------
Table 2: Sensor response to contamination reported as absolute change (top), percent change (middle)
and signal-to-noise ratio (bottom)
Injected Agent
E. coli
(in Terrific Broth)
B. globigii
(in sporulation media)
Wastewater (Secondary
Effluent)
Control Blank
(Dl Water)
Pi
uuMieiiiiduuM
(cfu/mL) (mg/L) (mg/L) (mg/L) (FTU) (mNTU) (counts) (#/mL) (UVA)
1.0E+02
6.0E+02
1 .OE+03
2.5E+04
1.0E+02
6.0E+02
1 .OE+03
2.5E+04
10L
(0.026 v/v)
0
-0.01
-1.3%
-3.5
0.00
0.0%
-0.1
-0.01
-0.9%
-4.2
-0.08
-7.0%
-19
-0.01
-0.5%
-1.5
-0.01
-0.8%
-2.3
0.00
0.1%
0.1
-0.04
2.9%
-9.6
-0.13
-11%
-32
0.01
1.5%
3.6
0.00
0.4%
1.0
0.01
1.0%
5.9
0.02
2.2%
13.7
0.41
36.1%
319.3
0.01
0.9%
1.2
0.01
0.7%
1.2
0.01
1.1%
2.1
0.16
14.8%
42.0
0.07
6.4%
36.3
0.01
1.9%
0.9
0.01
0.5%
1.1
0.01
0.8%
2.0
0.01
1.1%
2.4
0.08
7.6%
16
0.00
0.2%
0.4
0.01
0.5%
1.0
0.01
0.6%
1.2
0.03
2.6%
5.9
0.07
6%
12
0.04
4.4%
17
0.07
44%
1.4
0.10
43%
2.3
0.08
30%
1.8
0.09
65%
1.6
0.10
86%
2.3
0.06
43%
1.5
0.08
63%
1.4
0.08
64%
1.7
0.15
239%
3.1
0.09
15%
2.3
1.7
6%
3.2
2.9
12%
3.6
1.7
6%
2.7
13
45%
21
1.1
4.1%
2.0
0.9
3.3%
2.9
1.1
4.1%
1.9
7.6
27%
13.9
34
116%
47
1.5
6.6%
3.3
173
22%
9.4
256
32%
17
316
41%
16
4616
696%
263
86
12%
5.9
75
10%
5.6
123
17%
10.4
1706
255%
153
3356
538%
249
29
3.8%
1.7
No Data
No Data
446
230%
2.4
2099
7591%
84
No Data
No Data
1.7
120%
1.6
5.8
388%
5.6
4079
1158%
9.5
1907
4496%
40
No Data
No Data
0.0003
1.7%
2.1
0.0002
1.4%
1.5
No Data
No Data
0.0003
1.6%
1.9
0.0004
2.4%
2.4
0.0028
25.4%
13.2
0.0002
1.0%
1.4
-------
B.globigii injected at 1 x10 cfu/ml
B.globigii injected at 6x10 cfu/ml
E
1 15 -
E 15 -
B.globigii injected at 1x10 cfu/ml
;5
'•e
35 -
30 -
^ 20 -
i
! 15 H
^ 2000 -
- JMAR
- Laser Turbidimeter
- S::CAN TOC (mg/L)
- S::CAN Turbidity (FTU)
I
8
50 100 150 200
Time (min)
B. globigii injected at 2.5x104 cfu/ml
Figure
Time (min)
1: Sensor baseline and response to contamination with Bacillus globigii spores at four inch pipe densities
- 0.8
- 0.6 5
l—
u_
t
-------
=. co//at 1x102 cfu/ml
jS 10 H
100 150
Time (min)
r0.4 Z
o
E.co// at 6x102 cfu/ml
E.co// at 1x10 cfu/ml
40 -,
•
ID
s
€
20 -
10 -
0 J
ce
<
1400
1200
- 1.0
- 0.8
- 0.2
100
150
200
250 300 350
i
;._.
0.4 ^ ^ 20 -
o
6}
10 -
E.co//at 2.5x104 cfu/ml
Time (min)
Figure 2: Sensor baseline and response to contamination with Escherichia coll K-12 at four inch pipe densities
100 150
Time (min)
-------
Table 3: Formulation for terrific broth and sporulation media
Pancreatic Digest of Casein... 12 g
Yeast Extract...24 g
Dipotassium Phosphate...9.4 g
Monopotassium Phosphate...2.2 g
of water)
Nutrient Broth...8 g
Manganese Sulfate...40 mg
Calcium Chloride...100 mg
s::can spectro::lyser™ TOC yielded a response forE. coli
at 2.5xl04 cfu/mL and a weak response for B. globigii at
the same cell density. This is not surprising, since the s::can
spectro: :lyser™ detection principle is based on absorption
of UV and visible light. Table 3 shows that the media in
which the E. coli were cultured (Terrific Broth) is much
more enriched in sugars and amino acids compared to the
sporulation media in which the B. globigii spores were
suspended. The higher concentration of compounds that
absorb/emit light will lead to a greater response. The Hach
FilterTrak™ 660 sc Laser Nephelometer responded in similar
manner to the s: :can spectro: :lyser™ TOC, with a peak
emerging during injection at the 2.5xl04 cfu/mL density for
E. coli and B. globigii.
Real Tech Real UVT Online
The Real UVT measures percent transmission (i.e., the
light that is not absorbed) of UV light at 254 nm, so it was
hoped that either light absorbing functional groups on the
cells or the nutrient broth injected with the cells would elicit
a change. The Real UVT showed little response to either
E. coli, B. globigii or the control injection. This is likely
due to either the functional groups in the nutrient broth or
on the microorganism not absorbing light at 254 nm, or
the functional groups were not concentrated enough. The
wastewater injection caused a noticeable 25% (S/N 13.4)
change from the baseline.
Fluid Imaging Technologies FlowCAM®
FlowCAM® response to the biological agents and secondary
effluent was similar to the responses of an online turbidity
sensor: percent changes were large, but signal-to-noise ratio
was low due to high baseline variation. However, unlike the
responses of a turbidity sensor, large baseline changes were
sporadic and not always explainable. Large variability in the
baseline made changes difficult to detect. The exception was
the B. globigii injection at 2.5xl04 cfu/mL, where the signal-
to-noise ratio of 84 indicated a large, discernable change.
A good example of the FlowCAM® baseline variation
is the control blank data where GAC filtered tap water
elicited a change larger than some of the contaminants
injections. The control blank should not cause a change
larger than a contaminant injection. In later experiments,
it was observed that touching or bumping the instrument
during testing resulted in a spike in counts, which were
likely due to the release of accumulated particles in the
instrument plumbing. Also, flow to the optics for this
device is controlled by a peristaltic pump, and this flow
varied widely depending on how the length of the tubing
in the pump. If flow is cut off from the optics, no detection
will take place. Flow that is too fast might force particles
through the flow cell too quickly to be counted. These
instrument design/operational limitations could have led
to the changes recorded during the control blank injection.
The changes recorded by the FlowCAM® were likely real.
Attributing changes to a contaminant - or other random
event - proved difficult. Finally, images of the bacteria
taken by the FlowCAM® were visible merely as small
"dots." The highest available resolution of the 20X objective
was not enough to visualize the B. globigii or E. coli.
JMAR Technologies BioSentry®
Of the tested devices, the multiple angle light scattering
device (JMAR Biosentry®) performed best, with an obvious
response at 6xl02 cfu/mL for E. coli and B. globigii. A
change was not visible at the 1 x 102 cfu/mL level due to
the noise in the baseline. Figure 3 shows the change in
counts from baseline plotted against the increasing density
of both biological agents. The output from the BioSentry®
is in "unknown counts," or the number of particles that the
machine is counting. At any concentration, the response is
larger for E. coli compared to B. globigii since the vegetative
cells are larger and easier to detect than the spores.
-------
1.E+04 i
O)
TO
o
1.E+03
1.E+02
1.E+01
1.E+00
y=0.1787x+146.94
•B. globigii
E. coll
= 0.0659x + 57.296
R= 0.9995
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1 .E+05
Initial Concentration (cfu/m I)
Figure 3: Multiple angle light scattering device response to Escherichia coli K-12 and Bacillus globigii
Discussion and Significance of Sensor Response Results
B. globigii and E. coli were visibly detected at
2.5xl04 cfu/mL by all of the sensors tested. At the 1,000
and 600 cfu/mL cell densities, the Biosentry® provided
the only visually significant response. None of the sensors
provided any significant visual response at the 100 cfu/mL
cell density. It should also be noted that 10 L of secondary
effluent triggered an easily detectable response from each
sensor used in this study due to the fact that it contains
large particles and has a relatively high organic content
compared to tap water. This is particularly interesting when
considering that wastewater cross connections in drinking
water distribution systems may be detected by standard and
specialized sensors since raw effluent would presumably
be more concentrated in solids, organics and nutrients.
The sensor response results have been presented as what
can be visually detected as a change from a baseline.
Visual changes can be useful when manually evaluating
sensor data, but whether someone can discern a change
depends on the variability of the baseline data. The
experimental conditions in this study were ideal since
baseline variation was minimal, but variability can increase
in the field. Therefore, event detection algorithms should
be considered for use in conjunction with standard or
specialized water quality sensors to lower detection levels
(McKenna et al., 2008). Although the human eye may not
be able to discern a change from baseline data, an event
detection algorithm may detect it. However, the data
provides a glimpse of the water quality changes that could
be caused by biological contamination. Even if an event
detection algorithm indicates an unusual change, the data
that caused the alarm will likely be manually examined.
Finally, adding a specialized sensor could benefit a
chloraminated drinking water system. Inactivation of
microorganisms will be slower in chloraminated water,
especially Bacillus spores, so the results presented for
chlorinated water should transfer to chloraminated
systems. Furthermore, including devices like the s::can
spectro::lyser™ or Sievers® 900 will add TOC as an online
monitoring parameter in an online water quality monitoring
station. This is especially important for chloraminated
systems since total chlorine is not as effective at responding
to contamination in chlorinated water as in chlorinated water
(Szaboetal, 2008).
Sensor Cost Considerations
Table 4 lists the costs of each piece of equipment. The cost of
the JMAR BioSentry® unit used in these tests was $46,215.
Should there be a particularly sensitive point in a utility's
distribution system the BioSentry® may be an effective tool
for detecting low level biological contamination. However,
should a utility wish to deploy more units, they may look
to standard online sensors that cost less, but cannot detect
biological suspensions as low as the BioSentry®. If biological
contamination is a concern to a water utility, then this
tradeoff between detection and cost must be considered.
-------
Table 4: Sensor detection and purchase data
JMAR BioSentry®
s::can spectro::lyser™ TOG
Spectro::lyser™ turbidity
Hach FilterTrak™ 660 sc Laser Nephelometer
RealTech Real UVT
FlowCAM®
Sievers® 900 On-Line Total Organic Carbon Analyzer
Hach CL17 Total Chlorine Analyzer
kity at which sensor responded
(cfu/mL)
2.5xl04, IxlO3, 6xl02
2.5xl04
2.5xl04
2.5xl04
-
2.5xl04
2.5xl04
2.5xl04
Approximate Purchase Price
(year 2007 $)
50,000
25,000
5,000
5,000
35,000
25,000
5,000
-------
Page Intentionally Blank
-------
Conclusions and Future Work
The data from the studies indicate that some specialized
sensors can detect biological suspensions at lower densities
in drinking water than standard online water quality sensors.
The JMAR BioSentry®, which uses multiple angle light
scattering (MALS), detected the lowest concentration
(600 cfu/mL), while the s::can spectro::lyser™ TOC and
Hach FilterTrak™ 660 sc Laser Nephelometer performed as
well as the Hach CL17 Total Chlorine Analyzer and Sievers®
900 On-Line Total Organic Carbon Analyzer. An important
component to any future work would be using an event
detection algorithm in concert with the online water quality
sensors. Even though an obvious visual change did not occur
at low levels of contamination (100 to 1,000 cfu/mL) for all
sensors, there may be a subtle change that an algorithm could
detect that the human eye cannot. This is especially important
if a specific water quality parameter is "noisy" at the location
it is being monitored. Other conclusions are as follows:
• The operational and maintenance costs of all the
specialized optical devices tested were favorable.
There are no reagents to buy and replace and no major
maintenance issues were observed during the testing.
• As with all online turbidity and particle counting sensor
equipment, control of bubble formation is needed to
prevent false alarms.
• There is a wide range of capital costs for the equipment
tested. This is good news for potential consumers of this
equipment, since multiple sensors of varying cost can be
used together to create a more comprehensive detection
network. Furthermore, it is anticipated that the cost of
the multiple angle light scatter device can be reduced as
larger market demand is generated and cost efficiencies
are identified.
• The lower detection capability of the more expensive
equipment should be weighed against the higher cost
of the equipment. For example, one expensive device
with lower detection capability could be deployed to
a sensitive area while multiple less expensive devices
could be more widely distributed.
Finally, although some of these devices did not respond to
injections of microbial suspension, it should be noted that the
manufacturers might not have designed them for detection of
low density biological suspensions. Results from this study
should not be used to evaluate the detection capabilities
of these devices when used in other scenarios or for their
intended purpose.
-------
Page Intentionally Blank
-------
References
Allmann, T.P. and Carlson, K.H (2005). Modeling intentional distribution system contamination and detection. J. AWWA, 97(1),
58-71.
Byer, D. and Carlson, K.H. (2005). Real-time detection of intentional chemical contamination in the distribution system.
J. AWWA,91(1), 130-133.
Hall, J.S., Zaffiro, A.D., Marx, R.B., Kefauver, P.C., Krishnan, E. R., Haught, R.C. and Herrmann, J.G. (2007). Online water
quality parameters as indicators of distribution system contamination. J. AWWA, 99(1), 66-11.
Homeland Security Presidential Directive (HSPD) 7, 2003. Critical Infrastructure Identification, Prioritization, and Protection.
www.whitehouse.gov/news/releases/2003/12/20031217-5.html.
Homeland Security Presidential Directive (HSPD) 9, 2004. Defense of United States Agriculture and Food.
www.whitehouse.gov/news/releases/2004/02/20040203-2.html.
Kessler, A., Ostfeld, A. and Sinai, G. (1998). Detecting accidental contaminations in municipal water networks. J. Water Res.
Pl-ASCR, 124(4), 192-198.
King, K.L. and Kroll, D. (2005). Testing and verification of real-time water quality monitoring sensors in a distribution system
against introduced contamination. Proceedings of the AWWA Water Quality Technology Conference, Quebec City, Canada.
Kroll, D. and King, K.L. (2005). Operational validation of an online system for enhancing water security in the distribution
system. Proceedings of the AWWA Water Security Congress, Oklahoma City, OK.
Magnuson, M.L., Allgeier, S.C., Koch, B., De Leon, R. and Hunsinger, R. (2005). Responding to water contamination threats.
Environ. Sci. Technol. A-Pages, 39(7) 153A-159A.
McKenna, S.A., Wilson, M. and Klise, K.A. (2008). Detecting changes in water quality data. J. AWWA, 100(1), 74-85.
Szabo, J.G., Hall, J.S. and Meiners, G.C. (2006). Water quality sensor responses to injected contaminants in a chloraminated
pipe loop. Proceedings of the AWWA Water Security Congress, Washington, DC.
Szabo, J.G., Rice, E.W and Bishop, PL. (2007). Persistence and decontamination of Bacillus atrophaeus subsp. globigii spores
on corroded iron in a model drinking water system. Appl. Environ. Microbioi, 73(8), 2451-2457
Szabo, J.G., Hall, J.S. and Meiners, G.C. (2008). Water quality sensor responses to contamination in a chloraminated drinking
water distribution system simulator. J. AWWA, 100(4), 33-40.
USEPA, 2005a. Evaluation of Water Quality Sensors as Devices to Warn of Intentional Contamination in Water Distribution
Systems. EPA/600/R-05/105, Washington, DC. (Available through the secure WaterlSAC site: www.waterisac.org).
USEPA, 2005b. WaterSentinel Online Water Quality Monitoring as an Indicator of Drinking Water Contamination.
EPA/817/D-05/002, Washington, DC.
USEPA, 2005c. WaterSentinel System Architecture. EPA/817/D-05/003. Washington, DC.
USEPA, 2006. Water Quality Sensor Responses to Potential Chemical Threats in a Pilot-scale Water Distribution System.
EPA/600/R-06/068. Washington, DC. (Available through the secure WaterlSAC site: www.waterisac.org).
USEPA, 2007. Water Quality Sensor Responses to Contamination in a Single Pass Water Distribution System Simulator.
EPA/600/R-07/001, Washington, DC. (Available through the secure WaterlSAC site: www.waterisac.org).
Yang, Y.J., Goodrich, J.A., Clark, R.M. and Li, S.Y. (2008). Modeling and testing of reactive contaminant transport in drinking
water pipes: Chlorine response and implications for online contaminant detection. Water Res., 42(6-7), 1397-1412.
-------
&EPA
United States
Environmental Protection
Agency
Office of Research and Development
National Homeland Security Research Center
Cincinnati, OH 45268
Official Business
Penalty for Private Use
$300
Recycled/Recyclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
PRESORTED STANDARD
POSTAGES FEES PAID
EPA
PERMIT NO. G-35
------- |