United States
Environmental Protection
Agency
Office of Research and
Development
Washington, DC 20460
EPA/600/K-98/001
June 1998
xvEPA
Selected Papers in
High Visibility Drinking
Water Research
(1996 -1998)
-------
-------
EPA/600/K-98/001
June 1998
Selected Papers in High Visibility
Drinking Water Research (1996-1998)
National Risk Management Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, Ohio 45268
Printed on Recycled Paper
-------
FOREWORD
»•
This document is the result of a collaborative effort between the Water Supply and Water Resources
Division (WSWRD) and the Technology Transfer and Support Division (TTSD) of the USEPA's
National Risk Management Research Laboratory (NRMRL) both located in Cincinnati, Ohio. WSWRD
has the responsibility for conducting research in the field of water supply, wet weather flows and
watershed management for the USEPA and TTSD has the responsibility for ensuring that EPA's research
products are disseminated to a wide segment of the user community. For example, TTSD has prepared a
CD ROM version of a series of presentations from a nationally televised conference on source water
protection and made it available at the 1998 AWWA Annual Conference and Exposition in Dallas. This
document has been prepared in the spirit of disseminating research conducted by the USEPA staff as
widely as possible. It contains a selection of papers by the staff of WSWRD, in the field of water supply,
that have been peer reviewed and appear in a modified form in scientific journals over the last year.
These papers are only a sample of the papers, reports and other publications generated by the WSWRD
staff.
WSWRD has the responsibility for a broad program which includes helping prepare the primary and
secondary regulations for drinking water. This program integrates chemistry, engineering, microbiology,
and cost analysis to provide effective, reliable and cost-effective techniques for assuring the delivery of
safe drinking water. Other research activities include the responsibility for developing technology and
strategies for controlling contaminants such as:(l) agricultural and rural storm runoff;(2) combined
sewer overflows;(3) urban storm water and sanitary sewer overflows;(4) wastewater from small
communities; and (5) contaminated sediments. The goal of this research is to reduce the risks from
chemically and microbiologically induced public health risks resulting from these sources.
Environmental restoration strategies and technologies are also being evaluated.
The Division conducts some of the Nation's highest priority and most visible environmental programs. It
provides guidance to be used by the USEPA, States, EPA Regional Offices and drinking water utilities
in implementing various provisions of the Safe Drinking Water Act. For example, among its current
responsibilities are the evaluation of treatment technologies that will minimize the risks from
microbiological contaminants such as Cryptosporidium and other pathogens. The papers in this
document by Fox and Lytle, Clark, et al., Rice, et al. and Sethi, et al. explore these themes..
• Fox and Lytle describe the results of their investigation of the Cryptosporidiosis outbreak in
Milwaukee Wisconsin. These EPA on-site investigators led the field study and, along with other
members of the EPA team, received EPA's Gold Medal for their work.
• Clark, et al. present the details of a Salmonella outbreak in Gideon Missouri in which 600 of
approximately 1100 residents became ill and seven people died due to bird contamination of a
municipal storage tank. This study, conducted jointly with CDC and
the Missouri Department of Health and Natural Resources, marked the first time that water
quality modeling was applied to identify the source of a waterborne outbreak.
• Rice, et al. describe the results of a research project that demonstrated the usefulness of
indigenous aerobic bacterial endospores for assessing the performance of filtration. The
investigators showed that aerobic spores are a conservative indicator of filtration performance
with regard to removal of Cryptosporidium. Endospores are present in virtually all source waters
11
-------
in high quantities, are nonpathogenic and easy to analyze. Many drinking water utilities in the
United States and abroad are currently using this technique.
• Sethi, etal. deal with one of the most important current issues in drinking water, that of particle
counting and measurements. They explored the application of Multiple Angle Light Scattering
Techniques for characterizing particles in drinking water larger then one micron in diameter.
They also identify some of the deficiencies in turbidimeters that are currently in use.
The Division is conducting targeted research in the area of small systems technology for drinking water.
Small drinking water systems are among the most significant compliance problems associated with the
Safe Drinking Water Act. Small systems that serve less then 3300 hundred people represent more then
85 % of all the community drinking water supplies in the US. In addition, there are approximately
140,000 water systems that serve parks, schools, hotels etc, The Division is testing and developing
technology designed specifically to serve small communities in a cost effective manner. A spin-off from
this work has been the Division's successful involvement in EPA's Environmental Technology Initiative
program and the development of a drinking water Environmental Technology Verification program. The
papers by Li, et al. and two papers by Lytle and Schock describe technology and research that can be
applied to small systems.
• Li, et al. describe experimental work done at EPA's Test and Evaluation Facility in Cincinnati
Ohio in which several non-hazardous surrogates were evaluated for their equivalency in
predicting the performance of bag filters for removing Crytosopridium. The authors examined
turbidity, particle counting and bead removal.
• Lytle and Schock explore the sensitivity of pH to dissolved oxygen content in drinking water and
demonstrate the potential of using reaeration for corrosion control. This technology might have
special application to small water supplies.
• Lytle and Schock present the results of a very extensive experimental study that document the
effect of alloy composition and pH on the leaching of brass. It has major implications for the
potential of lead leaching from brass household fixtures.
The Division is also conducting research intended to insure the safe delivery of drinking water to the
consumer by conducting research that will provide solutions to the Nation's failing drinking water and
waste water infrastructure. For example, Clark, et al. have developed models that describe the mixing
characteristics of distribution storage tanks.
• Clark, et al. deal with an emerging issue in drinking water, that of the possible deterioration of
water quality in distribution systems. One of the components of drinking water distribution are
tanks which provide storage for fire protection and insure the reliability of service. They are also
reaction vessels in which disinfectant residuals can dissipate and microorganisms can multiply.
This paper presents a model that can be used to analyze the mixing characteristics of these
storage tanks..
WSWRD is conducting technology research that will minimize the risks from exposure to-the by-
products of disinfectants while protecting against microbiological threats. Technology is being
evaluated that will be utilized to support promulgation of the USEPA's Enhanced Surface Water
Treatment Rule and the proposed Disinfectant and Disinfection By-product Rules. These Rules and the
ill
-------
regulations promulgated under these Rules will affect virtually every water supply in the United States
and the public health of virtually every citizen in the US. Papers by Clark, et al. and Clark present
kinetic models for disinfection by-product formation under various experimental conditions.
• Clark, et al. explore the effect of bromide on the kinetics of chlorination by-product formation.
Experiments were conducted at three different pH levels and at two concentration levels of
chlorine.
• Clark proposed a chlorine demand and TTHM formation model based on second order kinetics.
The paper shows that TTHMs formation can be modeled as a function of chlorine consumption.
Other high visibility programs include research into the fields of wet weather flows, small waste water
treatment systems and source water protection.
We would like to acknowledge the assistance of Carol Grove, Corliss Straus, Peggy Heimbrock,
Thomasine Bayless, and Jean Dye of NRMRL's TTSD and Steven Waltrip of WSWRD for their
assistance in preparing this document.
We hope that you will find this collection of papers useful and informative.
Walter A. Feige, Technical Assistant*
James E. Smith, Research Sanitary Engineer**
Robert M. Clark, Director*
*U.S. Environmental Protection Agency
National Risk Management Research
Laboratory, Water Supply and Water
Resources Division
26 W. Martin Luther King Drive
Cincinnati, Ohio 45268
**U.S. Environmental Protection Agency
National Risk Management Research
Laboratory, Technology Transfer and
Support Division
26 W. Martin Luther King Drive
Cincinnati, Ohio 45268
IV
-------
TABLE OF CONTENTS
Foreword ii
Paper Title
The Milwaukee Cryptosporidiosis Outbreak: Investigation and
Recommendations - Fox, K.R. and D.A.Lytle 1
Tracking a Salmonella Serovar Typhimurium Outbreak in Gideon,
Missouri: Role of Contaminant Propagation Modeling - Clark, R.M.,
Geldreich, E.E., et al 14
Evaluating Water Treatment Plant Performance Using Indigenous Aerobic
Bacterial Endospores - Rice, E.W., Fox, K.R., et al 34
Evaluation of Optical Detection Methods for Characterizing Suspensions
in Drinking Water - Sethi, V., Patnaik, P., et al 51
Reliability of Non-Hazardous Surrogates for Determination of
Cryptosporidium Removal in Bag Filtration Systems - Lij S.Y.,
Goodrich, J.A., et al .79
The Use of Aeration for Corrosion Control - Lytle, D.A.,
Schock, M..R., et al 95
An Investigation of the Impact of Alloy Composition and pH on the
Corrosion of Brass in Drinking Water - Lytle, D.A. and M.R. Schock 130
Mixing in Distribution System Storage Tanks: Its Effect on Water
Quality - Clark, R.M., Abdesaken, F., et al 158
Modeling the Kinetics of Chlorination By-Products Formulation: The
Effects of Bromide - Clark, R. M., Pourmaghaddas, EL, et al 190
Chlorine Demand and TTHM Formation Kinetics - A Second Order
Model - Clark, R.M. . 203
APPENDIX A Related Professional Journal Articles 236
APPENDIX B
Expertise/Point-of-Contact List 238
v
-------
-------
The Milwaukee Cryptosporidiosis Outbreak:
Investigation and Recommendations
Kim R. Fox and Darren A. Lytle
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Cincinnati, Ohio 45268
Abstract
In March/April 1993, the largest U.S. waterborne disease outbreak occurred in Milwaukee, Wisconsin. In this
outbreak, more than 400,000 people became ill and the etiological agent responsible for the bulk of the cases was
Cryptosporidium. The City of Milwaukee was served by two different water treatment plants and during the
outbreak, one of the plants produced a water effluent that had a turbidity that approached 2.5 NTU. The high
turbidity of the effluent water indicated that large numbers of particulates were passing the water treatment plant.
The increase in particulate passage may have also meant an increase in passage of Cryptosporidium oocysts. The
source water for the Milwaukee treatment plants is Lake Michigan. Prior to the outbreak, the Milwaukee area
experienced severe spring storms and the Lake's turbidity and bacterial counts rose dramatically. This paper
discusses the investigation done by the Water Supply arid Water Resources Division (formerly the Drinking Water
Research Division) of the U.S. EPA about what may have happened to allow the oocysts to pass the water treatment
plant.
-------
Introduction
In April of 1993, Milwaukee, Wisconsin became the major focus of the drinking water industry. The focus of
attention centered on the large increase in reported cases of diarrheal patients throughout the city. Shortages of over-
the-counter medications for diarrhea control were also seen at local pharmacies. Although other organisms may have
been involved, the major increase in diarrhea was determined to be caused by the organism Cryptosporidhim.
Ensuing investigations were conducted by City, State, and Federal officials. These investigations suggested that
dririRfng water may have been partially responsible for distributing the organism hi Milwaukee and the surrounding
area. Random digit phone dialing surveys indicated that 403,0001 people were affected (had diarrhea) during the
incident. The Milwaukee Water Utility pumps water to approximately 800,000 people2.
In early April, 1 preliminary data (turbidity data and fecal specimens)1 had begun to implicate Cryptosporidium
in the drinking water as a contributing factor to the diarrheal outbreak. State and City officials contacted individuals
experienced with Ctyptosporidium analytical techniques and water treatment practices. The Water Supply and
Water Resources Division (WSWRD) formerly know as the Drinking Water Research Division (DWRD) of the
United States Environmental Protection Agency (USEPA) was contacted (April 8, 1993) for technical assistance.
As a result of this contact, the WSWRD sent an investigative engineering team to Milwaukee. The team was to
assist the Centers for Disease Control (CDC), State of Wisconsin, and the City of Milwaukee in determining how
Cryptosporidium may have passed through the water treatment plant. The WSWRD team conducted a preliminary
rapid engineering assessment of the two water treatment plants. They also evaluated plant operational and
laboratory data available for the time frame of concern. This paper discusses the efforts that the team took during the
investigation. The recommendations that were made for improving the operation of the water treatment plants are
also discussed.
Cryptosporidiosis Outbreak
On April 6, a local Milwaukee doctor ordered a parasitic analysis on a fecal specimen from a patient.
Cryptosporidhtm was detected in the fecal smear. At that time, local and state officials were notified of the
Ciyptospotiditim detection. A concurrent survey of diarrhea cases in local nursing homes was conducted. This
study indicated that the residents in nursing homes in the southern part of the city were fourteen times more likely
to have had diarrhea than those in the northern part of the city. Nursing home populations were used in this study
because nursing home residents were thought to be less likely to have traveled around the city. Thus, they would
have been exposed only to the environment within the nursing home.
The City of Milwaukee is served by two water treatment plants2. The Linwood Plant serves primarily the
northern half of the city, while the Howard Avenue Plant serves the southern half. Concurrent data on treatment
efficiency at the Howard Avenue Plant and the Nursing Home study suggested that the Howard Avenue Plant was
suspect in the disease outbreak.
The Howard Avenue Water Treatment Plant (HWTP) was ordered closed during the ensuing investigation and
a boil water order was issued for the service area. The Linwood Plant was capable of treating enough water to serve
the community during the late winter and early spring months. Both plants were needed during high water usage
periods such as in the summer.
As a result of the boil water order, drinking water fountains, ice machines and other devices that used municipal
drinking water were shut down. There was a run on bottled water supplies at local markets. Local restaurants
served canned soft drinks and juices without ice. Bottled or on-site processed water (boiling or filtration) was used
for washing salad ingredients and made available to customers. Telephone random-digit dialing follow up surveys
indicated that as many as 403,000 individuals experienced watery diarrhea between March 1 and April 28, 1993 L
Turbidity changes and hospital records suggested the survey tune frame.
-------
Treatment Facilities
The source water for Milwaukee, Wisconsin's water treatment plants was Lake Michigan. The city's water was
treated at two water treatment facilities. The two facilities were the Linwood treatment plant (located in the northern
half of the city) and the Howard treatment plant (in the southern half of the city). Both treatment facilities treated
Lake Michigan water by conventional water treatment processes (coagulation, sedimentation, filtration, and
disinfection). The Linwood facility (put into service in 1939) had a water treatment capacity of 275 million gallons
per day (MGD) (1.04 x 106 cubic meters per day) and an on-site clearwell storage capacity of 30 million gallons
(MG) (1.13 x 105 cubic meters). The Howard treatment facility (put into service in 1962) had a filtration capacity of
100 MGD (3.785 x 105 m3 per day)and an on-site storage capacity of 35 MG (1.325 m3) . During typical operation,
the Linwood plant supplied the northern 2/3 of the water district. The Howard plant supplied the remaining 1/3. A
large mixing zone existed in the distribution system between the two treatment facilities. The water from the
treatment plants flowed westward and then into southern and northern service areas.
The initial investigation linked the cryptosporidiosis outbreak to drinking water treated at the Howard treatment
facility, resulting in a temporary shutdown of the facility. The EPA team focused its efforts on understanding the
operation of the Howard treatment plant. The following is a description of the infrastructure and operation of the
Howard facility at the time of the outbreak (Figure 1). (There have since been considerable upgrades/modifications to
both treatment facilities).
Lake Michigan water was pumped 3.5 miles (5.63 kilometers) to the Howard Avenue treatment plant from the
Texas intake pump station. This station was located on the shore of Lake Michigan. A water intake pipe extended
approximately 7600 feet (2.32 kilometers) from the pump station into the lake to a submerged steel crib intake, 42
feet (12.8 meters) beneath the lake surface. Chlorine and/or potassium permanganate were occasionally fed into the
water at the Texas intake station for taste and odor control. There were some initial indications of zebra mussels at
the intake. That prompted action toward obtaining the ability to feed chlorine and/or potassium permanganate at the
intake site (a steel crib) to control the zebra mussels.
At the Howard treatment facility, the water entered a 0.13 mg (4.92 x 102 m3) capacity rapid mix chamber. Rapid
mixing was done by injecting air through perforated pipe at the bottom of the chamber. Maintenance staff at the
plant indicated that injection holes were checked regularly for clogging and proper ah- flow. At this location, free
chlorine, powder activated carbon (PAC), and the coagulants were added. Alum (aluminum sulfate) was the
coagulant used until August of 1992. In August of 1992, the facility switched to polyaluminum chloride (PACL).
The switch to PACL was done with the belief in three benefits 1) higher finished water pH for corrosion control, 2)
reduced sludge volume and 3) improved coagulation effectiveness in cold raw water conditions. Before switching to
PACL, the city consulted with the chemical manufacturer, Wisconsin DNR, and other communities receiving Lake
Michigan water and using PACL. -
The PACL flocculent produced a smaller, and typically denser floe than alum floe hi Milwaukee. The floe
characteristics suggested that PACL floe that did carry over from the sedimentation basins onto filters tended to
penetrate further into the filter bed than.alum floe. From a chemical dosage standpoint, PACL flocculation required
1/2 to 1/3 the amount of chemical as compared to alum coagulation to perform and meet the same effluent water
quality. PACL uses charge neutralization as the primary mechanism of particle destabilization (alum is normally
used in sweep floe coagulation).
After the rapid mix stage, the water went into one of four 1 MG (3.785 x 103 m3) parallel baffled coagulation
basins. The coagulation basins were followed by 4.5 mg (1.703 x 104 m3) sedimentation basins.
The water then passed through eight double sand/gravel media filters. Each filter was 42 feet (12.8 meters) by
69 feet (21 meters) and was capable of filtering 12.5 MGD (4.73 x 1Q5 m3 per day). Each filter held 0.19 MG
(7.19 x 102 m3) and was rated to filter 3 gallons/minute/square foot (122 L/m2/minute). Filter construction
consisted of steel filter underdrains covered by 29 inches (0.73 meters) of gravel and 22 inches (0.56 meters) of sand.
The effective diameter of the media was not known for sure, but plant personnel indicated that 0.35 mm effective size
stood out in their minds as the size. The media had been in place since plant construction. Sand were added to the
filters periodically as the filter media level dropped during normal operation or backwash procedures.
-------
Flow rates and headless were monitored across each filter and were the basis for initiating each backwash
procedure. The filter backwashing procedure was done manually. The operator watched the filter flows during the
backwash for filter short-circuiting due to mudballs or larger media hi the sand layer. Following backwashing, filter
to waste procedures were not followed nor are there the facilities for doing so.
The effluent water from the filters was combined before it entered a weir room. Chlorine and ammonia were
added to the water in the weir room for disinfection. The water then traveled through a 250 foot (76.2 meters) 72
inch (1.83 meters) pipe into a series of two clearwells (East Clearwell and West Clearwell). The East and West
Clearwell had capacities of 15 MG (5.68 x 104 m3)and 20 MG (7.57 x 104 m3) respectively. The water was then
pumped into the distribution system.
Turbidity was monitored at the following stages along the treatment process: raw water at the Texas intake
pump station, raw water entering the Howard treatment plant, filter influent (settled) water, filter effluent
(combination of 8 filters) water, and ciearwell effluent water. Monitoring the effluent turbidity of each filter was not
done. Although filter effluent turbidimeters were in place, they were not operating at the tune of the outbreak.
Turbidity as well as taste, odor, and bacteria, were measured in the plant's laboratory from each site once a shift
(every 8 hours).
Coagulant and/or flocculent dosages were adjusted as needed based on laboratory jar tests. A streaming current
detector (SCO) had been installed at the plant. The plant operation staff indicated that the instrument's data had not
assisted them in determining the proper coagulant dosages and was not being used. The reason for the non-use of
the SCO was not determined.
Water Supply and Water Resources Division's Team's Activities
The United States Environmental Protection Agency's (USEPA) Water Supply and Water Resources Division
(WSWRD) vyas contacted on April 8, 1993. The request was to provide technical assistance during the disease
outbreak. Initial telephone conversations were held between EPA personnel and representatives from both the State
of Wisconsin and Milwaukee City Officials. The conversations centered around how drinking water may have
played a role in transporting the organisms throughout Milwaukee. At the time of the contact with the WSWRD,
the boil water order had been issued and the Howard Avenue plant had been shut down. The WSWRD was asked
to go to Milwaukee and provide technical assistance during the follow-up investigation.
The team arrived in Milwaukee on Monday (April 12, 1993). The first signs of entering a city with a
potentially contaminated water supply were observed at the airport. All drinking water fountains had signs taped to
them stating not to drink the water. Water coolers with bottled water were placed along side all of the shut down
drinking water fountains. The EPA team met with the Chief Medical Officer and State Epidemiologist for
Communicable Diseases for the State of Wisconsin Department of Health who was in charge of coordinating the
epidemiological studies associated with the outbreak. The Chief Medical Officer was also the first to request the
WSWRD assistance and gave the team a brief update on the status of the outbreak. This update included the
number of reported cases of cryptosporidiosis, location of cases, and current monitoring efforts and studies.
During the course of the five days (April 12-16, 1993) that the EPA Team was in Milwaukee, both Mayor's
briefing and Press briefings were daily occurrences. The Mayor's briefing assembled all groups involved with any
related investigations to keep all parties informed of all information. The Press briefing normally followed to keep
the public informed of any and all progress involved in the outbreak investigation and recommendations.
The WSWRD Team's role in the investigation was to evaluate the water treatment facilities and the water
quality conditions before and during the time of the outbreak. This assessment included a visual inspection of the
treatment plants and a review of all operational and treatment data available.
The visual inspection of the treatment plants included both walkthroughs and discussions with plant personnel
concerning plant operational and maintenance practices. The records inspected included all monitoring data such as
-------
pH, turbidity, heterotrophic plate counts, coliform counts, disinfection dosages and residuals, coagulant doses, and
flow at various locations throughout the facility. The inspections were conducted by the EPA Team, consulting
engineers and representatives from the State of Wisconsin Department of Natural Resources (DNR). Two
representatives from the supplier of polyaluminum chloride (PACL) (the chemical under scrutiny at the time) were
also in attendance to answer technical questions about the PACL and assist in any way.
In reviewing the plant's operational and treatment data, the teams focused on the treatment scheme used to
remove particulates. The teams examined source water monitoring data, looking for shifts in source water quality.
Questions were asked to the treatment plant operators such as "How was the coagulant added to the raw water and
how often were dosages changed (what criteria were used to initiate a change in dosage)?" The rapid and slow
mixing basins and mixers were screened for basic problems such as mixing intensity and short circuiting.
Sedimentation was reviewed for baffling to prevent short circuiting and for sludge removal equipment. The teams
also looked at flow rates across each of the steps to determine if they were within acceptable limits.
The plant's filters and filter operations were then evaluated. Questions such as "What were the filters loading
rates (discussed in the plant description section) and quality of water going onto the filters?" were asked. Backwash
procedures were evaluated by examining the guidelines followed by the plant operators (the backwashing procedure
was automated, but was initiated by the operators based on time, headloss, or turbidity). The plant was shut down
at the tune so observing a backwashing procedure was not possible. The next procedure to be looked at was the
restart procedure for a filter after a filter backwash. Questions were asked about how quickly the filters were brought
back to full flow rates and filter to waste procedures (the Howard Avenue Plant did not have filter-to-waste
capabilities). Turbidity reduction and monitoring were checked in the lab records (Figures 2 & 3).
The last step checked was the disinfection chemicals and practices. How was the disinfectant added, what
contact time was available, how much disinfectant was added, and what was the residual were just a few of the
questions asked (the Howard Avenue Plant used both pre and post chlorination with ammonia addition in the final
weir room).
The above-mentioned groups met to discuss the data and discuss thoughts on the problems or deficiencies at
the plant. Operational practices that may have allowed Cryptosporidium to enter the water supply were also
evaluated. One of the requests made by the mayor was to develop a plan to reopen the Howard Avenue plant and
recommend future improvements to both of Milwaukee's treatment facilities. All involved parties were a part of the
discussions about what may have happened to allow the Cryptosporidium to pass through the water treatment
facility. These discussions also centered on what was needed to be done at the Howard Avenue facility before
restarting the plant.
Another topic of discussion was for making recommendations for the lifting of the boil order. Cryptosporidium
samples were being taken in the distribution system and a review of that information was requested. The discussion
among all parties at that time had decided that negative findings of Cryptosporidium in terminus distribution water
samples and Linwood treated water would justify lifting the order. Two consecutive sampling days that were
negative for Cryptosporidium were suggested as the controlling criteria.
The investigating teams also visited the Linwood water treatment plant. The Linwood plant was not
associated with passing Cryptosporidium oocysts. The Linwood tour was important in identifying potential
problems in the future and for making suggestions for improvements.
A discussion was held with the city officials where the investigative teams pointed out water treatment plant
deficiencies and described how Cryptosporidium may have passed the treatment process. The consulting
engineering team also presented requirements for putting the Howard Avenue treatment plant back on-line. They
also made suggestions for future improvements to both the Howard Avenue and Linwood treatment plants.
Following the discussion and question period with city officials, a press conference was held to present the findings
to the press and public. Both the EPA and the consulting engineering reports were made available to the public.
. 5
-------
In addition to the discussions described above, reports and phone conversations to USEPA officials were daily
events. Communication with all involved parties was essential to order to conduct the investigation.
Press Conference Presentation and Findings
The following information paraphrases the presentation made by the USEPA Water Supply and Water
Resources Division's investigative team to the City on the investigation of the Howard Avenue Water Treatment
Plants:
In response to a request from the City of Milwaukee, the U.S. Environmental Protection Agency's Water
Supply and Water Resources Division (WSWRD) sent a team to Milwaukee to pro vide-technical assistance. This
assistance was to both State and City officials during the suspected waterborne Cryptosporidium disease outbreak in
Milwaukee. The WSWRD team consisted of engineers familiar with filtration and with removal of
Cryptosporidium oocysts by filtration processes.
The review of the Howard Avenue Water Treatment plant turbidity and operational data was completed based
on filtration studies that WSWRD has done on removing Cryptosporidium oocysts from drinking water. At the
time of the outbreak, these studies were not available in the open literature as the research was ongoing. Preliminary
studies indicated that Cryptosporidium oocysts could be removed effectively by filtration. Stringent controls on
operating conditions and removal of turbidity are required for removing Ciyptosporidium.
The team from EPA entities concentrated its efforts on evaluating the operational and monitoring data available
from die Howard Avenue Water Treatment Plant (HWTP). In the tune frame immediately preceding March 1993,
the HWTP was consistently producing a low effluent turbidity water that routinely approached 0.1 NTU or less
(daily averages). During the period of March 18 thru April 8 (plant was shut down April 8), the effluent turbidity
from the HWTP was highly variable and ranged from between 0.1 and 2.7 NTU (Figure 2). The team was asked to
look at what caused the higher turbidity levels to occur. They also looked at what may have been done to have
maintained lower turbidity effluents. At all times during this period, effluent water samples were negative for
coliforms and met the Wisconsin DNR regulations for turbidity. Cryptosporidium had not been monitored. There
were no prior indications that monitoring was needed. The EPA team approach was to look at what may have
happened during this time interval.
The HWTP has been using polyaluminum chloride (PACL) for coagulation since September 1992 (the prior
coagulant was alum). The HWTP intake receives a highly variable quality of influent water (from Lake Michigan)
to be processed through the plant. During the time period of March 18 through April 8, the raw water turbidity
levels ranged from 1.5 to 44 NTU (Figure 3). Total coliforms ranged from <1 CFU/lOOmL to around 3200
CFU/lOOmL in the raw waters. All effluent water samples were negative for coliforms. Average raw water turbidity
levels for previous months were around 3 or 4 NTU and influent coliform levels <20 CFU/lOOmL.
During the time period being investigated, plant facility personnel responded to fluctuations in turbidity
throughout the treatment processes by adjusting coagulant dosages. Additionally, potassium permanganate, and
powdered activated carbon dosages were adjusted primarily to control taste and odor. Coagulation dosage
adjustments were also made to compensate for coagulation demands resulting from taste and order controlling
treatment. Throughout the time period, coagulant dose adjustments were continuously being made to meet the
demands of raw water quality.
Although the coagulant dosages were being adjusted, filter effluent turbidities on several occasions exceeded
turbidity values mat were achieved in previous months. As the coagulant doses approached what might be an
optimum dosage for the particular raw water conditions, improvements in settled water turbidities and filter effluent
turbidities were achieved. This pattern was seen twice. The first time was when polyaluminum chloride was the
primary coagulant at the HWTP. The second time was when the primary coagulant was switched back to alum on
April 2, 1993. The improvements in filter effluent turbidity demonstrated that the plant is fully capable of
-------
producing low turbidity water under optimal chemical coagulation conditions (i.e. dosages applied) even when
challenged with high turbidity raw water.
There are several factors that may have increased the time to reach the optimum coagulant dosages for low filter
effluent turbidities during this time. One of these factors may include a lack of historical use records for the coagulant
used. The previous coagulant had been used for almost thirty years. Records are available to show what dosages
were needed under many operating conditions. The new coagulant had only been used for a short time and the
historical records have not been fully developed. The optimum chemical dosages were sought through laboratory
testing and consultation with DNR and chemical supplier to achieve the lowest turbidity possible. The coagulant
adjustments were made based on all available data.
Another important factor is the time required to see a result in the treated water quality after chemical
adjustment. With a short residence time for the water in the plant and a rapidly changing influent water quality,
dosage optimization was difficult. Laboratoiy jar tests are good predictors of the proper chemical dosages and fine
tuning of the dosage is done in the plant operations. Other water utilities treating Lake Michigan water had been
using PACL for coagulation for over a year. Milwaukee Water Works consulted the other utilities before changing
coagulants and during the use of the coagulant since September 1992. .
During the higher effluent turbidity episodes, greater numbers of particulates passed through the HWTP as
evidenced by the higher turbidity values exiting the plant clearwell. Although a greater number of particulates
passed through the plant, this does not necessarily mean that Cryptosporidium oocysts passed through the plant.
This does, however, suggest that if a large number of oocysts were present in the source water at this time, the
likelihood of passage would increase. Monitoring for this organism was not a common practice in the drinking
water supply industry, nor required at that time in Wisconsin by Wisconsin Administration Code.
Recent research conducted by the WSWRD and by WSWRD sponsored projects and others4'5 have information
regarding removal of Cryptosporidium through filtration. This information may be of useful to the city of
Milwaukee. These studies indicated that strict control of filter effluent turbidity is necessary to achieve good
removal of Cryptosporidium oocysts. The WSWRD research projects involved spiking Cryptosporidium oocysts
into pilot plant facilities. The pilot filters were monitored for particulate removal (by particle counting),
Cryptosporidium oocysts removal, and turbidity reduction.
Particle counting was used to provide additional, more specific information for the evaluation of the filtration
process. Particle counting is a quick and easy way to provide particle distribution information of a water sample.
The Cryptosporidium oocyst is 3 to 5 microns in diameter. Particle counts of particles in the 3 to 5 microns range
were monitored throughout the runs. The filters used in these studies were challenged with water containing high
concentrations of oocysts. Various coagulation conditions were set up prior to the filters to simulate typical water
treatment practices. In every pilot plant study WSWRD conducted, good turbidity reduction was required to
achieve 99.9% reduction of the particulates in the 3 to 5 microns range.
The pilot plant experiments also demonstrated that when effluent water turbidity increased, particles in the 3 to
5 micro range increased along with other particle sizes. The increase in these particles were seen with turbidity
increases as little as from 0.1 NTU or 0.4 NTU. hi several of the filter runs, increases to Cryptosporidium oocysts
were also seen as the filter effluent turbidity increased from 0.1 to 0.4 NTU. This increase in oocysts filter passage
_was not observed in every filter run (in conjunction with the turbidity and particle count increases) but was observed
often enough to suggest that good coagulation/filtration control is necessary for Cryptosporidium removal. These
pilot plant studies were done with the WSWRD inhouse pilot plants using alum as the coagulant.
In one of the WSWRD sponsored field projects4, Cryptosporidium oocysts removal was evaluated at a 600
gpm water plant used solely for experimentation. The plant configuration was designed for two distinct studies, one
of the studies was done with the plant set up for conventional coagulation/settling/filtration. The other study was
done with the plant set up to bypass sedimentation (resulting in a direct filtration mode of operation). For both of
these studies, polyaluminum chloride (PACL) was the coagulant used. In the conventional treatment tests, the raw
water turbidity ranged from 2.5 NTU to 2.8 NTU during the four filter runs tested. During these tests, high levels
of Cryptosporidium oocysts were spiked into the raw water and the filter effluent was monitored for particle counts,
7
-------
turbidity and Ciyptosporidinm. The effluent turbidity for the conventional filtration runs ranged from 0.05 to 0.4
MTU and Ctyptosporidium oocysts removal ranged from 97.8 to 99.6%.
In the direct filtration studies, the influent raw water turbidities ranged from 1 to 10 NTU and the effluent water
turbidities were always below 0.4 NTU. In these filter runs, Cryptosporidium removal ranged from 99.3 to 99.9%.
The inhouse research projects and the field trials have indicated that good (>97%) removal of Cryptosporidium
oocysts is accomplished if proper coagulation control and filtration is done. Controlling filter effluent turbidity was
necessary to achieve good Cryptosporidium removal. In these studies, when effluent turbidity increased, the increase
did not always show higher effluent oocyst concentrations. Although, higher oocyst concentrations were only seen
in the filter effluents that showed a decrease in turbidity removal. In all of these studies, the filters were challenged
with waters artificially spiked with Cryptosporidium.
This research information was provided to the City of Milwaukee to' assist in an understanding and solution of
the Cryptosporidium outbreak and suspected waterborne relationship.
Other Plants Along Lake Michigan
The Howard Avenue Water Treatment Plant was not the only water plant along Lake Michigan that experienced
turbidity problems during March/April 1993. The source water quality in Lake Michigan at that time created
problems for a number of plants. Table 1 lists 7 different plants along Lake Michigan with the type of coagulant
used, the normal effluent and raw turbidities and the peak effluent and raw turbidities hi March/April 1993. Five of
the seven plants listed exceeded 1.0 NTU and one plant even exceeded 5.2 NTU. There were no reported recorded
increases in diarrhea observed at these communities in March and April 1993. The likely assumption is that the
source water at these plants did not contain as much Cryptosporidium-as may have been present in the source water '
at Milwaukee.
Recommendations
The team from the EPA's Water Supply and Water Resources Division presented information that suggested
that the passage of Cryptosporidium through the Howard Avenue Water Treatment Plant happened at the same time
as passage of other particulate matter (increase in turbidity). Actual passage of oocysts during this time frame will
be difficult to be proven analytically. Samples were not taken during that time. There were some ice samples that
were taken during the outbreak period that were analyzed for Cryptosporidium^. The two ice samples contained
CtyptospoHdium concentrations of 13.2 and 6.7 oocysts per 100 liters, respectively. The available data does
suggest that oocyst passage could have occurred. In WSWRD's earlier research, good turbidity removal was
essential to achieve good Cryptosporidium oocyst removal. The recommendation made to the City of Milwaukee
by the research team emphasized the importance to strive for optimal turbidity reduction at all times.
In order to achieve good reduction in turbidity at all times, stringent controls on coagulant/flocculant dosages
are required. This control could be automated or done by operator attention and that determination would be at the
discretion of the water utility. Subtle changes in effluent turbidity from a filtration plant may result in large changes
in particulates passing through the filters that may or may not be associated with pathogens. The goal therefore,
should be to remove these particulates and not have to worry about whether or not they are of concern.
Other suggestions made to Milwaukee included improving the monitoring of each filter's effluent turbidity on a
continuous basis. Particle counting was also suggested as a tool for use in monitoring treatment performance.
Evaluating alternative coagulation chemical, alternate disinfection techniques and stringent turbidity requirement
were presented to the City as means to improve treatment performance.
The suggestions made to the Milwaukee Water Treatment Plants are simple ones, but are ones that can have
significant impact if they are not acted upon.
-------
Acknowledgments
The authors want to acknowledge the cooperation and assistance provided by all individuals involved with the
investigation of the disease outbreak. These individuals include the many City of Milwaukee employees that were
available to answer any questions we posed and provided all data we requested. Assistance was also provided by
representatives from the CDC, the FDA, and the Wisconsin State Agencies. The investigation that was done could
not have been completed without the genuine cooperation of these groups.
References
1. MacKenzie, W. R., et al. "A Massive Outbreak in Milwaukee of Cryptosporidium Infection Transmitted
through the Public water Supply". The New England Journal of Medicine, 331:3:161 (1994).
2. City of Milwaukee. Water Engineering Division, 1991 Annual Report. Public Works Office (1992).
3. Fox, K. R. &: Lytle, D. A. Press Conference Presentation. Milwaukee, Wisconsin, (1993).
4. Nieminski, E. C. &: Ongreth, J. E. "Removing Giardia and Cryptosporidium by Conventional treatment and
Direct Filtration". JAWWA, 87:9:96 (1995).
5. Patania, N. L., et al. optimizing of Filtration for Cysts Removal. Denver, Colorado, (1995).
-------
m
tn
-H
A
u
0)
^
O
3
(0
i
H
O<
H
JM
•H £3
A H
M 55
^^
l| | «J
H Oi
•H jp
•O H
-rl K
"S *"*
3 iH
c^ nj
g
I|,J Q
(4 S
^^i
*rt w
-H S
flj (j)
P« fM
!i
M
3H
•^ fu
R
^ H
gj O
« S5
Coagulant
i
.u
o
w
^
ra
o
in
H
o
o
CO
in
H
i
vo
1
H
t-
H
H
O
1
tn
o
o
•^
•*
CM
CO
I
in
*
H
T-f
1
CNJ
00
a\
rH
in
H
o
CO
9>
co
i
yj
|
m
CM
in
in
o
o
in
CT|
in
CN
1
in
1
«
c-
CM
in
o
o
i
CM
O
O
00
•*
o
in
i
o
rH
1
•H
in
in
H
in
H
O
O^
o^
CM
m
i
**
J
g
UD
^
00
O
P1?
0
1
in
o
o
in
o
H
0
H
i
H
J
r-
I
m
P!
•H
n
en
X!
O
M
al
M
O
U-4
a
a)
T3 rt .
8>-
t)
m
o
•H
111
(4-1
CO
s
a)
"
S
U
10
-------
Steel Crib
Intake
Coagulation
Sedimentation
Basins
Filters
East Clearwell
15 MG
Texas Ave.
Pumping Station
Rapid mix
0.127 MG
West Clearwell
20 MG
FIGURE 1 HOWARD AVE WATER TREATMENT PLANT
11
-------
2.5-
I
1-
0.5-
Turbidity (MTU)
Coagulant Dose
250
-200
-150=1
-100-
(V
a
-5O
Mar 1.93 Mar 12, 93
Mar 23, 93
Date
Mar 31.93 Apr?, 93
Rgure 2 Combined Filter Effluent Turbidity and Coagulant Dose
12
-------
Plant Influent
Settled Water
i ii imnimnii mi 111 in i IIIIIM n in iniiiin 1111 in IIIIIIIIIIIIMII IIM
3/1/93 3/8/93 3/15/93 3/22/93
Date
3/28/93 4/5/93.
Rgure 3 Howard Ave Raw and Settled Water Turbidities
13
-------
Tracking a Salmonella Serovar Typhimurium Outbreak
in Gideon, Missouri:
Role of Contaminant Propagation Modeling
Robert M. Clark , Edwin E. Geldreich, Kim R. Fox,
Eugene W. Rice , Clifford H. Johnson,.
James A. Goodrich, and Judith A. Barnick
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Cincinnati, Ohio 45268
Farzaneh Abdesaken
DynCorp
Cincinnati, Ohio 45268:
Abstract
In early December of 1993, a waterborne disease outbreak was identified in Gideon, MO (U.S.A.). Initially 6-9
cases of diarrhea were identified at a local nursing home. By January 8, 1994,31 cases with laboratory confirmed
salmonellosis had been identified. Seven nursing home residents exhibiting diahrreal illness died, four of whom
were culture confirmed. It was estimated that approximately 44% of the 1,104 residents, or almost 600 people, were
affected with diarrhea between November 11 and December 27, 1993. A system evaluation was conducted in which
s. computer model (EPANET) was used to develop scenarios, to explain possible contaminant transport in the
Gideon system. It was concluded, based on this analysis, that the outbreak resulted from bird contamination in a
municipal water storage tank.
14
-------
Introduction
In early December of 1993 a waterbome disease outbreak was identified in Gideon, MO, U.S.A. Initially 6-9
cases of diarrhea were reported at a local nursing home raising the possibility of a waterborne outbreak. After an
initial Missouri Department of Health (DOH) investigation, the Missouri Department of National Resources (DNR)
was contacted and water samples were taken at various points in the system. Several samples were positive and
yielded 1-6 total coliforms (TC) per 100 ml and a few samples were fecal coliform (FC) positive. Several other
samples yielded results that were too numerous to count (TNTC) for coliforms and were also FC positive.
Original speculation regarding the cause of the outbreak focused on a water tank located on private property.
This tank, connected via a backflow prevention valve to the city water system, was used for fire protection at the
"Cotton Compress", a local cotton baling industry.
The municipal system has two elevated tanks (189 m3 [50,000 gal] and 378 m3 [1 00,000 gal]). A tank
inspector observed birds roosting on the Cotton Compress Tank (378 M3 [100,000 gal]) and the largest city owned
tank which has a broad flat roof and is an ideal roosting location.
On January 4, 1994, an EPA field team, in conjunction with CDC and the State of Missouri initiated a field
investigation which included a sanitary survey and microbiological analyses of samples collected on site. A system
evaluation was conducted in which a computer model (EPANET) was used to develop various scenarios to explain
possible contaminant transport in the Gideon system. An analysis of the propagation of chlorine residuals after the
outbreak was also developed. The results of this investigation are presented in this paper.
The Gideon, Missouri Outbreak
Gideon, Missouri is located in Anderson Township, in New Madrid County, which is in Southeastern
Missouri in the United States. The topography is flat and the predominant crop is cotton. In 1990, the population
of Gideon was 1,104 with a median income of $14,654 (25% of the population was below the poverty level). The
unemployment rate was 11.3%. Major employers in Gideon are a nursing home with 68 residents and a staff of 62
and the Gideon schools with 444 students (kindergarten through 12th grade).
The Gideon municipal water system was originally constructed hi the mid-193 Os and obtains water from two
adjacent 396 m (1,300 ft) deep wells. The well waters were not disinfected at the time of the outbreak. The
distribution system consists primarily of small diameter (5,10, 15 cm [2, 4 and 6 inch]) unlined steel and cast iron
pipe. Tuberculation and corrosion are a major problem hi the distribution pipes. Raw water temperatures are
unusually high for a ground water supply (14°C [58°F]) because the system overlies a geologically active fault.
Under low flow or static conditions the water pressure is close to (3.5 kgf/cm2 [50 psi]). However, under high flow
or flushing conditions the pressure drops dramatically as will be discussed later. These sharp pressure drops are
evidence of major problems in the Gideon distribution system. In the Cotton Compress yards, water is used for
equipment washing, in rest rooms and for consumption. The pressure gradient between the Gideon system and the
Cotton Compress system is such that the private storage tank will overflow when the municipal tanks are filling.
To prevent this from occurring a valve was installed in the influent line to the Cotton Compress tank. This same
pressure differential keeps water in the Cotton Compress tank unless there is a sudden demand hi the warehouse area.
The entire Cotton Compress water system is isolated from the Gideon system by a back flow prevention valve.
There are no residential water meters hi the Gideon system and residents pay a flat service rate ($11.50 per month)
for both water and sewage service. The municipal sewage system operates by a gravity flow with two lift stations
and as of December 31, 1993 served 429 households.
Identification of the Outbreak
On November 29th, the Missouri Department of Health (DOH), became aware of two high school students from
Gideon who were hospitalized with culture confirmed salmonellosis. Within two days five additional patients
15
-------
living in Gideon were hospitalized with salmonellosis (one student, one child from a day care, two nursing home
residents, and one visitor to the nursing home). The State Public Health Laboratories identified the isolates as
dulcitol negative Salmonella and the CDC laboratories identified the organism as serovar typhimurium. Interviews
conducted by the DOH suggested that there were no food exposures common to a majority of the patients. All of
the ill persons had consumed municipal water.
The Missouri Department of Natural Resources (DNR) was informed that the DOH suspected a water supply link to
the outbreak. Water samples collected by the DNR on December 16th were positive for FC. On December 18th,
the city of Gideon, as required by the DNR, issued a boil water order. Signs were posted at city hall, in the grocery
store, and two area radio stations announced the boil water order.
Several water samples collected by DNR on December 20th were also found to be FC positive. On December 23rd,
a chlorinator was placed on line at the city well by DNR, and nine samples were collected by the DOH and DNR
from various sites in the distribution system. None of the samples contained chlorine but one sample collected from
a fire hydrant was positive for dulcitol-negative Salmonella serovar typhimurium (Table 1). Figure 1 shows the
location of the sampling points and identifies those points that yielded fecal positive results from the DNR/DOH
survey. It is interesting to note from Table 1 that most of the sampling points which were coliform positive, were
also fecal coliform positive. Multiple entries in a column indicate repeat samples.
The Missouri DOH had informed the CDC about the outbreak in Gideon in early December and requested
information about dulcitol negative Salmonella serovar typhimurium. On December 17th, DOH informed CDC that
contaminated municipal water was the suspected cause of the outbreak and (on December 22nd) invited CDC to
participate in the investigation. A flyer explaining the boil order, jointly produced by DOH and DNR, was placed
in the mailbox of all of the homes in Gideon on December 29th and the privately owned water tower was physically
disconnected from the municipal system on December 30th. DNR mandated that Gideon permanently chlorinate
their water system. At the end of the study EPA provided input to DNR on the criteria necessary to lift the boil
water order.
Through January 8, 1994 the DOH had identified 31 cases with laboratory confirmed salmonellosis associated
with the Gideon outbreak. The State Public Health Laboratories identified 21 of these'isolates as dulcitol negative
Salmonella serovar typhimurium. Fifteen of the 31 culture confirmed patients were hospitalized (including two
patients hospitalized for other causes and who developed diarrhea while in the hospital). The patients were admitted
to 10 different hospitals. Two of the patients had positive blood cultures, 7 nursing home residents exhibiting
diarrhea! illness died, four of whom were culture confirmed (the other three were not cultured). All of the culture
confirmed patients were exposed to Gideon municipal water.
Ten culture confirmed patients did not reside in Gideon but all traveled to Gideon frequently to either attend
school (8), use a day care center in town (1), or work at the nursing home (1). The earliest onset of symptoms in a
culture confirmed case was on November 17 (this patient was last exposed to Gideon water on November 16). A
CDC survey indicated that approximately 44% of the 1,104 residents, or almost 600 people, were affected with
diarrhea between November 11 and December 27, 1993 in Gideon, MO. Non-residents who drank Gideon water
during the outbreak period experienced an attack rate of 28% (1).
Possible Causes
The investigation clearly implicated consumption of Gideon municipal water as the source of the outbreak.
Speculation focused on a sequential flushing program conducted on November 10 involving all 50 hydrants in the
system. The program was started in the morning and continued through the entire day. Each hydrant was flushed
for 15 minutes at an approximate rate of 2.8 m3/min (750 gal per min). It was observed that the pump at Well 5
was operating at full capacity during the flushing program (approximately 12 hours) which would indicate that the
municipal tanks were discharging during this period. The flushing program was conducted in response to taste and
odor complaints.
16
-------
It was hypothesized that taste and odor problems may have resulted from a thermal inversion that may have
taken place due to a sharp temperature drop prior to the day of the complaint. If stagnant or contaminated water were
floating on the top of the tank, a thermal inversion could have caused this water to be mixed throughout the tank
and to be discharged into the system resulting in taste and odor complaints (2). As a consequence, the utility
initiated a city-wide flushing program. Turbulence in the tank from the flushing program probably stirred up the
tank sediments which were transported into the distribution system. It is likely that the bulk water and/or the
sediments were contaminated with Salmonella serovar fyphimurium.
As mentioned earlier, during the EPA field visit, a large number of pigeons were observed roosting on the roof
of the 378 m3 (100,000 gal) municipal tank. Shortly after the outbreak, a tank inspector found holes at the top of
the Cotton Compress tank, rust on the tank, and rust, sediment and bird feathers floating in the water. According to
the inspector, the water in the tank looked black and was so turbid he could not see the bottom. Another
inspection, conducted after EPA's field study, confirmed the disrepair of the Cotton Compress tank and also found
the 378 m3 (100,000 gal) municipal tank in such a state of disrepair that bird droppings could, in the opinion of the
inspector, have entered the stored water. Bird feathers were in the Vicinity or in the tank openings of both the cotton
compress and the 378 m3 (100,000 gal) municipal tank.
It was initially speculated that the backflow valve between the Cotton Compress and the municipal system
might have failed during the flushing program. After the outbreak, the valve was excavated and found to be working
properly. The private tank was drained accidently after the outbreak during an inspection so it was impossible to
sample water in the tank bowel. However, sediment in the private tank contained Salmonella serovar typhimurium
dulcitol negative organisms as did samples found in a hydrant sample and culture confirmed patients. The
Salmonella found in the street hydrant (304 6th Street) matched the serovar of the patient isolate when analyzed by
the by the CDC laboratory comparing DNA fragments using pulse field gel electrophoresis. The isolate from the
tank sediment, however, did not provide an exact match with the other two isolates.
EPA Field Study
During the course of a sanitary survey and cross connection survey, several locations were identified that were
potential cross connections and it was observed that the well head areas were not properly protected. An extensive
microbiological survey was conducted and the system was found to have high levels of heterotrophic plate count
bacteria (HPCs) in various portions of the system.
Sanitary Survey
Although the investigation did not constitute a complete sanitary survey, it did suggest obvious areas of
possible concern which are listed below. Figure 1 shows the sampling locations utilized for EPA's microbiological
survey.
Well Head Protection
It was found that agricultural areas surrounding the wells drain across the well heads subjecting the area to large
amounts of runoff. The wells (Wells #5 & #6) are deep (>396.2 m [1300 ft]) artesian and thought to be protected by
the underlying geological formations. This protection could not be checked during the field study, but suggested
that the integrity of the aquifer should be investigated. The two wells operate alternately on a monthly basis.
When one well is pumping, the other is turned off. There did not appear to be any backflow devices (air gap
breaks)to prevent a siphon from being placed on the unused well pump when the other pump is turned on. This is
of concern because the water meter and valve for Well #6 is located in a pit that is reported to flood routinely. The
pit fills with runoff from the surrounding fields and must be pumped out before the meter can be read. Water in the
pit could be pulled into the distribution system and the contaminated water distributed throughout the system.
When Well #6 is pumping, a venturi effect could also pull some of the pit water into the system.
17
-------
There were some inconsistencies with regard to the free chlorine residual found at several locations. At the
nursing home, the residual was 0.2 mg/L, while only a block away on Whiterow, the measured amount was 1.08
mg/L indicating a closed pipe or hydraulic obstructions in the system.
Microbiological Survey
In an effort to characterize the microbial quality and to evaluate the effectiveness of applied disinfection as a
remedial measure, samples were taken on Jan. 5 and Jan. 6, 1994 from selected sites in the system. Sampling
locations included the source water (two artesian wells), fire hydrants near the water tanks, fire hydrants in areas of
static water, fire hydrants near homes where illness occurred, tap water from a residential service line and taps in the
restrooms of the nursing home and elementary school (Figure 1). In addition, surface drainage around well meter
boxes, water from a 30-foot deep private well, drainage water and sediment from the private tank and one historical
sample collected during the outbreak period were also examined. The samples were collected and shipped on ice via
overnight carrier to DWRD's Laboratory and were analyzed on the day of receipt. Table 2 summarizes the sampling
history.
Samples were analyzed for total coliform bacteria by the membrane filter technique using m-Endo LES agar and
by the presence absence procedure employing PA broth (3). Fecal coliform levels were determined by the elevated
temperature method using mFC agar in the membrane filter procedure (3). The spread plate technique using R2A
agar incubated at 25*C for 7 days was used to enumerate heterotrophic plate count bacteria. All turbid enrichment
cultures exhibiting an orange to red color were streaked onto bismuth sulfite agar, brilliant green agar and xylose
lysine deoxycholate agar (3). Target colonies were characterized biochemically and by serological examination.
Biochemical characterization of coliform and Salmonella isolates was performed using the API 20E multi-test kit
(bio Merieux Vitek, Inc., Hazelwood, MO 630422395) (3). Two liter samples for Salmonella analysis were
concentrated by the membrane filter or diatomaceous earth procedure. Selective enrichment of the concentrated
samples was conducted in selenite cystine broth incubated at 41.5°C (3,4). Serological identification was
accomplished using Salmonella polyvalent somatic (0) and polyvalent flagellar (H) antisera (Difco Laboratories,
Detroit, MI 48232-9830) (5). Appropriate positive and negative controls were included for all analyses.
Coliform Sampling Protocol
Source Water
Microbiological samples collected on January 6, 1994 from the wells contained no detectable coliforms per 100
mL and less than 200 heterotrophic bacteria per mL (Table 3). A sample of sump water in the meter well was
examined bacteriologically to characterize the extent of this contamination threat. As anticipated, the sample
contained a very high density of heterotrophic bacteria, estimated to be greater than 100,000 organisms per mL.
This excessive bacterial population interfered with establishing the density of coliform bacteria present in the
sample. A P/A coliform test yielded positive results and indicated that coliform bacteria were present.
Tank Storage
Water quality in the tanks was investigated by sampling at the nearest fire hydrant during a draw-down period
from both the 378 m3 (100,000 gal) and 189 m3 (50,000 gal) municipal tanks (samples 14 and 15 in Table 3).
Enterobacter cloacae and a strain ofAeromonas hydrophila were not detected at the nearest fire hydrant (Sample No.
15) to the 378 m3 (1 00,000 gal) tank by standard coliform procedures but were detected in the Salmonella-selective
differential medium incubated at an elevated temperature (41 °C.). These results suggest heterotrophic bacteria could
have interfered with coliform detection. The nearest hydrant to the 189 m3 (50,000 gal) storage tank contained a
significant amount of particulates that limited the membrane filter sample to 20 mL. Three isolates ofKlebsiella
oxytaca were recovered in the sample. The presence of particulate matter suggests that either this hydrant was not
18
-------
adequately flushed or that the tank water supply was static. This finding of course was consistent with the concept
of stagnant water in the tanks.
Pathogen Detection
All samples were also examined for Salmonella since this pathogen was the suspected agent in the Gideon
outbreak having been found in the stool of one patient and on December 16, 1993 in water from the fire hydrant on
6th Street (304 - 6th Street, Sample No. 2 in Figure 1). No Salmonella was detected in any of the samples collected
on January 5 and January 6 in the public water system, most likely because disinfection was initiated on December
23, 1993. Salmonella was found in the sediment collected on January 5 from the riser pipe of the Cotton Compress
water storage tank. As mentioned previously, the water tank had been physically disconnected from the public water
supply on December 30, 1993 and accidentally drained in the process. Residual water in the riser pipe did not
contain detectable Salmonella but both residual water and sediment samples did contain the coliform Enterobacter
cloacae. Sources of Salmonella and coliforms may have been from feces of pigeons observed to be roosting in the
tower vents. Among bird populations, there are always a few individuals who are shedding these organisms
(6,7,8,9).
System Evaluation
The purpose of the systems evaluation was to study the effects of distribution system design and operations,
demand, and hydraulic characteristics on the possible propagation of contaminants in the system. Given the
evidence from the survey and the results from the valve inspection at the Cotton Compress, it was concluded that
the most likely contamination source was bird droppings in the large municipal tank. Therefore, the analysis
concentrated on propagation of water from the large municipal tank in conjunction with the flushing program. This
does not rule out other possible sources of contamination, such as cross connections.
The systems layout, demand information, pump characteristic curves, tank geometry, flushing program, etc. and
other information needed for the modeling effort was obtained from maps and demographic information and
numerous discussions with consulting engineers and city and DNR officials. EPANET, a hydraulic/water quality
program, was used to conduct the contaminant propagation study (10).
Network Layout
Overlays were produced from the waterworks distribution system map (May 1981) obtained from the City of
Gideon with which to begin identifying links and nodes, according to the EPANET format. Links were first
identified utilizing a city map. Nodes were then identified based on pipe intersections, major changes in alignment,
or dead ends. Hydrants were identified as separate nodes in order to provide the basis for replicating the flushing
program. Houses and businesses were aggregated to provide system demand. A few nodes were inserted along
uninterrupted links to accommodate hydraulic demands from the houses. The three tanks and the well in operation
in November were identified separately according to EPANET requirements.
Node Characteristics
Elevations for the nodes were derived from 1978 U.S. Geological Survey (USGS) Madrid topographic map.
Hydraulic demands for each node were calculated based on the number of houses assigned to each node. Household
usage was calculated based on CDC reports of 1104 people living in 429 homes (2-6 people/house) and an
assumption of 0.28 m3 (75 gallons) of water used per person per day. The daily water use patterns (drinking,
bathing, washing, cooking, lawn watering) developed in an earlier study in Cabool, MO were used for the residences
and the schools (11). Daily demands for the nursing home were based on the CDC report of 68 residents and 62
staff assuming the same consumption rate. A lower rate of 0.076 m3 (20 gallons) used per student per day was
estimated for a node serving the school area. The calculated consumption rates are very close to those used in
19
-------
Cabool, MO, where meter and water use information was available for each residence, the nursing home, and
schools. Pump records indicate an average of 492 m3 (130,000 gal) of water pumped daily during the outbreak in
November. Daily consumption estimates using the per capita assumptions were approximately 77% of the daily
ptimpage. This difference is not unexpected nor excessive given experience in other communities. Water leaks
and/or unaccounted for users could account for this discrepancy as well as low per capita estimates. Table 4
summarizes some of the key values used as part of the EPANET program.
Tank and Pump Characteristics
Based on the EPANET format, the tanks are identified by number. Tank 200 (T200) is the 189 m3 (50,000
gal) municipal tank; Tank 300 (T300) is the 378 m3 (100,000 gal) municipal tank; and Tank 400 (T400) is the 378
m3 (100,000 gal) cotton compress tank.
It was learned that when pressure from a guage near T200 drops below 4.22 kgf/cm2 (60 psi), the pump at the
well is automatically turned on. Several pressure studies conducted at selected hydrants in the system were used to
help calibrate EPANET. A pump head curve was provided by DNR. Table 5 contains the pressure drop readings
obtained from tin's study by attaching a pressure gauge to an outside faucet near the hydrant to be tested. The
hydrants were flushed for 15 minutes at 2.8 m3/min (750 gpm) and the pressure recorded.
Link Characteristics
Links were calculated from the 2.54 cm to 61 m (1" to 200') scale map. Diameters were derived from the 1951
system map. Roughness coefficients were estimated based on the general age and location of the pipes and from the
pressure readings shown above.
System Performance
EPANET was calibrated by simulating flushing at the hydrants shown in Table 5, assuming a discharge of 2.8
m3/min (750 gpm) for 15 minutes. The "C" factors were adjusted until the head loss in the model matched head
losses observed in the field.
The hydraulic scenario was initiated by "running" the model for 48 hours. The water level reached 122 m (400.59
0) in T400 (Cotton Compress tank), 122 m (400.63 ft) in T300 and 122 m (400.66 ft) in T200. At 8 o'clock in the
morning of the third day the simulated flushing program was initiated by sequentially imposing a 2.8 m3/min (750
gpm) demand on each hydrant, 1 through 50, for 15 minutes. The entire process consumed 12.5 hrs. Utilizing the
TRACE option in EPANET the percentages of water from both municipal tanks were calculated at each node over a
period of 72 hrs. Based on the findings from excavating the backflow prevention valve the impact of flows from
T400 (Cotton Compress) were not considered in the simulation.
It was found from the simulation that the pump operates at over 3 m3/min (800 gal/min) during the flushing
program and then reverts to cyclic operation thereafter. The tank elevation for both municipal tanks fluctuate, and
both the tanks discharged during the flushing program. At the end of the flushing period nearly 25% of the water
from T300 has passed through T200 where it was again discharged into the system.
Pressure drops during the flushing program were simulated at the hydrants used for calibration. The model
predicts dramatic pressure drops during the flushing program and nodes 4 and 49 show negative pressures which can
be considered as zero. It was felt that, based on the information available, these results replicated the conditions that
existed during the flushing program closely enough to provide the basis for an analysis of water movement in the
system.
20
-------
Water Movement and Outbreak Pattern
Perhaps the key to the Gideon outbreak pattern and a major clue as to the most likely source of contamination
is given by tracing the movement of water from both municipal tanks during the flushing program. Figure 2 shows
the hourly movement of water from T300 in the system during the first four hours of the flushing period.
Water movement from T200 tends to dominate the area immediately around T200 during the first few hours of
the flushing period until it is drawn down. Water from T300 initially supplies most of the northern and western
portions of the system. Almost all of the water in the southern and eastern portions of the system is supplied by the
well.
Water movement was determined at 6 hour intervals for both T200 and T300 respectively starting at a point 24
hours into the simulation period or when normal operation has resumed. Water from both T200 and T300 reaches
virtually all of the system under normal operation with approximately 25% of the water in T200 passing through the
presumably contaminated T300.
The percent of water from T300 was determined at selected nodes in the system. During the flushing program
there were periods where 100% of the water at nodes near the school, the hydrant yielding the positive Salmonella
serovar lyphimurium (Sample No. 2 in Figure 1) and the nursing home are from T300. At all nodes water from
T300 is present at some time during the 3-day simulation period.
Contaminant Propagation
Data from the simulation study, the microbiological surveillance data in Tables 1 and 2 and the outbreak data
can be utilized to provide insight into the nature of both general contamination problems in the system and into the
"outbreak" itself. Table 1 and Figure 1 and the water movement patterns showed that the majority of the special
samples which were coliform and fecal coliform positive occurred at points that lie within the zone of influence of
T200 and T300. During both the flushing program and for large parts of normal operation, these areas are
predominately served by tank water which might lead us to believe that the tanks are the source of the fecal
contamination since there were positive FC samples prior to chlorination.
Data from the early cases in combination with the water movement data was utilized to infer the source of the
outbreak. Using data supplied by CDC and the water movement simulations, an overlay of the areas served by
T200 and T300 during the first 6 hours of the flushing period and the earliest recorded cases was created as shown in
Figure 3. As can be seen in Figure 3, the earliest recorded cases and the positive Salmonella hydrant sample were
found hi the area that was primarily served by T300, but outside the T200 area of influence, during the flushing
period. One can conclude that during the first six hours of rne flushing period the water, which reached the residence
and the Gideon School, was almost totally from T300. Therefore, it is logical to conclude that these locations
should experience the first signs of the outbreak, which makes a strong circumstantial case for T300 as the
contamination source.
Figure 4 displays the increase in the number of absentees from the Gideon schools during the outbreak period.
As can be seen, there was a sudden rise in absentees on Nov. 12, two days after the hydrant flushing program.
Figure 5 shows the progress of outbreak cases in the first few days of the outbreak. As mentioned, the disease
progressed in an apparent random manner after the first occurrences in the center of the city.
These arguments support the hypothesis that the sudden drop in temperature on the night of November 9, 1993
caused a "turnover" of water in the tank thereby mixing the contaminated portion of the tank water with the
relatively clean portion of the water column. This probably caused taste and odor complaints which resulted in the
rigorous flushing program on November 10th. Taste and odor problems had been reported previously but only a
limited flushing program in the area of complaints was attempted. The sudden discharge of the water column
probably stirred up sediments in the bottom of this tank. Mixing of the water column and the contaminated
sediments no doubt resulted in the outbreak.
21
-------
Given the history of coliform violations and taste and odor complaints, it is reasonable to question whether or
not this outbreak was a one-time event or whether the contamination had been occurring routinely or at least
periodically. These results may also explain some of the previous coliform violations found in the utility records.
; , I '. •
Based on the results of the DNR/DOH sampling program, it is likely that the contamination had been occurring
over a period of time, which is consistent with the possibility of bird contamination. If the cause was a single event
the contaminant would most likely have been "pulled" through the system during the flushing program.
Summary and Conclusions
The density of a pathogenic agent in contaminated water supply is subject to much variation because of
contaminate input, dilution effect of water, and any available disinfectant residual. In this case history, there was no
applied disinfectant and the density of Salmonella in the stratified water supply in the tanks was unknown. However
persistence of this pathogenic agent was extended by the static nature of the stratified water. Laboratory study of the
persistence of this Salmonella strain in the Gideon water supply demonstrated that the pathogenic agent was only
reduced in density by 30% during a 4-day period at 15°C. Thus, with repeated new inputs of S. typhimurium from
infected pigeons, there could be a continuing high level of this pathogen present. Obviously, this situation provided
a sufficient cell density to be an infective dose to many people in the community once the tank water was destratified
by abrupt drop in air temperature and released to the pipe network through the flushing activity near the tanks.
Infective dose levels for different Salmonella species may vary from 101 (S. typhi) to more than 10s cells (S.
typhimurium). In addition, there are a number of intervening factors which explains why not all people exposed to a
contaminated water source will become ill. Part of this variation in human response relates to water intake per day
and the nature of individual body defense against pathogen colonization. Individuals at greatest risk are infants,
senior citizens, people taking excessive stomach antiacid medicines, alcoholics, persons exposed to radiation or
chemical therapy and those with acquired immune deficiencies (AIDS victims).
The largest waterborne outbreak caused by S.typhimurium (phage type 2) contamination of a public water
supply occurred during late May and early June 1965, in Riverside CA. In this outbreak, more than 16,000 persons
became ill with acute gastroenteritis, at least 70 patients were hospitalized and three of these individuals died
(12,13). There were several similarities between this outbreak and the Gideon episode: source water was
groundwater; water supply was not disinfected; initial cases appeared in one defined area of the distribution system;
S. typhimurium was detected in six distribution samples in Riverside and one fire hydrant in Gideon; the same
Salmonella found in the water supply was also detected in patients with acute gastroenteritis; and routine coliform
sample collections were not representative of water quality over the entire pipe networks. Few samples contained
coliforms in either system. In Riverside, there was some evidence of a high heterotrophic bacterial count that could
have interfered with coliform detection. While there was no quantitation of the Salmonella isolated from the fire
hydrant sample in Gideon, one sample collected from the Riverside distribution system did contain 17 Salmonella
per liter (12-15). While this figure is conservatively low because of recognized limitations with Salmonella
quantitative methodology, it does demonstrate that pathogen infective doses cited for food outbreaks may not be
realistically applied in waterborne outbreak occurrences. For example, few residents of Riverside would have
ingested a liter of drinking water each day, so it seems likely that the dose of salmonellae to cause infection among
susceptible individuals in the community in both waterborne outbreaks may be much smaller than the dose often
cited in the literature (12).
It is reasonable to assume that given the pattern and rapidity of the spread of Salmonella serovar lyphimurium in
Gideon in November and December of 1993 that it was, in fact, a waterborne disease outbreak. The cause of the
outbreak is not obvious but was likely associated with bird contamination in the storage tanks. Given that the only
valve connecting the Cotton Compress and the Gideon distribution system was found to be closed, it is most likely
associated with the largest municipal tank which provided an excellent roosting place for pigeons. A tank inspector
observed that the tank was covered with bird feathers, dirt and droppings. The inspectors also observed that the
vents in the tank were designed in such a way as to allow for the possibility of contamination. It was also found
that the strain of Salmonella serovar typhimurium isolated from a patient survived for several days. This finding
22
-------
supports the hypothesis that the source of organisms was in the public water supply tanks and pulled through the
system during a vigorous flushing program that started near the tanks.
The authors believe that the lack of proper support for drinking water infrastructure repair and maintenance was a
contributing factor to the outbreak. In an effort to fund the utility operation, the City should implement a proper
billing system based on water use. F
Acknowledgments
The authors would like to acknowledge Ms. Jean Lillie, Ms. Diane Routledge, Ms. Nancy Frazier, Mr. Steve
Waltrip and Mr. Richard Findsen of the Drinking Water Research Division for their assistance in preparing this
manuscript. Mr. Terry Covert of the Bacteriology Branch of the Microbiology Research Division^ Environmental
Monitoring Systems Laboratory assisted in the bacteriological analysis conducted on the Gideon samples.
Many individuals were involved in obtaining data for the Gideon outbreak analysis contained in this paper
especialty Dr Frank Angulo of the Centers for Disease Control and Prevention (CDC) in Atlanta, Georgia and Mr
John E. Hill of the Missouri Department of Natural Resources (DNR).
During the investigation many citizens of Gideon itself and owners of industrial properties (active and
abandoned), were extremely cooperative during our cross connection surveys. The authors would like to
acknowledge the entire Gideon city staff, including the Mayor, City Clerk, police, etc. and especially Mr Bill
?™\ ^°TS th~ W56r Tmt°r f°r Gideon' Eddie Waldrid§e' Sue Tippen of the Missouri Department of Health
(DOH) and Jim Bradley of the New Madrid County Health Department were some of the first state officials to
respond to reported illnesses in Gideon. Dr. Don Sharp on loan to the Missouri DOH from CDC assisted in these
investigations. Messrs. Jack Baker, Jeff Pinson, Environmental Specialists and Mr. Gary Gaines Regional
Director from the Missouri DNR South Eastern Regional Office spent numerous hours in Gideon assisting the City
L/c.
investigating the water system. Delbert Silman of Silman Painting and Contracting was the first person to climb
the water towers. He verbally reported conditions, and added bleach to the tanks. •
* )£.' J°MU Gilfert' retired city employee and Mr. Charles Church, local carpenter/plumber and 'past employee of
toe City utility provided many useful details on the city distribution system. Mike Logsdon of the Jefferson City
Office of the Missouri DNR who is a specialist in investigation of cross connections made several trips to Gideon
from Jefferson City. The authors also owe special thanks to Mr. Clyde Zelch, a local tank inspector who climbed
inspected and photographed all three water tanks, and cleaned, disinfected and made temporary repairs to the two '
city owned towers.
The authors wish to thank KFVS-TV, the local CBS affiliate in Cape Girardeau, which was very cooperative
and factual during the coverage of the outbreak. uupct
-------
4. Spino, DF. Elevated temperature technique for the isolation of Salmonella from streams. Appl. Microbiol.
1966: 14:591-596.
5. Bacteriological Analytical Manual 7th ed. 1992. Food and Drug Administration, AOAC International
Arlington, VA.
6. Fennel, H. James, DB. and Morris, J. Pollution of a storage reservoir by roosting gulls. J. Soc, Water Treat.
Exam. 1974: 23:5-24.
7. Koplan, JP. Deen, RD. Swanston, WH. and Tota, B. Contaminated roof-collected rainwater as a possible
cause of an outbreak of salmonellosis. J. Hyg. (Lond); 1978: 81:303-309.
8. Jones, F. Smith, P. and Watson, DC. Pollution of a water supply catchment by breeding gulls and the
potential environmental health implications. Jour. Institution Water Engrs. Sci. 1978: 32:469-482.
9, Mutler, DF. The pharmacokinetics of dihydrostreptomysin sulfate in domestic pigeons. Tierarztl Prof. 1990:
18:377-381.
10. Rossman, LA. 1994. EPANET Users Manual, Drinking Water Research Division, USEPA, Cincinnati, OH
45268.
11. Geldreich, EE. Fox, KR. Goodrich, JA. Rice, EW. Clark, RM. And Swerdlow, DL. Searching for a water
supply connection in the Cafaool, Missouri disease outbreak of Escherichia co/z 0157:1-17, Water Research
1992: 84:49-55.
12. Collaborative Report. 1971. A Waterborne Epidemic of Salmonellosis in Riverside, California, 1965.
Epidemtological Aspects. Amer. Jour. Epiderniol. 93:33-48.
13. Boring 111, J.R., Martin W.T. and Elliott L.M. 1971. Isolation of Salmonellatyphi-murium from Municipal
Water, Riverside, California, 1965. Amer. Jour. Epidermiol 93:49-54.
14. Ross, E.G., Campbell K.W., and Ongerth H.J, 1966. Salmonella Typhimurium Contamination of Riverside,
Calif., Supply. Jour. Amer. Water Works Assoc., 58:165-174.
15. Greenberg, A.E. and Ongerth HJ. 1966. Salmonellosis in Riverside, Calif. Jour. Amer. Waterworks
Assoc., 58:1145-1150.
24
-------
1
CO
1
O
Z
CO
i
^
UJ
m
8
Q
o:
S }H
UJ 4
i °
8
LU
Q.
U_
CO
§
^
^
2
0
u.
§
r-'
UJ
n
".
jgj
~5>
co S
W U.
«^
ii . .
2* u.
o
J
2 «2 V
w 2 v
0
Q) O
O ^ i i
;«j
i§
if '7
§1
!O ^f
S ^ ' '
0
ii
i§ ' '
u
_i
ii
1§
o
So
\1 1- ii
« ^
*~O
_i
8 9
5§ , ,
*"o
S ^^ ir
J 8 II ||| i
S O « r». £2 §2 > • i
e (n •»- Q «j -r 5
^"1 •"•! V^] ««4
• i i • i -i «- ,• • • ••• i i . , , ,,
r-
1 ''''s^11*^111 V*VV
V
. -r- « £>
CO V ^ 1 1 "y I 1 • 1 1 1 1 1 1 1 Illl
1 • 'y" 1 1 1 1 1 1 1 1 1 1 1 Illl
i T7 » i i i i i i i i i i i iiii
"
' v I i i i i i i i i i i i 1111
0>
I 111 Illifiililillllili
a
-------
TABLE 2. GIDEON, MISSOURI - SUMMARY OF SAMPLE ANALYSIS*
Number of Samples
10
8
1
Dates Collected
January 5,1994
Januarys, 1994
January 12,1994
Dates Received & Analyzed
Januarys, 1994
January 7,1994
January 13,1994
* All samples were analyzed for total coliform bacteria, fecal coliform
bacteria, heterotrophic plate count bacteria and Salmonella.
-------
TABLE 3. MICROBIOLOGICAL CHARACTERISTICS OF THE GIDEON, MO PUBLIC WATER SUPPLY
Sampl
No.
12
13
15
14
1
2
3
4
5
22
20
21
10
8
6
23
24
(1)
N.D.=
* =
le Sample FreeCI2
Location Date MgVL
SOURCE WATERS
Well #5 Jan. 5 0.00
Well #6 Jan. 5 N.D.
MUNICIPAL WATER TANKS
100,000 gal Jan.5 0.11
50,000 gal Jan.5 0,11
DISTRIBUTION SYSTEM HYDRANTS
So. Anderson Jan. 5 0.04
304-6th St. Jan. 5 0.02
Total Fecal
Coliform Coliform
HPC per 100 mL per 100 mL
permL MF P/A
<10 <1
140 <1 - <1
560 <1 - <1
7.000 20** - <5**
45 <1 - <1
20 <1 - <1
120 Haven Ave. Jan.5 0.02 25,000 1 - <1
Jefferson St. Jan.5 0.11 24,000* 10 - <1
Street Hydrant
2nd Street Jan.5 0.04 24,000* <1 + <1
SERVICE TAPS - RESIDENCE. PUBLIC BUILDINGS
122WhiterowSt. Jan.6 1.08
Nursing Home Jan.6 1.12
Grade School Jan.6 0.21
COMPRESSOR WATER TANK
Drainage Jan.5 0.00
Sediment (1) Jan.5 0.00
Fire Hydrant Jan.5 0.15
SURFACE WATER
Payne Well Jan.6 N.D.
Meter Box Sump Jan. 12 N.D.
Salmonella serovar tvphimurium isolated
Not Done
Chromobacterigrn detected in HPC counts
5 <1
20 <1 - <1
9,000 <5** - <5**
320 <100** + <100**
95 <1 - <1
6,600 1 + <1
>100,000 <1 + <1
- soil organism
Species
Kleb. oxvtoca
C. freundii
C.freundii
E. cloacae
C. freundii
E. cloacae
E. cloacae
E.agglomerans
E.intermedium
C. freundji
** _
sample volume was limited (1,10 or 20 mL) due
to heavy particulates
27
-------
TABLE 4. ASSUMPTIONS USED IN ANALYSIS
Item
No. of Homes in Gideon
No. of Residents in Gideon
Persons/Households in Gideon
Average Daily Consumption*
189 m3 (50.000 Gal) Tank CT200)
Height
Diameter
378 m3 (100,000 Gal) Tank (T300)
Height
Diameter
Cotton Compress
378 m3 (100,000 Gal) Tank (T400)
Height
Diameter
Value
429
1,104
2.6
492 m3 (130,000 Gals)
8.8 m (29 ft)
5.5m(1BH.)
7J5m (24.5 «0
9.1m (30 ft)
10.1m (33 ft)
7J3m(7.3ft)
drinking,'bathing, washing, cooking, lawn watering, etc.
TABLE 5. PRESSURE TEST RESULT
Hydrant
Number
4
9
49
Pressure
Static Dynamic
psi kgf/cm* psi kof/cm2
58 0.22 7 0.026
53 0.20 8 0.033
50 0.19 18 0.068
2.8
-------
Park
Gideon Airport
Swiping Point
Figure 1.
Indicates Fecal and Tdal Conform Posioves
From DNR/DOH Survey
Cotton Compress Tank
100,000 Galon llunUpal Tank
/\ ^/ \\ ' C 50,000 Galon Munkafeal Tank
'' I"! Cotfonnf\i»Wve but not Fecal CoUtomi Positive
I—J From ONR/DOH Survey
^\ Sample results from EPA survey
DNR, DON, and EPA Sampling Results
29
-------
(Hour 1)
(Hour 2)
1- Well No. 5
2 • T200 (189 cu.m (50,000 gti) Mun Tank)
3 • T300 (37S cu.m (100,000 gal) Uun Tank)
4 - T400 (378 cu. m (100,000 gal) Cotton Compress Tank)
5 - Nursing Home
6 • Schools
$• Hydrants
O Nods
— Link
(Hour 3)
(Hour 4)
Figure 2. Water Movement From T300 During Hours 1-4 of 72 Hour
Simulation Period (T400 Valve Closed)
30
-------
""" U O HefflM called u part of CDC sunny
NOV. IS VlX
Residences wflh oonflmned case
Hydrant with confirmed Salmonella
Gideon Schools - reflects Increase In absentee level
20* or more of Tank 200 water
20% or more of Tank 300 water
Figure 3. Comparison of Early Confirmed Cases and Salmonella Positive Sample
versus Penetration of Tank Water During First Six Hours of
Flushing Program
*. Case Occurrence Data Was Obtained From Dr. Angulo of CDC
31
-------
o
o
o
C0
c
o
-------
CM
•a
a>
to
CM
c
a>
o o
(A
CO jf
o §
50
<« c
0) —
Es
o
in
£
3
01
33
-------
Evaluating Water Treatment Plant Performance
Using Indigenous Aerobic Bacterial Endospores
E.W. Rice, K. R. Fox, R. J. Miltner,
D.A. Lytle and C. H. Johnson
Water Supply and Water Resource Division
National Risk Management Research Laboratory
Cincinnati, Ohio 45268
Abstract
This manuscript describes the use of spores of aerobic spore-forming bacteria as microbial surrogates for
evaluating drinking water treatment plant performance. A method is proposed for assaying for the microbial
surrogate and the results of a survey of various water sources are presented. Data are presented for coagulation and
chlorine inactivation studies. Comparison evaluations for spore removals and turbidity and particle removals are
presented for pilot-scale and full-scale water treatment plants.
34
-------
The emergence of parasitic protozoa as etiological agents of waterborne disease has prompted renewed interest in
methods for evaluating the efficacy of unit drinking water treatment processes. Waterborne outbreaks of giardiasis •
and cryptosporidiosis have focused increased attention on the validity of various water quality parameters for
assessing treatment plant performance. Direct pathogen monitoring of encysted forms of protozoa is very
problematic, owing both to their random occurrence in ambient waters as well as methodological problems
associated with poor recovery and specificity. Hence, these methods do not lend themselves to routine monitoring
for determining treatment effectiveness. Levels of traditional bacterial indicators such as coliform bacteria, used as
general indicators of fecal pollution, have been reported to show significant correlations with parasite density in
source water (1). This correlation however, does not extend to finished drinking water. This study describes the
use of a microbial surrogate system, using endospores of aerobic spore-forming bacteria, which may be used to
evaluate removal efficiencies of biological particles.
Turbidity measurements and particle counting are two of the most valuable water quality parameters used in
assessing treatment plant performance. A comparison between source water and filtered water using these procedures
has been proposed as a reliable method for determining treatment plant performance (2). Turbidity appears to be an
adequate predictor the removal of cyst-sized particles when source water turbidities are greater than 5 ntu; however,
in less turbid source waters particle counting appears to be a more reliable indicator (2).
The proposed microbial indicator system uses endospores (often referred to simply as spores) of mesophilic,
aerobic spore-forming bacteria as surrogate organisms for evaluating treatment performance. In analyzing for spores,
the indigenous vegetative bacterial cells are inactivated by heat treatment. The surviving bacterial spores are
analyzed by cultural methods which permit the spores to germinate and produce bacterial colonies. These aerobic
spore-forming bacteria do not pose a public health risk and are best classified as benign, saprophytic organisms
whose principal habitat is the soil. This group of organisms consists primarily of species of the bacterial genus
Bacillus. The spores are ellipsoidal to spherical in shape and on average measure approximately 0.5 X 1.0 X 2.0
(Am. The spores are noted for being very resistant to various environmental pressures. These organisms are very
easy to culture and are present throughout most drinking water treatment trains.
Materials & Methods
Spore enumeration. Both naturally occurring aerobic spore-forming bacteria and pure cultures of Bacillus
subtilis spores were used in this study. The B. subtilis spores were purchased from a commercial laboratory (Raven
Biological Laboratories, Omaha, NE 68106). The protocol for enumeration was a modification of the procedure
recommended the detection of aerobic bacterial spores in milk (3).
The sample to be analyzed is collected aseptically 'placed in a sterile pyrex bottle or flask, and covered with a
cap or some sort of closure. Care must be taken to ensure that the cap does not come off during the heating process.
Aluminum foil placed over the top of a ground glass closure is suitable for this purpose. Each individual flask or
bottle should be labeled in such a manner that the identifying markings are not removed during the heat treatment
process. A pilot flask must be prepared containing a volume of water equivalent to that which is to be heat treated
in the individual samples and should contain a thermometer for determining temperature. The thermometer is held
in place by a loose fitting foam plug. The flasks are then placed in a thermostatically controlled water bath,
equipped with a shaker for constant agitation, equilibrated to 82°C. The water level in the bath should be equal or
slightly higher than the water level in the flasks. The flasks are agitated throughout the heat treatment The entire
water bath opening may be covered with aluminum foil to minimize heat lost and to decrease the time required for
the flasks to reach the desired temperature. The temperature of the water bath should be lowered to 80°C when the
temperature of the water in the flasks reaches 79°C. When the contents of the flasks reach a temperature of 80°C,
begin the timing sequence and keep samples in the bath for an additional 12 minutes. Care must be taken to
prevent water in the bath from entering the flasks containing the samples. At the end of this tune period, remove the
flasks from the water bath and cool the samples immediately in an ice bath.
After cooling, the samples are analyzed using the membrane filtration method. Appropriate dilutions are filtered
onto a 0.45 jam porosity membrane filter. The membranes are then placed on nutrient agar containing the dye
trypan blue. The medium is prepared in commercially available dehydrated nutrient agar. Appropriate amounts of
35
-------
the dehydrated medium are weighed and trypan blue is added to the dry powder (0.015 gm/L). The medium is
hydrated, heated to boiling, and sterilized by autoclaving at 121°C for 15 minutes. The sterile medium is then
ascptically dispensed in 7-8 ml into 60 x 15 mm petri dishes with loose fitting lids. Petri dishes with loose-fitting
lids are used to facilitate the growth of the strictly aerobic spore-formers. The final pH should be 6.8 +/- 0.2 at 20-
25°C. The membranes are placed on the agar surface and the plates are then inverted and incubated for 20-22 hours
at 35°C. After incubation, all colonies on the plates are counted using a binocular dissecting microscope. In this
procedure all colonies are considered to be derived from bacterial spores which were present in the sample and
capable of surviving the heat treatment. A minimum of two replicates were analyzed for each dilution examined.
Picking isolated colonies and determining the Gramstain reaction (4), provides a good quality control check to
ensure that the organisms present on the membrane are Gram positive rod-shaped bacteria.
Analytical methods. Total coliform bacteria were enumerated by the membrane filter procedure using m-Endo
LES agar, and heterotrophic plate count (HPC) analyses were conducted by the spread plate technique using R2A
agar with incubation at 28°C for 7 days (4). Turbidity was measured using a model DRT 100B turbidimeter (HF
Scientific, Inc., Fort Meyers, FL). The turbidimeter was standardized on each day of operation using turbidity
standards supplied by the manufacturer. An electronic particle counter, HIAC/ROYCO model 9064 (Pacific
Scientific Co., Montclair, CA) was used for particle sizing and counting analysis. The particle counter was
equipped with a laser detector, HRLD 150 sensor (1-150 (im), and the unit was connected to a personal computer
system. In this study total particles are defined as all particles in the 1-150 |im range.
Measurements of zeta potential were conducted using a Malvern ZetaSizer 4 system (Malvern Instruments, Ltd.
Worcester, Great Britain). Minusil solutions were used for daily calibrations of this instrument. A local lake water
source was used as the suspending medium. The water was filtered through a series of 0.40 urn porosity membrane
filters.The zeta potential of the filtered lake water was determined and the water was spiked with increasing
concentrations of B. subtilis spores, Giardia lamblia cysts and Cryptosporidium parvum oocysts to establish the
concentration of each organism required to make an accurate and distinguishable measurement of the respective
organism's zeta potential, free from interference of background particles in the lake water. A concentration of at least
120,000 organisms/mL was required for all organisms. Two 250 L samples of the filtered lake water were spiked
with the predetermined amount of the subject organism. The pH of samples was measured and approximately 10 ml
were removed for zeta potential measurements. The pH was incrementally decreased (0.2-0.3 pH unit increments)
by the addition of 2% (V/V) hydrochloric acid. A five-minute mixing interval, employing a magnetic stirrer, was
allowed to elapse between each pH increment prior to sampling for zeta potential readings. The same procedure was
followed for the higher pH values, which were obtained by the incremental addition of 8 N sodium hydroxide. The
experiments were conducted in duplicate. G. lamblia cysts were derived from Mongolian gerbils and C. parvum
oocysts were derived from neonatal bull Holstein calves. The C. parvum oocysts were not preserved in potassium
dichromate. The cysts and oocysts were prepared as previously described (5). Briefly, fecal slurries were
sequentially passed through a series of sieves, followed by centrifugal flotation using sucrose gradients.
Coagulation jar tests. The effectiveness of aluminum sulfate (18 waters of hydration) and ferric chloride (6
waters of hydration) coagulation for the removal of turbidity, particles and spores was studied using a local lake
water source. The impact of coagulant type, dose and raw water quality, i.e. initial turbidity, and temperature were
investigated to determine turbidity, particle count and spore removal efficiencies. The standard jar test apparatus
(Coffman Industries, Inc., Westford, MA) was used as previously described (6). Briefly, an inoculum of a
suspension of B. subtilis spores was added to the lake water to yield a concentration of approximately 1 x 103
spores/mL. The water was dispensed into 1.5 L rectangular plexiglass jars. Coagulant additions were made
simultaneously to the jars. Following coagulant addition, the waters were rapidly mixed at 100 revolutions per
ntinute (RPM) for 2-3 minutes, followed by 20 minutes of slow mixing at 30 RPM. The paddles were removed
from the jars and the water was allowed to settle for 60 minutes. The supernatants were then removed at a level 10
cm below the surface and analyzed.
Chlorine disinfection. Chlorine inactivation studies were conducted using indigenous aerobic spores found in
a local river water source. The river water was diluted in a ratio of 1:10 parts using 0.05 M chlorine, demand-free
phosphate buffer. The mean level of spores in the water was 15,250 CFU/100 mL (4.18 logic CFU/100 mL). The
water had a final pH of 6.86. The experiments were conducted at 23°C. The reaction beakers were stirred
continuously during the exposure times using an overhead stirring apparatus. Chlorine was added in the form of
36
-------
dilute sodium hypochlorite and chlorine measurements were made at the end of each exposure time using the N,N-
diethyl-p-phenylenediamine (DPD) procedure for determining free and total chlorine (4). The action of the oxidant
was neutralized at the desired times using sodium thiosulfate (4). Duplicate experiments were conducted for each
contact time. The mean free chlorine residual was 1. 60 + /-0.2 mg/L and the mean total chlorine residual was 1.75
+/-0.2 mg/L during the course of the experiments.
Pilot plant study. Spore removals were compared with the removals of other water quality parameters as part of
a pilot plant study designed to compare the effectiveness of conventional and enhanced coagulation procedures for
controlling disinfection by-products (7). A lake water served as the raw source water and was trucked daily to the
pilot plant. The lake water had a pH of 8.16, an alkalinity of 99 mg/L as calcium carbonate and a total organic
carbon level of 4.81 mg/L. The two parallel pilot-scale treatment plants, operated at 6.4 L/min, 24 hours per day,
employed rapid mixing, flocculation and sedimentation followed by chlorination and sand filtration. Aluminum
sulfate, Al2 (804)3-14 B^O (alum) was used as the coagulant, with the mean dose for the conventional plant
(operating for turbidity control) being 44 mg/L resulting in a mean settled water pH of 7.3. The mean alum dose for
the enhanced plant (operating for optimized control of disinfection-by-product precursors and total organic carbon)
was 152 mg/L yielding a mean settled water pH of 6.68. Chlorine was applied before the filters and resulted in a free
chlorine residual in the filter effluents of 2.69 mg/L and 2.44 mg/L for the conventional and enhanced plants,
respectively. Samples for spores, total coliforms, HPC, and turbidity, were collected daily for 5 days during the 14
day period. Particle count samples were collected over a 40 hour period during the 5 days of sampling.
Spore survey. A variety of water types from various geographical locations were examined to determine the
levels of spores of indigenous aerobic spore-forming bacteria. On one occasion the same site was sampled prior to
and immediately following a major precipitation event to determine the contribution of spores derived from soil run-
off. Four full-scale water treatment plants, all using surface water supplies (one aqueduct and three river sources),
were also surveyed (one time each) to compare spore and particle count removals through the various treatment
trains. One of the 4 utilities was sampled on a monthly basis for a 12-month period to study seasonal effects. The
following are descriptions of the utilities which were surveyed:
Utility 1, located in California, used pre-ozonation followed by alum coagulation and direct filtration
through deep bed anthracite filters and disinfection.
Utility 2, located in Iowa, used conventional treatment with alum coagulation, settling, sand filtration and
disinfection.
Utility 3, located in Louisiana, also used conventional treatment with alum coagulation followed by sand
filtration and disinfection.
Utility 4, located in Ohio, was sampled monthly for one year. This plant used alum coagulation followed by
settling, and filtration through sand filters and granular activated carbon filters, and disinfection.
Results and Discussion
Jar tests and zeta potentials. The results of the coagulation and sedimentation experiments are shown in
Figures 1 and 2, For the alum coagulant studies at 25°C, the pH of the source water was 9.5 and the turbidity was
22.1 ntu; at 6.5°C, the pH was 8.1 and the turbidity was 7.36 ntu. For the ferric chloride experiments at 25°C, the
pH of the source water was 8.1 and the turbidity was 5.76 ntu; at 6.5°C the pH was 8.2 and the turbidity was 8.11
ntu. The data clearly show that log reduction of spores was similar in magnitude to that of turbidity, total particles
and particles in the Cryptosporidium oocyst size range. More importantly, spore removals closely paralleled
particle and turbidity removal in response to coagulant dose under all of the water quality conditions examined.
Even when conditions were less favorable for particle removal, such as low coagulant dose or hi cold water treated
with alum, these observations held true. The only discrepancy was that spore removals were typically lower when
coagulant was under dosed or under other conditions where particle removal was impaired. This characteristic
would be considered a positive attribute. These responses hi spore removal relative to particles and turbidity may
serve as useful indicators of declining treatment effectiveness.
37
-------
The surface charge characteristics of B, subtilis spores were measured and compared to the charge characteristics
of (?. lamblia cysts and C. parvum oocysts (Figure 3). The zeta potential for the spores ranged from -16 mv to -20
mv as the pH of the spiked pond water was increased from 5.0 to 10.5. These values are similar to those observed
for G, lamblia cysts (-12 mv to -20 mv) over the same pH range. The zeta potential for C parvum oocysts ranged
from -5 mv to -13 mv. Zeta potential readings are not a direct measure of removability, but can be used to
determine coagulant dosages necessary for optimizing coagulation.
Chlorine disinfection. The results of the chlorine inactivation of an indigenous aerobic spore population are
shown in Figure 4. An approximate two-log reduction occurred after an exposure time of 65 minutes. This amount
of inactivation would equate to a simple disinfectant residual concentration (mg/L) multiplied by .the exposure time
(min.) or CT (mg-min/L) value of approximately 114 for total available chlorine. A three log reduction occurred
after 180 minutes, yielding a simple CT value for total available chlorine of 315. Previous chlorine inactivation
studies, conducted under similar conditions of pH and temperature using pure cultures of B. subtilis spores reported
a three log reduction after five minutes of exposure to 100 mg/1 of available chlorine, yielding a CT value of 500 (8).
The increased resistance observed in the pure culture studies may be attributed to the more consistent nature of the
purified spore preparation. Under natural conditions indigenous spores would exist in different stages of maturity
and metabolic dormancy, and would thus exhibit differences in response to oxidative and other environmental
stresses (9).
The comparison of CT values of a given oxidant for different microorganisms is subject to many variables, not
the least of which being the manner in which CT values are calculated The method used for determining viability is
an important consideration when analyzing protozoan inactivation studies. Under similar conditions of pH and
temperature in the Guidance Manual for the Surface Water Treatment Rule, CT values required for a two log
reduction in viability were near 31 for Giardia lamblia cysts (10). These CT values are contrasted with the
extremely high chlorine CT values of 7,200 for a two log reduction in viability for C. parvum oocysts (1 1). Based
upon these findings, and noting the above mentioned caveats regarding CT value comparisons, the indigenous
aerobic spores exhibited a chlorine resistance greater than that reported for G. lamblia cysts, but considerably less
than that for C. parvum oocysts.
The chlorine CT values obtained for the indigenous spores indicate that the removal of these organisms as the
result of inactivation would be minimal compared to physical/chemical removal when using them as an indicator of
treatment effectiveness through coagulation, sedimentation and filtration. Only if a utility were chlorinating or using
another oxidant at the beginning of the treatment train might disinfection play an important role in the inactivation
of the surrogate.
Enhanced coagulation pilot plant study. The water quality parameters examined during the course of the
pilot plant study are listed in Table 1. In all cases removals of microorganisms, turbidity and particles were
improved in the enhanced coagulation process. Under conventional and enhanced coagulation conditions, both total
coliform bacteria and HPC bacteria were below the level of detection in the chlorinated filter effluents. The aerobic
spores were the only microbiological parameter which persisted throughout the entire treatment train.
Table 2 lists the log removals for aerobic spores, 3-5 ]j.m particles and total particles in the settled and filtered
samples from the two types of coagulation. The spore removal trend was similar to that of the particles. The
additional removal of the spores during filtration was largely a physical phenomenon, because the chlorine contact
time during filtration was less than 20 minutes.
Survey. A survey of the indigenous levels of aerobic spores found in environmental waters from different
, geographic locations is summarized hi Table 3. All surface waters examined, with the exception of the one lake
water sample from Wisconsin, contained indigenous levels of-spores between 102-104 CFU/lOOmL. The ground
water samples, all taken from wells used as drinking water sources, consistently exhibited low levels of spores. The
presence of spore-forming bacteria (Bacillus spp.) in ground water has previously been proposed as an indicator of
the degree of purification or contamination from surface impurities (12)., Based upon the data in Table 3 it appears
that the levels of these organisms in most surface water supplies would be sufficient to allow then- use as a surrogate
to evaluate treatment efficiency.
38
-------
On one occasion a local river source was sampled prior to a major rainfall. The initial water sample contained
aerobic spores at 120 CFU/mL, turbidity of 5.50 ntu, and contained 1.85 x 106 total particles/mL and 5.90 x 105
particles/mL in the 3-5 /jam size range. Following a rainfall of 1.24 inches within a 24-hour period, the spore level
increased to 2,500 CFU/mL and the turbidity increased to 106 ntu. The corresponding particle counts increased to
6.9 x 107 total particles/ml, and 1.1 x 107 particles/mL in the 3-5 Jim size range.
A comparison of the levels and the removal efficiencies of spores, 3-5 |J.m particles (oocyst sized particles) and
total particles between source water and filtered water for the four utilities, is shown in Table 4. Differences between
spore and particle removals ranged between 0.02 0. 14 log units for utilities 1 and 4. For utilities 2 and 3 there was
approximately a 1 log difference between spore removal and removal of 3-5 |j.m particles and a 0.40-0.68 log
difference between total particle and spore removal. These data indicate that spore removal was either closely
matched or served as a conservative indicator of treatment effectiveness for removal of particles.
Levels of aerobic spores present in the source water, settled water and sand and granular activated carbon filter
effluents during the one-year sampling period for utility 4 are shown hi Figure 5. The removal efficiencies for spores
and particles by the three unit processes compared to levels in the source water are listed in Table 5. Given the
premise that treatment plant performance is best evaluated by comparing source water to final filter effluent, it can be
seen that spore removal closely paralleled particle removal. Spore removal most closely mirrored particle removal
during the winter and summer seasons. Again, these data illustrate the close relationship between spore and particle
removal. Four samples of backwash water were also obtained from utility 4 (one sample in the winter, one in the
summer and two during the autumn period). The samples averaged 35,000 CFU spore/100 mL (range 1,100-
115,000 CFU spores/100 mL. These levels of spores were comparable to those found in the source water (Figure
5). This finding is in keeping with previous reports that spores can accumulate in sand filters and a sudden increase
in numbers can indicate filter breakthrough (13). It should also be noted that spores were found hi the clear wells of
all four utilities, with utility 1 having 2 CFU/100 mL, utility 2 having 70 CFU/100 mL, utility 3 with 245
CFU/lOOmL, and utility 4 averaging 21 CFU/100 mL for the 12-month sampling period. These data as well as the
pilot plant data, where spores in the clear wells averaged 3 CFU/100 mL, indicate that the spore-forming organisms
can be monitored through the entire treatment process.
The use of particle counting, especially in reference to utilities routinely treating low turbidity waters (< 5 ntu),
has been recommended for determining log reduction of cyst-sized particles (2). It has been noted, however, that
cyst-sized particles can occur as agglomerates of smaller particles which may be sloughed off during a filter cycle.
The measurement of particle removal as a surrogate for cyst or oocyst removal in such cases would give misleading
values. This problem is accentuated when comparing removal efficiencies by particle counting for coagulated and
settled waters. One advantage of the spore procedure is that the analytical method measures only aerobic spores;
other particulate matter is not counted.
Indigenous spores do not serve as indicators of the presence of protozoan parasites. However, monitoring for
spore removal, coupled with monitoring for turbidity and particle counts would allow a utility to optimize unit
processes and thus provide more efficient treatment. The spores are smaller in size than cysts or oocysts and thus
are a conservative indicator of removal efficiency. Their presence throughout the treatment train allows actual
removal rates to be calculated. It should be noted that the use of spores as surrogates for evaluating treatment plant
performance neither precludes nor negates the continued use of turbidity and particle counting. The electronic
measurements provide near real time data. Monitoring for spores provides next day data as an additional tool to
assist in fine-tuning treatment processes.
Summary
Monitoring for indigenous spores of aerobic spore-forming bacteria represents a viable method for determining
treatment plant performance. Comparison of levels of spores between source water and filter effluents provides an
indication of biological particle removal efficiency. Unlike other microbiological parameters, levels of spores can be
detected throughout the treatment process, and the endospores themselves do not propagate in the various unit
processes. The aerobic spores appear to occur at levels in most source waters which would permit their use as a
microbial surrogate for evaluating treatment plant efficiency. These organisms do not present a public health threat,
39
-------
and being primarily of soil origin would tend to increase during periods of run-off. The use of the membrane
filtration method allows the analyst to increase the volume examined as numbers of these organisms and turbidity
decrease through treatment. Larger volumes of turbid water may be examined by dividing the sample into smaller
aliquots and combining the counts of the individual aliquots. Spore removal closely parallels particle removal.
These organisms are very resistant to disinfection. The proposed procedure is very straightforward and provides a
useful routine method for evaluating treatment plant performance. Further studies designed to directly correlate
parasite and spore removal coupled with more extensive surveys of occurrence in various source waters would
provide additional information regarding the general application of this procedure for various types of waters.
References
I. Lechevallier, M.W. & Norton, W.D. Examining Relationships Between Particle Counts and Giardia,
Ctyptosporidium, and Turbidity. Jour. AWWA, 84:12:54 (Dec. 1992).
2. Bellamy, W.D. et al. Assessing Treatment Plant Performance. Jour.AWWA 85:12:34 (Dec. 1993).
3, Standard Methods for the Examination of Dairy Products. APHA, Washington, D.C. (16th ed., 1992).
4, Standard Methods for the Examination of Water and Wastewater. APHA, AWWA, WEF, Washington, D.C.
(ISthed., 1992).
5. Owens, J.H. Et al. Pilot-Scale Ozone Inactivation of Cryptosporidium and Giardia. Proc. Water Qual.
Technol. Conf., AWWA, pp. 1319-1328 (Nov. 1994).
6, Lytle, D.A. & Fox, K.R. Particle Counting and Zeta Potential Measurements for Optimizing Filtration
Treatment Performance. Proc. Water Qual. Technol. Conf, AWWA, pp. 833-856 (Nov. 1994).
7. Miltner, R. J. Et al. Evaluation of Enhanced Coagulation for DBP Control. Proc. Nat. Conf. Env. Engr.,
ASCE, pp. 484-491 (July 1994).
8. Williams, N.D. & Russell, A.D. The Effects of Some Halogen-Containing Compounds on Bacillus subtilis
endospores. Jour. Appl. Bacteriol., 70:427 (1991).
9. Knott, A.G., Russell, A.D. & Dancer, B.N. Development of Resistance to Biocides During Sporulation of
Bacillus subtilis. J. Appl. Bacteriol. 79. 492 (1995).
10. U.S. Environmental Protection Agency. Appendix E. Guidance Manual for Compliance With the Filtration
and Disinfection Requirements for Public Water Supplies Using Surface Water Sources. Office of Drinking
Water. Washington, D.C. (1990).
11. Korich, D.G. Et al. Effects of Ozone, Chlorine Dioxide, Chlorine and Monochloramine on Cryptosporidium
parvitm Oocyst Viability. Appl. Environ. Microbiol.. 56:1423 (1990).
12, Schubert, R.H.W. Der Nachweis von Sporen der Bacillus - species im Rahmender Hygienischen
Wasserbeurteilung. Zbl. Bakt. Hyg. I. Abt. Orig. B, 160:155 (1975).
13. Anon. Indicator Systems for Microbiological Quality and Safety of Water. J. Environ. Path. Tox. and Oncol.
7:5/6:132 (May-Aug. 1987).
40
-------
c
o
h*
a.
O
+^
s
2.
E
t»
<2
«4"d
to
J3
>
C
s
«— 1
U
2
H
"«
o
H
e
£
in
"3
5
E
I
co
§5
3 G
0 j
2 si
c^ S2
o 1-1!
1|
u!s
U
<*L s—
g^
.0 c
il
U ^
rt ff.
00
H —
*
e
1
-»
g
^
t-
co-
O
r- 1
O
oo '
•<*
s
c
.2'
•*"*
••2
M)
O
U
- e
.0
*4-»
C
£ >
«J O
PC O
o o
O OS
OO CM
^^
O O
O •*!•
CO -M
OS ^
o c5
o ts
T— 1
1— 1 >"•
f» v
1
^^
1
•4-*
l1
4~* "
£ u
o o
§N
O 1-1
CO CM
co r-
t- «—
CD C5
oc *O
0 i—
o \y
rH
VO
i
I
0 "i
— V!
C :
• S ;
* "d.
"3 i>
60 S
O C^-
u ^
U a
1 ll
c ^ —
W V5 0
41
-------
Table 2. Removals of Aerobic Spores and Particles Across Pilot Plants
1 Location
Conventional Coagulation
Settled
Chlorinated/Filtered
I Enhanced Coagulation
i
Settled
Chlorinated/Filtered
Log Reduction1
Spores
0.85
2.12
1.51
2.42
3 -5 Am Particles 1 Total Particles
1.11
2.10
1.86
2.91
0.91
1.70
1.39
2.87 1
Source water to unit process effluent
Note: % Reduction=100%*(l-10-(LogReduction))
42
-------
Table 3. Levels of Indigenous Aerobic Spores in Various Source Waters
Water type
Location
Spores
CFU/100 mL
Surface
River
Lake
Ground
California
Indiana
Iowa
Kentucky
Ohio
Louisiana
Utah
Ohio
Wisconsin
Montana
Ohio
Puerto Rico
Texas
Vermont
210
4400
4000
23000
14000
37000
350
1600
15
1
7
3
12
43
-------
I
flu
*-*
§
*•*
s
I
S
i
cu
•a
CO
U3
I
I
o
I
c
o
O
rt
£
I
B.S
cu
sn O
-H E
<£
'I
co _j
£
J
o
P
U
fP
oo vo en
O O •*; t;
VI &>
+ *
44
-------
Table 5. Seasonal Cumulative Log Removals Through Unit Processes of Utility 4
Log Removal
Particles, log
Season
Winter
Spring
Summer
Autumn
Unit Process
Settling
Sand Filtration
GAC Filtration
Settling
Sand Filtration
GAC Filtration
Settling
Sand Filtration
GAC Filtration
Settling
Sand Filtration
GAC Filtration
Spores, log
1.19
2.19
2.89
1.39
2.57
2.70
0.90
1.69
2.22
0.85
2.09
2.06
3— 5/un
1.08
2.12
2.92
0.99
2.00
2.19
0.73
1.68
2.22
0.32
1.36
1.76
Total
1.22
2.02
2.96
1.32
2.19
2.30
0.86
1.64
2.24
1.79
1.49
1.81
Winter(1994)- December(1993), January, February
Spring(1994)- March, April, May
Summer(1994)— June, July, August
Autumn(1994)— September, October, November
45
-------
a
*•*-»
o
3
1
60
3
2.00
1.75
1.50
^^ i * ~^*
'// (a> •
-/
1.25 \/l
LOO:
0.75
0.50
0.25
n rv\
-
-
• Turbidity
v Total Particles .
T 3-5 /tm Particles
n B. subtilis Spores"
1,1,1,1,1,1
0.00 s
0
0 20 40 60 80
Alum Dosage, mg/L
100
20 40 60 80
Alum Dosage, mg/L
100
Figure 1. Log reduction of turbidity, particles,
and spiked B. subtilis spores after alum coagulation and
sedimentation of lake water at (a) 25 C and (b) 6.5°C.
46
-------
• Turbidity
v Total Particles .
3-5 /tin Particles
B. subtilis Sporels
20 40 60 80 100
Ferric Chloride Dosage , mg/L
20 40 60 80 100
Ferric Chloride Dosage, mg/L
Figure 2. Log reduction of turbidity, particles,
and B. subtilis spores after ferric chloride coagulation
and sedimentation of lake water at (a) 25°C and (b) 6.5°C
47
-------
0
> -5
2
1 -10
3
Cu
-15
-20
-25
O Raw water
n Bacillus subtilis
T Cryptosporidium parvum oocysts
• Giardia lamblia cysts
T TT
4 5 6 7 8 9 10 11
pH
Figure 3. The effect of pH on filtered lake water spiked with
Bacillus subtilis, Cryptosporidium parvum, and Giardia lamblia.
48
-------
4.5
4.0
3.5
3.0
e
o
2.5
1.5
1.0
0.5
0.0
0
pH6.9
23 °C
1.60mg/LfreeCl2
1.75 mg/L total C12
50 100 150
Time, (min)
200
250
Fig. 4. Inactivation of indigenous spores in diluted
river water by chlorine.
49
-------
•1-4
S
3
"8
OS
i
.*->
cu
"o
O)
PJj
CQ
CD
IS
Wi
•g
o
a,
I
(U
I"1 OOT/QJ3 §0I
50
-------
Evaluation of Optical Detection Methods for
Characterizing Suspensions in Drinking Water
Virendra Sethi, Robert M. Clark and Eugene W. Rice
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Cincinnati OH 45268
Priyamvada Patnaik, Pratim Biswas
Aerosol and Air Quality Research Laboratory
Department of Civil and Environmental Engineering
University of Cincinnati
Cincinnati OH 45221-0071
Abstract
Turbidimeters and optical particle counters (OPC's) are currently used in the water supply industry to monitor
particulate matter in water. The response from these optical instruments is governed by the shape, structure,
refractive index and size distribution of particles in the suspension. The recommended design criteria for
turbidimeters allows for large tolerances in the design parameters, which leads to variation in measurements from
different instruments. OPC's provide size specific information but are limited to detection of particles larger than
0.5 (J,m. More importantly, OPC's are calibrated with homogeneous spherical particles of known refractive index,
which may lead to inaccurate sizing of microorganisms or particles with optical properties different from those of the
calibration particles. Fundamental principles of particle light scattering were used in the present work to evaluate
the effects of optical design parameters in turbidimeters, and the dependence of OPC sensor response on particle
refractive index and the angle of detection. Performance data from two OPC's for sizing monodisperse, spherical
polystyrene latex (PSL) and SiO2 particles, and several microorganisms in the 0.5-10 jim size range are presented.
A more broad-based multiple angle light scattering measurement technique was developed to obtain angular
distributions of scattered light intensity from suspensions of PSL, SiO2 and several pure cultures of microorganisms.
Such angular distribution patterns of scattered light may also be used to improve the sizing performance of OPC's for
microorganisms, or to uniquely identify specific types of microorganisms.
51
-------
Introduction
Light scattering and light obscuration techniques are used successfully to characterize suspended matter in
drinking water, food beverages and process fluids. Turbidity and, increasingly, particle counting are used to detect
waterborne particulate matter for compliance with drinking water regulations, and to monitor the effectiveness of
water treatment at various process stages. Recent changes in water quality regulations and concerns regarding
waterborne cryptosporidiosis1.2 are driving the water supply industry towards re-evaluation of the particle detection
capabilities of instruments presently used in the industry. Design and performance evaluation of particulate control
processes require size specific information of suspensions in raw waters and in the finished product, respectively.
Such information is also useful at low concentrations for early detection of particulate breakthroughs3.
Because of simplicity and ease of operation, turbidimeters are used routinely in water treatment plants to
monitor finished water quality. More importantly, turbidity serves as a surrogate indicator for the occurrence of
microorganisms in water. Recent studies indicate that the assumption of a correlation between turbidity and
microbial quality may either not be true4, or may be inadequate in preventing outbreaks of waterborne diseases.
Design differences among turbidimeters pose a difficulty in comparing responses from different turbidimeters5, and
rafse the issue of a need for standardization6.7.8. Further, turbidity is an aggregate optical property, and lacks size
specific information for detection of microorganisms in a specific size range. The recent availability of optical
particle counters for liquids, however, has provided a means of acquiring size distributions for waterborne
suspensions.
Turbidimeters and particle counters operate differently in that the former responds to a "cloud" of particles in a
sample view volume, while the latter counts and sizes individual particles as they pass through a view volume.
The use of OPC's is currently limited by the high initial cost and need for operator training. Comparative studies in
performance evaluation show mat OPC's, because of higher sensitivity, respond to particle breakthroughs much
sooner man turbidimetersS.10.11. Ideally, given an OPC measured size distribution, it should be possible to obtain
turbidity as a sum of the scattered light signal response from each of the particles in a suspensions I2, provided
particle characteristics such as size, shape, structure and refractive index are known. Experimental correlations of
changes in turbidity with corresponding changes in particle counts have been successful in some studies13.14.
However, in another study4, no relationship between turbidity, particle counts and bacteriological quality of water
was found. The lack of such agreement may be caused by one or several reasons: (a) lack of information on optical
properties of suspensions and microorganisms; (b) limitation of OPC's in detecting particles smaller than 0.5-1.0
Jim; and (c) differences in optical design of the instruments. Other discrepancies may arise due to the dilution that is
usually required to reduce particle concentrations to prevent coincidence errors in OPC's. These discrepancies may
include dilution errors, or the changes in particle size distribution caused by swelling and disaggregation of
agglomerate structures during dilutions15.
The success of OPC's in providing size specific information tends to obscure the fact that there is a large fraction
of suspended matter that is smaller than the detection limit of OPC's. The undetected fraction of particles may carry
nutrients that can support biological activity later, and may also exert a disinfectant demand4. This fraction of
particles that goes undetected in OPC's, however, contributes to turbidity and is possibly missed only by the lack of
sensitivity of turbidimeters at the currently recommended 90° detection angle. Further, the interpretation of a
physical particle size from the optical response obtained from OPC's requires a known optical response for the
particles that is specific to the angle of detection of the sensor. In the absence of such information for the variety of
inorganic, organic and microbial particles being viewed, the sizing is at best an estimate of an optical response
relative to the calibration PSL particles. Particles that have an index of refraction smaller than PSL, for example,
may either go undetected or be counted hi a smaller size class. Previous studies that have attempted to correlate the
performance between turbidimeters and OPC measurements5.14 have not addressed the problem of optical detection
with respect to particle optical properties and the effects of instrument design. Evaluating and identifying the causes
for common problems reported in previous studies could possibly lead to improvements in instrument design and/or
compliance criteria.
52
-------
Both turbidimeters and OPC's are single angle detection methods, and are a subset of a more broad based
multiple angle light scattering (MALS) measurement method. The MALS technique has been used successfully in
previous studies for characterization of liquid borne macromolecules16, and gas borne particles17=18>19. The
availability of the electrodynamic balance made it possible to study light scattering properties of individual particles
in the micrometer size range. Such studies have been limited to gas suspended particles or droplets, and to our
knowledge, there are no MALS techniques for study of individual particles suspended in liquids. The only
exception to this is a method used in flow cytometry where scattered light can be measured at one near-forward angle
and also at 9'0°.
Ideally, simultaneous measurement of scattered light at multiple angles as individual particles pass through a
light beam could provide more detailed information about size and shape of individual particles hi water. Such a
device is not currently available, however, and the MALS measurement method for a collection of particles hi a
small view volume (similar to turbidimeters) comes closest to such an approach. For dilute suspensions of
monodisperse spherical particles, scattered light from a collection of particles in a view volume is reinforced at all
angles. A multiple angle measurement from such a view volume can therefore be used to represent single particle
behavior. If the particles are not spherical but instead have other geometric shapes (cylindrical rods, ovoid), random
orientation of the particles in the view volume causes a loss of information that could have been available from
single particles aligned at a specific angle with the incident beam. The averaging effect in such a case, however, may
still be unique and can be used to characterize the optical properties of particles in monodisperse suspensions20. As
an extreme case, a suspension may be made up of particles of irregular geometric shapes (single cells, chains,
agglomerates), and/or different sizes (polydisperse). In such a case, the optical response would represent a statistical
average of the size distribution and optical properties of the suspension. In the absence of prior information on the
optical properties, multiple angle measurements can be used to infer the absolute or mean optical characteristics.
For microbial suspensions, multiple angle measurements can provide insight into then- optical properties and
internal structure, and can also serve as unique "optical signatures" to identify specific microorganisms. MALS
measurements also offer a method to study suspensions that are not resolvable under the microscope for reasons of
lack of refractive contrast, or due to constant particle motion2!.
Objectives
The objective of the present study was threefold. The first was to evaluate the extent of variations in optical
response that could result due to the tolerances allowed in the recommended design criteria for turbidimeters. A
theoretical simulation was used to investigate the effects of specification of wavelength of light, and the collection
angle at the detector. The effect of refractive index of particles on turbidimeter response was also demonstrated
theoretically. The second objective was to evaluate the sizing performance of OPC's for microorganisms. A
theoretical simulation of a light scattering type OPC sensor was used to show the significance of minor changes hi
refractive index on the sizing accuracy of spherical particles. Sizing performance of two OPC's was then evaluated
experimentally using spherical homogeneous monodisperse particle suspensions of known refractive indices, as well
as microbial suspensions with unknown optical properties. And the third objective was to present a case that the
single angle detection method used hi turbidimeters and OPC's is limited, and a multiple angle light scattering
technique could be developed for better characterization of non-spherical particles with unknown optical properties.
An experiment was developed to obtain angular intensity distribution of scattered light from several suspensions of
monodisperse spherical particles and microorganism isolates. A size inversion scheme was applied to infer particle
sizes from these angular distribution measurements.
Optical Detection - Background
A brief background on the theory of light scattering for optical detection of particles is presented in this section.
Light scattering by single homogeneous spherical particles is described by the Rayleigh and Lorentz-Mie
theories20*22.23. The value of dimensionless size parameter X (X = 7t.dp / 1, where dp is the particle diameter, and 1
is the wavelength of the incident light) is used as guiding criteria for the choice of the appropriate theory. The
Rayleigh scattering theory is applicable for small particles (X < 0.3). The Lorentz-Mie theory extends the size range
to include larger particles (X > 0.3), and reduces to the Rayleigh theory for smaller particles. The interaction of an
53
-------
incident light beam of intensity IQ with a particle in the Lorentz-Mie regime, results in a dissymmetric angular
distribution of the scattered light intensity I(q), and is a unique function of particle size, shape, and refractive index.
The angular distribution of scattered light intensity from a spherical particle, measured at a distance R in the
scattering plane, is described as24:
1(9) =-
8?c2R2
(1)
where Cw and Cup are the Lorentz-Mie parameters for the vertical and horizontal polarized components of the
scattered intensity. CyV and CHH are a function of the size parameter X; particle refractive index m,' and the scatter
angle q and can be calculated by using computer subroutines23-25.
'- [lr
The refractive index of particles, m, is a key property that affects the calculation of GW and CHH m Equation
(1), Waterborne suspensions vary widely in composition and content, viz. silt, clay, minerals, inorganic and
organic matter, microorganisms, and their optical properties remain largely unknown. Refractive index may be real
(non-absorbing) or complex (absorbing), and may be affected by the structural inhomogeneities that might be
associated with agglomerates or microbial cells. As an estimate, the value of the real part of refractive index for
waterbome particles, relative to water (mw), could be expected to lie in the 1.05 (estimate for microorganisms) to
1.50 (inorganic minerals26) range. Turbidimeters are calibrated in nephelometric turbidity units (NTU) using
suspensions of formazin (mw, unknown) prepared under controlled conditions27, or using suspensions of styrene
divinyl benzene copolymer microbeads8 (mw =1.17) prepared to match NTU scales. OPC's are calibrated using
monodisperse PSL suspensions (mw = 1.195). Sensitivity to refractive index variations in OPC sensors is
minimized by limiting the collection angle at the photosensor to a narrow cone of fight at a near-forward angle. In
spite of such design features, particles with refractive indices very different from that of PSL may not be sized
correctly, the extent of agreement varying from sensor to sensor28. For turbidimetric measurements of the
heterogeneous suspensions encountered in drinking water, the use of an "average" or "effective" refractive index may
be applicable. Such information for raw and treated drinking waters, however, is conspicuously missing in the
literature.
In addition to parameters m and X for single particles, the angular scatter intensity distribution for polydisperse
suspensions is a function of particle size distribution parameters. The Lorentz-Mie theory (Equation 1) may be
emended for polydisperse suspensions of spherical particles provided that the particles are sufficiently apart so as to
prevent multiple scattering - a condition usually satisfied by the low particle concentrations encountered in drinking
water. For a suspension with size distribution function n(dp), the response at any angle is an aggregate contribution
from aH the particles in the suspension, and can be written as:
: "j
Qvv(9) = / C« (m'6>M) • "<
-------
by the measurement method (e.g. optical microscopy ~ 0.5 jam; OPC ~ 0.5 - 1.0 pm). Ideally, it may be desirable
to develop measurement capability to detect viruses, implying a lower detection limit of 0.001-0.1
Jim. The presence of Rayleigh scattering from the water molecules, and noise from photo electronics30, however,
have limited the present practical detection capability in OPC's to ~ 0.5 u,m.
The power law representation for particle distributions in drinking water has been based on particle size and
count data obtained primarily from OPC's. It should be noted that the power law form of representation may be a
subset of a more complete log normal or Gaussian form of distribution usually found in natural systems^. In a study
on particles in oceanic waters by Harris31, an electron microscope was used .to measure particle size distributions in
the 0.02 to 8.0 \un size range and the data was found32 to conform to a power law representation. Since the
contribution to light scattering from particles smaller than 0.02 pm is expected to be negligible, power law may be
assumed to represent the optically detectable size range.
The scatter intensity as described by Equation (1) is a function of the angle of detection. Dissymmetry in
angular intensity distribution has been used in the past to characterize macromoleculeslS, and particles in the
Lorentz-Mie regime33,3! A data inversion procedure has been used by Chang and Biswas35 and Chang et al.36 to
estimate the lognormal size distribution parameters and refractive index for combustion aerosols. The procedure
uses a least square fitting technique to match scattering intensity measurements made at multiple angles with the
theoretical response for spherical particles using particle size (dp) and the refractive index (m) as two independent
variables.
Experimental Methods
The focus of the experimental work in this study was to establish the effect of optical properties of particles on
the sizing performance of OPC's, and further, to obtain MALS patterns for suspensions of microorganism isolates.
Two commercial OPC's, one based on the principle of fight scattering (Met-Onea: LS-211) and the other based
on the principle of light obscuration Hiac/Roycob : HRD-150), were used to size suspensions of PSLc (1 020 1 500
2.032, 3.500, and 5. 100pm), and Si02c(l. 100 and 5.000 pm). '
Several suspensions of microorganisms were prepared to study the sizing response from OPC's and multiple
angle light scattering patterns. Escherichia coli and Enterobacter cloacae cultures were grown in brain heart
infusion broth for 18-20 hours at 35° C. The cultures were concentrated and washed twice in phosphate buffer2?.
The cultures were resuspended in 10 mL of buffer and further diluted for the experiments. The Streptococcus bovis
and Enterococcus faecium cultures were grown in brain heart infusion broth supplemented with 1% (w/v) yeast
extract and prepared in the same manner as the E. coli and E. cloacae cultures. The Giardia muris cysts and
Cryptosporidium muris oocysts were prepared as described elsewhere3?. Diluted suspension of Streptococcus bovis,
Bacillus subtilis, Cryptosporidiwn muris and Giardia muris were sized using the two OPC's.
A multiple angle light scattering (MALS) experiment was developed for studying the optical response for
waterborne suspensions. A brief description of the experimental set-up shown in Figure 1 follows. A 5.0 mW,
He-Ne (1 = 632.8 nm) plane polarized laser beam of intensity I0 vv, chopped with a high frequency chopper, was used
as the incident light source. A vertical polarizer, a 632.8+1.0 nm line filter and a variable aperture were used to
receive only the vertical polarized component (Ivv) of the scattered light at the same wavelength as the incident
source beam. A lock-in-amplifier, locked to the reference frequency from the chopper, was used to eliminate
background optical noise and amplify the photocurrent. The photomultiplier tube assembly (variable slit, line filter,
and polarizer) was mounted on a rotator with an angular range of 30° to 120° with a resolution of 1°. A 35 mL
cylindrical glass vial with an outer diameter of 27.0 mm was used to hold the sample.
a Met-One, Grants Pass, OR 97256
b Hiac/Royco, Division of the Pacific Scientific Company, Silver Spring, MD 20910
c Bangs Laboratories, Inc., 979 Keystone Way, Carmel IN 46032-2823
55
-------
Tllp experimental set-up was calibrated using standard monodisperse PSL suspensions. De-ionized filtered (0.22
jim Nuclcopore) water was used for all dilutions. The angular distribution of scattered fight from a blank sample
was comparable to that reported by Black and Hannah^. AU measurements were made at 5° intervals, and the
angular distribution patterns were consistently replicated at different dilutions to ensure absence of multiple
scattering. MALS measurements were made for suspensions of PSL (0.460, 0.560, 0.750, 0.993, 1.020, 1.500 and
2.032 jim), and were used to test the effectiveness of the size inversion scheme35. Similar measurements were also
made for SiQ2 (1-100 Jim), S. bovis, Enterococcus faecium, B. subtilis, Enterobacter cloacae and Escherichia coll
A summary of experiments and results are listed in Table 1.
Results and Discussion
As stated earlier, the focus of the present work was to understand the effects of optical design parameters and
particle characteristics on optical detection of waterborne suspensions. The next two sections describe the results of
theoretical simulations for turbidimeters and experimental results of sizing from OPC's. The third section describes
the results of the MALS approach for optical characterization of particles.
Turbidimeters. The variability in experimentally obtained turbidity measurements has been established in a
previous study5. The objective of the analysis here was to demonstrate the effect of design parameters for
turbidimeters on the variation in optical response. Manufacture of turbidimeters is guided by recommendations
provided by the United States Environmental Protection Agency (US EPA)2?>38 in USA, and the International
Organization for Standardization (ISO) in Europe3^. The recommendations are provided to ensure consistency
among instruments for the purpose of regulatory compliance. Performance studies5, however, show that though all
manufacturers meet the US EPA suggested design criteria, variation is observed in measurements from different
instruments. Such variations are likely to be a result of the tolerancel5 levels permitted by the specifications. In
this section, the influence of the source light wavelength (1), and the angle of detection (q) on the scattered intensity
is analyzed theoretically. The effect of particle refractive index on the optical response is also simulated to highlight
the importance of optical properties of suspensions.
Turbidity measurements are affected by the optical, mechanical and electronic design of the instrument. The
design criteria suggested by US EPA and ISO are listed in Table 2, and the terms are illustrated in Figure 2. The
requirement that scattered light be measured at 90° was probably influenced by one or all of the following
factors6.7-40: (a) stray light from dust particles, gas bubbles, and diffusion from edges and surfaces; (b) sensitivity to
vibrations; and (c) placement of optical components. Hach et al.15 have discussed the features of turbidimeters
currently used in the industry. Sigrist? has studied the effects of particle size on the scattering efficiency, the
distribution of the scattered light, and suggested improvements in design of turbidimeters.
The intensity of scattered light from a standard calibration suspension, measured at 90° to the incident beam, is
used to define the Nephelometric Turbidity Units (NTU). For a suspension that can be represented by a size
distribution function, the turbidity (NTU) value can be calculated using Equations (2) and (3). The angular
distribution of scattered light intensity was simulated using a typical set of values of a (10s) and 6 (3.82) reported in
literature29, and a relative refractive index of 1.195 (PSL). The source wavelengths were varied in the range specified
by the two design criteria listed in Table 2. The limits of integration dj and &2 in Equation (2), were chosen to be
0.01 and 10 ^m respectively. The theoretical response was scaled to nephelometric turbidity units using a turbidity
calibration standard with known particle concentration^ (0.269 p.m styrene divinyl benzene copolymer microbeads,
40 NTU = 4.85 x 10s $/mL). Figure 3(a) shows the plot of the theoretical turbidity response as a function of the
angle of detection q from 20" to 130". Such an angular variation is important to ascertain the sensitivity of optical
response to the detection angle (the design specification, being limited to a solid angle centered at 90°). The
400-600 nm allowable range of wavelength in the US EPA specification (as indicated in the spectral response range
for the detector) allows for as much as 40% variation compared to a 15% variation from the 860+60 nm range
allowance by ISO, both when measured at 90°, and throughout the range of angles investigated. Further, the size of
the solid angle, over which the scattered signal is integrated, can cause an additional uncertainty. This is attributed
to the variations in the non-linear scattering response for the calibration suspension and the sample suspension. The
tighter specification of 90±2.5" in ISO is likely to cause lesser variation as compared to the larger 90+30° angle range
specified by US EPA. The effect of particle refractive index on the turbidimetric measurement is shown in Figure
56
-------
3(b). A 15% change in the average refractive index of the suspension, can change the turbidimetric response by a
factor of 10. The absence of prior information on the optical properties of waterborne suspensions, therefore, makes
meaningful comparison between inter-regional or inter-seasonal turbidimetric measurements difficult.
While there are limitations that result from variations in the turbidimetric technique, the method itself can be
used to bridge the gap between the OPC detectable particles ( > 0.5 |-im), and the submicron sized particle (0.02-0.5
pm). For 13 > 3, the fraction of particles in the submicron range has a greater contribution to light scattering than
the larger size fraction9. The implication is significant because this submicron fraction is not detectable by the
OPC's, but may be detected by the turbidimetric method. To investigate the relative contributions from these two
size fractions, another set of simulations was made with two values of 13 (B = 2 - where the proportion of larger
particles is higher, and B = 4 - where the proportion of smaller particles is relatively larger). The limits of
integration were set for the two size ranges as 0.01 to 1.0 |im, and 1.0 to 10 |j,m. Figure 4 shows the four plots for
the fraction of the total light scattered for the two values of B and the two size ranges.. For the case of a suspension
with B = 2, and a 90° turbidity value of say 1.0 NTU, 99% of the 1.0 NTU is contributed by particles in the 1-10
p,m size range, and only 1% is contributed by the smaller size fraction. In such a case, measurements made with
OPC's may be sufficient in characterizing the distribution completely. On the other hand, for B = 4, 80% of the 1.0
NTU is contributed by the smaller size fraction, and only 20% by the larger (OPC detectable) size range. Thus,
while the OPC data would miss the large number of particles in the submicron range, the turbidimetric response
may be used to detect the presence of particles in this size range. The use of OPC's is effective for finished waters,
but the influent stream, with higher proportion of submicron particles, would still require use of the turbidimetric
approach. This has also been pointed out by Kavanaugh9.
Figures 3(a) and 3(b) also show that the intensity of scattered light for a given suspension decreases with
increasing angles, making the choice of 90° as the angle of detection appropriate to prevent errors due to vibrations.
The cost of such a choice, however, is severe. A loss of an order of magnitude in the signal strength implies a loss
of detection capability for concentration changes by a factor often. For the range of particles encountered in drinking
waters, as discussed above, the measure of turbidity is still useful, and motivates the investigation of increasing the
sensitivity of turbidimeters by using smaller detection angles. The maximum allowable concentration to prevent
coincidence error in OPC's is currently below the detection range of the turbidimeters. As an estimate, a change of
0.02 NTU, the least detectable limit of turbidimeters, is equivalent of a concentration change of ~3.5 x 104 #/mL of
1.0 |J,m PSL monodisperse particles. The maximum allowable concentration to prevent coincidence error in a
Hiac/Royco OPC '(HRLD-150 sensor) is 1.8 x 104 #/mL ; and 3.5 x 103 #/mL for Met-One OPC (LS-211 sensor),
both of which are lower than the least detectable concentrations in turbidimeters. A simple change to a smaller
detection angle would make the turbidimeters more sensitive to concentration changes, and at the same time bridge
the present gap between the lower detection limit of turbidimeters and upper concentration limits of OPC's.
From the discussion above, it follows that the difficulty in comparison of turbidity responses from different
instruments may arise from the ample allowances within the design criteria, and lack of information on the optical
properties of the suspensions. The turbidimeter design criteria seem to have evolved by agreement between the
available instrument manufacturers, and/or technical and cost constraints at the time of its inception5, and continues
to remain so from a need for regulatory compliance. Given the analysis described here, and the advances in laser and
photo detection technologies, it may be appropriate to re-evaluate these features of the design criteria with an intent
to upgrade the detection capability to meet the increasingly stringent demands.
Optical Particle Counters._Optical particle counters use the principles of light obscuration and/or light scattering to
size and count particles individually. The choice of the principle used for a specific sensor design is a function of
particle size of interest, particle concentrations, background noise, instrument ruggedness and cost. The sizing
response is influenced by particle properties such as size, shape and refractive index, and design of the optical sensor.
The performance of two particle counters (Met-One and Hiac/Royco) was evaluated in the present study to size
monodisperse PSL, SiO2, and several microorganisms.
To illustrate the effect of detection angle and particle refractive index on the instrument response, light scattering
intensity distribution was simulated theoretically for the Met-One fight scattering sensor (LS-211). The LS-211
detector collects scattered light over an angle range of 5-33°. The Lorentz-Mie scattering coefficient Cvv, was
57
-------
integrated over these angles as a function of particle diameters and three possible values of refractive indices to obtain
the sensor response shown in Figure 5. Calibration measurements using various PSL sizes are also plotted with the
theoretical curves in Figure 5. The plots indicate that particles in the 0.5 - ~ 4 |im size range with a refractive index
less than that of PSL, are likely to be undersized by the OPC. A loss in resolution in the ~ 2.5 - ~ 5.0 |im size
range for the LS-211 with PSL is predicted theoretically, and also observed in the calibration data.
The principle of obscuration type sensor (HRLD-150) used in the Hiac/Royco was reported by Sommer41. The
theoretically obtainable obscuration signal intensity is restricted by the size of the photodetector, which collects a
narrow, but finite, cone of scattered light in the forward direction. The forward cone also causes a loss of resolution
similar in nature to that for LS-211 sensor discussed above. Scattering intensity at forward angles, however, is
relatively insensitive to refractive index, and therefore, the response from an obscuration type sensor is less likely to
be influenced by the refractive index of the suspension particles.
The HRLD-150 sensor has a lower detection limit of 1.0 um and a coincidence limiting concentration of
I.SxlO4 #/mL. The sensor response for 1.500, 2.032, 3.500 and 5.100 (J.m PSL suspensions are plotted in Figure
6(a) as cumulative percentage concentration. The steepness of the curves is an indication of the size resolution. The
use of a 50% cutoff size (50% particles less than), an often used criteria for aerosol sizing instruments42,
underestimates the actual particle size, an artifact possibly resulting from the optical noise at the lower end of the
detectable range. A 90% cutoff size was found to better match the actual sizes.
The LS-211 sensor has a lower detection limit of 0.5 Jim, and a coincidence limiting concentration of 3.5x103
fe'mL. Two monodisperse suspensions of similar sizes but different refractive indices, 1.020 |a.m PSL (mw = 1.195)
and 1-100 Jim SiC>2 (mw = 1.10),were analyzed using this sensor (Figure 6(b)). The sensor respondedwith clearly
defined peaks for both the samples. The response from HRLD-150 for 5.100 urn PSL and 5.000 (J,m SiC>2
suspensions is also shown in Figure 6(b).
OPC response was next investigated for suspensions of B. subtilis, S. bows, and C muris oocysts on LS-211,
and additionally G muris cysts on the HRLD-150 (Figure 6(c)). The curves for S. bovis on both the instruments,
and B. subtilis on the HRLD-150, remained unresolved, as indicated by the least size being greater than the 50%
cutoff line.
Table 1 shows the microscopically determined sizes for the suspensions investigated in this study along with
the 90% cutoff sizes determined from the two OPC's. As anticipated, the OPC determined sizes for PSL samples
match well with the microscopic sizes reported by the manufacturer. It is interesting to note, that while the optical
siting of 1.020 f.tm was accurate, the larger 1.100 p.m SiC>2 particles, with refractive index lower than that for PSL,
were sized as 0.9 Jim (Figure 6(b)). The reverse was observed for the larger 5.100 |im PSL and 5.000 |j,m SiO2
particles measured with the obscuration sensor HRLD-150. The agreement between OPC measured sizes and the
microscopic sizes vary for the different microorganisms on each OPC due to shape and refractive index effects. The
LS-211 indicated a size of 1.0 (am for the rod shaped 6. subtilis (the microscopic size being 0.5-1 |im x 1.2-5.0
JJUn). The G. murfs response on the HRLD-150 indicated a peak at 7.8 Jim at the 90% cutoff (the expected size
range being 7-12 |im). A comparison of the responses for C muris from HRLD-150 and LS-211 highlights the
influence of optical sensor design and particle characteristics on sizing of particles. For the HRLD-150, the peak
appeared up in the ~ 4 (im range, as also reported by Rossi43. The LS-211, however, responded with d peak at
about 2 Jim as opposed to the microscopically observed 4-7 \ua diameter. The difference may have been a result of
the influence of angular dependence of optical response for the oocysts and/or the loss of size resolution in the ~ 2.5 -
-* 5 Jlni range as discussed above. Detailed information on the angular scatter intensity distribution from such
microorganisms is essential for accurate sizing. In the following section, the results of multiple angle light
scattering measurements for the microorganisms performed in this study are presented.
Multiple Angle Light Scattering (MALS). The objective of the present MALS study was to first develop and
validate the experimental technique and the data inversion method using PSL suspensions, followed by
measurement of angular scatter intensity distributions for monodisperse suspensions with known and unknown
Optical properties. The experimental setup described above was calibrated using 0.993 (im PSL suspensions, and
used to obtain multiple angle measurements for 0.460, 0.560, 0.750, 1.020, 1.500 and 2.032 urn PSL suspensions.
58
-------
The data inversion scheme referred to earlier was used for these measurements at 40750°, 50760° and 60770° angle
pairs to estimate the particle sizes. The inferred sizes were plotted against the sizes determined by the manufacturer
using electron microscopy in Figure 7. The high sensitivity of the scatter intensity to angular position at some
angles makes the choice of some angles for determining sizes more preferable than others, as also reported by Kerker
eta!44.
The MALS patterns for a 1.02 jam monodisperse PSL suspension, a 1.100 um monodisperse SiC>2 suspension,
and five suspensions of microorganism isolates are shown in Figures 8(a) and 8(b). The plots represent an average
of several replicates run at various concentrations to ensure absence of multiple scattering. The intensity scale is
arbitrary, and the plots have been separated out vertically for clarity. The change in angular distribution due to
refractive index differences for similar sized PSL and SiC>2 suspension is evident in Figure 8(a). The patterns for S.
bovis and E.faecium (both ovoid) are also shown in Figure 8(a) for comparison with the spherical PSL and SiO2
suspensions. Figure 8(b) shows the response for rod shaped microorganisms. The random orientation of the large
number of rods in the view volume tends to "wash-out" the finer structures (peaks and valleys as in the case for
spheres); however, the patterns in Figure 8(b) still reflect distinct shapes for the three microorganisms. These
"optical signatures" offer a potential approach for real time characterization for specific microorganisms.
Additionally, such angular distributions of scatter intensities, integrated over the band of collection angles of any
given OPC photosensor design, could provide a more valid scheme for accurate sizing. A summary of previous
studies on light scattering from microorganisms has been presented elsewhere32.
The data inversion scheme was applied to the angular distribution data for the microorganism suspensions, with
the assumption that the microorganisms can be approximated as homogeneous (uniformly distributed refractive
index) spherical particles. The assumption is crude, and was made only to test the applicability of the size
inversion scheme for the measurements obtained for microorganisms. The results were within the range of the
reported microscopic sizes and are listed in Table 1. A theoretical simulation with these diameters and the average
refractive index assumed for inversion, however, did not reproduce the experimentally obtained patterns. To
accomplish such a match-up, it is necessary to include cell shape effects and spatial refractive index variations, which
again remain largely ill-understood. A pattern recognition scheme from a library of known signatures, may provide
a more feasible route for characterizing microorganisms as suggested by Wyatt45.
MALS measurements were also made for two calibration suspensions used for turbidimeters : a formazin
standard, and a non-formazin standard (Jenwayd). The results are plotted in Figure 9 and show the enhanced
scattering response at smaller angles. The experimental results are comparable to the theoretical simulations of
turbidimetric response shown in Figures 3 (a & b).
The full potential of the MALS technique, either to uniquely characterize polydisperse suspensions using signal
from a "particle cloud", or to characterize individual particles using multiple sensors hi OPC's, has yet to be
realized. Flow cytometry, a technique used frequently in medical sciences for example, uses some of the principles
of MALS successfully46. The application of cytometry in drinking water industry has been limited primarily by
instrument sophistication and high capital cost. The availability of low cost lasers, photosensors, charge coupled
devices, and present day computational capabilities, however, can lead to development of multiple angle
measurement techniques, and thus make available the advantages of real time particle characterization using the
MALS method.
Summary and Conclusions
A brief review of the optical particle detect ion methods has been presented with applications hi drinking water
as the focus. A large body of literature exists regarding the problem of particle detection in liquids. An integrated
approach to the problem of turbidity measurement and particle detection by optical counters has been presented
using fundamental elastic fight scattering principles.
There is a need to re-evaluate the US EPA design criteria for turbidimeters with the purpose of making them
more stringent, possibly following the path of ISO. In our judgment,"a change to a smaller angle would improve
the sensitivity of the instrument without hampering the stability. No special effort was made to dampen the
59
-------
fl-i
I; jiff ['
J\ "If
vibrations on our setup, and no problems with stability were experienced for angles up to 30°. Besides,
commercially available ratio turbidimeters already use the second angle of detection to be as small as 15°.
Raw water feeds with power law parameter 13 > 3 (high proportion of particles smaller than the size detection
limit of most OPC's) are candidates for characterization using the turbidity method. A combination of turbidity and
particle counting methods is needed to obtain complete information on the suspended matter. An increase in
turbidimeter sensitivity by a change in detection angle may be used to bridge the lower detection limit of
turbidimeters and coincidence limited concentrations in OPC's. There is also a need to establish a database of
refractive indices of source water suspensions to permit a better comparison of inter-instrument, inter-regional, and
inter-seasonal turbidity measurements.
The sizing performance of the two commercial OPC's varied with the type of particles. Refractive index of the
particles is a key issue in accurate sizing of spherical particles. Sizing discrepancies could be explained on
theoretical grounds using known optical properties of spherical particles as in the case of SiO2 and PSL. In the
absence of structural and optical information as in the case of C. muris samples, however, the observed sizing errors
cannot be explained. Electromagnetic characterization of the species being sized is essential, and the application of
MALS method in one form or another may be the only method to fulfill the need.
A multiple angle light scattering experiment was developed, and calibrated using monodisperse PSL particles.
A size inversion procedure was used to infer PSL sizes in the 0.460 to 2.032 |j.m size range, and comparison was
made with the manufacturer specified sizes. Fair agreement was found, the results being sensitive to the choice of
angles used in the inversion. "Optical signatures" were obtained for pure cultures of B. subtilis, S. bovis, E.
faechtm, E. coli, and E. cloacae, and were found to be unique for each of the cultures. Such angular distribution
contains all the information that is needed to uniquely characterize suspensions of pure cultures of microorganisms,
or to accurately size microorganisms in optical particle counters.
Acknowledgments
The support of Kim Fox, James Goodrich, Clifford Johnson, Larry Lobring, Richard Miltner, James O'Dell,
James Owens, Lewis Rossman and Thomas Speth of the USEPA with equipment and materials is gratefully
acknowledged.
This work was supported in part by US EPA Cooperative Agreement CR816700, W4-A. The conclusions
represent the views of the authors, and do not necessarily represent the opinions, policies or recommendations of the
US EPA. Part of this work was completed during the first author's appointment to the Postgraduate Research
Participation Program administered by the Oak Ridge Institute for Science and Education through an inter-agency
agreement between the US DOE and the US EPA.
The mention of commercial products is not to be construed as an endorsement of such products.
d Jenway LTD., Felsted, Dunmow, Essex, England CM6 3LB.
60
-------
References
1. Moore, A.C., Herwaldt, B.L., Craun,G.F., Calderon, R.L., Highsmith, A., and Juranek, D.D., "Waterborne
Disease in the United States, 1991 and 1992", JAWWA, 86(2):87-99, (1994).
2. Pontius, F.W., "Protecting the Public Against Cryptosporidium", JAWWA, 84(8): 18-22,122-123 (1993).
3. Hargesheimer, E.E., Lewis, C.M., and Yentsch, C.M., Evaluation of Particle Counting as a Measure of
Treatment Plant Performance, American Water Works Association Research Foundation, Denver (1992).
4. McCoy, W.F., and Olson, B.H., "Relationship Among Turbidity, Particle Counts and Bateriological Qualith
within Distribution Lines", Water Research, 20(8): 1023- 1029, (1986).
5. Hart, V.S., Johnson, C.E., and Letterman, R.D., "An analysis of Low-Level Turbidity Measurements,"
JAWWA, 84(12:40-45, (1992)
6. Black, A. P., and Hannah, S. A., "Measurement of Low Turbidities", JAWWA, 57:901-917, (1965).
7. Sigrist, W., "An Assessment of the Latest Discoveries in the Measurement of Turbidity", VOM Wasser,
Verlag Chemie, GmbH., Weinheim/Bergstr, (1975).
8. Papaeosta, K., Spair, J.A., and Katz, M., "The Rationale for the Establishment of a Certified ReferenceStandard
for Nephelometric Instruments," AWWA/WQTC, San Diego, November 11-15, 1299-1332, (1990).
9. Kavanaugh, M.C., Tate, C.H., Trussel, A.R., Trussell, R.R., and Treweek, G., "Use of Particle Size
Distribution Measur ements for Selection and Control of Solid/Liquid Separation Processes," in P'articulate in
Water, Ed. Kavanaugh et al, Adv. In Chem. Ser., Vol 189, (ACS), (1980).
10. Ginn Jr., T.R., Amirtharrajah, A., and Karr, P.R., "Effects of Particle Detachment in Granular-Media
Filtration;" JAWWA, 34(2):66-76, (1992).
11. Lewis, CM., Hargesheimer, E.E., and Yentsch, CM., "Selecting Particle Counters for Process Monitoring ",
JAWWA, 84(12):46-53, (1992).
12. Sommer, H.T., "Correlation of Particle Counts and Turbidity: The Effect of Raw Water Particle Size
Distribution", AWWA/WQTC, San Francisco, CA, November 6-10, (1994).
13. LeChavellier, M.W., and Norton, W.D., "Examining Relationships Between Particle Counts and Giardia,
Cryptosporidium and Turbidity ", JAWWA, 84(12):5460, (1992).
14. Letterman, R.D., "Comparison between Particle Counters and Other Measurement Instruments ", Tech.
Report #502, Pacific Scientific LTD., Silver Spring, MD, November, (1993).
15. Hach, C.C., Vanous, R.D., and Heer, J.M., "Understanding Turbidity Measurement", Technical Information
Series - Booklet No. 11, Hach Company, Loveland, Colorado (1990).
16. Wyatt, P.J., "Light Scattering and the Absolute Characterization of Macromolecules ", Analytica Chimica
Acta, 272:1-40, (1993).
17. Ray, A.K., Souyri, A., Davis, E.J., and Allen, T.M., "Precision of Light Scattering Techniques for Measuring
Parameters of Microspheres ", Applied Optics, 30(27):3974-3983, (1991).
18. Wyatt, P.J., and Phillips, D.T., "A New Instrument for the Study of Individual Aerosol Particles" J Coll
Inter. Sc. 39(1): 125-135, (1972).
61
-------
19. Dick, W.D., McMiirry, P.M., and Bottiger, JR., "Size- and Composition-Dependent Response of the
DAWN-A Multi-angle Single-Particle Optical Detector", Aerosol Science and Technology, 20:345-362, (1994).
20. van de Hulst, H. C., Light Scattering by Small Particles, Dover, New York, (1981).
21, Livesey, P.J., and Billmeyer Jr., F.W., "Particle Size Determination by Low Angle Light Scattering: New
Instrumentation and a Rapid Method of Interpreting Data", J Coll. Inter. Sc., 30(4):447-472 (1969).
22. Kerker, M., The Scattering of Light and Other Electromagnetic Radiation, Academic Press, New York,
(1969).
23. Bohren, C.F., and Huffman, D.R., Absorption and Scattering of Light by Small Particles, John Wiley &
Sons, New York (1983).
24. Hinds, W.C., Aerosol Technology-Properties, Behavior, and Measurement of Airborne Particles, John-Wiley
& Sons, New York (1982).
25. Dave, J.V., Subroutines for Computing the Parameters of the Electromagnetic Radiation Scattered by a
Sphere, IBM Scientific Center Report 320-3237, (1968).
26. Feklichev, V.G., Diagnostic Constants of Minerals, Mir Publishers, Moscow (1992).
27. Greenberg, A.E., Clesceri, L.S., and Eaton, A.D., Standard Methods for the Examination of Water and
Wastewater, American Public Health Association, Washington DC., (1992).
28. Liu, B. Y. H., Szymanski, W. W., and Ann, K. H., "On Aerosol Size Distribution Measurement by Laser &
White Light Optical Particle Counters", J Enviro. Sci., pp. 19 - 24, May/June (1985).
29. Goldgrabe, J.C., "Particle Counting as a Method of Evaluating Conventional and Biological Filter
Performance", MS Thesis, University of Cincinnati, Cincinnati, OK (1992).
30. Sommer, H.T., "Examining Molecular Background Scattering and the Trade-Off between Sensitivity and
Sample Flow Rate of Optical Particle Counters", Hiac/Royco Technical Literature: Reprint from
Microcontamination, March (1991).
31. Harris, J.E., "Characterization of Suspended Matter in the Gulf of Mexico-II: Particle Size Analysis of
Suspended Matter from Deep Water", Deep-Sea Research, 24:1055- 1061,(1977).
32. Sethi, V., "Experimental Studies in Transport of Dissolved and Suspended Contaminants in Drinking Water
Distribution Systems", Ph.D. Dissertation, University of Cincinnati, Cincinnati (1996).
33. Vavra, J., and Antalik, J., "Particle-size Analysis of Latexes by Classical and Small-angle
Spectrodissymmetry", Coll. Polym. Sci., 269:1060-1063, (1991).
34. Antalik, J., Vavra, J., and Liska, M.," Application of Spectrodissymmetry Method by Characterization of
Poly(styrene)", Coll. Polym. Sci., 271:95-99, (1993),
35. Chang, H.S., and Biswas, P., "In Situ Light Scattering Dissymmetry Measurements of the Evolution of the
Aerosol Size Distribution in Flames", J Coll. Inter. Sc.,
153(1): 157-166, (1992).
36. Chang, H.S., Lin, W.Y., and Biswas, P., "An Inversion Technique to Determine the Aerosol Size
Distribution in Multicomponent Systems from In Situ Light Scattering Measurements", Aerosol Science and
Technology, 22:24-32 (1995).
62
-------
37. Owens, J.H., Miltner, R.J., Schaefer, F.W. M and Rice, E.W., "Pilot-Scale Ozone Inactivation of
Cryptosporidium and Giardia", Proc. Water Quality Technol.
Conf. (AWWA), pp. 1319-1328, (1994).
38. USEPA Method 180.1, EPA/600/R-93/100, EMSL, USEPA, Cincinnati, OH 45269, August, (1993).
39. ISO, International Standard 7027, International Organization for Standardization, Geneva, Switzerland, (1990).
40. O'Konski C.T., Bitron, M.D., and Higuchi, W.I., "Light Scattering Instrumentation for Particle Size
Distribution Measurements", Symposium on Particle Size Measurement, Sixty-first Annual Meeting, ASTM
Boston (1958).
41. Sommer, H.T., "Performance of Monochromatic and White Light Extinction Particle Counters", IES
International Conference of Particle Detection, Metrology and Control, Arlington, VA, February 5-7, (1990).
42. Liu, B.Y.H., Yoo, S.H., and Chae, S.K., "Lower Detection Limit of Aerosol Particle Counters", Proceedings
of the 40th Annual Technical Meeting, Institute of Environmental Sciences, Chicago, IL, May 1-6, (1994).
43. Rossi, P., "Particle Counters: New Tools for Monitoring Drinking Water Quality and Plant Performance",
Tech. Report #505, Pacific Scientific LTD., Silver Spring, 1@0, November, (1993).
44. Kerker, M., Daby, G.L., Cohen, J.P., Kratohvil, J.P., and Matijevic, E., "Particle Size Distribution in La Mer
Sulfur Sols", J Phys. Chem., 67:2105-2111, (1963).
45. Wyatt, P.J., "Differential Light Scattering: a Physical Method for Identifying Living Bacterial Cells", Applied
Optics, 7(10):1879-1896, (1968).
46. Lloyd, D., Ed. Flow Cyometry in Microbiology, Springer-Verlag, New York, (1993).
63
-------
"' I," I" .I"!1
£
5
o
3
I
13
§
E
5
s
en
•88
u
"3
«tt
1
e/j
1
Q,
U
-------
Table!. USEPA and ISO Design Criteria for Turbidimeters
USEPA
ISO
Light Source
Tungsten Lamp
2200-3000 K
860nm
Beam path length
through the sample
<10cm
Parallel beam with
maximum divergence/
convergence of 2.5°
Detection Angle
90° + 30°
90 ±2.5°
00 = 10°-20°
Spectral Response (X) 400-600 nm
Not specified
Sensitivity
0.02 NTU for waters
with 1.0 NTU or less
0.01FNU for waters
with 1.0 FNU of less.
Calibration
Formazin or Polymeric Formazin
Standard
NTU Nephelometric Turbidity Units
FNU Formazin Nephelometric Units
light source wavelength
aperture angle
65
-------
I
«
I
i
I
.!>
66
-------
Sample
Trap
Detector
Figure 2. Schematic diagram for turbidimeter design specifications.
67
-------
Figure 3. Theoretical optical response from bubidimeten: (a) effect of USEPA and ISO
design criteria for X and collection angle, and (b) effect of refractive index variations.
100
10.00 -
o
CM
It
68
-------
100 -q
05
Effective
Refractive
Index
20 40 60 80 100 120 140
0
Angle, 8
69
-------
1
s
§
•J3
a
i
u<
Figure 4. Fractional contribution to turbidimetric response as a function of P
for two size ranges.
0.1 -
0.01
p = 2
V"
20 40
0.01-1.0 nm
1-10 jun
0.01-1.0 jim
—f-
60
~T~
80
Angle, 9
T
100 120 140
70
-------
Figure S. Optical Response for LS-211 sensor on Met-One OPC.
10s -i
o o
Theoretical Response
PSL (m,= 1.197)
8102001,= 1.10)
Microorganism (n^ assumed as 1.05)
Experimental Calibration Response
O Pro-Gain LS-211 Response
from PSL calibration
10
Particle Diameter (urn)
100
71
-------
Figure 6. Size response from OPC's for (a) PSL suspensions,
(b) PSL and SiOr and (c) suspensions of microorganisms.
100
1 \J\J
80 -
-
60 -
40 -
20 -
0
(a)
ff^/^ r~*
I\ if 1
// / /
*1 1 l
.' / i
< /
9- i , 1
III /
If/ /
1 / /
I / /
1 i ^ '
1 / /
I I / '
Ij / /
J i f l
11X
If /
J/-*^"""^
90% cutoff
50% cutoff
HRLD-150 Sensor
« 1.500PSL
• 2.032 PSL
» 3.500PSL
* 5.100 PSL
34567
Particle Diameter (jam)
8
10
72
-------
100-
80 -
60 -
40 -
-1—i—•—r
0 1 2
0-
5.000 SiOj
5.100 PSL
T
3
T
4
HT
5
_1—,—r
6 7
~r
8
~r
9
10
Particle Diameter (nm)
73
-------
100-
in
HRLD-1SO
O B.fubtilis
a S. bovfs
Ctyptosporidiwn muris
Giardia muris
LS-211
• B. subtilis
• S. bavis
Cryptosporidium muris
20-
0
1456
Particle Diameter (n
8
10
74
-------
Figure?. Comparison of MALS inferred PSL diameters with manufacturer
specified lizes.
2.50
2.00-
Ratio angles
40°/50°
50C/60"
A 60a/70°
0.50
0.00
Manufacturer specified particle diameter (jim)
75
-------
Figures. Angular distribution of scattered intensity for (a) spheres and ovoids, and
(b) rod shaped microorganisms.
.£
A
s
00
1
q. ti
''•a''
Oi
V
.•-a.
(a)
E.faecium
1.020 PSL
1.10SIO2
*
• i i ] i i i | i i i ] i—i—i—,—i—,—,—|—,—,—r
20 40 60 80 100 120 140
Angle, 6
76
-------
I
rt
u
00
I
3
\
(b)
1
V
\
\
\
B. subtilis
Eco//
E. cloacae
1 ' I '
80
Angle, 8
20
40
60
100
1 ' I '
120
140
77
-------
Figure 9. MALS measurements for fonnazin and non-fonnazin calibratioa standards.
i
•3
e
.§>
CO
0.1 -
0.01
20
40
Forraazin2.86KTU
Non-fonnazin 5.0 NTU Standard (Jenway)
I
60
80
Angle, 9
100
1 I ' '
120
140
78
-------
Reliability of Non-Hazardous Surrogates for Determination of
Cryptosporidium Removal in Bag Filtration Systems
Sylvana Y. Li, James A. Goodrich, James H. Owens,
Frank W. Schaefer IE, and Robert M. Clark
Water Supply and Water Resources Division
Cincinnati, Ohio 45268
GeneE.Willeke
Institute of Environmental Sciences, Miami University
Oxford, Ohio 45056
Abstract
Testing of field-scale bag filtration systems yielded results indicating that 4-6 |J,m polystyrene microspheres can
be used as a reliable surrogate for determination of Cryptosporidium oocysts removal in bag filtration process. A
nearly perfect linear correlation was observed between log removals of 4-6 |j.m polystyrene microspheres and
Cryptosporidium parvum oocysts. The correlations between Cryptosporidium removal and other potential
surrogates including 4-6 |im particle counts, 1-25 |im particle counts, and turbidity produced slopes that were
substantially greater than 1, and may vary geographically due to variations in raw water quality. The similarities
and dissimilarities in log removal between Cryptosporidium oocysts and the potential surrogates for the bag
filtration system were found to be the result of their differences in size distribution.
79
-------
Introduction
Ctyptosporidium, a waterborne agent causing gastroenteritis, is a known microbiological contaminant in
drinking water responsible for diarrhea and death among children, and immune-compromised individuals.1.2
Regional surveysS.'* have shown that Cryptosporidium oocysts are widespread in drinking water sources and have
been detected in insufficiently treated effluent from water treatment plants and in swimming pools.5
Cryptosportdium-Klated diseases have been also reported throughout the world.6.7-8
Investigations have shown that drinking water disinfectants such as chlorine or monochloramine at typical
dosages have virtually no effect on the inactivation of Cryptosporidium oocysts.9 High-dose ozonation appears to be
effective for inactivating Ctyptosporidium oocysts in drinking water,10*11 but its application may result in generation
of disinfectant by-products that exceed proposed MCLs.
Commonly employed in water plants, filtration is likely to be the most practical treatment technology used for
Cryptosporidium removal in the near future.12 At the present time, however, there are no accurate and precise
methods for determination of Cryptosporidium removal rates in filtration systems. Furthermore, the recovery of
Cryptosporidium varies significantly between different experimental methodologies and procedures,!2 and the direct
use of Cryptosporidium in treatment studies would pose a potential health risk. Therefore, finding more reliable,
non-hazardous surrogates is necessary.
Studies developed to determine the pliability of Cryptosporidium oocysts indicated that during the filtration
process, Cryptosporidium oocysts may pass through filtration membrane pores which are smaller than the diameter
of the organism and a fraction of these oocysts remain viableJ4 An understanding of this characteristics are important
for evaluation and optimization of filtration-based physical treatment systems.
LeChevallier and Nortonls proposed that turbidity and >5 urn particle counts are useful indicators of filtration-
based plant performance. In this study, four surrogates were compared with Cryptosporidium oocysts in
determinating removal rates of drinking water bag filtration systems. The surrogates tested were turbidity, 1-25 \im
particle counts, 4-6 p,m particle counts, and 4-6 jam polystyrene fluorescent microspheres. These results are
discussed with relation to the similarity and dissimilarities between the size distributions of Cryptosporidium
oocysts and the potential surrogates.
Test Methods
Three field-scale bag filtration systems acquired from different vendors were evaluated for the removal of
Cryptosporidium oocysts and potential surrogates. Bag filter #1 was supplied by US Strainrite Inc., Lewiston,
Maine; bag filter #2 was manufactured by Filtration Systems, a division of Mechanical Mfg. Co. in Sunrise,
Florida; and bag filter #3 was obtained from the 3M company in St. Paul, Minnesota. Made of polypropylene
fabrics, bag filters #1 and #3 consist of multiple layers and bag filter #2 appears to be a single layer. The nominal
porosity of bag filter #1 and #2 is 1 Jim. According to the fine silica testing results from the manufacturer, bag filter
$3 had a 99% and 95% removal efficiency for 2.5 )J.m and 1.5 p.m dust, respectively.
Cryptosporidium parvum oocysts were isolated from sieved (10-, 20-, 60-, and 100 mesh sieves) feces of
Holstein bull calves by centrifugation (1100 x g) through a step gradient of Sheather's sucrose.16 Purified
Cryptosporidium oocysts were stored in phosphate buffered saline (PBS) with penicillin (100 unit/ml) and
streptomycin (100 Hg/ml) at 4 °C for up to 6-8 weeks. These oocysts are referred to as antibiotic preserved
Cryptosporidium oocysts (APO) in this research. Another type of Cryptosporidium oocysts was obtained from the
University of Arizona. These oocysts, which were stored in potassium dichromate at 4 °C for approximately one
month, are referred to as dichromate preserved Cryptosporidium oocysts (DPO).
A pre-determined number of fluorescent 4-6 p.m polystyrene microspheres, supplied by Polysciences, Inc.,
Warrington, PA 18976 or Cryptosporidium oocysts suspended in a fixed volume of 0.01% (v/v) Tween 20
solution, was fed into the raw water upstream of the inlet pump (Figure 1). Approximately 5% of the total effluent
volume was filtered with a 1 urn pore size polycarbonate membrane (diameter=293 mm), supported in a stainless
steel filter manifold. The filter was removed from the manifold and eluted with a squeegee using approximately 200
80
-------
ml 0.01% (v/v) Tween 20 solution. The eluant was concentrated to 0.5 - 7.5 ml via centrifugation at 1200 x g.
Cryptosporidium parvum oocysts in both influent and effluent samples were stained with an indirect fluorescent
monoclonal antibody (IFA), supplied by Waterborne, Inc., New Orleans, LA 70118. Polystyrene microspheres and
oocysts were enumerated using a hemacytometer under a ultraviolet (UV) microscope using epifluorescence at 400x
total magnification.
The influent loading rate ranged from 9.17xl03 to l.SOxlO5 per liter with an average of 5.53xl04 per liter for
the polystyrene microspheres. For Cryptosporidium oocysts, the loading rate ranged from 7.04xl03 to 1.34xl05 per
liter and averaged 3.24x104 per liter.
Particle count analyses were conducted using a Met One® particle counter. The light scattering liquid 211
sensor provided measurements for particle size ranging from 1 to 25 u.m. Turbidity measurements were performed
with a Hach® 2100P portable turbidimeter. Unlike the polystyrene microspheres and Cryptosporidium oocysts,
turbidity and particle counts in the raw water were subject to natural variations in raw water quality. The measured
raw water turbidity during the study period average 10.59 NTU with a maximum of 59.4 NTU, and a minimum of
1.81 NTU.
The microspheres studies, turbidity and particle count measurements were performed at pressure drops of 0,
0.35 0.70, 1.05 and 1.76 kg/cm2 (0, 5, 10, 15 and 25 psi, respectively) for bag filter #2, and 0, 0.49, 1.05 and
1.76kg/cm2 (0, 7, 15 and 25 psi, respectively) for bag filters #1 and #3. Average inlet pressure of 3.52kg/cm2 ( 50
psi) and influent rates of 47, 95 and 151 liter/min (12.5, 25 and 40 gpm, respectively) were used. Log removal
studies comparing APO and DPO Cryptosporidium oocysts for the bag filters was examined at a flow rate of 95
liter/min (25 gpm) using the same procedure as for the potential Cryptosporidium surrogates. The duration of each
Cryptosporidium or microsphere test was approximately 70 minutes.
Results
Table 1 presents the observed log removal for turbidity, 1-25 (J.m particle counts, 4-6 (im particle counts, and
4.5 fim polystyrene microspheres at various operational conditions. The log removal for APO Cryptosporidium for
the three bag filters are summarized in Table 2.
Inspection of Table 2 shows that without application of bag filters into the filter vessels, the three blank tests
yielded consistent APO Cryptosporidium log removal of 0.12, 0.13, and 0.19 for the vessels of bag filter #1, #2,
and #3, respectively. In comparison, a blank test using the DPO Cryptosporidium shows a lower removal rate. Of
2.70xl08 DPO Cryptosporidium oocysts spiked into the influent, 2.57xl08 were recovered in the system effluent.
The corresponding log removal was 0.02, lower than that of the APO Cryptosporidium for the same bag filter vessel
(Table 2). The difference was likely to reflect the effect of oocysts adhesion to surface walls of the system. This
phenomenon is significantly decreased when oocysts are stored in potassium dichromate. Previous studies
conducted by US EPA Water Supply & Water Resources Division also show that the potassium dichromate may
change surface characteristics of the oocysts such as zeta potential.17
In order to account for the influence of system loss on the removal rate of tested bag filters, the determined log
removal for APO and DPO Cryptosporidium oocysts are corrected with their respective averaged removal rates from
the blank tests using the following equation:
~ LR
app
(1)
where LR is log removal, and subscripts act, app, and blk refer to actual (corrected), apparent (observed), and blank,
respectively. The corrected log removal rates are presented in Table 1.
As shown in Table 2, the corrected log removal of bag filter #1 for APO Cryptosporidium ranges from 1.35 to
1.48 with an average of 1.41+0.066. Bag filter #3 has the highest log removal ranging from 3.00 to 3.63 with an
average of 3.29+0.32, while bag filter #2 yields the lowest log removal in the range of 0.26 to 0.64 (Table 2).
Operational parameters investigated included pressure drop across the bag filter, flow rate, Cryptosporidium and
81
-------
surrogate influent spike levels. The concentration level of the spike may positively or negatively affect removal rates
of the bag filters. In order to evaluate these possibilities for the bag filters, log removal rates for the 4-6 jam
polystyrene microspheres were determined at various operational conditions (Table 2). Figure 2 shows the log
reduction variations with varying pressure drops at each flow rate tested for the bag filters. Figure 3 displays the
correlation between log reduction and loading of the polystyrene microspheres, turbidity, 1-25 \lm particle counts,
and 4-6 jxm particle counts.
As shown in Figure 4, the reduction of 4-6 jam polystyrene microspheres at a given flow rate exhibits no
significant dependence on pressure drop for bag filter #1 and #2, but decreases with pressure drop for bag filter #3.
The log removal ranges from 1.14 to 1.88 with an average of 1.39±0.19 (Is, n=23) for bag filter #1, and from 0.14
to 0.72 with an average of 0.46±0.17 (Is, n=7) for bag filter #2 (Table 1). In contrast, bag filter #3 shows
significant dependency of log removal on the pressure drop (Figure 4). Log removal for the 4-6 jam polystyrene
microspheres varies from 0.93 to 3.42 with an average of 2.08±1.05 (Table 1). The log removal drops significantly
as the bag becomes fouled and pressure drop increases. The large standard deviation reflects the effect of pressure
drop on performance of the bag filter #3.
As shown in Figure 5, bag filter #1 displays fairly constant log removals for the 4-6 fim polystyrene
microspheres and turbidity. It is apparent from the lack of correlations that the log reduction for bag filter #1 is not
dependent on the number of 4-6 |im polystyrene microspheres and turbidity loaded in the influent. Additionally,
relative standard deviations of the corrected log removal for bag filter #1 (Table 1) are 13.67%, 20.93%, 42.73%,
and 36.84% for 4-6 [im polystyrene microspheres, turbidity, 1-25 (am particle counts, and 4-6 (0,m particle counts,
respectively. Also shown in Figure 5, log removal for bag filter #3 varies substantially. However, the variation
appears to result from significant dependance of bag filter #3's performance on other operational parameters including
pressure drop and flow rate (Figure 4).
Discussion
Reliability of Potential Surrogates
Based on the corrected log removals for APO Cryptosporidium and the potential surrogates shown in Tables 1
and 2, the reliability of each surrogate for determination of Cryptosporidium removal rates was evaluated in terms of
accuracy and precision. The removal rates for APO Cryptosporidium and its surrogates were averaged for each of
the three bag filters, the average log removal of the surrogates for bag filter #3 include only those trials with a
pressure drop of less than 7 psi, the same pressure drop range at which removal rates for APO Cryptosporidium were
examined. As mentioned earlier, bag filter #3 performance is a function of pressure drop; the removal rate
significantly decreases with increase of pressure.
As shown in Figure 4, correlations in log reduction between APO Cryptosporidium and its potential surrogates
can be sufficiently described by the following linear models:
Log (Cijptosportdiuni) = 1.040 Log (microspheres) - 0.041; R2=0.999
Log (Cryptosporidium^ 1.145 Log (4-6 |Jin p.c.) + 0.134; R2=0.980
Log (Cryptosporidium') = 1.510 Log (1-25 |J.m p.c.) + 0.036; R2=0.979
Log (Cryptosporidium) = 1.913 Log (turbidity) + 0.080; R2=0.969
The slope of the linear correlation increases from nearly 1 for the 4-6 \im polystyrene microspheres to 1.913 for
turbidity. The 1:1 correlation in log reduction between 4-6 |im polystyrene microspheres and APO
Cryptosporitttuin is associated with the highest R2. Although all of these models have high R2 values, the results
suggest that 4-6 urn polystyrene microspheres may be the most accurate among the surrogates examined. This
relationship may vary somewhat at high pressure drops because of the Cryptosporidium's ability to contort, however
current research being performed by the authors indicates this difference may not be statistically significant. Of
greater importance to removal efficiencies is the bag filter gasket and vessel design and integrity. One other note is
that bag filter manufacturers recommended that the pressure drop should not exceed those used in this study.
Pressure drops any larger than those reported here, will most likely prevent the bag filter from operating.
82
-------
Figure 4 also shows that the standard deviation associated with averages of log removal varies among the
Cryptosporidium surrogates. The 4-6 |0,m polystyrene microspheres and turbidity exhibit a standard deviation for
each filter bag similar in magnitude with those of the APO Cryptosporidium (Figure 4a, d). On the other hand,
particle counts, particularly the 1-25 (im range, have higher standard deviations (Figure 4b, c). These results
suggest that both the 4-6 (im polystyrene microspheres and turbidity are more precise than particle counts in
determination of bag filter performance for removal of Cryptosporidium oocysts.
Surrogate Reliability and Size Distribution
The bag filter study results suggest that reliability of a Cryptosporidium surrogate for physical filtration
evaluation (without the addition of chemical coagulants) depends on the similarity of its size distribution with that
of Cryptosporidium. As shown in Figure 5, the 4-6 pm polystyrene microspheres have a particle size distribution
ranging from 3.51 to 7.00 |im, which is very similar to that of the Cryptosporidium oocysts. Both are characterized
by a single size peak centered at 4.00-6.00 |J,m. The similar particle size distribution yields a nearly perfect linear
correlation in log removal between APO Cryptosporidium and 4-6 pm polystyrene microspheres.
On the other hand, 1-25 Jim particle counts in the raw water exhibit a significantly different size distribution
from that of Cryptosporidium and the 4-6 [im polystyrene microspheres (Figure 5). As shown in Figure 7, 1-25 |im
particle counts have two size distribution peaks; one peak occurs in the range of 1.01 to 2.00 |4,m for nearly 70% of
the particles and the other peak occurs at 5.01 to 6.00 jim. Because the bag filters have an average nominal pore
size of 1 urn, the greater proportion of the 1.01-2.00 pm particles in the raw water leads to an apparently lower log
removal for the 1-25 |im particle counts. Consequently, use of natural surrogates such as 1-25 Jim particle counts
will be most likely subjected to limitations imposed by natural variations in the source water and seasonal change.
Conclusions
Because of similar particle size distribution with Cryptosporidium oocysts, 4-6 um polystyrene microspheres
are an accurate and precise surrogate for determination of Cryptosporidium log removal in physical bag filtration
process without the addition of chemical coagulant. One log removal of 4-6 urn polystyrene microspheres is
equivalent to 1.040 log removal of Cryptosporidium. Because of their uniform size distribution over time, the
determined removal rates for 4-6 Jim polystyrene microspheres are associated with the least variance that is
comparable with the variation for APO Cryptosporidium. Initial result also indicates that potassium dichromate
used in oocyst preservation results in significant changes of oocyst surface characteristics and adhesion ability onto
walls of treatment system. Compared with DPO Cryptosporidium, APO Cryptosporidium are more representative
of the oocysts in natural environment.
Other examined Cryptosporidium surrogates, including naturally occurring 4-6 pm particle counts, 1-25 um
particle counts, and turbidity, are less accurate and precise. Because of the differences in particle size distribution,
the surrogates and APO Cryptosporidium show significantly different log removals for the three bag filters. The
statistically significant linear correlations suggest that 1 log removal of APO Cryptosporidium is indicated by
1.145, 1.510, and 1.913 log removal, respectively, for 4-6 fim particle count, 1-25 u.m particle count, and turbidity.
Additionally, the particle counts have greater variation of log removals for given bag filters than APO
Cryptosporidium, 4-6 |J,m polystyrene microspheres, and particle counts. The greater variation appears to be the
result of natural variation in water chemistry of the raw water with time, leading particle counts to be the least
precise Cryptosporidium surrogates for physical filtration evaluations. Filtration processes that rely on chemical
addition, for particle adsorption to filter media will need to be evaluated in future research.
Acknowledgment
The authors acknowledge Dr. John Cicmanec and Mr. Hector E. Moreno of the Microbial Contaminants
Control Branch, Water Supply and Water Resources Division, U.S. EPA for providing purified Cryptosporidium
oocysts in this research.
83
-------
References
1 Pontius, F.W. Protecting the Public Against Cryptosporidium. J. AWWA, 85:4:122 (Aug., 1993)
2 MacKenzie; W.R., Hoxie, N.J.; Proctor, M.E.; Gradus, M.S.; Blair, K.A.; Peterson, D.E.; Kazmierczak, J.J.;
Addies, D.G.; Fox, K.R.; Rose, J.B. & Davis, J.P. A massive outbreak in Milwaukee of Cryptosporidium
infection transmitted through the public water supply. The New England J. ofMedicine, 331:3:161 (1994)
3 LeChevailier, M.W.; Norton, W.D. and Lee, R.G. Occurrence of Giardia and Cryptosporidium spp. in Surface
Water Supplies. Applied and Environmental Microbiology, 57:1620 (1991)
4 LeChevallier, M.W.; Norton, W.D. and Lee, R.G. Giardia and Cryptosporidium spp. in Filtered Drinking
Water Supplies. Applied and Environmental Microbiology, 57:2617 (1991)
W '.: ,
5 Sorvillo, F.J.; Fujioka, K.; Nahlen, B.; Torney, M.P.; Kebabjian, R. and Mascola, L. Swimming-associated
Cryptosporidiosis. American J. Public Health, 82:5:742 (1992)
6 Molbak, K.; Hojlyng, N.; Gottschau, A.; Correia Sa, J.C.; Ingholt, L.; Jose Da Silva, A.P. & Aaby, P.
Cryptosporidiosis in infancy and childhood mortality in Guinea Bissau, West Africa. BMJ, 307:417 (1993)
7 Chen, Y.; Yao, F.; Li, H.; Shi, W.; Dai , M. & Lu, M. Cryptosporidium infection and diarrhea in rural and
urban areas of Jiangsu, People's Republic of China. J. Clinical Microbiology, 30:2:492 (1992)
8 Roach, P.D.; Olson, M.E.; Whitley, G. and Wallis, P.M. Waterborne Giardia Cysts and Cryptosporidium
Oocysts in the Yukon, Canada. Applied and Environmental Microbiology, 59:67 (1993)
9 Korich, D.G.; Mead, J.R.; Madore, M.S.; Sinclair, N.A.; and Sterling, C.R. Effects of Ozone, Chlorine
Dioxide, Chlorine, and Monochloramine on Cryptosporidium parvum Oocyst Viability. Applied and
Environmental Microbiology, 56:1423 (1990)
10 Finch, G.R.; Black, E.K.; Gyurek, L.; and Belosevic, M. Ozone Inactivation of Cryptosporidium parvum in
Demand-free Phosphate Buffer Determined by in Vitro Excystation and Animal Infectivity. Applied and
Environmental Microbiology, 59:4203(1993)
11 Owens, J.H.; Miltner, R.J.; Schaefer, F.W. Ill and Rice, E.W. Pilot-Scale Ozone Inactivation of
Cryptosporidium and Giardia. Proc. 1994 AWWA WQTC, San Francisco, Calif.
12 Chapman, P.A. and Rush, B.A. Efficiency of Sand Filtration for Removing Cryptosporidium Oocysts from
Water. J. Med. Microbiol, 32:243 (1990)
13 Nieminski, E.G.; Schaefer, F.W. Ill and Ongerth, J.E. Comparison of Two Methods for Detection of Giardia
Cysts and Cryptosporidium Oocysts in Water. Applied and Environmental Microbiology, 61:1714 (1995)
14 Li, S.Y.; Goodrich, J.A.; Owens, J.H.; Clark, R.M.; Willeke, G.E. and Schaefer F.W. III. Potential
Cryptosporidium Surrogates and Evaluation of Pliable Oocysts. Proc. 21st Annual RREL Research
Symposium, EPA/600/R-95/012 p.305-308. (1995)
15 LeChevallier, M.W. & Norton, W.D. Examining relationships between particle counts and Giardia,
Cryptosporidium, and turbidity. J. AWWA, 84:12:54 (December, 1992)
16 Finch, G.R.; Daniels, C.W.; Black, E.K.; Schaefer F.W. Ill and Belosevic M. Dose response of
Cryptosporidium parvum in outbred neonatal CD-I. Applied and Environmental Microbiology, 59:3661 (1993)
84
-------
17 Lytle D.A. and Fox, K.R. Particle Counting and Zeta Potential Measurements For Optimizing Filtration
Treatment Performance. Proc. 1994 AWWA WQTC, San Francisco, CA.
85
-------
§
i
s
UJ
W
I
'S.
if
||
— — 00 O
90
•«•
W W | U Id | Id
O O *•• ^ "^
?!
gw , was , w
«n 25 r* 5?
S p , 8 £ , $
O —t O O O
.79E+05 3J7Et04
.94E-f04 I.02E+03
89E+04 I.09E+03
22Et04 2.69E+04
9JE+04 1.7SE+04
5 »•? | 2 ? i ?
00 O O 0
•C — •» o» —
OB "i I d !>• I A
— — — e4 —
*» "> S S 2
«i «n i ^ ^ i "!
•• •*> 22 2
a s °, ?; s a 2
0000000
td Ed id td id id Id
^f OB >A QO ^^ (^ ^s
—7 >o o r» o» O
— — — — — — K
3333333
r» vo «A ri
SSsSSoSSSSSS
A .••; .*. 1 X it A A "t A A A
SSSS^S^o^oo
tx] IT! ttj (tj ||3 uj lij t»3 [J3 tn £ri
SCN ^h ^^ »•* rfl ff^ |^ ^f »•* »••
.**. — — VB r*. *. p * ««• p
c* pi r» •»
N vei ft — — vq i«; 4r- i
SS332SSS2SSS
SSS3S 3SSSSSS
>n 0> O. — * «n •* «#
— r-oo— "
S 5 a 55 S 2 S? R S
— o o o o o" e o o o o
. 0-»-°?99$?9?
id w (d 11 td w td id (d w Id
oi — oo c*{ — i»: m -• i* ci P«'
ooof-i^t-22 — «3p3
1 1 1 I
ti
1 1 1
I I I I
till
till
o o o o
n vi ea v
sm
Id W (d (d
r» «o >n ci
OO I™* ^0 *£!
" oC r* m
oo oo ee oo
oooo
us lu u ui
!S ee B •?
i"» « in vo
S333
i wwuj
— ri ob oo
o«-£8
86
-------
S » 2 8 8
o o o o o
f*^ ^^ (^
o o o
o
111 Is]
••« 9 vt
S o 8 o o
1 1
~ <
•«} T r«;
is ri o
A vi ri rj
— r*
O
S
v\ pi
5.64E+04
6.73E+03
0 O O O O
•
lu ii] Id ti) |*^
^^ ^h Ot m v^
% «n
ri •— ri
i 1
1 I
1 i
r« — ae
m O O? —^: r*
— — o o
o o d b o
r» t»> »^ »»• r»
0 0 -0 0
- t
O O O OjO O O
•+*•»• +
1*3
P* m •« id
ssss
w
333S
*• + •»• +
II
til Id Id
33
WWW
$$3
«ri 08
r
M
S'
is
S
S!
87
-------
I
CO
•s
§
57E
oo
§
vo r-j t»i
,-__,! o O O:
O" OJ «-£ — ~j O O Of »»i i»J «•>
1 i 1
?t~ 00: OO
22} 2
m uJ ros rii
K w? 3j P
f>| ^ »Tj vpf OO P- S
ml ao r4 — ao r*': tr>
so
s
!
S|
o\ o\i o> Os pv
in
os
r
i
O O O:
I
go
0 V". O
so!
A A JS M- "7 *? *?l T *? «?« T
= ill iiii liil iii
CO CO £Q| i I
88
-------
c
E
CD
O
CO
CD
CO
CO
O)
cd
m
Figure 1
cd
cd
/*§&\
f .-.:•**!..•••:., ?\
'•. ••• ••ly-i,--.-; -A
•r' SSs^fc'-li1
' -i'-'A': WV;-t---;''^
-Sj^:';
• V . •' '-• • •'*..: -.-'. •:•*-•
' ;- 'r'.»•;/; '.. ..*•*--.. ':-:
..'. ',..;.'•• »-•.. .-.•;•:•••
'••^^i^%i^^:>
LL
89
2
Q
-------
Figure 2. Log removal of 4-6 Jim polystyrene microspheres at various
pressure drops and flow rates for (a) bag filter #1, (b) bag filter #2,
and © bag filter#3. The error bars indicate one standard deviation
• associated with the average.
1
2.5J
1.5-
I
0.5 J
(A)
2.5-
1.5-
0.5-]
(B)
4.5
3.5-
2.5-
5
1.5-
0.5
= 47UMin
-O- Q = 95L/Min
-»- Q = 151 L/Min
5 10 15 20 25 30
Pressure drop (psi)
Q = 47UMin
-O Q = 95L/Min
0 5 10 15 20 25 30
Pressure drop (psi)
Q = 47UMin
-O- Q = 95UMIn
0 5 10 15 20 25 30
Pressure drop (psi)
90
,;,:; .jt.'i,. ;jjiii"l .lit
-------
v- ca n
III
m
o top .
*~* ra ro
^
e>
T
n
-i ' ' ' ' i
CM «-
6tT|
•fe »
(5
O T^
T
CM
uojpnpaj 601
o |
91
-------
to
f
-CM
eu
(uinpuodsoidAJo) Boq
-CT Q.
1
-CM g
OJ
«
CM
-co *-?
o
cl
£
-CM in
92
-------
Figure 5. Size distribution of Cryptosporidium oocysts, 4-6 jim polystyrene
microspheres, and average 1-25 um particle counts of the raw
water.
100
10-
1
100
10
o
0.1
4-6 \an polystyrene
microspheres
1-25 pm particle count
8888888888
Sizs range (van)
93
-------
Hll ' ;>':':! •
.Hi
-------
The Use of Aeration for Corrosion Control
Darren A. Lytle, Michael R. Schock
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Cincinnati, OH 45268
Jonathan A. Clement
Black and Veatch Engineering
100 Cambridge Park Dr.
Cambridge, MA. 02140
Catherine. M. Spencer
Wright-Pierce Engineers
Topsham, ME 04086
Abstract
Aeration is a useful drinking water treatment process. Aeration has been used to remove hydrogen sulfide,
methane, radon, iron, manganese, and volatile organic contaminants (VOCs) from drinking water. Aeration is also.
effective in removing carbon dioxide which directly impacts pH and dissolved inorganic carbon (DIG), the most
influential parameters on lead and copper solubility. As a result, aeration can be an effective corrosion control
strategy presuming the initial pH and DIG are appropriate. Mineral precipitation brought about by water quality
changes resulting from aeration may present operational constraints under some conditions. Aeration produces very
consistent water quality, and may be advantageous especially to smaller utilities, because of the relatively low costs,
and simple operational and maintenance needs.
95
-------
Introduction
Aeration is a useful water and wastewater treatment process. Historically, aeration has been used in drinking,
water treatment to remove carbon dioxide, hydrogen sulfide, methane, radon, iron, manganese, and volatile organic
contaminants (VOCs) responsible for foul taste and odor from water. Recent research indicates that aeration is also an
effective corrosion treatment technology. Aeration systems are well established and understood, relatively simple,
low maintenance, and treatment does not require the addition of chemicals. Under the proper set of raw water
conditions, aeration may be particularly attractive to smaller-sized water utilities or those utilities burdened with
staffing and budgetary constraints.
Recent research on cuprosolvency in drinking waters has demonstrated the importance of dissolved inorganic
carbon (DIG) and pH to copper levels hi drinking waters.1-2 The relationships among DIG, pH and lead solubility
has been known for some time.3-5 in many water supplies, corrosion control could be achieved by increasing the pH
and/or reducing DIG of a water. Depending upon the pH, DIG includes some fraction of dissolved carbon dioxide in
the forms of aqueous carbon dioxide (CC>2) molecules or carbonic acid (H2CO3°). Thus aeration serves as a
potential corrosion control treatment alternative through CO2 removal and subsequent pH increase and DIG
reduction. Corrosive waters (low pH, high CC>2 concentration, and low hardness) of several regions of the United
States are particularly suited for aeration treatment.
This paper presents a brief review of aeration literature with emphasis on application to corrosion control.
Through relationships among carbon dioxide concentration, DIG, and pH before and after treatment, the theoretical
impact of aeration on copper and lead corrosion is examined. This is borne out through field data from case studies
of two small systems using aeration for corrosion control. This paper also presents several chemical and operational
constraints on aeration and both negative and positive secondary impacts of aeration.
Historical Corrosion Control Experience
Over the years, numerous studies have tried to examine and discuss the corrosion of metals by the "attack" of
dissolved COi or carbonic acid. Two especially confusing issues were encountered while reviewing the literature.
First, research documentation is ambiguous on whether added corrosivity is mechanistically caused by specific
reactions with molecular aqueous CC>2 or carbonic acid, as opposed to pH and carbonate/bicarbonate complexation
effeets.6 Secondly, the diversity of units used by investigators to express CO2 made comparisons among studies and
interpretations difficult. Units of mg CO2/L, mg H2GO3/L, mg CaCO3/L, and mg C/L have all been used to
express CC-2 concentration. And in the case of some references the units were never made clear by the author. Table
1 presents conversion factors for commonly expressed units.
The impact of dissolved CC-2 on corrosion of distribution system materials and the benefit of its removal from
water by aeration has been recognized and documented for at least 150 years. One reportTdescribed the occurrence of
lead poisoning in a wealthy royal family who had recently moved into a new estate in 1848. The previous owners
had not encountered such poisoning. Upon further investigation, it was discovered that the spring that fed the
residence, which had previously been open to the air, had been covered with an iron dome. The dome prevented the
escape of CC>2 which subsequently increased lead levels in the water by increasing the corrosion of lead service lines
through the higher DIG and lower pH.
In 1913, WhippleS reported on the addition of soda and lime, and the application of aeration to neutralize or
remove carbonic acid from corrosive New England groundwaters. He stated that neutralization of carbonic acid
above 5 mg/L (presumably expressed as CC>2) with lime becomes costly and hardness increases are considerable, and
suggested considering aeration to remove CO2. He did investigations into the effect of capacity, water depth, and
ratio of exposed surface to volume of water. The removal of carbonic acid from water drops falling through the air
was found to depend most on the time of exposure rather than the distance dropped. Field experiences showed that
the minimum practical carbonic acid limit attainable through fountain or air spray aerators in water a water with
96
-------
initial concentration of 25 to 30 mg/L was 5 to 6 mg/L (again, presumably as CO2). Lower levels were achievable
by the addition of hydroxide- or carbonate-based chemicals or the use of tray aerators and trickling beds.
Mullen et al.9 examined the impact of a packed tower aeration (PTA) system used for VOC removal in New
Jersey on water quality and corrosion rates of mild steel coupons. They found that air stripping reduced coupon
corrosion rates (5.4 to 3.5 mils per year [MPY]) despite increasing the dissolved oxygen level from 4.9 to 11.7
mg/L. This apparently resulted from the pH increase caused by the removal of CO2. Initial pH and calculated CC>2
levels of 7.1 and 26 mg CC>2/L, respectively, were adjusted to 8.0 and 4 mg CO2/L following aeration.
Guertin et. al. 10 investigated the use of "venturi-aeration" technology and the benefit of CO2 removal and pH
increase in low pH waters of New England. Engineering and economic issues associated with aeration were briefly
discussed, as well as benefits of aeration for simultaneously controlling radon, VOC's, and corrosion. O'Brien11
presented performance data and further elaborated on the use of the venturi-aeration process in reducing the
coixosivity of potable waters.
Kirmeyer et. al.12 observed decreased copper levels at distribution system monitoring sites in Vancouver, WA
following the completion of packed-tower aeration systems installed at two wells for TCE removal. Carbon dioxide
levels decreased from 7.3 to 1.8 mg CO2/L and 6.2 to 0.9 mg CO2/L, and pH levels increased from 6.9 to 7.9 and
7.4 to 8.1 in the two respective wells. Dissolved oxygen levels remained nearly unchanged following aeration.
Edwards et. al.6 presented a brief theoretical comparison of caustic, lime and aeration for raising the pH of a
water as a copper corrosion control strategy. They showed that for a hypothetical initial water having a calcium
hardness of 100 mg/L CaCO3, alkalinity of 250 mg/L CaCO3, and a pH of 6.5 or 6.7, higher pH values were
achievable by aeration without precipitating calcite, and copper levels were more substantially reduced because of the
lower DIG. This occurs because alkalinity is a conservative property (remains constant) during aeration, and the
higher pH is balanced by reduced DIG.
Several investigators have suggested that aeration processes will increase corrosion rates of distribution system
materials as a result of the addition of dissolved oxygen (DO) to the water. Dissolved oxygen has been known to
directly impact corrosion rates of a variety of distribution system materials for years.5,13-16 Aerating waters
containing no or little initial dissolved oxygen may increase corrosion rates by creating a higher oxidation/reduction
potential in the water, thereby oxidizing the metals in the plumbing materials. For example, copper solubility in
waters with very low dissolved oxygen levels will likely be limited by copper(I) solids.1.2 Copper levels based on
copper (I) solubility are significantly lower relative to copper levels based on copper(II) solubility, which would be
expected in waters with significant levels of dissolved oxygen or that are disinfected.
Nelson and Powell17 examined the role of dissolved oxygen on copper corrosion in two subdivisions by
supplying one with aerated water and the other non-aerated water. They found that on average, water aerated to a
dissolved oxygen level of 6.0 mg/L lead to copper levels in homes of 3.5 times greater than non-aerated water
(dissolved oxygen level of 0.4 mg/L). While examining the secondary impacts of a full-scale packed tower aeration
system used for TCE removal over a 3-month period, Umphres and Wagner18 observed only slightly higher
corrosion rates of mild steel and no increases were observed with the copper. They suggested that the corrosive
nature of increasing dissolved oxygen levels following aeration was offset by calcium carbonate deposition that had
been also experienced. A pH increase of approximately 7.6 to 8.0 and an average calculated CO2 loss of 5 mg/L
resulted from aeration.
The Minnesota Department of Health conducted a copper corrosion control study19 after they had discovered a
large number of medium-sized water utilities in their state failed the copper action level during the initial monitoring
period in 1992 under the Lead and Copper Rule (LCR).20-23 Their investigation showed that iron removal plants
were more likely to have elevated copper levels, fn most cases, iron removal involved aeration, filtration, and
chlorination/fluoridation. Experimental studies and additional utility monitoring suggested that copper levels in
systems with iron removal plants having major points of aeration in their treatment had higher average copper levels
(1.7 mg/L) than non-aerated plants (0.8 mg/L). They concluded that elevated copper levels observed in the aerated
waters were related to elevated dissolved oxygen levels (4.90 mg/L in aerated versus 1.45 mg/L in non-aerated
97
-------
systems). Lead and Copper Rule exceedence data collected by the Illinois Environmental Protection Agency24
showed that there was an action level exceedence, in 39% of the 357 water utilities in the state that applied aeration
for some phase of treatment (typically for iron and manganese). Hie individual exceedences were copper (25%), lead
(10%) and utilities that failed both lead and copper (3%).
The. use of de-aeration is practiced by some industries to reduce the corrosiveness of their water supply.
Industrial plants with cooling towers that are continuously re-aerated are particularly concerned. Powell et. al.25.26
discussed design considerations and the application of de-aeration in different industrial applications.
Chemistry of Aeration
Impact of Aeration on Carbonate Chemistry and pH. The direction and degree aeration impacts corrosion is
dependent on raw water quality (pH, DIG, DO, Ca, etc.) and process COa removal efficiency. Metal corrosion
principles, carbonate water chemistry, and metal solubility are important in understanding the water changes that
may occur as a result of aeration and the impact those changes may have on corrosion and metal release to the water.
The benefits of aeration relative to corrosion control are gained from the reduction of dissolved inorganic carbon
(DIG) concentration and increase in pH resulting from CC>2 removal from corrosive waters. DIG is comprised of
carbonic acid (HjCOj* = COa + t^CC-s), bicarbonate ion (HCCV), and carbonate ion (CDs2-), plus any
carbonate-containing metal ion pairs and complexes (e.g. CaHCO3+, MgCCV). The concentration of dissolved
COj (assuming H2CO3*is equivalent27 in a water can be calculated from measurements of total alkalinity and pH
under equilibrium conditions.27"31 Aqueous COj is only the dominant carbonate species below pH = pKj', pKi'
being the first dissociation constant for carbonic acid adjusted for temperature and ionic, strength. Therefore, the
ability of aeration to significantly remove CC>2 will continuously decrease as the pH raises above this value.
Given adequate time during aeration, the transfer of CC>2 between air and water will continue until the CC>2
concentration between the two is at equilibrium. The concentration of dissolved CO2 in a water at equilibrium with
the atmosphere is given by Henry's Law.27*29*32
Figure 1 represents the solubility of CC>2 for various partial pressures and temperatures and shows that the
solubility is relatively low (rarely above 2.0 mg CO2/L). The driving force for CC>2 removal is the difference
between the initial calculated CC>2 concentration and the equilibrium CO2 concentration. The driving force can in
some instances be quite high (upwards of 80 mg CC>2/L) for many ground waters across the United States.
Providing that CaCOa precipitation does not occur, as the CC>2 concentration hi water changes, the DIG and pH
also change but alkalinity remains constant (a conservative property).27>29,33 An expression for the pH of a water hi
equilibrium with atmospheric CC>2 can be derived through equilibrium equations from the measured alkalinity and
partial pressure of atmospheric CC>2. From approximately pH 7.5 to 9.5, the concentration of bicarbonate ion is
essentially equal to the alkalinity. Therefore, the pH change resulting from aeration can be closely approximated
from:
(1)
where pHi is the initial pH, pHf is the pH of the aerated water, CC>2f is the CO2 concentration of the aerated water,
and CC>2i is the CO2 concentration of the initial water.
Both the initial DIG and the theoretical DIG at CC>2 equilibrium can be calculated by equation (2) using the pH
and corresponding total alkalinity for either initial or theoretical conditions of equilibrium with atmospheric CC-2
from the following equation:32
98
-------
DIG = (1 +
Alkalinity
rr '
(2)
[If] Kt
1 +
[#*]
where [ ] represents concentrations in mol/L and Alkalinity is expressed as eq/L. The difference between the initial
and equilibrium DICs represents the theoretical DIG removed during aeration. Values of pH and DIG following
aeration calculated as above represent a "best" case scenario. Full-scale aeration units are capable of achieving as
high as 80 to 90% removal CO2 from some waters. As a better guide, Figures 2 and 3 show anticipated transfer of
CC>2 and DIG (respectively) following aeration for given initial alkalinities and pH values, assuming 90% removal of
CO2.
Impact of calcium on carbonate chemistry and pH. The concentration of Ca2+ in a water is important to
consider before applying aeration to drinking water applications. If CaCOa, is precipitated from the water, a
reduction in alkalinity and buffering capacity will occur. This will reduce the final pH attainable through aeration.
Additionally, CaCC>3 deposition can present operational and maintenance problems with aeration devices, causing
plugged diffusers, clogging of air jets, and a buildup of deposits on trays and on surfaces. The initial saturation state
of CaCC>3 can be predicted using the Langelier Index (LI)30,31,34 or calcium carbonate precipitation potential.3^34
Pretreatment by ion-exchange softening can reduce these problems.
Computer modeling. Estimates of pH, CO2, and DIG conditions following aeration may be readily obtained by
programming the relevant equations into a commercially-available spreadsheet software package. However, there are
some cases where there could be significant inaccuracy in the prediction of post-aeration conditions using these
simplified equations. For waters in which there is considerable ion-pairing or where calcium carbonate precipitation
could be substantial, the final pH will tend to be underestimated. It would also be important to have some idea of
the magnitude of the scaling problem that might be encountered. More complicated calculations are also needed
where there could be other precipitation or redox reactions that could impact the final pH or other operational aspects
of the system. Mathematical solution of these kinds of problems can become very complex. In these cases, there are
a variety of equilibrium geochemical modeling computer programs that can operate on personal computers and
provide necessary complexation corrections, ionic strength corrections, and estimates of equilibrium speciation and
mass transfer, such as PHREEQE, MINEQL+, and MINTEQ .series35-39; but for some systems kinetic
considerations could also be important.
The impact of aeration on corrosion. The chemical changes that occur during aeration, particularly DIG and pH,
strongly impact lead and copper solubilities and corrosion rates. 1-6,31,40-53 Because DIG serves to control the buffer
intensity in most water systems, sufficient DIG is necessary to provide a stable pH throughout the distribution
system for corrosion control,1-43.45 even if conditions are not purely optimal from a solubility standpoint.
Increasing the pH during aeration of a corrosive water will generally significantly reduce lead and copper corrosion.
In addition, a reduction in metal release could also help wastewater treatment plants meet state or local wastewater
discharge requirements.
Oxidizing agents (i.e., O2, HOC1°, OC1-, C12) can affect corrosion and metal dissolution in two major ways.
Oxidizing agents may affect the nature of passivating films on a pipe by altering the crystalline characteristics and
porosity of corrosion product films. Usually of more importance, higher concentrations of the oxidizing agents may
accelerate the rate of corrosion and metal uptake into the water during short stagnation periods. For example, only
under extremely oxidizing conditions can metallic lead hi solder, brass and pipe be oxidized to Pb(IV). However,
even under low dissolved oxygen or chlorine residual levels, sufficient oxidant should be present to drive the
reaction of metallic lead to Pb(II).42.46 Increases in oxidant levels will increase the driving force of the oxidation
reaction, but does not necessarily affect the equilibrium solubility. This is particularly so when the metal tends to
take on one dominant oxidized valence state (e.g. Pb2+, Zn2+) under normal conditions.
99
-------
The situation with metals that can exist in more than one oxidation state under potable water conditions (e.g.
Cu, Fe) becomes more complicated. Copper can readily exist under the wide variety of redox potential conditions
observed in drinking waters as copper metal, Cu(I) or Cu(II). A model for the reversible transformation of copper
between Cu(I) and Cu(II) has been postulated by Werner and Sontheimer53, which 'is consistent with observed
trends in copper levels versus time reported by a variety of international researchers, and analysis of solids on copper
pipe indicating the simultaneous presence of layers including Cu(I) and Cu(II) oxides.1'2.52
The transformation of copper's valence state following the addition or depletion of an oxidant in water can
drastically impact copper equilibrium concentrations. Figure 4 shows solubility curves for Cu(II) controlled by
Cu(OH)2(s) at three levels of DIG superimposed on solubility curves of the two possible Cu(I) solids, CuOH(s) and
CU2O(s). The addition of DO by aeration to a ground water with a low EH could drastically impact copper levels,
because the difference in solubility between cuprous hydroxide or oxide solids and corresponding cupric solids
exceeds a factor of 100 to 1000 over the common pH range of 6 to 8.1.2.52
Figure 5 also broadly illustrates approximate changes in copper solubility that could result from aeration. The
EH (or pE) of ground waters depends on many factors, such as the oxygen content of recharge water, the residence
time of the water, the distribution and reactivity of organic matter and other potential reductants, and the presence
and quantity of potential redox buffers in the aquifer.33 Consider, however, a scenario for a ground water devoid of
oxygen and without sulfate reduction that is likely to be in equilibrium with a variety of iron minerals. A plausible
Ejf-pH zone is superimposed on the En-pH stability diagram for high DIG situation (96 mg C/L). Much of the
zone lies where either CU2O is stable, or where no copper oxidation would occur. If aeration is initiated, the water
would move toward much higher EH, and likely higher pH. Thus the water could readily pass into the stability
fields of dissolved species (e.g., CuCOa0, Cu(OH)2°) or more soluble solids [CuO, Cu(OH)2].
When to consider aeration as a corrosion control solution. The previous sections introduced the fundamentals of
corrosion, metal solubility, and water chemistry changes associated with aeration. These are intended to be useful
as an initial design guide in deciding the applicability of aeration for corrosion control. Lead and copper solubility
diagrams should be consulted to estimate lead and copper solubility changes corresponding to estimated DIG and
pH changes, keeping in mind that these, too, are not absolute.i,2>5,6,3l,40,4i,43,45-47,49,52
The viability of aeration as a corrosion control technique must be evaluated case by case considering all of the
water quality conditions. A suitable water must have a significant CO2 concentration and a corresponding relatively
low pH. As the pH approaches 7, the CO2 fraction of DIG available for stripping rapidly drops and benefits are
greatly reduced. As a guide, several references8.3!.54 have suggested that waters with CO2 levels ranging between
approximately 4 to 10 mg CC-2/L are not suited for aeration. At such low CO2 levels, the driving force for CO2
removal, pH increase, DIG decrease, and economic benefits resulting from aeration are minimal.
Limestone contactors may be suitable for some water qualities amenable to aeration, and they provide another
relatively simple treatment option. Considerable guidance exists for identifying compatible water chemistries and
designing the systems.55-59 The dissolution of limestone in a limestone contactor to increase the pH and alkalinity
of high DIG groundwaters is hindered by interference with excess carbonic acid from the dissolved carbon dioxide.
flic excess carbon dioxide must be removed before limestone can dissolve. Effective operation of contactors is also
dependent upon the pH and hardness of the waters, and materials that can foul the limestone surface.
Table 2 is a summary of data gathered from aeration plants where the primary goal was not necessarily CO2
removal, with the exception of the Vancouver, B.C. sites. Carbon dioxide levels are given for several plants and can
be considered within the same general range for the other plants given the similarities in alkalinity and pH values.
Most of the waters before treatment fall within the 4 to 10 mg CO2/L suggested guidance range and have a pH
bqjwcen 7.2 and 7.7, T,he data shows that some CO2 removal is typically observed, along with small increases in
pH and decreases in DIG, but not always enough to be efficient for corrosion control. If DIG removal is critical in
reducing metal solubility, the addition of an acid prior to aeration would lower the pH and increase the available
CC-2 for removal, resulting in greater DIG reduction.
100
-------
Source waters having low DIG levels should not be considered for aeration treatment. A DIG guideline of 2 mg
C/L has been suggested5 as a minimum DIG to provide adequate pH buffering and to form protective lead carbonate.
Subsequent discussion among corrosion researchers has been leaning towards pushing that limit upwards to as much
as 5 to 10 mg C/L primarily to assure adequate buffering ability under a variety of distribution system conditions.
In general a low DIG water (2 to 5 mg C/L) will have a low CC-2 concentration which would likely result in failure
to meet the CC-2 source water guideline for aeration of 4 to 10 mgs>31=54. If the predicted final DIG of an aerated water
falls below 5 mg C/L, alternative corrosion control techniques such as chemical pH and DIG adjustment chemical,
or possibly corrosion inhibitors should be considered more suitable. In some waters, a limestone contactor could be
added post-aeration to supplement pH and DIG.
Evaluation of the chemistry feasibility of aeration should also include the occurrence of other contaminants that
would either support or complicate the use of aeration. These include calcium carbonate precipitation, presence of
iron at levels above 0.2 mg/L, presence of manganese at levels above 0.05 mg/L, the presence of VOC's, and radon
levels. These other water quality parameters can limit the use of aeration treatment by precipitating calcium
carbonate when the treated water pH exceeds the saturation pH value (calcium carbonate), can be oxidized in the
presence of the oxygen in the air (iron, manganese) or can be beneficially removed by aeration (radon, VOC's). A
general summarization of chemical considerations in selecting aeration is given by Figure 6.
Some secondary reactions may complicate predicting or achieving the desired pH through the aeration process,
although the occurrence of such problems is not generally documented. For example, theoretically, in waters with
high concentrations of ferrous iron, initial pH increases as the CC>2 is stripped from the water could be followed by a
pH drop induced by the slower oxidation of ferrous to ferric iron60. The increase in iron oxidation rate with
increasing pH is well-known61, but it is typically three to ten times slower than the pH response to the CC>2
removal60. Oxidation of hydrogen sulfide to sulfate would be another potential secondary reaction that could lower
the pH. In any case, the significance of the impact of the secondary reaction would be a strong function of the
concentration of the secondary reaction species and reaction products relative to the buffering intensity exerted by the
carbonate system at that particular combination of pH and DIG.
Adverse side effects. Problems associated with calcium carbonate precipitation include precipitation on packing
materials in aeration towers, clogging of aeration diffiisers, and problems with calcite deposition on pump impellers
and distribution piping. It may be possible to design some types of aeration units to operate under the saturation
pH. LaMotta62, for example, addresses engineering and chemical considerations in CO2 removal in tray aerators.
He presents aeration design calculations that enable a unit to produce a water at any state of finished water pH,
calcite saturation, and carbon dioxide removal to equilibrium values. The equilibrium pH, DIG, and calcium
carbonate saturation values should be determined before modifications to the design of the aeration systems are
evaluated. If calcite precipitation is an issue with aeration, it will likely also be an issue with pH adjustment and
may place an upper pH limit on any type of treatment.
A water with little or no dissolved oxygen and no disinfectant will probably not be corrosive, particularly to
copper, unless there is microbiologically induced corrosion or hydrogen sulfide attack. In such a water, copper
solubility would likely be controlled by Cu(I) solids, leading to lower copper levels than a water having a higher
dissolved oxygen level and redox potential, where Cu(II) solubility dominates. The more practical issue is a
system that has not been aerated and aeration is installed for an alternative reason such as iron removal. One result,
they may be a significant increase in metal release from plumbing materials in the system.
Opening a water system to the atmosphere increases the risk of microbiological contamination. Umphres and
Wagner18 evaluated the microbiological quality of non-disinfected water treated by packed tower aeration. They
found that standard plate counts (using either R2A or plate count agar media) were approximately 2 logs greater in
aeration effluent waters. Effluent densities were typically 103/mL following aeration. No coliforms, however, were
detected in any of the samples. When prechlorination was applied to produce effluent residuals greater than 0.6
mg/L, little difference in microbiological activity was noted. Adding an aeration unit to a previously closed system
will lead to a pressure break. Appropriate engineering measures (such as re-pumping) may need to be taken to
repressurize the system. The oxidation of iron and manganese may lead to red and black water complaints in the
distribution system. Filtration may be required in such cases.
101
-------
Beneficial side-effects. Aeration is often used as the means of oxidizing iron from a raw water for removal. The,
levels of aeration required to oxidize iron are quite low so aeration treatment used for corrosion control will also
oxidize iron present in the raw water to form visible iron hydroxide if the raw water iron levels exceed 0.2 mg/L.31
The iron can stain aeration media and vessels and cause customer complaints in the distribution system unless the
excess iron is removed by filtration. Tray aerators have been successfully used for aeration of iron containing raw
waters.
Aeration can also oxidize manganese present in raw water. The reaction rate of oxygen with manganese is very
slow at pH values less than 8.5 so it is possible that manganese precipitation would not be observed until the water
is in the distribution system. If manganese at levels above 0.05 mg/L is observed in the raw water, manganese
removal would be necessary to prevent manganese staining and customer complaints.
The presence of radon in raw water at levels above 3,000 pCi/L provides an added incentive for a utility to
consider aeration treatment for radon removal as well as corrosion control. Removal of excess carbon dioxide and
radon occur simultaneously during aeration. The Henrys constants of radon and carbon dioxide are similar, thus
aeration treatment designed for carbon dioxide removal will give a predictable reduction in radon concentration.
Aeration Technologies
There are a variety of standard and commercially available technologies for stripping gases from water. Aeration
technologies are typically divided into two major categories; waterfall aerators and injection (also referred to as
"bubble" or "diffusion") aerators. As implied, waterfall aerators involve dropping through the air, either directly as
with tray systems, or after air injection during flow in venturi-based systems (including Mazzei® systems).
Injection aerators bubble compressed air through the water. The goal of choosing an aeration technology is to
achieve maximum gas transfer at a minimum cost. The rate at which the volatile compound is removed from water
depends on a number of variables including air-to-water volume ratio, contact time, temperature of both the air and
water, and the physical and chemical properties of the contaminant. The selection of an aeration technology is
dependent upon the gas to be removed, desired gas removal efficiency, ease of operation, spacial limitations, noise
control, cost, and the esthetic appearance of the equipment for treatment plants located in residential areas. Kinner et
al,63 evaluated aeration alternatives for small systems removing radon from groundwater. They identified four
special qualifications of designing a aeration unit for a small water system: (1) easily retrofit into existing facilities,
(2) require little initial costs, (3) require little maintenance, and (4) have low relative operating costs to conventional
technologies.
There are a number of specific aeration technologies available including packed-tower aeration, spray tower
aeration, diffused aeration, multiple tray aeration, and surface aerators and aeration pumps. A short description of
some of the major aeration technologies follows, and some advantages and disadvantages of different aeration
approaches are broadly summarized in Table 3.
Waterfall Type
Tray aerators and packed towers/columns. These are common waterfall aeration types. They can be constructed
to give removals dissolved carbon dioxide approaching equilibrium by increasing the number of trays or the height
of the packing in the tower or by blowing air countercurrent to the water flow. These units can be simple to
construct and can be placed outside in warm climates. Packed tower systems have been developed by a number of
manufacturers. Depending on the packing or media used in the tray system, they resist iron and calcium carbonate
fouling. Typical heights are 15 to 30 feet and typical ainwater ratios are 20:1 to 200:1. Water flow rates can range
from 10 gpm to 1,000 gpm. Benefits are low power consumption, particularly if no blowers are used, and high
transfer efficiencies. Drawbacks are the size of the units and lack of accessibility for maintenance of the packed
towers.
Eductor systems. Venturi aeration systems can educt air into water and permit some transfer of gases as long as
provision is made to release the dissolved gases soon after eduction. Transfer efficiencies are low under ambient
102
-------
conditions with air.water ratios ranging from 1:1 to 4:1. These systems may be categorized as "waterfall" types,
because the water has to free-fall into a basin or chamber to allow the gases to release. Mazzei® injection systems
are a kind of hybrid, operating similarly to venturi systems, but with air infused as fine bubbles. Air to water ratios
are similar to those noted for venturi-based systems.
Spray aerators. Spray aerators work by forcing water through nozzles to create fine droplets which fall through the
air to provide the ainwater contact. Air contact time is generally quite short, less than 10 seconds. Spray aeration
may be used in open basins in warm climates. Typical discharge heads are 20 feet. Nozzle design is critical to the
production of fine and uniform droplets without clogging. Drawbacks are similar to packed towers in that a large
aeration area is required and the nozzles are difficult to access for maintenance,
Diffuser Type
Fine bubble aeration systems have been developed as packages over the past few years to permit air stripping of
carbon dioxide and other volatiles from water. Fine bubble aerators are best suited to small systems where ainwater
ratios of 10:1 to 100:1 can be developed economically. Most fine bubble aeration systems have a maximum water
flow near 200 gallons per minute. If a high ainwater ratio is required, the permissible water flow rate through a unit
must be reduced. Fine bubble aeration systems are compact, lightweight, and accessible for maintenance. They are
not suitable to treat water with iron or manganese above the secondary standards and may lose efficiency in hard
water as calcium carbonate precipitates on the diffusers.
Coarse bubble diffuser systems have also been developed as package treatment systems over the past few years.
The coarse bubble diffusers are usually less efficient than fine bubble aeration systems but are less susceptible to iron
or hardness fouling.
A hybrid aeration system has been developed that has characteristics of both tray aeration and diffused bubble
aeration. It consists of flat trays stacked above a collection basin. Large volumes of air are blown countercurrent to
the water flow to suspend the water in trays. Ainwater ratios from 25:1 up to 1,000:1 are possible. The hybrid is
compact, accessible for maintenance, can treat water with iron or hardness and is highly efficient. It is also one of
the more expensive aeration systems to purchase and operate.
Case Studies
Two recent studies illustrate successful application of aeration systems to remedy corrosion control and other
water quality problems, and illustrate several engineering considerations in choosing and implementing the
appropriate technology.
Case Study 1: School System. A water system serving a school with a population of 200 investigated the use of
aeration to raise the pH of its water to control levels of lead and copper. The water system conducted monitoring of
lead and copper in accordance with the requirements of the 2121, and the 90th percentile levels for both lead copper
were above the action levels. The water system decided to participate in a research study which is part of a
cooperative agreement between the New England Water Works Association and the U.S. EPA to evaluate the ability
of aeration to reduce lead and copper in under full-scale conditions. The water system's source is a bedrock well
with a water quality which well suited for aeration (Table 4). Given the low pH (6.1 to 6.3), moderate DIG (14 to
18 mg C/L), and carbon dioxide concentration (30 to 40 mg CC>2/L), aeration was deemed appropriate.
The aeration device used for the evaluation was a diffused bubble SysternTf. The system was selected since it
is very compact (0.60 in wide, 1 m high and 1 m long) and space to house treatment equipment was limited. It
was also selected for its ability to remove radon, as the raw water radon concentration for the system ranges from
9,000 to 11,000 pCi/L. The device contains three chambers in which the flow passes through in series. At the base
tLowry Engineering, Unity, ME.
103
-------
of each chamber is a diffuser, which bubbles air through the depth of the liquid. The unit is capable of treating up to
25 gpm (1.6 Us) and has an air to water ratio of approximately 100:1.
Over the course of the study the aeration unit removed averaged of 96% of the carbon dioxide present in the raw
water. The DIG was reduced from a range of 15 to 18 mg C/L to 6 to 7 mg C/L. The pH was increased from an
average of 6.24 ± 0. 11 (confidence interval represents one standard deviation) to 7.47 + 0.09. As shown in Figure
7, the finished water pH was very consistent over tune, which appears to be a major advantage associated with
aeration based pH adjustment.
Copper and lead levels were reduced significantly as expected with the increase in pH and reduction in DIG
(Figures 7 and 8). This is further illustrated in Figures 9 and 10 for copper and lead, respectively, which show
conceptually how the solubilities drastically shifted downwards by decreasing DIG and increasing pH of the actual
aeration system. The diagrams can only show approximate metal concentration changes, because they do not
directly correspond to the temperature, ionic strength, and standing times of the samples. Further, the exact
quantitative correspondence is especially unlikely for soldered joints relative to pure lead systems, and copper pipes
under non-equilibrium conditions present in the pipe loop system. Clearly, however, the relative changes in copper
and lead levels are characterized well by the theory and diagrams.
Seasonal fluctuations in metal levels were observed, especially prior to aeration, caused by the decrease in water
usage during summer months when school was not in session. Though the metal level trends were downward at
the time of initiation of treatment, comparison of posttreatment points with those corresponding to the same season
before treatment show significant benefits to aeration. Additionally, both lead and copper levels remained very
stable after treatment. The system has achieved full compliance with the requirements of the LCR. In addition.
radon was reduced from an average of 9,867±329 to < 174 pCi/L.
Case Study 2: Small Municipal Water System. The Harrison Water District is a small public water system in
Maine. In accordance with the first six-month monitoring requirement, the District collected 20 first draw water
samples for lead and copper analyses from high risk houses. The 90th percentile lead level was 38 (ig/L and the
90th percentile copper level was 5.91 mg/L.
iii ' ' . i! ' M I'*1! ; i ' ' ,i ,
The data available for the gravel pack well source reveal an average pH of 6.25 to 6.28, alkalinity of 18 to 26
mg/L and calcium hardness values ranging from 19 to 23 mg/L. Dissolved inorganic carbon values are estimated at
II to IS mg C/L indicating a relatively high level of dissolved carbon dioxide. Initial radon levels ranged from
3,500 to 5,400 pCi/L.
Aeration treatment was designed to remove 95% of the influent radon. For the combined 350 gallon per minute
(22 L/s) flow from the two wells, this required two six-stage diffused bubble aeration units. Each unit was designed
to treat a 175 gpm (11 L/s) flow at an air: water ratio of 11.9:1 using a 5 horsepower blower.
It was estimated that at 95% removal of radon, reduction of carbon dioxide would result in treated water DIG
values of 6.5 to 8 mg C/L and the resultant water pH would be 7.1. In practice, the resultant treated water pH value
has been approximately 7.3 and the resultant DIG values are 7 to 8 mg C/L.
Check samples collected for lead and copper at five of the former worst case residences after only three months of
aeration revealed that copper levels had been reduced to an approximate 90th percentile level of 0.39 mg/L (0.30
mg/L average) and the lead levels had been reduced to an approximate 90th percentile value of 22 |j,g/L (12 p,g/L
average). Treated water radon levels were less than 200 pCi/L.
'I;1 i ' . '
l|r
Conclusions
'iii • ' . •" ' , ;
Aeration is a suitable corrosion control alternative for lead and copper under the proper set of initial water
quality conditions. Benefits are gained by the increase in pH and reduction in DIG resulting from the removal of
CC-2 during aeration. In some instances, secondary benefits can be gained from aeration such as VOC, radon, or iron
removal. A suitable water has a relatively low pH and hardness, and high DIG (CC>2). Aeration technology is
104
-------
well-established, simple, and relatively inexpensive. It is particularly practical and attractive to small utilities
because of ease of operation and minimal maintenance requirements. There are a number of concerns and limitations
to aeration, however, including low DIG waters, calcium carbonate deposition, and potential for increased
microbiological activity. The guidelines and calculations presented are not exact and are to be an engineering
screening tool. Bench or pilot-scale testing is still advisable under some circumstances. Tables 5 and 6 summarize
the information presented in this document and can be used as guides to assessing the feasibility of aeration for
corrosion control.
Acknowledgments and Disclaimer
The authors are grateful for the cooperation of the Harrison Water District Board of Trustees and Superintendent
Bill Winslow for sharing data and enabling the use of their system for an example case study. The authors would
also like to thank David Roth and Kelly Mayhew of the U.S. EPA for their assistance hi preparing this document,
and two anonymous reviewers for very constructive comments on the content and organization of the paper. ,
Mention of trade names or commercial products is for explanatory purposes only, and does not constitute
endorsement or recommendation for use by the U.S. EPA.
References
1. Schock, M. R., Lytle, D. A. & Clement, J. A. "Effect of pH, DIG, Orthophosphate and Sulfate on Drinking
Water Cuprosolvency", U. S. Environmental Protection Agency, Office of Research and Development,
Cincinnati, OH, EPA/600/R-95/085, (1995).
2. Schock, M. R., Lytle, D. A. & Clement, J. A. Effects of pH, Carbonate, Orthophosphate and Redox Potential
on Cuprosolvency. NACE Corrosion/95, Orlando, FL, (1995).
3. Schock, M. R. "Response of Lead Solubility to Dissolved Carbonate in Drinking Water". Journal of the
American Water Works Association, 72:12:695 (1980).
4. Schock, M. R. "Response of Lead Solubility to Dissolved Carbonate in Drinking Water". Journal of the
American Water Works Association, 73 - 3:36 (1981).
5. AWWARF. Lead Control Strategies. AWWA Research Foundation and AWWA, Denver, CO, (1990).
6. Edwards, M., Meyer, T. E. & Schock, M. R. "Alkalinity, pH and Copper Corrosion By-Product Release".
Journal of the American Water Works Association, 88:3:81 (1996).
7. Weston, R. S. "Lead Poisoning by Water, and Its Prevention". Journal of the New England Water Works
Association, 34:4:239 (1920).
8. Whipple, G. C. "Decarbonation as a Means of Removing the Corrosive Properties of Public Water Supplies".
Journal of the New England Water Works Association, 27:193 (1913).
9. Mullen, E. D., Johansen, G. M. & Abdelrahman, A. A. Effects of Packed Tower Air Stripping on Corrosion
Rates. Elizabethtown Water Company, (1986).
10. Guertin, P. D., O'Brien, J. & Martin, J. A. Raising pH by the Venturi-Aeration Process. The NEWWA
Symposium on Solutions to Controlling Lead and Copper Corrosion, Nashua, New Hampshire, (1993).
11. O'Brien, J. E. "Reducing Corrosivity and Radon by the Venturi-Aeration Process". Journal of the New
England Water Works Association, 109:2:105 (1995).
12. Kirmeyer, G. J., Kreft, P. & Hoffman, R. Practical Full Scale Demonstrations to Address Copper Corrosion
105
-------
Including Aeration to Remove Co2. Proc. AWWA Water Quality Technology Conference, San Francisco, CA,
(1994).
13. J. M. Montgomery Consulting Engineers. Water Treatment Principles and Design. John Wiley and Sons,
New York, (1985).
14. AWWARF. Internal Corrosion of Water Distribution Systems. AWWARP-DVGW Forschungsstelle ,
(1985).
15, Larson, T. E. "Bacteria, Corrosion, and Red Water". Journal of the American Water Works Association,
31:1186(1939).
16, Larson, T. E. "Deterioration of Water Quality in Distribution Systems". Journal of the American Water Works
Association, 58:1307 (1966).
17, Nelson, J. A. & Powell, R. M. Copper Corrosion Studies in Pinellas County. AWWA Water Quality
Technology Conference, Miami Beach, Florida, (1980).
18. Umphres, M. D. & Van Wagner, J. H. "An Evaluation of the Secondary Effects of Air Stripping", U. S.
Environmental Protection Agency, Risk Reduction Engineering Lab, Cincinnati, OH, 600/S2-89/005, January
(1990).
19. Minnesota Dept. of Health Report: Corrosion Control Treatment Survey for Small-size Community Public
Water Systems. (1994).
20. Maximum Contaminant Level Goals and National Primary Drinking Water Regulations for Lead and Copper.
Final Rule, U. S. Environmental Protection Agency, Fed. Reg., 56:110:26460 (June 7, 1991).
21, Drinking Water Regulations; Maximum Contaminant Level Goals and National Primary Drinking Water
Regulations for Lead and Copper. Final Rule; correction, U. S. Environmental Protection Agency, 40 CFR
Parts 141 and 142, Fed. Reg., 56:13 5:32112 (July 15, 1991).
22. Drinking Water Regulations: Maximum Contaminant Level Goals and National Primary Drinking Water
Regulations for Lead and Copper. Final Rule; correcting amendments, U. S. Environmental Protection
Agency, 40 CFR Parts 141 and 142, Fed. Reg., 57:125:28785 (June 29, 1992).
23. Drinking Water; Maximum Contaminant Level Goals and National Primary Drinking Water Regulations for
; Lead and Copper. Final Rule; technical corrections, U. S. Environmental Protection Agency, 40 CFR Parts
141 and 142, Fed. Reg., 59:125:33860 (June 30, 1994).
24, Byus, L. A. Illinois Environmental Protection Agency, Personal Communication. (1995).
25. Powell, S. T., Bacon, H. E. & Lill, J. R. "Recent Developments in Corrosion Control". Journal of the
American Water Works Association, 38:7:808 (1946)
26. Powell, S. T. & Bacon, H. E. "Corrosion Control by Deaeration". Water Works and Sewerage, 84:4:109
(1937).
27. Stumm, W. & Morgan, J. J. Aquatic Chemistry. John Wiley & Sons, Inc., New York, New York, (Second
Edition ed., 1981).
28. Kemp, P. H. "Chemistry of Natural Waters-I. Fundamental Relationships". Water Research, 5:297 (1971).
' il|! • ' '" ' ',,',''«'" ' ' ' ' , • ™
29. Snoeyink, V. L. & Jenkins, D. Water Chemistry. John Wiley and Sons, New York, (1980).
30. Schock, M. R. "Temperature and Ionic Strength Corrections to the Langelier Index (Revisited)". Journal of the
106
ilia
-------
American Water Works Association, 76:8:72 (1984).
31. AWWA. Water Quality and Treatment. McGraw Hill, New York, (3rd ed., 1990).
32. Loewenthal, R. E. & Marais, G. v. R. Carbonate Chemistry of Aquatic Systems: Theory and Application.
Ann Arbor Science Publishers, Ann Arbor, Michigan, (1976).
33. Drever, J. /. The Geochemistry of Natural Waters. Prentice-Hall, Inc., Englewood Cliffs, NJ, (1982).
34. Rossum, J. R. & Merrill, D. T. "Evaluation of the Calcium Carbonate Saturation Indexes". Journal of the
American Water Works Association, 75:2:95 (1983).
35. APHA-AWWA-WEF. Standard Methods for the Examination of'Water and Wastewater. American Public
Health Association, Washington, DC, (18th Edition ed., 1992).
36. Allison, J. D., Brown, D. S. & Novo-Gradac, K. J. "MINTEQA2/PRODEFA2, A Geochemical Assessment
Model for Environmental Systems: Version 3.0 User's Manual", U. S. Environmental Protection Agency,
Environmental Research Laboratory, Office of Research and Development, Athens, GA, EPA/600/3-91/021,
March (1991).
37. APHA-AWWA-WEF. Standard Methods for the Examination of'Water and Wastewater, American Public
Health Association, Washington, DC, (19th Edition ed., 1995).
38. Schecher, W. D. & McAvoy, D. C. MINEQL+: A Chemical Equilibrium Program for Personal Computers,
User's Manual Version 3. 0. Environmental Research Software, Hallowell, ME, (1994).
39. Parkhurst, D. L., Thorstenson, D. C. & Plummer, L. N. PHREEQE - A Computer Program for Geochemical
Calculations. Water-Resources Investigations 80-96, U. S. Geological Survey, (1980).
40. Gregory, R. & Jackson, P. J. Central Water Treatment to Reduce Lead Solubility. Proc. AWWA Annual
Conference, Dallas, TX, (1984).
41. Schock, M. R. & Gardels, M. C. "Plumbosolvency Reduction by High pH and Low Carbonate—Solubility
Relationships". Journal of the American Water Works Association, 75:2:87(1983).
42. Schock, M. R. & Wagner, 1. "The Corrosion and Solubility of Lead in Drinking Water". Ch. 4 In: Internal
Corrosion of Water Distribution Systems. AWWA Research Foundation/DVGW Forschungsstelle, Denver,
CO, (1985).
43. Schock, M. R. -"Understanding Corrosion Control Strategies for Lead". Journal of the American Water Works
Association, 81:7:88 (1989).
44. Schock, M. R. "Causes of Temporal Variability of Lead in Domestic Plumbing Systems". Environmental
Monitoring and Assessment, 15:59 (1990).
45. Schock, M. R. "Internal Corrosion and Deposition Control". Ch. 17 In: Water Quality and Treatment: A
Handbook of Community Water Supplies. McGraw-Hill, Inc., New York, (Fourth ed., 1990).
46. Schock, M. R., Wagner, 1. & Oliphant, R. "The Corrosion and Solubility of Lead in Drinking Water". Ch. 4
In: Internal Corrosion of Water Distribution Systems. AWWA Research Foundation/DVGW
Forschungsstelle, Denver, CO, (Second ed., 1996).
47. Sheiham, 1. & Jackson, P. J. "Scientific Basis for Control of Lead in Drinking Water by Water Treatment".
Journal of the Institute of Water Engineers and Scientists, 35:6:491 (1981),
48. Edwards, M., Meyer, T. & J.P, R. Effect of Various Anions on Copper Corrosion Rates. Proc. AWWA Annual
107
-------
Conference, San Antonio, TX, (1993).
49. Ferguson, J. F., von Franque', 0. & Schock, M. R. "Corrosion of Copper in Potable Water Systems". Ch. 5
In: Internal Corrosion of Water Distribution Systems. AWWA Research Foundation/DVGW
Forschungsstelle, Denver, CO, (Second ed., 1996).
50. Rehring, J. P. & Edwards, M. The Effects of NOM and Coagulation on Copper Corrosion. Proc. ASCE
National Conference on Environmental Engineering, Boulder, CO, (1994).
51. Rehring, J. P. The Effects of Inorganic Anions, Natural Organic Matter, and Water Treatment Processes on
Copper Corrosion. Master of Science, University of Colorado at Boulder, (1994).
52. Schock, M. R., Lytle, D. A. & Clement, J. A. Modeling Issues of Copper Solubility in Drinking Water.
Proc. ASCE National Conference on Environmental Engineering, Boulder, CO, (1994).
53. Werner, W., GroB, H.-J. & Sontheimer, H. "Corrosion of Copper Pipes in Drinking Water Installations".
Translation of- g\vf-Wasser/Abwasser, 135:2:1 (1994).
54. AWWA. Water Treatment Plant Design. AWWA, New York, (1969).
55. Benjamin, L., Green, R. W. & Smith, A. "Pilot Testing a Limestone Contactor in British Colombia".
Journal of the American Water Works Association, 84:5:70 (1992).
56. Letterman, R. D., et at. "Limestone Bed Contactors for Control of Corrosion at Small Water Utilities", U. S.
Environmental Protection Agency, Water Engineering Research Laboratory, Cincinnati, OH,
EPA/600/S2-86/099, February (1987).
57. Letterman, R. D., Haddad, M. & Driscoll, C. T., Jr. "Limestone Contactors: Steady State Design
Relationships". Journal of the Environmental Engineering Division, ASCE, 117:3:339(1991).
58. Letterman, R. D. & Kaftan, S. "A Computer Program for the Design of Limestone Contactors". Journal of
the New England Water Works Association, 110:1:42 (1996).
59. Letterman, R. D. "Calcium Carbonate Dissolution Rate in Limestone Contactors", U. S. Environmental
Protection Agency, Risk Reduction Engineering Laboratory, Cincinnati, OH, EPA/600/SR-95/068, May
(1995).
60. Spencer, C. M. & Brown, W. E. pH Monitoring to Determine Aeration Effectiveness for Carbon Dioxide and
Radon Removal. Proc. AWWA Water Quality Technology Conference, Denver, CO, (1997).
61. Stumm, W. & Lee, G. F. "Oxygenation of Ferrous Iron". Industrial and Engineering Chemistry, 53:2:143
(1961).
62. LaMotta, E. J. "Chemical Analysis of Co2 Removal hi Tray Aerators". Water Resources Bulletin, 31:2:207
(1995).
63. Kinner, N. E., etal. "Radon Removal Techniques for Small Community Public Water Supplies", U. S.
Environmental Protection Agency, Office of Research and Development, Cincinnati, Ohio, EPA/600/2-90/036,
Feb. (1990).
108
-------
Table 1. Conversion factors for commonly expressed units.
To convert from units of:
to units of mg CC^/L multiply by:
CO2
C
H2C03
CaCO3
1.0
0.27
1.41
2.27
109
-------
1 ! '. J ", Til . '-;:, "III!1 I Nil i 'IH I' I '.;',. V.i •.'(••' .': "'•"!' ' i •'•;'( . '• ' ™ •'" " . . : :
. „ 'III '' ,,: '
Tabk 2. Water quality of aerated water reported for some systems.
pH
Location Reason units
Lake Forest; Fla.1
Before
After
Ortega Terrace, Fla,1
Before
After
Fonts VedraBeach Fla.1
north system
Before
After
Porvte VedraBeach Fla.1
south system
Before
After
Arlington Manor1
Before,
After
Venetia Terrace1
Before
After
LakeLudna1
Before
After
Magnolia Gardens1
Before
After
Lake Shore Terrace1
'
Before
After
Saratoga Point1
Before
After
Roosevelt Gardens1
Before
After
Milmur Manor1
Before
After
Glyrdea Pirlc1
Before
After
CedarSfaores1
Before
After
AppJe Valley1
Before
After
Eagan*
Before
After
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
hydrogen sulfide
7.5
7.9
7.7
7.9
7.3
7.3
73
7.5
7.4
8.1
7.8
7.9
7.5
7.9
7.5
7.9
7.7
8.1
7.5
7.5
7.6
7.7
7.5
7.6
7.3
7.4
7.6
7.8
6.8
7.8
7.3
7.5
".".!-. ill" ". 'iii: v
Alkalinity
mgCaCO,
140
140
114
114
126
126
142
142
140
140
110
110
144
144
138
138
118
118
138
138
116
116
134
136
132
. 132
124
124
CO2 Dissolved DIG
mg/L Oxygen, mg/ mg/L
7
2.8
4.2
2.7
13.5
8.5
15.2
9.7
8.6
1.8
3.4
2.7
9.8
3.7
7
Z8
4-3
1.8
9.5
9.5
6.5
5.3
9
8.7
14
11
7
4.2
0
4.3
0
4.5
0
3.1
0
Z8
0
5.1
0
5.4
0
4.8
0
23
0
4.9
0
5.7
0
4.5
0
7.2
0
5.7
0
0.4
0.3
9.1
0.6
8.2
36
34
29
28
34
32
38
36
37
34
27
27
37
35
35
34
30
29
35
35
29
29
34
34
35
34
31
31
110
-------
Faribault2
Litchfield2
Maple Grove2
Minne tonka2
Plymouth2
Savage2
Waconia2
Before
After
Before
' After
Before
After
Before
After
Before
After
Before
After
Before
After
Wayzata2
Before
After
Water Station #1, Vancouver, B.C3
Before
After
Water Station #4, Vancouver, B.C.
Before
After
Wells, S.W. "Hydrogen Sulfide Problems of Small Water Systems". —iJAWWA", 2:46:160-170 (1954).
2 Slat Tower Aerators", Proceedings, llth AWW A Water Quality Technology Conference,
Norfolk, VA. (Dec. 1983).
Kirmeyer, G.J., et aL Practical Full Scale Demonstrations to Address Copper Corrosion Including Aeration to
Remove CO2", Proc. AWWA Water Quality Technology Conference,an Francisco, CA., Nov. 6-10 (1994).
7.1
735
7.3
7.3
7.1
7.3
73
735
73
7.4
7.2
7.5
7.3
7.25
7.35
7.35
6.9
7.9
7.4
8.1
0.2
7.5
0.2
3.4
0.3
0.82
0.1
6.35
0.3
0.82
0.3
8.7
0.2
1.4
1.2
2.1
7.3 7.3
1.8 7.4
6.2 9.3
0.9 9.5
111
-------
Table 3. Advantages and Disadvantages of Various Aeration Technologies for Corrosion
Control,
TXEE
ADVANTAGES
DISADVANTAGES
Waterfall Type
Packed Tower Effectively removes a variety of
volatile substances including
ammonia, hydrogen sulfide, and
carbon dioxide
Requires very high air flow rates to
maintain moderate stripping factors; may
require supplemental air source
Secondary impacts may include: air
pollution, mineral scaling, and
microbiological growths
Limited ability to remove semi-volatile or
non-volatile compounds including taste and
odor
Requires large packing depths
Freezing in colder climates
if not enclosed
Cascade Fifty percent higher air to water ratio Requires supplemental air source
Packed Tower than traditional packed tower increasing cost and maintenance
Better removal of non-volatile or Secondary impacts may include: air
semi-volatile compounds including pollution, mineral scaling, and
odor and taste causing compounds microbiological growths
Lower-packing depths than traditional Freezing in colder climates
packed tower aerators if not enclosed
Existing stripping towers can be
modified relatively easily
Higher removal efficiencies than
traditional packed towers
Can accommodate high water
loading rates
Spray
High air to water ratio
Can require large areas
112
-------
Table 3. Advantages and Disadvantages of Various Aeration Technologies for Corrosion
Control.
Has aesthetic value
Duration of exposure of water to air is very
short
Nozzle prone to clogging by iron,
manganese, or hardness
Cascade
Freezing in colder climates if not enclosed
Duration of exposure to air can be Rarely operating in colder climates
adjusted to fit requirements
Obstructions can be used to increase Can require large areas
turbulence
Generally inefficient
Trays Large surface areas for prolonged
water to air contact
Algae and slime growth problems
Requires ventilation or a supplemental air
source
Can operate at high air to water
ratio for more difficult to strip
compounds
Freezing in colder climates if not enclosed
Trays can be lined with coke, stone, Tray materials can sometimes corrode
or ceramic balls as needed
Can be operated over a wide range
of flow rates
Designed for easy access to and rapid Algae and slime growth problems
removal of scales and other fouling
that may occur
Eductpriype
Simple to operate Low ainwater ratios, 1:1 to 4:1
High water flow rates possible
Air Injection Type
Fine Bubble Package units available
Higher cost than most aerators
113
-------
Table 3. Advantages and Disadvantages of Various Aeration Technologies for Corrosion
Control.
Requires less space than spray Requires longer contact time
or cascade
Improved gas exchange compared Requires blowers, increasing operational
to waterfall aerators cost and maintenance
Not subject to freezing
Negligible water head loss
Does not adapt easily to existing aerators
114
-------
Table 4. Mendon School System- Representative Water Quality
Parameter Value
pH 6.0-6.3
Alkalinity, mg CaCO3 25-30
DIG, mg C/L 14_18
CO2, mg/L 30-40
Ca, mg/L 5_10
Pb 90th Percentile, ug/L 29
Cu 90th Percentile, mg/L 2.79
TDS, mg/L 135
115
-------
Table 5. Data gathering and analysis steps related to the feasibility of aeration for corrosion control.
Step 1. Acquire complete water chemistry data to characterize water sources. Many of these may be
parf of normal yearly state requirements. Four groups of analytes may be needed to properly
characterize the source water for a system.:
• Major constituents (e.g. Ca, Na, Mg, Cl, SO.,, total alkalinity, temperature, pH, etc.)
» Possible interferents (e.g. Fe, Mn)
• Carbon dioxide, plus any other volatile constituents of interest e.g. radon, VOC's.
Step 2. Analyze unpresentable or sensitive constituents, such as temperature, pH, carbon dioxide and
alkalinity, on-site.
Step 3. Determine water production needs (volume to be treated, flow rates, maximum peak demands,
well output, etc.) and storage requirements (if any).
Step 4. Compute the theoretical COZ (H2CO3*) concentration from tiie initial pH and alkalinity (or
acidity), eg. Equation 3, Table 6. This is also a cross-check on analyses in Step 1, and should
be equal to or higher than that COZ concentration,
'. ' ...... I!!' <• '. , ' ' , ,..•'.: • ..
Step 5. Determine desired values for aerated water pH, alkalinity, DIG, temperature, and other
parameters based on evaluation of corrosion control and other treatment needs.
Step 6, Calculate the final pH obtainable by 100% efficient aeration with respect to carbon dioxide,
using Equation 1, Table 6. CO2rcan be approximated from Equation 8, Table 6. Calculate the
corresponding DIG from Equation 2, Table 6 using the final pH value. These values can be
compared to needs determined in Step 5. If close, proceed to in-depth analysis of feasibility
and comparison to other treatment alternatives for pH/DIC adjustment.
Step 7. Evaluate costs and feasibility of other alternative means of adjusting pH or DIG (if necessary),
such as lime, caustic (NaOH or KOH), sodium silicate (Type N or Type D), soda ash or
Step 8. Proceed with detailed engineering evaluation of aeration systems, determining needed
air/water ratios from manufacturers11 specifications to meet pH target, size of blowers,
treatment units, packaged systems, double pumping needs, environmental constraints (eg.
temperature restrictions) and other relevant design parameters.
Step 9. Compare full aeration process to other alternatives determined in Step 7, including
consideration of source water pretreatment issues, calcium carbonate scaling, and any other
post-treatment needed, such as the supplemental addition of corrosion inhibitors.
116
,! hi:
-------
1
0
t>
a
O
*n!
0
o
|H
a
|
"ra
1
.3
-*-»
•55
u
!3
0,
1
1
^3
O
V-i
o
calculations i
i_
£
f
§
1
2
*•— *
Ǥ
CO
^)
vd
,2
^
«
H
g
ts
8
<3
oi
.9*
1
o
•i
i
ii
cd
_B
§
a,
g
o
?
&
H
u
i
^
o
09
a
1
5
£,
I
FT
00
1
g
(S g
8 1
-^ i
J1 1
ex
af a-
" §
III^N ^
D, cT
u
1
b.
O
- L_
*
X
* O
E S^
-n U
4. ** <+_i
8"
a,
!s"
^
•A "2>
1 s
e* St* Jb
^^ _^ ^^
~ ^ S
™ CB
Q
|
T 1^
3C C *J
£ to a
M
•f' iS "^
^ .ti
w
II
11
^^
f7
"
S
s
I
""•
r- ^
—
^ ^^T r-
t "
|
«ji-
*
•— «
11
a
D
||l.||
liifl
«s
in ri
§ 2 i
^5 S rt
-&> A >2 g.
fiT ^ J '^"a I
"5S o •£. "^* *5 ^a O
a •* -B ^ "* 3 ""^
* *-* ^ _ O .w
2 -A2* g
0 vi^ § Q.
5 u n T3 "^
O O bD 3
U « 0 0
— 3^2
x IE
•1 = s
2 OT iS
0
o
•*"
A
C4
5/5- |
1-1 7
^s^"*
^ °
I8U
II
o"
y,
o* ,
'i *s M ^ g 1
m > B S M rt
ro
117
-------
K- ;
g.
£2.
r—
a
G?
o
o
§
•a
i
.a
&
f
I
m
£
i
•81
2
o
CS
i
u
vo*
O
S
i
£•9
1
•8
-------
8"
<3 ,«
j-|
-S
H
5"
v§
^
:§
& .
IIP 1
S • ^ §* o1 «
5^u/i -* -* n o m o
mg/L [COJ in
equilibrium with the
atmosphere at
different
temperatures
CO
(S vo ci oo in w
v> ""5J* ^r t*^ t*^ c*i
\o va v^ vo vo »o
A 6 el? & 6 &
J~ o o o o o
jU o ""» o o
10 —i — • es «s tn
Equilibrium
constant for
carbonic acid
dissociation at zero
ionic strength, log
K,
o\
119
-------
3.5
3.0
I"
BD
S 2.0
aT
T3
1 L5
i
.a
1.0
0.5
0.0
i i I i r
I I I I I I I
Partial pressure of CC>2 (atm)
0.0003
0.0004
0.0006
0.0008
Atmospheric equilibrium
i i i i i
i i i i
0
10 15 20
Temperature, "C
25 30
Figure 1. Solubility of aqueous carbon dioxide in equilibrium
with atmospheric carbon dioxide at various temperatures and
partial pressures (1=0.001).
120
-------
• 11
1 ' 11
Starting Alkalinity
(mg CaCO3/L)
5
50
100
200
300
400
Initial pH
Figure 2. Predicted change from initial CO2 concentration to that in equilibrium
with the atmosphere (Pa)2=10"3'52 atm) assuming 90% CO2 stripping efficiency,
1=0.005, at 25°C.
121
-------
! I I ' Starting
Starting Alkalinity
\ \ \ \\
\ \ \ \\\
\ \ \
\ \ \\
Initial pH
Figure 3. Predicted reduction of initial DIG concentration to that in equilibrium
with the atmosphere (PCO2=10~3'52 arm) attributable to loss of CO2, assuming
90% CO2 stripping efficiency, 1=0.005, at 25°C.
122
-------
100.00000
10.00000
1.00000
o
Q 0.10000
0.01000
0.00100
Cu(OH)2(s) at DIG level
4.8 mg C/L
14.4 mg C/L i
96mgC/L ;
8
10
pH
Figure 4. Predicted copper(II) solubility at different DIG levels compared to copper(I)
solubility1, assuming new pipe with Cu(OH)2,1=0.001, and 25° C.
123
-------
1.50
4 5 67 8
pH
9 10 11 12 13 14
Figure 5. EMF-pH diagram for copper in water containing carbonate, assuming
formation of cupric and cuprous hydroxide at DIC=96 mg C/L, 1=0, copper
species concentrations of 1.3 mg/L, and 25 °C. Stippled area is common zone for
iron-rich groundwaters (Drever, 1982, reference [33]).
124
-------
Figure 6. Aeration Decision Tree
Ground water source
pH<7,2
DIG > lOmg C/L
Yes
1
Calcium hardness
undersaturated at
target pH?
Yes
Manganese < 0.05
mg/L
Yes
I
Iron < 0.2 mg/L
Yes
Radon > 3,000 pCi/L
r
— Yes
Aeration
favored
No
Not an incentive
for aeration
125
No
Consider other
treatment
No
Pilot investigation
needed for use of
anti-sealant pre-
softening, or
management of
carbonate sludge.
No
Aeration possible,
oxidize & filter
manganese
— No
Aeration possible,
filter iron after
aeration
-------
Aeration Started
*§.
P-
8.0
7.5 -
7.0 -
6.5 -
6.0 -
5.5
(b)
i i
Figure 7. Copper (a) and pH (b) results for the H.P. Clough School in Mendon.
126
-------
Aeration Started
0
Figure 8. Lead results for the H.P. Clough School in Mendon.
127
-------
100.0000
bfl
e
10.0000
1.0000
0.1000
Before aeration
After a£patioh
V
0.0100
X
/ . - •
= 6.0
= 7.0
= 8.0
pH=10.0
0.0010
0
25 50 75 100
mgC/LDIC
125 150
Figure 9. Schematic example of the approximate change in copper solubility
caused by aeration for the H. P. Clough School. Diagram applies to copper(IT)
solubility assuming equilibrium with Cu(OH)2 solid at 25° C and I=0.02.50
128
-------
100.0000
10.0000
i.oooo
0.1000
0.0100
0 25 50 75 100
nag CfL DIG
125 150
Figure 10. Schematic example of the approximate change in lead solubility
caused by aeration for the H. P. dough School. Diagram assumes equilibrium
withPbC03 or Pb3(OH)2(CO3)a solids at 25"C and 1=0.02."
129
-------
An Investigation of the Impact of Alloy
Composition and pH on the Corrosion of
Brass in Drinking Water
Darren A. Lytle and Michael R. Schock
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Cincinnati, Ohio 45268
Abstract
A better understanding of brass corrosion may provide information and guidance on the use of the safest
materials for the production of plumbing fixtures, and optimization of corrosion control treatments. The effect of
alloy composition and pH on the metal leached from six different brasses and the pure metals that make up brass
(lead, copper, and zinc) in Cincinnati, Ohio tap water was studied. Results demonstrated that the amount of
composite metal leached from the alloys corresponded well with alloy composition. At the pH values tested (7.0
and 8.5), the impact of pH on the lead leached from the alloys increased as the amount of the lead in the alloy
increased.
130
-------
Introduction
Lead service lines, tin:lead solder, and brass fixtures are considered major contributors of lead to drinking water.
The practice of using lead pipe and lead-based solders in home plumbing systems was eliminated in 1986.(1) In
addition, all pipes and fittings were required to be constructed of "lead-free" material which, by legal definition,
contains less than 8% lead.
The metal composition of brass ranges from approximately 60 to 80% copper, 4 to 32% zinc, 2 to 8% lead, <
6% tin, and trace amounts of iron and other alloying elements depending on its application. Therefore, by previous
definition, plumbing products such as faucets and valves constructed of brass and similar alloys are considered
"lead-free" and are the only lead-containing components still permitted for use in drinking water systems. For this
reason, understanding the mechanism(s) of metal release from brass and techniques to control that release are
important in meeting Lead and Copper Rule (LCR)(2-4) requirements and limiting metal exposure to water
consumers. The first objective of this paper is to establish relationship(s) between the composition of six brass
alloys and the metal (Cu, Zn, and Pb) concentrations leached from them in Cincinnati (Ohio) tap water. The impact
of the same water on the metals leached from the pure metal components of brass (Cu, Zn, and Pb) was also
explored. The second objective is to evaluate the impact of pH (7.0 and 8.5) on metal leached from brasses, lead,
copper and zinc.
The experiments were conducted using standard l"x2"xl/8" metal coupons and a "fill-and-dump" test protocol in
chemical variations of Cincinnati tap water. Conclusions were based on graphical and statistical data comparisons.
Developing a better understanding of brass corrosion will provide additional guidance to water utilities, consultants
and governmental agencies on strategies to reduce lead release into drinking water.
Background
A growing public awareness of lead toxicity has highlighted the contribution of lead by leaded brass plumbing
fixtures and faucets to drinking water. A recent preliminary LCR survey(5) of 46 large systems and 7 medium and
small systems conducted by the Association of Metropolitan Water Agencies stated: "By far the major reason is or
is suspected to be lead leaching from new (newer) faucets and fixtures", referring to reported causes of high lead
levels in drinking water samples.
In California, the Natural Resources Defense Council and the Environmental Law Foundation filed a complaint
with the California Superior Court on December 15, 1992. The complaint stated that a number of major faucet
manufacturers marketed faucets that leach lead into the drinking water at levels exceeding the limit of 0.5 \ig/L set
under the state's Proposition 65 Standard. (6) The lawsuit targeted 13 major faucet manufacturers. The case was
settled in August of 1995 when a number of major manufacturers agreed to meet stipulations which included
warning labels and a lead reduction program. (7)
Similarly, the Environmental Defense Fund, Natural Resources Defense Council and the California Attorney
General recently filed suit under California Proposition 65 against four manufacturers of submersible water pumps.
(8) Some of these pumps, which are used to draw water from private wells, are constructed with brass components
and, as a result, can provide a point source of lead contamination.
Brasses are defined as copper alloys that contain zinc (5 to 40% Zn) as the principal alloying element with or
without other alloying elements, such as tin, iron, aluminum, nickel and silicon. Brasses containing more than
20% zinc are refereed to as "yellow brass", while brasses containing less than 15% zinc are referred to as "red brass".
Lead is added to brass to improve the machinability of the alloy and make castings pressure tight by filling the
voids created as the casting cools. Lead is insoluble in copper-zinc alloys and therefore, during solidification, lead
precipitates and forms a dispersion of second phase particles or globules both at the grain boundaries and within the
matrix. The globules serve to improve the alloy's machinability by acting as chip breakers, reducing tool clogging
and allowing increased cutting speed. (9) Lead is found in brass from 0.1 to 12.0%(10), however, brasses most
commonly used in household fixtures contain 1.5 to 7.5% Pb(l 1). The reference standard for machinability is
131
-------
"free-cutting " or "free-machining" brass, composed of 61.5% Cu, 3% Pb, and 35.5% Zn. This alloy, designated
C36000, is given a reference machinability rating of 100% and is the standard to which all other brasses are related.
Tflie majority of published research that has examined the effect of drinking water on brass, bronze and other
zinc-containing alloys was focused on the corrosion phenomenon referred to as "dezincification". Dezincification has
received special attention from the drinking water industry primarily because of the physical and visible damage to
plumbing systems that results, leading to costly pipe failures and plumbing blockages. Dezincification of small
valves and fittings in the United States has not been nearly the problem it has in Great Britain and some other
places in northern Europe primarily because standard materials used in the U.S. tend to be low-zinc-containing red
brasses.(12)
Dissolution refers to the solubilization of the alloy metals into solution. As mentioned, the majority of past
brass corrosion research was focused on dezincification and unfortunately in nearly all of those cases metal leaching
levels or dissolution was not addressed. Literature pertaining to metal dissolution from brass alloys is typically
presented in a general fashion.
, • ., '
Samuels and M<§ringer(13) evaluated short-term metal leaching from 8 kitchen faucets in 4 different types of
wafer and aqueous fulvic acid. The study involved measuring metal concentrations in standing water samples taken
from mverted water faucets for two 24-hour stand periods. Results showed that lead as well as copper, zinc,
chromium and cadmium leached in varying amounts depending on the type of faucet used and the leach solution.
The source of chromium and cadmium contamination was not identified, introducing the possibility that some of
the water was in contact with the plating material on the external surface of the faucet.
',:!' • . .:' ; •'' ' ' '•• ' ., ' ' '' '« ' . " i. : •, ;
Njelsen(14.,lS) found that lead could be picked up in water from household plumbing fixtures composed of brass
such as water meters and mixer fittings, and main and stop valves composed of gunmetal (copper alloy containing
5% Pb).
Birden et al. (16) found lead contamination hi copper pipe loop systems constructed with no lead containing
parts other than brass compression fittings. Lead leaching was attributed to the lead hi the brass used to make the
fittings.
Schock and Neff(17) obtained data from a 2-year field and laboratory study that implicated brass valves and
fittings as potentially significant sources of lead, copper and zinc in drinking water. The field study indicated that
new chrome plated sampling faucets on pipe loops in galvanized and copper systems were associated with 125 mL
water samples that exceeded the lead maximum contaminant level (MCL) that was 0.05 mg Pb/L at the tune of the
study, As a result, they conducted trace metal leaching experiments on six new sampling taps using an inverted
faucet technique. The tests were performed using deionized and Champaign (Illinois) tap water (3 faucets for each
water). Samples were taken every 2 days for 2 weeks following 24-hour dwell times. The samples were analyzed
for lead, cadmium, zinc, iron and copper. The faucets subjected to deionized water leached about 100 mg Pb/L and
the amount leached dropped off rapidly to less than 10 mg Pb/L at the end of the 2 weeks. Faucets subjected to
Champaign tap water initially leached approximately 1 mg Pb/L and dropped to less than 0.3 mg Pb/L at the end of
2 weeks.
Gardels and Sorg(18) also evaluated metal leaching from faucets. They used a different metal extraction
approach than the fill and dump scenario implemented by the previous investigators. They mounted 12 different
faucets upright to a manifold system connected to a pressure activated pump and water storage reservoir. In this
position, the faucets were operated under pressure in a similar fashion to how they would be used in a home.
Leaching tests were conducted with deionized and Cincinnati (Ohio) tap waters over a 9 month period. Standing
i water samples were taken at a variety of standing times and sampling schemes. Samples were analyzed for lead,
copper, zinc, iron, chromium and cadmium. Results indicated that lead leached from new cast brass faucets could
contribute lead to drinking water in excess of 10 jig/L. They showed that as much as 75% of the lead leached from
common kitchen faucets could be collected in the first 125 mL of water from the faucet; more than 95% of the lead
was (follected in 200 to 250 mL.
132
-------
Experimental Methods
Test apparatus
The test apparatus consisted of a large water reservoir (100 gallon Nalgenel tank) connected to 12 parallel test
loops by 1/2" inside diameter (ID) Schedule 80 polyvinyl chloride (PVC pipe. Test water contained in the
reservoir was recirculated in the tank by a PVC pipe recirculation line and magnetic drive pump system. That water
was fed to the loops by a separate magnetic drive pump and line. AR pump components in contact with the water
were constructed of metal-free materials. Water that had passed through the loops' was discharged to a waste drain.
Each test loop was constructed of a variety of PVC Schedule 80 and high density polyethylene (HOPE) valves,
fittings, and pipe. The systems each contained a 3-way valve, a sample holding cell, and tubing made of Teflon®.
The holding cell was sized to tightly hold one l"x2"xl/8" metal coupon (Metal Samples, Inc., Munford, AL) in
order to maximize the ratio of coupon surface area to volume of water in contact with the coupon. The coupon
surface area was 4.75 in2 and the volume of water was 26 mL, resulting in a surface area to volume ratio of 0.18
in2/mL. Figure 1 is a photograph of the test system and Figure 2 is a schematic of an individual test loop.
Coupon material
Six differently composed brass coupons, plus a pure copper, pure zinc, pure lead, and 60:40 Sn:Pb solder
coupon were simultaneously tested. One brass and the lead coupon were tested in duplicate to ensure the results
were reproducible. The results presented here and in an earlier EPA report (19) indicate that this degree of
duplication was sufficient to satisfy the precision necessary for the objectives of this study. The brasses used were
chosen because they represent common materials used to manufacture brass faucets and other fixtures used in
drinking water systems. Several coupon finishes were available from the coupon supplier: 120-grit, milled, and
glass bead finishes. The 120 grit finished coupons were chosen for this study because it was the finest finish
available and was presumed to give the most consistent coupon surface.
The compositions of the brass coupons were reported by the coupon manufacturer as percentage ranges. In
many cases, relatively wide composition ranges, such as those given for lead, overlapped among different brasses.
The nature of the study, however, required more definitive metal coupon compositions. As a result, approximately
0.1 gram of each coupon type was microwave digested. (20-23) The digestates were analyzed for lead, copper, zinc,
iron and tin by flame atomic absorption spectroscopy (Table 1).
Problems with coupon contamination lead to the development of a thorough coupon precleaning procedure. (24)
The cleaning procedure was, in part, a combination of two American Society for Testing and Materials (ASTM)
coupon wash procedures: designations G31-72 (25) and D2688-83 (26), and basically consisted of a dilute acid
wash, ultrasonification, and an acetone wash.
Operating procedures
The daily operating procedure was as follows: During the first three test runs, the Nalgene® tank was filled with
approximately 30 gallons of Cincinnati tap water. Water pH was chemically adjusted, if necessary, by adding 6 N
HC1 or 8 N NaOH. The tank recirculation pump was in continuous operation during the make-up of the feed water
to assure complete mixing and dispersement of chemical additives. In the last two test runs, the water was
determined to be stable enough (i.e., did not significantly change chemically) that the storage tank was filled to 80
gallons and that water was used over the following several days. The free chlorine residual and pH were checked
daily and chemically adjusted as necessary. Free chlorine residual was maintained above 1 mg/L and the pH was
controlled within 0.10 pH units of the run goal. Free chlorine residual was maintained by adding sodium
hypochlorite solution (4 to 6% available chlorine).
133
-------
After the water was prepared, all valves in the loops were opened to the waste drain. The feed pump was
activated. Test water was flushed through the loops simultaneously at a flow rate of approximately 0.15 L/min for a
brief rinse period of 5 to 10 minutes for total flow of approximately 10 gallons. The valves in the loops were then
closed, holding the water in the cells in contact with the metal coupons. The water was held in the cells for
approximately 22 to 24 hours. Next, the cells were sampled by opening the labcock above the cell to the air and
then opening the 3-way Teflon® stopcock below the cell while simultaneously holding a 60 mL Nalgene® HOPE
sample bottle below the stopcock sample port. Air admitted through the top stopcock allowed the leach water to
drain by gravity into the sample bottle. The total sample volume was only about 26 mL, most of which was
contained in the Teflon® sample cell. The valves were then closed'and newly prepared source water was again
flushed through the cells, as mentioned, repeating the procedure. Initially, sampling from the cells was conducted
dally. After several weeks of operation, sampling was reduced to three times per week. Water was flushed through
the cells daily (Monday to Friday), even when sampling was not done. Air was in contact with the metal coupons
for a short period of time, however, the coupon surface was not exposed long enough to completely dry.
Test runs were generally terminated after approximately 150 days. Termination was based on apparent
stabilization of metal leaching levels over the relatively short exposure time. At the end of each test run the coupons
were removed from die cells, the system was cleaned as previously described, new coupons were installed, and the
Study was repeated with a different extraction water.
Sampling
Water samples taken in 60 ml bottles from the sample cells, as mentioned, were preserved in 0.15% ultrapure
HNO3(27) and analyzed for lead, copper, zinc, iron, and tin. Free and total chlorine and pH were measured daily
und dissolved oxygen was measured weekly in the tank feed water. These parameters were measured immediately,
prior to fresh water being fed to the loops. Additionally, three feed water samples were taken: a 250 mL, 0.15%
HNOj preserved sample for background metal analysis (Ca, Cu, Fe, K, Mg, Mn, Na, Pb and Zn); a 250 mL sample
for background wet chemistry analysis (alkalinity, Cl, NH3, NO3, PO4, SiO2, and SO4); and a 30 mL sample for
total inorganic carbon (TIC) analysis. All sample bottles, with the exception of the TIC bottle, were Nalgene®
HOPE bottles. TIC samples were taken in 30 mL glass vials having caps with conical polyethylene liners. Special
care was taken to insure that no air or air bubbles were trapped in the sample.
Analytical procedures
Unless otherwise specified, all chemicals used in this study were Analytical Reagent (AR) grade. Deionized
(Dl) water was prepared by passing building distilled water through a Milli-Q PlusP cartridge deionized water
system (Millipore Corp., Bedford), with a final resistivity > 18.2 MO.
Glassware (excluding pipets) used for the preparation of standards and solutions was cleaned using a 5%
solution qf Contrad 700, The glassware was thoroughly rinsed with deionized water. Glassware to be reused
immediately was cleaned by soaking hi 10% (v/v) concentrated HNC-3 and rinsed with DI E^O. Glass pipets were
cleaned by at least an overnight soaking in 5% Contrad solution, followed by rinsing with dilute AR-grade HC1 in a
plastic pipet washer. The final rinse was a minimum of 8 total volumes of deionized water cycled through the pipet
washer. Air displacement micropipets with disposable tips were used for handling and transferring solutions.
•"I , ;, , ;', . •' ;:;,', I., • • •'
Ultra-pure nitric acid, HNC-3 (Ultrex, J. T. Baker Chemical Company., Phillipsburg, NJ) was used to preserve
samples. The volume of nitric acid added to the sample was 0.15% of the sample volume. 6 N HC1 (Mallinckrodt,
Inc., Paris, KY) and 8 N NaOH (Fisher Scientific, Failawn, NJ) were used to chemically adjust feed water pH.
Sodium phosphate (NasPO^HaO) (Fisher Scientific, Failawn, NJ) was used to adjust the phosphate concentration in
the feed water.
Lead was analyzed with a Perkin Elmer Model 4000 Spectrophotometer (Perkin Elmer Corp., Norwalk, CT)
equipped with a Model HGA 400 furnace programmer and AS 40 autosampler. All other metals were analyzed with
a Perkin Elmer Model 5000 Atomic Absorption Spectrophotometer and an AS 50 autosampler. Comparison
134
,.i; if ,. . , - ' i. Ml , i! ' ! • .1, ' j. ,, .,li I' .'frig !' 'lit I I.;i(1.
-------
studies showed no statistically significant differences between measurements made by the instruments. The pH was
measured with a Orion Model EA 940 pH meter (Orion Research, Inc., Boston, MA) and an Orion Ross Sure-Flo™
electrode. Free and total chlorine were analyzed with a Hach DR/2000 spectrophotometer (Hach Company,
Loveland, CO). TIC was analyzed by a coulornetric procedure on a UIC Model 5011 CO2 coulometer (UIC Inc.,
Joliet, IL) with Model 50 acidification module, operated under computer control. A complete list of analytes and
analytical methods is shown in Table 2.
Statistical data analysis and interpretation
Differences in metal leaching trends (plot of metal release versus elapsed time) were determined using several
statistical approaches. Data normality was determined by the Kolmorogov-Smirnov Test. In most cases of this
study, data distributions were not normal (at 95% confidence), suggesting the use of the following non-parametric
statistical data interpretation techniques. The Kruskal-Wallace Analysis of Variance (ANOVA) on Ranks was used
to demonstrate differences in metal leaching trends from coupons. The Kruskal-Wallace Test is necessary when (a)
determining if three or more groups are affected by a single factor (e.g. brass composition), and (b) samples are not
normally distributed or do not have equal variances. (28) Since ANOVA statistics only indicate whether two or
more groups are different, the Student-Newman-Keuls test and the Dunn's test were used to make multiple
comparisons between all possible group pairs. The Student-Newman-Keuls test is used when group sizes are equal
and Dunn's test is used when group sizes are unequal. All statistical comparison tests were made at 95% confidence
(P=0.05), that is to say with 95% confidence that the groups differ significantly. In the few cases were the data were
normally distributed, the Kruskal-Wallace one-way ANOVA test on Ranks was used to demonstrate differences in
leaching trends.(29) All statistical calculations were made using Sigmastat™ (Jandel Scientific, San Rafael, CA)
statistical software.
Experimental Results
Test run 1
Non-chemically adjusted Cincinnati tap water was used in test run 1 (Table 3). As a result, larger than desired
pH fluctuations (0.18 pH units standard deviation) were observed throughout the test run, The solubility of copper,
in particular, is strongly impacted by pH and even small pH fluctuations (0.1 pH units) have been shown to
considerably influence copper solubility. (24) The mean sulfate concentration (1 16.3 mg S04/L) was higher during
this test run than in others. Researchers have suggested that sulfate can have a negative or positive impact on copper
(30,31) and lead (32) corrosion depending on initial water quality, however, the degree, mechanism, and conditions
favoring the impact have not been fully established. Free chlorine, a strong oxidizing agent and an influential
parameter on metal corrosion rates (rather than equilibrium metal concentration), was relatively high in
concentration, averaging approximately 2.0 mg/L. Perceived increases in metal levels associated with free chlorine
increases are often misinterpretations of increased oxidation rates. The increased sulfate and free chlorine levels
observed during this test run were likely related to seasonal water quality variations in the Ohio River (main source
of Cincinnati tap water). Dissolved oxygen and TIC were not directly measured during this test run. TIC,
however, was calculated using WATEQX(33), a computer FORTRAN 77 program, and source water quality
values. The calculated value was 14.3 mg C/L with a standard deviation of 1.2 mg C/L.
Lead leached from brass coupons
Lead leaching trends from the brass coupons (C36000 was tested in duplicate) and duplicate pure lead coupons
during test run 1 are shown in Figure 3. Throughout this paper, "trend" is defined as the plot of metal release versus
elapsed time. The greatest rate of lead leaching from the brass coupons occurred during the first 7 days of the test
run. During this time lead levels dropped off sharply and trends were difficult to differentiate. Initial high values
were thought to be related to new, clean coupon surfaces and thinly dispersed lead globules present on the coupon
surfaces at the start-up. The globules contribute initially because of the high surface area, made slightly higher by
any smearing during preparation not removed by the cleaning procedure. For cast brasses, lead tends to migrate
135
-------
•IF!"'
".'Ill
toward the mold surface, and these are somewhat preferentially segregated toward the water surface contact. Exposure
depletes lead at the surface, and passivation films (hydroxycarbonates, oxides, phosphates) fill voids between grains
and cover surfaces as exposure continues, decreasing oxidation and contact with underlying globules. Both
mechanisms serve to reduce the contribution later on. The concentration of lead leached during this period varied
among brasses by no more than approximately 80 \igfL and was generally a function of the lead content of the brass;
brasses with higher lead contents tended to leach higher amounts of lead.
From approximately 7 to 15 days into the test run, lead concentrations gradually leveled off and the difference in
magnitude between lead levels leached from the brasses narrowed. During this time period, the relationship between
a brass's lead content and the amount of lead leached from the brass became less apparent. Statistical comparisons
between the lead leaching trends during this time period showed that there was no significant difference between
most trends at 95% confidence.
After 15 days, the lead leaching trends appeared to split into two distinct groups. For future reference, the
groups will be labeled as the "low" group and the "high" group, in reference to the relative magnitude of lead leached
from the coupons. During the period from 15 to 60 days, the general trend of both groups continued to be a gradual
downward one. The trends appeared to level off or stabilize after 60 to 70 days.
The "low" group consisted of the duplicate free-cutting brass (C36000) coupons and one yellow brass (C85400)
coupon. The amount of lead leached from these brasses was nearly indistinguishable for the remainder of the run
(after 60 days). The similarity in lead leaching patterns was reflective of the similarity in compositions of the two
brasses (see Table 1). The quantity of lead in the two brasses overlapped while free-cutting brass contained slightly
more zfnc and slightly less copper than the C85400 coupon. As a group, the amount of lead leached at the
completion of the study was approximately 8.5 |ig/L. The brasses in this group were two of the three lowest
lead-containing brasses evaluated. Lead levels dropped by approximately 90% over the entire test run period.
The "high" group consisted of the three red brasses: C83600, C84400, and C84500, and yellow brass C85200.
From 60 to 125 days, these brasses followed parallel lead leaching trends but differed in magnitude by as much as
50 Hg/L. C85200 brass leached the least amount of lead in the group over this time period. The average lead
concentration leached from the brasses in the group during the last 21 days of the study was approximately 28 (ig/L.
Lead levels dropped by approximately 80% over the study period.
With the exception of the C85200 brass, the magnitude of lead leached from each brass (Figure 3) corresponded
well to the lead composition of the brass alloy after 60 days of leaching. In other words, brasses containing greater
amounts of lead tended to leach greater amounts of lead than brasses containing lower amounts of lead. C85200
brass is a yellow brass and has a composition similar to that of the brasses in the low group. Based upon the
similarity, it was expected that the amount of lead leached from C85200 brass would have been similar to the
amount of lead leached from brasses in the low group. The difference between the two brasses is that C85200 brass
contained slightly more copper and slightly less zinc than C85400 brass.
A white precipitate formed on several of the brasses. Upon close visual examination, it was noted that the
density of the coating increased as the zinc composition of the brass increased. A nearly identical appearing but
denser solid formed on the pure zinc coupon. Based on the observations and water chemistry, it was theorized that
an insoluble zinc film, probably basic zinc carbonate, had formed on the coupon surface. The film may have
provided corrosion protection to the brasses by acting as a diffusion barrier and reducing the amount of lead leached
from the brasses, This observation could explain why more lead was leached from the lower zinc-containing
G85200 brass than expected.
Localized upward and downward patterns in the leaching trends were occasionally observed in varying degrees
throughout the test run. In most cases, the trends appeared to closely parallel changes in water quality (pH and
chlorine residual).
The pure lead coupons leached more lead than any of the brasses during all stages of this test run. This
suggested that the amount of lead leached from the brasses was not controlled by simply the same diffusion
mechanism (i.e. diffusion of the metal cation into solution). Instead, other characteristics such as available surface
area or other alloy surface and surface film properties controlled metal dissolution kinetics.
136
-------
Statistical comparisons of the lead leaching trends made over the entire test run and after the lead trends
appeared to stabilize (after 60 days) showed that most of the trends were significantly different at the 95% confidence
interval. Good reproducibility between the lead leaching trends of duplicate free-cutting brass coupons was
observed.
Copper leached from brass coupons
Copper leaching trends (plot of copper release in mg/L versus elapsed time) were similar to those observed for
lead in that two groups, a "high" and a "low", were established (Figure 4). The major difference was that the
division into trend groups took place at the beginning of the test run.
The high group consisted of the three red brasses: C83600, C84400 and C84500. Unlike lead leaching trends,
a relatively constant amount of copper (approximately 0.22 mg/L) was leached from the brasses in the high group
throughout the entire test run. The leaching patterns and copper levels of the high group brasses were nearly
identical throughout the entire test run. These results suggested that small differences in copper composition in
similar alloys did not translate into significant differences in copper leaching levels.
The low group of brasses consisted of C3 6000, C85200 and C85400, which were also the brasses that
contained the lowest percentage of copper. The copper leaching trends among these brasses were also nearly
identical. Copper levels gradually decreased over the first 20 to 40 days followed by a leveling-off trend at a
concentration of approximately 0.03 mg/L.
After approximately 120 days, the pure copper coupon leached nearly the same amount of copper as the red
brasses. This suggested that a critical brass composition exists where the alloy's copper leaching properties become
similar to those of pure copper.
A number of local peaks and valleys were observed in the copper leaching trends. This suggested that copper
levels were very responsive to fluctuations in extraction water quality. Copper levels were particularly strongly
influenced by pH. For example, in one instance over a 35-day period (from 95 to 130 days into the test run), the pH
of the extraction water gradually dropped from approximately 8.7 to 8.2. A corresponding increase or inverse pattern
was seen in copper levels. This was especially evident with the high group brasses, where copper levels more than
doubled during this time period.
In most cases, trends within groups were not statistically different from each other. Also, copper concentrations
of duplicate samples taken from the copper (C36000) coupons were not statistically different over the entire test run
or after 60 days, thus good reproducibility was observed.
Zinc leached from brass coupons
Zinc leaching trends (plot of zinc release in mg/L versus elapsed time) were different than those observed for both
copper and lead in that there were not groups of distinct leaching trends (shown in Figure 5). Zinc levels leached
from the brasses were a function of the percentage of zinc in the brass throughout the entire study. Zinc
concentrations appeared to decrease very slightly over the first two weeks, then stabilized after about 60 days for the
remainder of the study in the range of about 0.05 to 0.30 mg/L. Zinc leachate concentrations fluctuated slightly
throughout the study. Fluctuations appeared to be due to changes in background water chemistry (e.g., pH and
chlorine residual).
The pure zinc coupon leached as much as 11.3 mg/L at the start of the study, but dropped rapidly to about 0.32
mg/L after 40 to 50 days. By the end of the study, zinc levels leached from the pure zinc coupon were lower than
levels leached from the yellow brasses and nearly the same as the red brasses. As was the case for copper, there
appeared to be a zinc content threshold at which the alloy resembled the leaching qualities of pure zinc. Zinc
content was low (maximum 34.7%) relative to copper and the pure (100%) zinc coupon. This suggested that factors
other than strictly diffusion control zinc dissolution kinetics.
137
-------
Statistical analysis of the zinc leaching data showed that nearly all coupon leaching trend comparisons were
statistically different. Zinc leaching trends observed after 60 days showed that only the duplicate coupon (C36000)
was not statistically different.
Test run 2
Table 3 summarizes the extraction water quality used during test run 2. The pH of the tap water was strictly
controlled through chemical adjustment with HC1 and/or NaOH and averaged 7.01 with a standard deviation of 0.05
pH units. Sulfate levels were lower during this test run than the first by approximately 50 mg/L, averaging 68 mg
SO.|/L. Free chlorine was also lower during this test run than the first, averaging 1.4 mg/L with a standard
deviation of 0.28 mg/L. TIC and DO were monitored during this test run, averaging 12.6 mg C/L and 8.9 mg
OI/L, respectively.
Lead leached from brass coupons
Figure 6 shows lead leaching trends from the six brass coupons (and one duplicate C36000 coupon) and
duplicate pure lead coupons during this test run. The lead leaching patterns during this test run resembled those in
test run I. In some cases, lead concentrations observed at the end of this test run were one order of magnitude greater
than levels observed for the same coupon during the test run 1. The greatest amounts of lead leached from the
coupons during the first 14 days of the run, with levels dropping off rapidly during this period. The magnitude of
lead leached from the brasses during this period differed by as much as 300 )J.g/L among the brasses. The relative
amount of lead leached from the brass during this period was generally related to the amount of lead in the brass.
From 1.4 to approximately 40 days, the lead leaching trends began to level off. At the beginning of this period,
the magnitude of lead leached from the brasses appeared to be random in that lead trend lines from the different
brasses crossed over each other and followed no distinct pattern. By 40 days, the direction of the trend and ranking
order with respect to lead concentration of the lead trends became consistent for the remainder of the study.
From 40 days to the end of the test run, the tendency of the majority of the lead leaching trends was to slightly
decrease. Yellow brass C85200 was the exception in that the decrease was more apparent. Also, C85400 brass,
oddly, exhibited a gradually increasing lead leaching trend before finally leveling-off. C85400 is a yellow brass,
containing the most zinc but the least lead of the brasses tested. Perhaps extreme and rapid zinc dissolution exposed
additional lead surface sites as time passed. However, this observation was not made with the similarly composed
C85200 brass, decreasing the likelihood of this explanation.
A distinct set of trend groups like those observed in test run 1 did not develop during this test run. By the end
study of the study, trends tended to be evenly separated by roughly 10 to 50 jag/L. The percentage of lead leached
from brass was directly related to the amount of lead in the alloy .(i.e., the more lead in the alloy the more lead
leached from the brass). As was the case in test run 1, the pure lead coupons leached more lead than any of the
brasses.
Statistical comparisons of the leaching trends made over the entire test run show that approximately half of the
trends are significantly different at the 95% confidence interval. Trends that are not significantly different are mostly
of the same group or brass type, yellow or red brasses. The duplicate C36000 trends were not significantly different
over the entire test run.
Copper leached from brass coupons
Copper leaching trends differed from lead leaching trends in that copper levels did not start out extremely high,
followed by a rapid drop and gradual leveling-off (shown in Figure 7). With the exception of C36000 brass, the
trends gradually increased over the first 60 days of the test run to a plateau at which they stayed for the remainder of
the study. This pattern suggested that a protective oxide film formed on the coupon surface during coupon storage
138
-------
prior to study initiation as a result of exposure to the atmosphere. The film provided some corrosion protection to
the coupons initially, however, with time dissolved and new films and equilibrium conditions were established.
The C36000 brass leaching trend decreased gradually during the first 60 days, after which the trend leveled off.
There was difficulty in accurately distinguishing or ranking leaching trends, with the exception of the two
C36000 brass coupons. The C36000 brass coupons stood out from the other brasses in that significantly lower
amounts of copper were leached from them. They contained the least amount of copper of the brasses tested. The
difficultly in detecting differences or ranking copper leaching trends among the remaining brasses was due to the
inconsistency or sporadic nature of the leaching trends despite strict control on pH of the test water. This behavior
could have been associated with particulate material containing copper.
Copper levels generally remained between 1 to 2 mg/L of copper throughout the duration of the study for the
group of coupons. The amount of copper leached from the pure copper coupon was initially higher than levels . .
leached from brasses, however, rapidly fell below those levels leached from the red brasses.
Leaching trends were determined to be statistically significantly different from each other at 95% confidence in
most cases. However, lines connecting data points crossed over each other frequently making it difficult to
distinguish or rank trends. Reproducibility of copper leached from the duplicate free-cutting brass coupons was
good.
Zinc leached from brass coupons
Zinc leaching patterns followed a slightly different trend than lead and copper (Figure 8). In almost all cases,
zinc levels gradually fell with time, in most cases leveling-off by approximately 60 days. The amount of zinc
leached from the coupons was directly dependent upon the amount of zinc in the alloy. C36000 leached the most
zinc, leaching as much as 2 mg/L more than the next closest alloy, and averaged about 4 to 5 mg/L by the end of
the study. The levels leached from C36000 were nearly the same as the levels observed with the pure zinc coupon.
C85400 and C85200 brasses were next in order of leaching. The two yellow brasses contained the next greatest
amount of zinc. Zinc leached from them appeared to be still decreasing at the end of the study. Red brasses C84500
and C84400 were next, and very close, averaging around 0.5 mg/L and C83600 leached at < 0.25 mg/L. In nearly
all cases (after 60 days and over the entire study), zinc leaching trends were statistically different. Reproducibility of
zinc leached from the duplicate free-cutting brass coupons was good.
Impact of pH on metal leached from coupons
The influence of pH on metal dissolution from the alloys and pure metals was evaluated by comparing metal
levels leached from the coupons exposed during test runs 1 and 2. The results, showed that the three red brasses all
behaved similarly with respect to the influence of pH on lead levels leached. As expected, the water considered to be
most corrosive towards lead, the lower pH water, leached the most lead from the red brass coupons. This was
particularly evident with C83600 and C84400 coupons, where a difference of approximately 100 |ig Pb/L was
maintained throughout the duration of their respective studies. The influence of pH on lead leached from C84500
was not as significant and at one stage, between 20 and 60 days, appeared to be insignificant. The most notable
difference was that C84500 brass contained the most zinc of the three red brasses tested. As a group the red brasses
contain the most lead, nearly twice the lead content of the yellow brasses.
The two yellow brasses (C85200 and C85400) exhibited different lead leaching responses to pH than the red
brasses. Interestingly, lead leached from C85200 showed that pH had no impact on lead concentration while the
impact of pH on lead leached from C85400 brass was small (10 |J.g Pb/L) relative to the red brasses. The major
composition differences between the red and yellow brasses are that red brasses contain more lead but far less zinc.
These observations suggested that the dissolution rate of lead from brass was not diffusion limited or pH dependent.
The difference between the amount of lead leached from C36000 at pH 7.0 and 8.5 were intermediate to these
observed with yellow and red brasses. Lowering the pH to 7.0 increased lead levels significantly (by approximately
139
-------
30 Hg Pb/L) but not to the degree observed with the red brasses. Free-cutting brass contains the most zinc of all the
brasses tested and contains less lead than the red brasses and more lead than the yellow brasses.
Copper leaching levels were significantly higher at the lower pH for all brasses except C36000 brass. The
difference was greater amongst the red brasses, with copper levels being more than 1.5 mg/L higher at pH 7.0. The
difference between copper leached from the coupons at pH 8.5 and 7.0 appeared to widen as the percentage of copper
In the alloy increased. Free-cutting brass leached the least copper and the differences between copper leached at pH
7,0 and 8.5 was insignificant.
Zinc leaching levels were significantly higher at the lower pH for all brasses. As with copper, the difference
between zinc leached at pH 8.5 and 7.0 appeared to increase as the percentage of zinc in the alloy increased. The
largest difference was seen with the free-cutting brass (which contained the most zinc) and the smallest difference was
red brass CS3600 which contained the least zinc.
Lead coupons
The pH 7.0 test water was more corrosive towards lead than the pH 8.5 water. At pH 7.0, lead levels were
highest at the start of the study and gradually dropped with time. The lead levels never appeared to stabilize, even
by the completion of the test run at 160 days. At pH 8.5; lead levels appeared to stabilize after only approximately
25 days. At the end of the test runs, lead levels were more than 200 ug/L higher at pH 7.0 than at pH 8.5. The
results demonstrated the anticipated effect of pH on lead solubility.
Copper coupon
The pH 7.0 test water was also more corrosive towards copper. The leaching pattern over time was unlike that •
for lead in that copper levels started low (approximately 1 mg Cu/L) but increased rapidly to a peak at
approximately 4 mg/L by 40 days. The copper leaching trend proceeded to drop gradually for the remainder of the
study, falling back to the original level of about 1 mg/L in 160 days. It appeared that at the termination of the test
run, the copper levels were still decreasing and that further time would be required to confidently state that the
copper levels stabilized. This suggests that it may take a great amount of time for copper levels from new copper
surfaces to stabilize under exposure to corrosive conditions.
The copper leaching pattern of the pure copper coupon exposed to pH 8.5 water differed from pattern at pH 7.0.
The copper levels were significantly lower under these conditions (pH 8.5) and the trends followed a different
pattern. Copper levels started out at about 0.5 mg/L and gradually dropped to approximately 0.25 mg/L by 120
days. It appeared that by 80 days into the test run the leaching trend had leveled off. The leaching trend was also
smooth in comparison to the lower pH trend.
Zinc coupon
The lowest pH (7.0) test water was also the most corrosive towards pure zinc. At pH 7.0, zinc levels started
high (approximately 9 mg/L) then dropped rapidly to about 4 mg/L by 20 days. The levels increased slightly, then
leveled off for the remainder of the test run after approximately 60 days. Final zinc values fluctuated randomly
around 4.5 mg/L.
Lower zinc levels were observed at pH 8.5. The zinc levels started out quite high, but decreased rapidly over
the first 30 days, falling to a more gradually decreasing slope. The trend appeared to level off at about 60 days, to a
concentration of about 0.2$ mg/L. The curve was relatively smooth.
140
-------
Impact of sulfate on copper dissolution
Although studying the impact of sulfate on copper corrosion was not stated as an objective of this study,
background sulfate levels, theoretical considerations, and observations made over the test runs warrant brief
discussion. Figure 9 (taken from Schock et al. [31]) shows copper levels (after concentrations were considered
stabile) leached from the pure copper coupons during the two test runs superimposed on theoretical solubility curves
based on cupric hydroxide solubility modeling. The theoretical curves correspond to average TIC concentrations
during the respective runs and represent "new" surface conditions. The figure also includes the computed solubility
line for CU4(OH)6SO4H2O, which is in closer agreement to the experimental data at higher pH values. The absence
of pH effect on copper levels in these test is at variance with trends of both soluble and total copper leaching over the
same pH/TIC range in other U.S. EPA laboratory experiments (31) conducted in synthetic waters without sulfate.
This suggests that a constituent or constituents like sulfate in the tap water is the cause of the different behavior.
Clearly more experimentation needs to be done to clearly identify the role of sulfate on copper corrosion.
Effect of oxidants on metal leaching from coupons
A detailed evaluation of the impact oxidants of metals leached from the coupons could not be conducted due to
the small sample volume within the test cell. The role of oxidants on the rate of metal dissolution is an important
one and should be considered at least form a theoretical standpoint. The dissolution of the metals from the brass
and pure metal coupons is initially facilitated by oxidation. Both free chlorine and dissolved oxygen were initially
present in the test waters. The levels the metals achieve for different lengths of stagnation tune are closely related to
the nature and concentrations of the oxidants present in the water, as well as the surface area of the metals exposed to
the water and available for oxidation and dissolution. Related research (31,34,35) has shown that free chlorine, the
stronger oxidizing agent, reacts faster with the metals than dissolved oxygen, and the reaction proceeds until either
metal solubility equilibrium is reached, or oxidant concentrations are depleted in the cells. With respect to the
described study, this means free chlorine reacts with the coupons first, and if solubility equilibrium is not achieved
by the time the chlorine residual is gone, the oxidation reactions will continue until the dissolved oxygen is gone.
Over time, corrosion scales develop on the surfaces, that can inhibit mass transfer of oxidant species and metal ions
to and from the metal surface, and reduce the reactivity at the water/surface interface.
In addition to the data presented ha this paper for 22 to 24 hours of stagnation, supplemental experiments were
performed with 72 hours of stagnation. (19,35) In many cases, metal levels continued to increase with stagnation
time to 72 hours or beyond. In systems where oxidant concentrations were completely depleted, metal levels
stabilized, or in the case of copper, sometimes decreased. These chemistry effects would certainly be expected to
pertain to the practical domestic use situations with faucets. Thus, the trends of metal concentration versus
stagnation time cannot be simply inteipreted in terms of a radial-diffusion model, and likely represent a combination
of kinetic and physical factors (35) that also would vary with the age of the faucet.
Discussion and Conclusions
The results of this study have demonstrated that the amount of lead leached from brass was generally dependent
on the amount of lead in the brass at pH 8.5 and 7.0. That is, the more lead in the alloy, the greater amount of lead
leached from the alloy. In nearly all cases, the highest amounts of lead leached from the brasses over the first two
weeks of exposure, and lead levels dropped the quickest during this period. At pH 8.5, the concentration of lead
leached from the brasses leveled off after 60 to 70 days. Under more corrosive conditions (pH 7.0), lead levels
appeared to be decreasing slightly at the end of the test run (155 days). The pure lead coupon leached considerably
more lead than the brasses. This suggested that the amount of lead leached from the brasses was not controlled by
solubility, but instead other factors such as surface area limitations or kinetic factors were important. Slight
fluctuations in leaching trends were often associated with pH and free chlorine fluctuations in the extraction water.
The impact of pH on the amount of lead leached from brass was dependent on the amount of lead in the brass.
In higher lead containing red brasses, more lead leached from the brasses at pH 7.0. This observation followed
predicted lead solubility response to pH and was also observed with the pure lead coupon. As the amount of lead in
141
-------
the alloy decreased, however, the impact of pH on the amount of lead leached from the alloy decreased to the point
where there was little or no impact on the yellow brasses. This observation reinforced the theory that dissolution
kinetic^ are nqt controlled by diffusion limitations.
The amount of copper leached from the brasses at pH 8.5 was clearly related to alloy composition. It appeared
that red and yellow brasses could be divided into two groups. The red brasses leached the most copper and by the
end of the study were leaching as much copper as the pure copper coupon. It was hypothesized that at some critical
copper content, the leaching properties of the alloy resemble the leaching properties of pure copper. Copper levels
were relative stabile throughout the entire study. At pH 7.0, copper levels were sporadic and difficult to order.
Participate material containing copper was thought to be partly responsible for this observation. Copper levels
Increased over the first 60 days of the study and leveled off. Once again, some red brasses leached as much copper as
the pure copper coupon. The amount of copper leached from the alloys was strongly influenced by relatively small
fluctuations in pH.
The impact of pH on copper levels leached from the brasses increased with increasing copper content of the
alloy. Largest differences in copper levels leached from the same brass at pH 7.0 and 8.5 were observed with pure
copper and the red brasses. The difference decreased as the copper content decreased. Lower copper levels were
observpd at pH 8.5.
The amount of zinc leached from the brass coupons at pH 8.5 and 7.0 was dependent on the amount of zinc in
the alloy. The highest zinc levels were observed at the start of the study. Zinc levels then dropped off slightly to
level off after approximately 60 days. By the end of the test runs, yellow brasses leached more zinc than the pure
zinc coupon. This suggested that a critical zinc composition exists where the brass leaching properties resembled
the leaching properties of pure zinc. Zinc levels were sensitive to pH fluctuations.
The impact of pH on zinc levels increased with increasing pH. The largest differences between zinc levels at pH
7.0 and 8.5 were observed with pure zinc and the yellow brasses. The difference decreased as the zinc content
decreased. Lower zinc levels were observed at pH 8.5.
"i'iii ' ' .1 ' " ;; ' ' ,
The results suggest that reducing the amount of lead in a brass faucet or other plumbing fixture or valve will
reduce the lead levels at the tap. In this study, the free-cutting and yellow brasses (C36000, C85200 and C85400)
Icache3 the lowest amount of lead. They also leached the least amount of copper since they also contained the
lo%vest amount of copper. They did, however, contain the highest amount of zinc and leached the greatest amount of
zinc. Although zinc levels at the tap are not a regulatory concern, high zinc containing brasses are at greater risk of
deztncifieation attack.
It Is important to stress that the conclusions of this study were based on tightly managed laboratory
experiments with identically machined coupons in several closely controlled waters using consistent, reproducible
operating procedures. Extrapolation of the results to field conditions where alloys are subjected to the distribution
system should be done with some caution. For example, brass faucets encounter mechanical operation which may
effect metal levels by influencing leaching trends or physically dislodging protective films. Physical features of alloy
fixtures such as the faucet structure and the alloy surface finish also likely play a role hi the degree to which metals
leach from them in the distribution system. In addition, if the parts of a faucet are made from different alloys, the
corrosion behavior may be affected by galvanic coupling effects. Future research needs to include exploring the
degree to which some of these variables influence the metals leached from alloys. Specific areas include evaluating
the effect of the following variables on metals leached from alloys: alloy machining and finishing techniques; faucet
structure; water flow pattern, velocity, and pressure through faucet structures; mechanical operation; fixture age; and
stagnation time. Kinetic issues related to metal dissolution rates also need to be further addressed.
Acknowledgments
The authors wish to express particular thanks to Stanford Tackett for his initiation of the study. We would also
like to thank the following members of the USEPA staff: Herb Braxton and Kenneth Kropp for daily operating and
sampling of the test system, and Keith Kelty, James Doerger, James Caldwell, Louis Trombly, and Patrick Clark
142
-------
for analytical support. And final thanks goes to Greg George, Steve Harmon, John Damman, and Roger Rickabaugh
of Technology Applications Inc. and Kim Brackett and Cory Demaris of IT Corp. for additional analytical support
and technical discussions.
Disclaimer
The information in this document has been funded wholly or in part by the United States Environmental
Protection Agency. It has been subject: to the Agency's peer and administration review, and it has been approved
for publication as an EPA document. Mention of trade names or commercial products are for explanatory'purposes
only, and does not constitute endorsement or recommendation for use.
References
1. Public Law 99-339. Safe Drinking Water Act Amendments of 1986. (1986).
2. Drinking Water Regulations; Maximum Contaminant Level Goals and National Primary Drinking Water
Regulations for Lead and Copper. Final Rule; correction, U. S. EPA, 40 CFR Parts 141 and 142, Fed. Reg.,
56:135:32112 (July 15, 1991).
3. Maximum Contaminant Level Goals and National Primary Drinking Water Regulations for Lead and Copper.
Final Rule, U. S. E. P. Agency, Fed. Reg., 56:110:26460 (June 7, 1991).
4. Drinking Water Regulations:' Maximum Contaminant Level Goals and National Primary Drinking Water
Regulations for Lead and Copper. Final Rule; correcting amendments, U. S. EPA, 40 CFR Parts 141 and 142,
Fed. Reg., 57:125:28785 (June 29, 1992).
5. Association of Metropolitan Water Agencies. Bulletin #92-35, Lead and Copper Rule Monitoring. (1992).
6. Water Policy Report. (1992).
7. News Release. National Resources Defense Council. "Faucet Settlement to Affect 90% of U.S. Market".
(1996).
8. News Release. Environmental Defense Fund and National Resources Defense Council. "Many Pumps in
Private Wells Pose Lead Threat; Env. Group. Attorney .General File Suit". (1994).
9. Plewes, J. T. & Loiacono, D. N. "Free-Cutting Copper Alloys Contain No Lead". Advanced Materials and
Processes, 140:4:23 (1991).
10. Uhlig, H. The Corrosion Handbook. John Wiley and Sons, New York, (1948).
11. AWWARF. Lead Control Strategies. AWWA Research Foundation and AWWA, Denver, CO (1990).
12. Jester, T. C. "Dezincification Update". Jour. AWWA, 77:10:67 (1985).
13. Samuels, E. R. & Meranger, J. C. "Preliminary Studies on the Leaching of Some Trace Metals from Kitchen
Faucets". Water Research, 18:1:75 (1984).
14. Nielsen, K. Measurements and Test Procedure for Cadmium and Lead Pick up in the Water from Drinking
Water Installations. 7th Scandinavian Corrosion Conference, (1975).
15. Nielsen, K. "Control of Metal Contaminants in Drinking Water in Denmark". Aqua, 4:173 (1983).
16. Birden, H. H., Calabrese, E. J. & Stoddard, A. "Lead Dissolution From Soldered Joints". Jour. AWWA
78:11:66(1985).
143
-------
1 7.
18.
19.
20.
21.
22.
23.
24,
25,
26.
27.
28.
29.
30.
3 1 .
32.
33.
34.
Schock, M. R. & Neff, C. H. "Trace Metal Contamination from Brass Fittings". Jour. A WWA, 80:11 :47
(1988).
Gardels, M.'C. & Sorg, T. J. "A Laboratory Study of the Leaching of Lead from Water Faucets". Jour.
81:7:101 (1989).
Lytle, D. A. & Schock, M. R. "Stagnation Time, Composition, pH and Orthophosphate Effects on Metal
Leaching from* Brass", 0. S. Environmental Protection Agency, Office of Research and Development,
Washington, DC, September (1996).
Anonymous. Microwave Digestion of Lead Turnings. Application Note NM-5 (Metals and Mining) CEM
Corp., P.O. Box 200, Matthews, NC. 28106 , (1990).
Anonymous. Microwave Digestion of Pb:Sn Solders. Application Note NM-4 (Metals and Mining) CEM
Corp., P.O. Box 200, Matthews, NC. 28106, (1990).
Rains, T. Personal communication. ACS Spectroscopy Short Course, Pittsburgh Conference, New Orleans,
LA, (1991).
Oilman, L. Microwave Sample Preparation Manual. Table 8, Acid Selection Guide, AquaRegia. Application
Note NM-4 (Metals and Mining) CEM Corp., P.O. Box 200, Matthews, NC 28106 , (1989).
Lytle, D. A., Schock, M. R. & Tackett, S. Metal Corrosion Coupon Study Contamination, Design, and
Interpretation Problems. Proc. AWWA Water Quality Technology Conference, Toronto, ON, (1992).
ASf M Standard G31-72. American Society for Testing and Materials (ASTM), A. S. f. T.a.M. (ASTM),
(1983).
ASTM Method 2688-83 Method B, Coupon Test. American Society for Testing and Materials (ASTM), A.
S. f, T. a. M. (ASTM), (1983).
USEPA. "Handbook for Analytical Quality Control in Water and Wastewater Laboratories", U. S.
Environmental Protection Agency, Environmental Monitoring and Support Laboratory, Cincinnati, OH,
(1979).
Jandel Scientific. Siginastat User's Manual. Jandel Scientific, San Rafael, CA, (1992).
Glantz, S. A. Primer of Biostatistics. McGraw-Hill Inc., New York, (Third ed., 1992).
Edwards, M., Meyer, T. & Rehring, J. Effect of Various Anions on Copper Corrosion Rates. Proc. AWWA
Annual Conference, San Antonio, TX, (1993).
Schock, M. R., Lytle, D. A. & Clement, J. A. "Effect of pH, DIG, Orthophosphate and Sulfate on Drinking
Water Cuprosolvency11, U. S. Environmental Protection Agency, Office of Research and Development,
Cincinnati, OH, (1995).
Dodrill, D. M. Lead and Copper Corrosion Control Based on Utility Experience. Master's Thesis, University
of Colorado, (1995).
van Gaans, P. F. M. "WATEQX - A Restructured, Generalized, and Extended FORTRAN 77 Computer Code
and Database Format for the WATEQ Aqueous Chemical Model for Element Speciation and Mineral
Saturation, for Use on Personal Computers or Mainframes". Computers & Geosciences, 15:6:843 (1989).
Schock, M. R., Lytle, D. A. & Clement, J. A. Effects of pH, Carbonate, Orthophosphate and Redox Potential
on Cuprosolvency. NACE Corrosion/95, Orlando, FL, (1995).
144
-------
35. Lytle, D. A. & Schock, M. R. Impact of Stagnation Time on the Dissolution of Metal from Plumbing
Materials. Proc. AWWA Annual Conference, Atlanta, GA, (1997).
145
-------
Figure 1. Photograph of coupon test loop system.
146
-------
I '
Upper Manifold
Nipple, 1/2 " HDPE
Check Valve
Nipple, 1/2" HDPE
Reducer Coupling, l/2"Xl/4", FVC
Labcock, 1/4", PVC, MPT X MPT
Labcock, 1/4", MPT
Tee, 1/4", PVC
Coupling, 1/4" MPT to 1/4" Tubing
-Sample Cell, Teflon
-Tubing, Teflon, 1/4"
- Stopcock, Teflon, 3-Way
-Tubing, Teflon, 1/4"
Rotameter, 1/4" Couplings
-Tubing, Teflon, 1/4"
-Connector, 1/4" FPT to 1/4" Tubing
• Reducing Nipple, 1/2" X1/4", HDPE
• Needle Valve, 1/2", PVC
Nipple, 1/2" HDPE
• Bottom Manifold
Figure 2. Schematic of individual test loop.
147
-------
1000
60
=i.
100 :
10
i—•—i—i—i—•—i—•—i
0 20 40 60 80 100 120 140 160
Elapsed Time, days
C36000
C36000
C83600
C84400
• C84500
• C85200
* C85400
* Pure lead
Figure 3. Lead leached from brass and pure
lead coupons during test run #1, pH =8.5.
148
-------
0 20 40 60 80 100 120 140 160
Elapsed Time, days
o C36000
• C36000
* C83600
- C84400
o C84500
• C85200
A C85400
* Pure copper
Figure 4. Copper leached from brass and pure
copper coupons during test run #1, pH =8.5.
149
-------
";• tj;"'; "i ^
10
I
1 -
0.1 -
0.01 ' '•'•'•'• r • '•'•'•
0 20 40 60 80 100 120 140 160
Time, days
C36000
C36000
C83600
C84400
" C84500
" C85200
* C85400
• Pure zinc
Figure 5. Zinc leached levels from brass and pure
zinc coupons during test run #1, pH =8.5.
150
-------
1000 r
tJ
^b
•a
Hi
H-}
0 20 40 60 80 100 120 140 160
Elapsed Time, days .
C36000
C
C
C84400
a C84500
• C85200
A G85400
» Pure lead
Figure 6. Lead leached from brass and pure
lead coupons during test run #2, pH =7.0.
151
-------
4.0 I—•—r
0 20 40 60 80 100 120 140 160
Elapsed Time, days
C36000
C36000
C83600
C84400
C84500
C85200
C85400
Pure copper
Figure 7. Copper leached from brass and pure
copper coupons during test run #2, pH =7.0.
152
-------
20
40 60 80 100 120 140 160
Elapsed Time, days
o C36000
• C36000
* C83600
- C84400
." C84500
- C85200
* C85400
* Pure zinc
Figure 8. Zinc leached from brass and pure
zinc coupons during test run #2, pH =7.0.
153
-------
10.0000
W)
i
1.0000
0.1000
0.0100
0.0010
Theoretical Solubility
Run 1 (T1014, pH=8.5)
Run 2 (TIO13, pH=7.0)
Experimental Data
Cu4(OH)6S04H2j>
• Runl
O Run 2
8
10
pH
Figure 9. Comparison of equilibrium theoretical and observed 24-hour stagnation copper levels for
coupon study, showing large deviation from expected value for system with high pH and high sulfate.
Thick line is predicted "langite" solubility.
154
-------
s
m
1
a
"8
I
r
I
fc.
*
S
OC9OOOC3OOOOOO
-H +1 -l-l +1 •« -H -H +1 -H' +1 -H 4-1
OOC3CDOOOOOOOCS
§g § 3
Sao*9
= -H
o
OOOO*-IOOO ., os" ° J.I
•H +1 +1 +1 +1 -H -f I -H II -1 +1 I
m
VO OO r-l f- Q VO
V» CO Q XO t-* « .-I
\O
o\Of-4vo<-iotnovo
ti~i«Mr4«noooo\
+1 +1 tl -H -H
I I g
mm
mcoca
fflpa
S SS
^
°B
tf^f 8
1^-8
I|ll
sSf^
u
3*
155
-------
Table Z Analytical Method. Uied far Chemical Analyrii of Water Sample*
AntlvvtA
* Met4*
Cakauen
""' Mafoextuni
.•''; Sodium
Pousoura
Iroo
Copper
Lead
Zinc
Manganese
Calcium
Mtfncasum
Sodium
Potuwum
Copper
lead
''' Hoc ,,
/• Manganese
StKooa(«lSiO,)
It,,'1 Sulftar(a»SO,-)
Aluminum
Iroo
Anioni
Chloride
Fluoride
Orthopbocphate
Total Fho^jbate
Nittate-:N
Sffieate
Sultite
Totai AOaBirity
Other*
Diatohred Oxygen
Temperature
Ammonia
Total loorf anie
Catfaoa
Total Chlorine
Free Chlorine
pH
Method
AA— Flame
AA-Flame
AA-Flaroa
AA-Flame
AA— Flame
AA-Flame
GFAAS
AA-Flame
AA-FUrae
ICAP
ICAP
ICAP
ICAP
ICAP
ICAP
ICAP '
ICAP
ICAP
ICAP
ICAP
ICAP
Automated Potentiometrie
Titration
Automated Standard
,:, ;, '. Addltioai ' ;
Poteatiaaetric KE
Automated Colorimetric
Automated Colorioetrio
Automated CotoniBetric
Automated Colorimetric
Automated Turtidimctric
Automated Fotentiometric
TitratioQ to EqiBvafeacePoint
"
Winkler(Aride Modification)
—
Automated Colorimetric
Coulotnetrie Titration
DPD CoJorimetric
PPD Colorimetric
Ooacd- Syflem Eteetroaetric
Method Number
7140
7450
7770
7610
7380
7210
7421
7950
7460
2017
200.7
200.7
200.7
200.7
200.7
200.7
2oa7
200.7
200.7
200.7
200.7
4SOO-Q- D.
—
340.2
I-2601-aS
1-2600-85
A303-5173-00
A303-5220-13
A303-5220
9038 '" .
2320B.4.6.
4500-0 D.
•
350.1
D513-92
8167
8021
— —
Reference
EPA'
EPA1
EPA1
EPA1
EPA1
EPA1
EPA1
EPA1
EPA1
EPA'
EPA"
EPA*
EPA'
EPA*
EPA*
EPA4
EPA*
EPA*
EPA'
EPA*
EPA*
Std. Method*7
Orion4
EPA1
USGS1
USGS*
Alpkem*
Alpkcm'
Alpkem'
EPA1
Std. Method*'
Std. Methods7
— — ,
EPA*
ASTM*
Hach'
Hactf
EPA(DWRD)"
Detection
Limit (m*/L) •
0.1
10
3.0
0.25
0,05
0.02
0.002
0.01
0.01
0.01
0.025
0.025
ZO
0.003
0.02
0.001
0.0004
0.053
0.045
0.025
0.002
1.0
.
<0.1
0.10
0.02 («P04)
O.OS(acPOJ
0.02 (asN)
0.4 (MSiO,)
~6.0(aaSO,)
~a3(a«CaCOJ
—
0.50
_-
0.03
f Uitekim EavbaiauaM S»ft«' EPA-«NU4-»-010(19M).
7'3>iwJ«4M.lVJ*l«li.Ea«i««io.rfWilKiUW«l«iNT,M8uBSli.
•imA«».lS»kWASTM S4.whnk-M0io.il. vnlwaH.ilWilM (I).
156
Kill
1
-------
<*•. ve ao
: cj
o o
§
DS
C9
V
VI
o o oo
V
•^ p
p
*V
^^ C^ C3 f^l f^ C*^ Cft 08 ^2 ^^
^^oqt^pp B Bpv^
»-<<»if^ooo «MOO
-N v v
o o o o
so oo
iri t-j
O O
V V
S §
V V
c4oo
vn S
V)
f
3
w
V
re
I
Su
s g w
g, y «« »
* f 5- » %£ ¥
?^8ig
1
g
I
rt
a
157
-------
Mixing in Distribution System Storage Tanks:
Its Effect on Water Quality
Robert M. Clark
Water Supply and Water Resources Division
National Risk Management Laboratory
Cincinnati, OH 45268
Farzaneh Abdesaken
DynCorp
Cincinnati, OH 45268
Paul F. Boulos and Russell E. Mau
Montgomery Watson
Pasadena, CA91101
158
::>• ..::i: ;i Jill,., i.,:;..:,-L,i: ;<•.,!• ";.i
-------
Introduction
Drinking water utilities in the United States (U.S.) have played a major role in protecting public health through
the reduction of waterborne disease. For example, in the 1880s for one year, the typhoid death rate was 158 deaths
per 100,000 in Pittsburgh, PA but by 1935 the typhoid death rate had declined to 5 per 100,000. These reductions
in waterborne disease outbreaks were brought about by the use of sand filtration, disinfection and the application of
drinking water standards (Clark, et al. 1991a).
Concern over waterborne disease and uncontrolled water pollution resulted in a dramatic increase in Federal water
quality legislation between 1890 and 1970. Even though significant advances were made in elimination of
waterborne disease outbreaks during that time period, drinking Water concerns began to change. By the 1970's more,
than 12,000 chemical compounds were known to be in commercial use with many more being added each year.
Many of these chemicals cause contamination of ground and surface water and are known to be carcinogenic
and/or toxic. The passage of the Safe Drinking Water Act of 1974 was a reflection of this concern.
The Safe Drinking Water Act of 1974 and its Amendments of 1986 (SDWAA) requires that the U.S.
Environmental Protection Agency (USEPA) establish maximum contaminant level goals (MCLGs) for each
contaminant which may have an adverse effect on the health of persons. Each goal is required to be set at a level at
which no known or anticipated adverse effects to health occur, allowing for an adequate margin of safety (Clark, et al.
1987). Maximum Contaminant Levels (MCLs), or enforceable standards, must be set as close to MCLGs as
feasible.
Most of the regulations -established under the SDWAA have been promulgated with little understanding of the
effect that the distribution system can have on water quality. However, the SDWAA has been interpreted as meaning
that some MCLs shall be met at the consumer's tap, which in turn, has forced the inclusion of the entire distribution
system when considering compliance with a number of the SDWAA MCLs, Rules and Regulations. Distribution
systems are generally designed to insure hydraulic reliability, which includes adequate water quantity and pressure for
fire flow as well as domestic and industrial demand. In order to meet these goals large amounts of storage are usually
incorporated into system design, resulting in long residence times, which in turn may contribute to water quality
deterioration. Many water distribution systems in the U.S. are approaching 100 years old and an estimated 26% of
distribution system pipe in this country is unlined cast iron and steel and is in poor condition. At current
replacement rates for distribution system components, a utility will replace a pipe every 200 years (Kirmeyer, 1994).
Conservative design philosophies, an aging water supply infrastructure, and increasingly stringent drinking water
standards are, therefore, causing concerns over the viability of drinking water systems in the U.S.
Storage tanks and reservoirs are the most visible components of a water distribution system but are generally
least understood in terms of their impact on water quality. Although these facilities can play a major role in
providing hydraulic reliability for fire fighting needs and in providing reliable service, they may also serve as vessels
for complex chemical and biological changes that may result in the deterioration of water quality.
This paper will examine the issue of residence time and the use of "compartment" models for describing mixing
regimes in tanks.
Previous Research on Tanks and Water Quality
Theoretical Studies
Danckwerts was one of the first investigators to discuss the concept of distribution functions for residence-times
and he explained how this concept can be defined and measured in actual systems (Danckwerts, 1958). When a fluid
flows through a vessel at a constant rate, either "piston-flow" or perfect mixing is usually assumed. However, in
159
-------
practice many systems do not conform to either of these assumptions, so that calculations based on them may be
inaccurate. Danckwerts illustrated the use of distribution functions by showing how they can be used to calculate the
efficiencies of reactors and blenders and how models may be used to predict the distribution of residence-times in
large systems.
Germeles has developed a model based on the concept of forced plumes and mixing of liquids in tanks
(Germelcs, 1975). He considered the mixing between two miscible liquids of slightly different density (<10%) when
one of them is injected into a tank partially filled with the other. A mathematical model for the mixing of the two
liquids was developed, from which one can compute the tank stratification. The model was also verified
experimentally.
Empirical Studies
Several investigators have conducted field studies and attempted to apply relatively simple models to distribution
storage tanks. Kennedy et al. (1993) have attempted to assess the effects of storage tank, design and operation on
mixing regimes and effluent water quality. The influent and effluent flows of three tanks with diameter-to-height
ratios ranging from 3.5:1 to 0.4:1 were monitored for chlorine residual. Chlorine levels were also measured within
the water columns of each tank. Although chlorine profiles revealed some stratification in tanks with large height-
todiarneter ratios, completely mixed models were more accurate than plug-flow models in representing the mixing
behavior of all three tanks. These investigators further indicated that the quality of the effluent from completely
mixed tanks deteriorated with decreasing volumetric change. The authors found that standpipes were the least
desirable tank design with respect to effluent water quality.
Studies conducted by Grayman and Clark (1993) have indicated that water quality degrades as a result of long
residence times in storage tanks. These studies highlight the importance of tank design, location, and operation on
water quality. Computer models, developed to explain some of the mixing and distribution issues associated with
tank operation, were successfully used to predict the effect of tank design and operation on various water quality
parameters. But because of the diversity of the effects and the wide range of design and environmental conditions, the
authors concluded mat general design specifications for tanks are unlikely. They also concluded that models will
most likely be refined and developed to facilitate site-specific analysis.
Infrastructure and Water Quality
Drinking water infrastructure is essential to public health, spurs productivity and profitability, increases the tax
base and creates skilled jobs in construction, engineering and manufacturing (Water Week, 1994). It is estimated that
57,000 jobs are created for every one billion dollars invested and yet this same infrastructure, if improperly designed
and maintained, can be detrimental to water quality.
Some of the factors that influence changes in water quality in the distribution system include: chemical and
biological quality of source water; effectiveness and efficiency of treatment processes; adequacy of the treatment
facility, storage facilities and distribution system; age, type, design, and maintenance of the distribution network; and
quality of treated water (Clark and Coyle, 1990). A factor, infrequently considered, that may influence water quality
in a distribution system is the effect of mixing of water from difference sources, which is a function of complex
system hydraulics (Clark* et al., 1991b; Grayman, et al., 19&8; Boulos, et al., 1994, 1995a). Since water
distribution systems frequently draw water from multiple sources such as a combination of wells, and/or surface
sources, this can be a major problem.
Recent experience has demonstrated that there may be problems with drinking water systems in the U.S. Table 1
summarizes some of the more notable recent water supply problems. Many more communities than those listed in
Table 1 have been placed on boil water orders or experienced coliform MCL violations over the last 3-4 years. In
160
-------
Milwaukee, WI, it is estimated that a cryptosporidiosis outbreak infected more than 400 thousand water consumers
and it is estimated that more than 100 immunocompromised people died. Boil water orders in Manhattan in New
York City and Washington, DC have drawn attention to drinking water problems in major metropolitan areas.
Two other recent examples of the importance of drinking water infrastructure in protecting public health occurred
in small communities in southeastern Missouri. Cabool, MO experienced a cross connection between sewage
overflow and two major distribution system line breaks caused by freezing temperatures, in December, 1989
resulting in a spread of E. coli 01571-1:7 infection in a town of nearly 2000 people. Six people died and 85
experienced bloody diarrhea. The town used a nondisinfected ground water (Geldreich, et al., 1992). Another
example of infrastructure failure occurred in Gideon, MO in November, 1993 when out of a population of 1000,
400-500 people contracted Salmonella Typhimurium and seven died. Bird droppings apparently contaminated a
storage tank. As with Cabool, the city used a nondisinfected ground water.
Storage Tanks
The direct impact of storage tanks on water quality was first documented during studies conducted by the U.S.
EPA in conjunction with the South Central Connecticut Regional Water Authority (SCCRWA). The first study
was conducted in the Cheshire service area and the second was conducted in the Brushy Plains service area of
SCCRWA.
During the Cheshire study, the fluoride feed was shut off and defluoridated water entering the system was used as
a water quality tracer. Initially during the sampling study, the wells were operated in order to minimize the variation
of water level within the Prospect Tanks which serve the Cheshire service area. After two days, this policy was
changed to allow a much greater variation of water leyel in the tanks. This policy was followed for the remainder of
the study (Clark, et al., 1991b).
The resulting fluoride profiles are shown in Figure 1. As can be seen, the difference between fluoride levels
entering and leaving the tank were very large. After approximately 7 days the levels of fluoride leaving the tank
approached the levels entering the tank. From this information, one could conclude that it would require
approximately 7-10 days for the tank to reach equilibrium as illustrated in Figure 1. During each period of tank
discharge, it required approximately 8 hours for the effluent fluoride level to reach a constant concentration level
indicating a significant amount of short circuiting.
A second study conducted in the Cherry Hill / Brushy Plains service area of the SCCRWA. As with the
Cheshire study, the fluoride feed was shut off at the Saltonstall plant, which supplies the system. Figure 2 shows
the results from the fluoride and chlorine sampling at the tank. High but decreasing concentrations of fluoride occur
on the discharge cycle and low concentrations prevail as the tank fills with water from the system. Figure 2 also
shows the chlorine residuals in the tank effluent. As can be seen the discharged water has very low residuals due to
the long residence times in the tank as implied from the previous study.
Model Development
In this paper compartment models will be used to characterize mixing in tanks. For purposes of model
development, we will consider two types of tanks. One is the traditional inflow and outflow storage tank in which
the water volume in the tank will expand or contract, depending on service demands. The Prospect and Brushy
Plains tanks mentioned previously are examples of this type of tank. The other type of tank is illustrated by a tank
located at the water treatment plant in Azusa, CA. It functions as a flow balancing mechanism and is used to impart
CT times to the water. Water flows in and out of the tank simultaneously.
Mass balance equations will be utilized to describe the mixing conditions in each of the tank models.
161
-------
Compartments will be assumed to represent different mixing conditions in the various segments that make up the
w^tcr volume in each tank (Mau, et al., 1995 c, b; Grayman and Clark, 1993).
For the traditional inflow and outflow tanks, the inlet and outlet pipes are near the bottom of the tank and
therefore water entering the bottom of the tank will remain at the bottom of the tank displacing older water present at
the beginning of the filling period. This general flow regime might be characterized as "first in", "first out".
The approach to compartmentalization discussed in this paper .is very similar to the approach suggested by Mau,
ct al., (1995a & b). However, Mau assumes steady state conditions for each inflow period (fluoride and inflow rate)
and for each outflow period (outflow rate). In this work the steady state assumptions have been relaxed and the actual
time varying values have been approximated by polynomials.
Tank Mixing Models
A compartmentalization approach will be used to represent different mixing zones in the
water column within a tank. For purposes of this analysis, a maximum of three compartments is assumed. Figure
3 illustrates the inflow and outflow compartmental configurations associated with classical inflow/outflow tank
systems. One (completely mixed), two and three compartment models are developed for the inflow/outflow and
continuously flowing tank in the following sections.
dt
=
Ql" "'
Inflow/Outflow Tank Model
' •
One Compartment Model
Inflow Conditions: Water is assumed to enter the tank with a flow rate of Chn and fluoride concentration of Qn.
is the concentration of fluoride in the tank and VT is the tank's volume.
:r dvT
T-+c* T-
r dt T dt M "'
Change in the tank volume is described by:
~Q,
, *Zin
dt
Solving equation (2) numerically, using Euler's Method, we obtain:
(2)
With the initial conditions of:
F//=0) = VTt
^r-^=Q,,lcl,,-CjQ,n
(3)
162
-------
By substituting equations (2) and (3) into (1) we obtain:
dt Vr
Outflow Conditions: Water flows out of the tank with a rate of Qout. Therefore,
' = -Q,,,,PT
dVT
(4)
dt
dCT
Vr~dt
Change in tank volume:
(5)
dt
dVr
_ !_=
dt
Therefore;
AH- 1) =
Qout(n)
Substituting (6) into (5) results in no change of fluoride concentration in the tank. Therefore,
dCr
dt
-=0
(6)
(7)
(8)
Two Compartment Model
Inflow Conditions: The volume of compartment A, VA, is fixed. Compartment B has a variable volume, VB.
Water enters compartment A with a flow of Qin, gpm and with fluoride concentration of Cin. Flow rate from
compartment A to B is indicated by QAB. Then inflow conditions for compartment A are as follows:
dt
Initial conditions are
Since compartment A has a fixed volume, we have:
dV.
dt
(9)
(10)
(11)
(12)
(13)
Therefore,
163
-------
••:' t"!» !.'•! '' I Bill
I" » '" "!' i. ' ' '"•?••'' '. «":S ' F'VIJI 'f [',
dt
(14)'
Inflow conditions for compartment B are:
d(CKVB)_
dt
_
^
But,
(15)
dt
(16)
164
-------
Therefore;
This implies:
Qin(CA-CB)
dt
(17)
(18)
dt
Outflow Conditions:
The mass balance equations for outflow from compartment B are:
where,
Therefore;
This implies:
&*.
dt
VB (n + '!) = VB(n) - h * Qout (n)
dCB_QBA(CB-CB)_
dt va(n>
B
The mass balance equations for outflow from compartment A are:
d(VACA)
— f*\ /*"* /*"! /""*
where QBA = gout, therefore,
dC. O (Ca-CJ
A_ xZ-ouP' B A'
dt ~ Vt
(19)
(20)
(21)
(22)
(23)
(24)
165
-------
Three Compartment Model
For this model it is assumed that the volumes of compartments A and C are fixed, volume of compartment B is
variable and a function of time, t, and there is a flow exchange between compartments B and C, namely QBC an Oec=QcB> an=VBM+h*Q!n
(n)
dt
The mass balance equations for inflow to compartment C are:
W
-------
Since volume of compartment C is fixed, then
dCc^Qsc(CB-Cc)
dt Vc
Outflow conditions for compartment B are:
dt
As before, volume of compartment B is variable and we have:
dV
dt
•=-QB
therefore,
which results in
M
dt
Finally, the mass balance equation for the outflow of compartment A is:
which implies,
dt
Flow Through Tank Model
(34)
(35)
(36)
(37)
(38)
(39)
(40)
As with the classical tank model, these compartments were utilized to represent the mixing regime in the flow
through tank model. Figure 4 illustrates the flow conditions associated with these assumptions.
One Compartment Model
In this case the tank is filling and emptying at the same time. Therefore, the change in the volume is a function
of the difference between inflow and outflow. This implies:
d(CAVA)
dt
dVt
(41)
(42)
167
-------
where,
dt
Integrating this equation numerically will result in:
Therefore,
Qia(C,-CA)
dt
(43)
(44)
(45)
T\vo Compartment Model
", i ' I'l: •" i
In this model, the volume of compartment A is fixed, therefore
we have:
and Qjn=QAB- Then for compartment A
dt
.(46)
The volume of compartment B is variable and is a function of inflow and outflow. Therefore, the change in fluoride
concentration is:
dt
-=QincA-Qou,cB
dVf
dt
-=Qln-Qaut
Numerical integration of this equation results in:
(47)
(48)
(49)
Substituting these formulae in the mass balance equation, we obtain:
dt V.M
(50)
Three Compartment Model
This model assumes constant and equal volumes for compartments A and C. Therefore, the inflow to compartment
A is the same as the flow form compartment A to B. For the same reason, flow from B to is the same as the outflow
from compartment C. Compartment B has a variable volume which is a function of the difference between inflow and
outflow.
For compartment A, we have:
4CAVJ
(51)
168
-------
Since dVA/dt=0, we have:
dCA Qin(Cin-CA)
dt
(52)
For compartment B, which has a variable volume, the conditions are:
dt
-=QABcA-QBCcB
On the other hand,
dVP
dt
z^in x-'out
which implies:
(n)
(nK
out >
YB ~~ TB
Substituting these values into the mass balance formula we obtain:
dCK^Q;n(CA-CB)
dt , V<">
B
For compartment C, with a fixed volume, the mass balance formula is as follows:
d(CcVc)
dt
-=QaMCB-Qoutcc
which results in:
dt
Qom(cR-cc)
K,
(53)
(54)
(55)
(56)
(57)
(58)
Model Validation
Data from field studies conducted at the SCCRWA and the City of Azusa were utilized to validate the models
discussed previously. Each of the systems and related field studies are described below.
SCCRWA
The SCCRWA is a large regional water system serving more than 100,000 customers in the New Haven,
Conn, region. The 16 service areas in eight pressure zones are fed by both surface water and ground water sources
(Clark, et al., 1991b). Though most of the system is tied together into two inter-connected systems, one service
area, the Cheshire area in the northern portion of the system, is largely independent and was used as the basis for the
study described.
The town of Cheshire is primarily a residential area. This service area is fed by two separate well fields: the
North Cheshire well field composed of four wells with a combined capacity of 3.6 mgd, and the South well field
169
-------
with two wells and a capacity of 2.5 mgd. Storage is provided by two adjacent tanks, Prospect tanks 1 and 2,
which float on the system, with 2.5 mil gal of storage each. The height of these tanks is 25 ft, and typically the
tanks operate in a range of 14 to 19 ft, with the operation of the wells manually controlled by an operator who
responds to the water levels in the tanks. Average daily demand is approximately 2.2 mgd during the winter and
slightly more in the summer. Water use does not vary significantly by day of the week. Highest water use occurs
during the morning and evening, with lowest water use occurring during the night.
Water entering and leaving the Prospect tanks was monitored frequently. An automatic sampler designed to
sariiple at a set interval (generally at a 1- or 2-hour interval) was attached to the outlet line from the tanks. A
continuous chart recorder monitored the flow in the combined inlet-outlet line. The variation between fluoride
levels entering and leaving the tank was very significant. Only after approximately seven days did the levels of
fluoride leaving the tank approach the levels entering the tank. From this information, it is concluded that it would
require approximately 10 days for the tank to reach an equilibrium.
The Cherry Hill/Brushy Plains service area covers approximately 2 sq mi in the town of Branford in the eastern
portion of tlie SCCRWA area. This service area is almost entirely residential, containing both single-family homes
and apartment-condominium units. Average water use during the sampling period was 0.461 mgd. The
distribution system is composed of 8- and 12-in. mains. The terrain in the service area is generally moderately
sloping, with elevations varying from approximately 50 ft mean sea level (MSL) to 230 ft.
Water is pumped from the Saltonstall system into the service area by the Cherry Hill pump station. Within the
Cherry Hill/Brushy Plains service area, storage is provided by the Brushy Plains tank. The pump station contains
two 4-in. centrifugal pumps with a total capacity of 1.4 mgd. The operation of the pumps is controlled by water
elevation in the Brushy Plains tank. Built in 1957, the tank has a capacity of 1 mil gal. It has a diameter of 50 ft, a
bottom elevation of 193 ft (MSL), and a height (to the overflow) of 70 ft. During normal conditions, the pumps are
set to operate when the water elevation in the tank drops to 56 ft and to switch off when the elevation reaches 65 ft.
City ofAzusa
The City ofAzusa Light and Water Department (City) purveys water to portions of the San Gabriel Valley in
Southern California. The City serves water to approximately 25,000 connections comprising an estimated
population of 80,000. In 1993, the City supplied 6.4 billion gallons of water to its customers from a combination
of treated surface water and ground water from wells (Boulos, et al., 1995b).
The Ed Heck Reservoir was recently constructed in the vicinity of the Canyon Filtration Plant in Azusa, CA.
The above-ground, circular reservoir is constructed of pre-stressed concrete. The interior diameter of the reservoir
measures 154 feet and the height measures 30 feet. The reservoir holds roughly 4 million gallons when full. City
staff anticipate the water level will normally be held between 15 and 27 feet above the reservoir bottom.
A 24-inch diameter inlet pipeline from the Canyon Filtration Plant Pump House feeds the reservoir. This
pipeline enters near the base and has a right-angle elbow at its end. This elbow directs water counter-clockwise
along the reservoir perimeter. The energy of the inflowing water is sufficient to cause, a visible rotation of the
reservoir water in the same, counterclockwise direction. The filtration plant is capable of producing 10 MGD.
A 30-inch diameter outlet pipeline carries water to the distribution system. This pipeline also lies at the
reservoir base, 5.25 feet clockwise of the inlet pipe. Distribution system operational practice makes it highly
unlikely that water would enter the reservoir from the outlet pipeline.
Fluoride Inflow/Outflow Model
A key feature in this analysis is to model for the changes in fluoride concentration in
the influent and to model the inflow and outflow rates for the tanks. Best fit
polynomials were utilized to describe these variations. Tables 2, 3 and 4 contain these
170
; i, „:",., i,: 1!
-------
polynomials.
Brushy Plains
Brushy Plains tank operates on alternating filling and emptying periods. The initial volume in the tank was
800,000 gallons. Initial fluoride concentration was 0.95 mg/L. In the three-compartment configuration, the
exchange rate between.compartments B and C was assumed as 10% of the inflow or outflow depending on the filling
or emptying period. Column 1 of Table 2 shows the inflow and outflow periods, column 2 contains the polynomial
for the fluoride entering the tank, and columns 3 and 4 contain the models for inflow and outflow in gpm. In the
three compartment model the exchange rates between compartments B and C were assumed to be 10% of the inflow
and outflow.
Prospect Tank
The Prospect tank is either filling or emptying. The initial volume of the tank was 2,961,788 gallons. The
initial fluoride concentration was 0.75 mg/L. In the three compartment model, the exchange rate between
compartments B and C was assumed to be 10% of the inflow, when filling, and 10% of outflow, when emptying.
Table 3 contains the equation describing inflow and outflow for the Prospect Tank. Figure 5 illustrates the use
of influence models for the Prospect Tank. Figure 5a shows the influent flow data and model for hours 129-140.
Figure 5b shows actual and modeled fluoride concentration for hours 13b-140, and Figure 5c shows actual and
modeled effluent flows for hours 140-150.
Azusa Tank
Azusa City tank operates on a simultaneous filling and emptying mode. The initial volume was 3,000,000
gallons with an initial fluoride concentration of 0.5 mg/L. Table 4 contains the models for each time period for the
Azusa Tank.
Solution Technique
Euler's Method was utilized to solve the differential equations numerically. This method produces an error of
order O(7z). To reduce this error, a small h=0.001 or 0.0001 was chosen. According to this method, at each step a
system of numerical solutions was obtained by solving:
where
y<"+1>=yM+h*f (^">, y<">) , rt=0,l...
y'(x)=f(x,y(x))
(59)
(60)
In this case, f(x(n), y(n)) would be the derivative of fluoride concentration or the mass
balance equation in each compartment (i.e., dCA/dt etc). Basically, final values at the end of each step was set as
initial values for the next step.
For one compartment model inflow, Qjn, outflow, Qout, and fluoride concentration entering the tank, fl, were
modeled as a continuous function of time. Then, at each time-step h, these functions were evaluated and/or updated.
There are two methods of operations:
a) Tank is either filling or emptying.
In this case we need to solve the equations in two parts:
Do over n = 1 to k;
171
-------
;v
>!* '!!'" !';
Evaluate Qtn(n):
Evaluate jl(");
Update V^n);
(61)
end;
Do over n= k+1 to m;
Evaluate Qoul(");
Update V/")";
b) Tank is filling an,d emptying simultaneously;
Doovern;
Evaluate Qit/");
Evaluate Q0,,J");
Evaluate JIM;
Update VA(")\
QinM*(f(n)-cAM)
(62)
end;
Qin^i Qou/")>flf")> Vjfn), andC^(") are the values of these functions at the nth step.
For two compartment model again we need to consider the working condition of the tank.
a) Tank is either filling or emptying. Then,
Do over n= I to k;
Evaluate Qjnf");
Evaluate fl(n);
Update
V M
VB
(63)
(64)
end;
Do over n = &+1, to m;
Evaluate Qput^
Update ¥&(");
Q M*( C.-Cjn))
X'nirt * R A '
(65)
end;
b) Tank is filling and emptying simultaneously:
172
-------
Do over n = 1 to k;
Evaluate Qjn(");
Evaluate fl(n);
Update VB(n);
Q | fot '"
CB =€„'">+h^-
Q.n<">*(CA-€„<">)
(66)
(67)
V <">
VB
The three compartment model has the same two cases as discussed before and:
a) Tank is either filling or emptying, with a flow exchange between compartments B and C at
all times. Flow from B to C, QBc> is a function of inflow when tank is filling, and flow between C and B, QBc, is
a function of outflow when tank is discharging. Therefore,
Do over n— I to k;
Evaluate QinM;
Evaluate Qout(n\
Evaluate fl(");
Update Vg(n);
M*( ftn>-C '"' )
(J A )
Q.n<">*(CA"'+1>-CnM)
CCf" + ')=CAW+/Z*^i
(68)
(69)
(70)
end;
Do over n = k+1 to m;
Evaluate Qout(") andQCB(n);
Update 7B(»);
V,.
v,M
(71)
(72)
173
-------
1' '"'I1 •" !«:' null'
C <*+'»=C "<>=
end;
Q
(73)'
b) Tank is filling and emptying simultaneously:
Do over n = / to k;
Evaluate Qin(") an
Update
t«*i)-
~
')— c tn)
cln>— c
Model Application
An objective of this analysis was to determine which compartment assumption best described the mixing
conditions for the three tanks. A variable was the percent of initial volume for each compartment in a given
"compartment model" For example, for the two compartment model in the Prospect Tank the assumptions ranged
from 10% for the first compartment and 90% for the second compartment, to 70% for the first compartment and 30%
for the second compartment. Calculated outcomes were regressed against the observed data from the effluent data and
the respective R-squares were calculated. The differences between the means predicted by each model were compared
to determine if they differed from one another. If they differed, then the model with the highest R-square was
assumed to be the best -model. If they did not differ, then the simplest model was assumed to be the best model.
Results from each tank are discussed below.
Prospect Tank
Table 6 contains the results of the analyses for the Prospect Tank. The one compartment model's best fit test
resulted in a R-square = 70%. For the two compartment model, several assumptions for the variable and fixed
compartments were considered. The solution started with 10% of the initial volume for the fixed part and 90% for
the variable portion. This amount was then increased by 10% for the fixed and decreased by 1 0% for the variable
compartment assumption. The 1O% and 2O% partitioning also did not result in a better fit than the one
compartment model, but the fit for 30% (30-70) partition improved and yielded an R-square = 82.6%. The 40% (40-
60) assumption had a very similar R-square of 82.7% but was not a major improvement over the 30% portion.
Three compartment models were run with 10% for each fixed and 80% for the variable compartment. This
amount was changed to 20% for the fixed and 60% for the variables compartment. The R-square for the first case
was 72.9%. The best fit test indicated that the 3 compartment model for the 20-60-20% partitioning was the overall
best fit with an R-square of 83.9%.
For this case, the three compartment model with a 20-60-20 portioning was assumed to be the "best" model.
This same appraisal was used for each tank.
Brushy Plains
174
-------
Table 7 contains results for the Brushy Plains Tank. The one compartment model yielded an R-square of 91.4,
indicating that the one compartment assumption was good.
The two compartment model was examined for various configurations, including 10-90 (i.e., 10% of the initial
volume for the fixed part and 90% for the variable part), 20-80, 30-70, 40-60, 50-50, 60-40, and 70-30 partitioning.
The 70-30 model resulted in the highest R-square of 91.96% and was assumed the best model.
For the three compartment models partitioning ranged from 10-80-10 to 30-40-30 (i.e., 30% of the initial
volume for the fixed compartments A and C and 40% for the variable compartment B.) The highest R-square was
89%, which indicated poorer fit than the one or two compartment models.
Although the one compartment model resulted in a good fit, the overall best fit model was the two
compartment model with 70-30 partitioning.
Azusa City Tank
Table 8 contains results from the Azusa City Tank.
' The best fit test for one compartment model resulted in R-square 94%, which indicates that one compartment
model is a very suitable configuration for this tank.
Several different partitionings for the two compartment model were examined. Partitions ranged from 10-90 to
60-40 for the fixed-variable compartments.
The 10-90 configuration which yielded in R-square = 90% was the best among the two compartment
partitioning.
For the three compartment model only two cases were examined: 1) 10% for each fixed compartments and 80%
for the variable compartment, 2) 15% for each fixed and 70% for the variable compartment. The first case resulted in
the better R-square = 91%.
It was concluded that the one compartment model was the best configuration for this tank.
Summary and Conclusions
Storage tanks and reservoirs are a major and necessary component of water distribution systems, however, they
may have a negative impact on water quality. These tanks may serve as complex chemical and biological changes
that may result in the deterioration of water quality. Key factors in this effect are residence times and mixing
characteristics.
In this analysis, compartment models were developed to provide this characterization. Clearly, there are an
almost infinite numbers of compartment modeling assumptions that could be applied to this type of analysis. For
purposes of this analysis, one, two or three compartment models were developed. It is clear that in some cases,
compartment models provide a superior representation of the mixing and residence times in tanks. It is, therefore,
important that this type of understanding be applied to further tank studies.
References
Boulos, P.P., Altaian, T., Jarrige, P.A. and Collevati, F. (1994) "An Event-Driven Method for Modeling
Contaminant Propagation in Water Networks." J. Applied Mathematic Modeling. Vol. 1-8, 84-92.
Boulos, P.P., Altman, T., Jarrige, P./A. And Collevati, F. (1995a). "Discrete Simulation Approach for Network
Water Quality Models.1 J. Water Resources Planning and Management. ASCE. Vol. 121, No. 1. 49-60.
Boulos, P.P., Grayman, W.M., Bowcock, R.W., Clapp, J.W., Rossman, LA., Clark, R.M., Deininger, R.A. and
175
-------
j l iiSl;;*'1 ....... !""i • ....... Ill -tW^' ' ''
' ..... it1!' "• ..... !' I ..... ...... : :i;"!i
Dhingra, A.K. (1995b). "Comprehensive Sampling Study of Storage Reservoir Water Quality: Characterization of
the Dynamics of Inner Mixing and Free Chlorine Residual Distribution." Accepted for publication in J. American
.Water Works Association.
W • ' 'Jlil .•,<•••:' „ ' ' " ' :
Clirk, R.M., Adams: JlQ., Miltner, R.M., "Cost and Performance Modeling for Regulatory Decision Making,"
Water. Vol. 28, No. 3, 20-27 (1987).
Clark, R.M. and Coyle, J.A., "Measuring and Modeling Variations in Distribution System Water Quality." Jour.
American Water Works Association. 82:46 (August 1990).
Clark, R.M., Ehreth, D.J. and Convery, J.J., "Water Legislation In the U.S.: An Overview of The Safe Drinking
Water Act" Toxicology and Industrial Health. Vol. 7, No. 516, 43-52 (1 991 a).
Clark, R.M., Grayman, W.M., Goodrich, J.A., Deininger, R.A. and Hess, A.F., "Field-Testing Distribution Water
Quality Models." Jour. American Water Works Association. 67-75 (1991 b).
Danckvverts, P.V., "Continuous Flow Systems." Chemical Engineering Science. Vol. 2, No. 1, 1-18 (Feb. 1958).
Gcldreich, E.E., Fox, K.R., Goodrich, J.A., Rice, E.W., Clark, R.M. and Swerdlow, D.L., "Searching for a Water
Supply Connection in the Cabool, Missouri Disease Outbreak of Escherichia Coli 0157:H7." Water Research. Vol.
26, No. 8, 1127-1137(1992).
Germeles, A.E., "Forced Plumes and Mixing of Liquids in Tanks." Fluid Mechanics Vol. 71, Part 3, 601-623
(1975).
Grayman, W.M. and Clark, R.M., "Using Computer Models to Determine the Effect of Storage on Water Quality."
Jour. American Water Works Association. Vol. 85, No. 7, 6777 (July 1993).
Grayman, W.M., Clark, R.M. and Males, R.M., "Modeling Distribution System Water Quality: Dynamic
Approach," J. WRPMD, ASCE, Vol. 114, No. 3 (May 1988).
Kennedy, M.S., Moegling, S., Sarikelli, S. and Suravallop, K. "Assessing the Effects of Storage Tank Design."
Jour. American Water Works Association. Vol. 85, No. 7, 78-88 (July 1993).
" " l!""l!i/ ••;••'. . ;;;,' ,';.••.•••
Kirmeyer, G.J., An Assessment of the Condition of North American Water Distribution Systems and Associated
Research Needs, Unpublished (1994).
!>il! ' ,; I,'"1 ; ...,':,,' >' '
Stevens, AA Moore, L.A. and Miltner, R.J., "Formation and Control of Non-Trihalomethane Disinfection By-
Products," Jour. American Water Works Association. 81, No. 8, 54-60 (August, 1989).
Water Week, Vol. 3, No. 10, 1 (1994).
176
-------
TABLE 1. SUMMARY OF RECENT WATER QUALITY PROBLEMS IN THE U.S.
CITY
Milwaukee, Wl
POPULATION DATE OF
AFFECTED ONSET
800,000 Apr. 7, 1993
CAUSE OF
PROBLEM
Gryptosporidiosis
RESULTS
Estimated 400,000
people ill and
6 people died
Washington, DC
Boca Raton, FL
Talent, OR
Carrollton, GA
Cabool, MO
Gideon, MO
New York, NY
(Manhattan)
Utica, NY
1,000,000
106,000
3,000
13,000
2,090
t
1,009
35,000
135,000
Dec. 8, 1993
Jan. 30, 1991
May 22, 1992
Jan. 30, 1987
Dec. 15, 1993
Nov. 29, 1993
JulyS, 1993
Nov. 1992
Turbidity
Violation
Coliform
Violation
Cryptosporidiosis
Cryptosporidiosis
E. coli
0157:H7
Salmonella
Typhimurium
Coliform
Violation
Coliform
Violation
Boil Water
Order
Boil Water
Order
3,000 people ill
13,000 people ill
6 people
died, 85 were
sick with bloody
diarrhea
400 people
ill, 7 people died
Boil Water
Order
State Confereno
Action Plan
177
-------
Table 2. Influent and Affluent Models for the Brushy Plains Tank
Time (hrs)
0-3
3-9
9-14
14-21
21-28
28-34
34-39
39-46
Fluoride In (mg/L)
O.OGx - O.OOOOSx3
-
7.3-0.2X
-
82 - 3x + O.OSx2
-
1 94.5 -6x + O.OSx2
-
Inflow (gpm)
0.07x3 - O.OOOOSx5
-
O.OOOSx4
-
0.4x - 0.000002X5
-
O.OOOOSx4
-
Outflow (gpm)
-
0.01x3
-
0.1 8x2
-
1597x-42x2 + 0.004x4
-
8(e-13)x8
178
-------
Table 3. Influent and Effluent Models for Prospect Tank
Time (hrs)
0-21
21-35
35-43
43-61
61 -80
80-98
98-114
114-129
129-140
140-150
150-162
162- 167
167-174
174-192
192-210
210-227
227 - 242
242 - 261
261 - 279
279 - 297
297 - 308
308 - 321
Fluoride In (mg/L)
1 1 - 0.7x + 0.02x2 - 0.0002X3
-
133-4x + 0.02x2
-
2(e-6)x3-2(e-10)x5
-
16-0.002x2 + (e-5)x3
-
3 - 9(e-5)x2
f
22 - 0.23x + 0.0006X2
-
2(e-8)x3
-
-4.3 + 0.02x
'
-90 + 0.7x- 0.001 x2
>
-4.8 + 0.02X
-
242 - 1 .5x + 0.002X2
Inflow (gpm)
789x - 49x2 +0.8x3 -7(e-5)x5
-
3106962-81385x+11x3-
0.001 Xs
-
0.14x3 - O.OOSx4 + 1.2(e-5)x5
-
10131904-194505X +
1049x2-0.01x4
-
-779239 + 91 28x-27x2
-
-91658 + 6790x1/2
-
-O.OOOSx3 + 2(e-6)x4
-
-285072 + 10x2 - 0.00009X4
-
-40357401 + 3491 Ox - 780x2
+ 3(e-11)x7
-
-23608794 + 135897x - 0.7X3
+ 1.2(e-11)x7
-
-4051835 + 24001 x - 36x2
-
Outflow (gpm)
-
8385 - 3x2 + 5(e-6)x5
-
-561047 + 12262x-x3 +
4(e-5)x5
•
2732x - 44x2 + 0.2x3
-
-653258 + 8373x - 27x2
-
-11 58074+1 2994x-36x2
-
-1851281+14050x-0.1x3
-
-15256431 + 132728X-
2x3 + 1.2(e-5)x5
-
-38452981 + 379978x -
1001x2+6(e-6)x5
-
-460040 +2459X- 0.01 x3
-
-2790770+1 7335X - 27x2
-
2(e-6)x4 - 6(e-9)x6
179
-------
Table 4. Influent and Effluent Models for the Azusa Tank
Time
(hrs)
0-1.2
1.2-2.8
2.8-5
5 - 6.8
9 - 9.4
9.4-10.2
10.2-12.4
12.4-13
13-13.6
13.6-14.8
14.8-16
16-17
17-18.2
18.2-19
19-20.2
20.2 - 21
21-22
22 - 24.6
24.6 - 26
26-28
28-30
Fluoride In (mg/L)
0.5
-693+1 48x-7.9x2
0.004X4- 0.0007x5 + 3(e-5)y?
O.OSx2 - O.OOSx3
0.002X3 - O.OOOOSx4
-283 + 48x - 3x2 + 0.09X3 -
0.0009X4
0.0002X3 -0.000005X4
Outflow
(gpm)
10298-1 581 8x+11628x2
47259X - 6391 3x2+ 30401x3-
4788X4
13421 959 -9094333X
+2302515X2 -258071X3 +
10805X4
13421 959 -9094333X +
230251 5x2- 258071 X3+ 10805X4
89x4-18x5+0.92x6
37298901 -8032286X + 107453x3
- 7207X4 + 7.7x6
1472133 - 225809X + 8689x2
1677445 - 235757X + 8309x2
1757327 - 228032X + 741 9x2
2990031 - 362294X + 10993x2
2197368 - 248943X + 7069x2
8308758 - 891234X +23908x2
2985626 - 306539x + 7880x2
7260595 - 70461 2x + 17108x2
51 94822 - 481 960x +111 91x2
1(e-6)x8-4(e-8)x9
O.OOOOOOSx7
50158565 - 49498 12x +
1 37361 x2- 31 x4
O.OOSx4
Inflow
(gpm)
5923 - 0.7x3 +
0.09x4 - 0.004X5 +
9(e-5)x6-1(e-6)x7
+ 4(e-9)x8
180
i L:
-------
Table 4. (Continued )
Time
(hrs)
Fluoride In
(mg/L)
Outflow
(gpm)
Inflow
(gpm)
30-31
1(e-5)x4-3(e-7)x5
31 -32
32-34
34 - 35=8
35.8 - 37.2
37.2 - 39.2
39.2 - 40.2
40.2-41.2
41.2-42.2
42.2 - 43
43-44
44 - 44.8
44.8 - 45.8
45.8 - 46.8
46.8 - 47.8
47.8 - 48.8
48.8 - 50.4
50.4 - 52.6
52.6 - 53.8
53.8 - 54.6
54.6 - 55.6
55.6 - 56.6
56.6 - 57.4
57.4 - 59.2
59.2 - 60.8
0.5
10793247 - 707438X + 11598x2
12431154 - 788490X + 12509x2
13004943 - 799565X + 12295x2
2(e-6)x7 - 5(e-8)x8
-0,001 Sx5 + 0.00004X6
353x2 - 0.02X5 + 6(e-6)x7
21820731 - 1099327X + 13848x2
23337896 - 1147050x + 14096x2
23020403 - 1104494X + 13250x2
37758587 - 1772876x + 20812x2
2177644 - 1005085X + 11546x2
48100956 - 2165256X + 24368x2
25101748 - 1107189X + 12210x2
24728874 - 1068258x + 11538x2
27900434 - 1182419x + 12529x2
22150451 -915628x+9464x2
12772259 - 513661x + 5165x2
1(e-10)x8
5923 - 0.7x3 + 0.9X4 - 0.004x5
9(e-5)x6 -1 (e-6)x7 + 4(e-9)x8
4x4-0.12x5+1.4(e-5)x7
22223081 - 604964X + 66x3
9.3(e-13)x9
2.6(e-9)x7
0.0005X4
2(e-9)x7
5923 - 0.7x3 +
0.09X4 - 0.004x5 +
9(e-5)x6-1(e-6)x7
+ 4(e-9)x8
181
-------
Table 6. Best Fit Results for the Prospect Tanks
Model
1 Compartment Model
2 Compartment Model
f
3 Compartment Model
10%-90%
20%-80%
30%-70%
40%-60%
10%-80%-10%
20%-60%-20%
R-Square
70%
68.7%
77.9%
82.6%
82.7%
72.9%
83.9%
182
-------
TABLE 7. Best Fit Results For Brushy Plains Tank
Model
1 Compartment Model
2 Compartment Model
f
3 Compartment Model
10%-90%
20%-80%
30%-70%
40%-60%
50%-50%
60%-40%
70%-30%
10%-80%-10%
15%-70%-15%
20%-60%-20%
25%-50%-25%
30%-40%-30%
R-Square
91.4%
68.9%
84.4%
89.4%
90.9%
91.6%
91.9%
91.96%
69.6%
79.8%
85.2%
88.2%
89.6%
183
-------
"':;"T' ! U?!1! I1"1
TABLES. Best Fit Results For Azusa City Tank
Model
1 Compartment Model
2 Compartment Model
f
3 Compartment Model
10%-90%
20%-80%
30%-70%
40%-60%
10%-80%-10%
15%-70%-15%
R-Square
94.3%
90%
89.5%
88.7%
88.1%
91%
90%
184
-------
CD
c?
I
OT
O
- o»
CM
CO
CO
8
£
CO
O
I
o
o
o
o
CD
5
o
01
I
D)
IE
1 "1 —
CO
0
1
o
I
0
1
w
o
o
1/Biu uo|jBJiu«ouoo
185
-------
1.2-
1 *~
0.8-
5^
o>
E 0.6-
d '
c: •
° 0.4-
(D
c
•e
1 0.2-
0
-
C
1 -
0.8-
|
6 04-
8
•i
•c 0.2-
o
ul
n _
.'•
•
|l
f
A
T
* •
V
1 1 1 1 1 —
) 5 10 15 20 25
Jm'X
" f
I 1
1 1
4
fc
1
f
1
H
Discharging I
•-M Filling V
r
,T '
1 •*
mglm T
jr |
i *
f F
1 «
1
1
I
i
1
]
I
V
30 35 40 45 50 55 60 65
•
^.-A.
"**"" Y
t f** *^| I
in
i
i .
• u
0 5 10 15 20 25 30 35 40 45 50 55 60 65
Time • hour
Figure 2. Chlorine and Fluoride Residuals at the
Brushy Plains Tank
186
-------
3a. Single Compartment Model
Qin
3b. Two Compartment Model
3c. Three Compartment Model
Figure 3. Compartment Model Configurations for Single Reactor Tanks
187
-------
4a. Single Compartment Model
Qoul
', 'f - 1;! ., ,i'::-;v .'j'.-
4b. Two Compartment Model
4c. Three Compartment Model
Figure 4. Flow Through Tank Model
188
.'"{» fj-ffij "i i, '. . j1.. :, i; ,';,;.;' i ,.' 'n.u'il'.f ; .• i..".,' hi .!.'f ' '''"' It.'« '. ":' i.. i;-li,:..iilli&lli j Uii fiill''!"
-------
5a.
O.M
O.W
04ft
0.90
140 Mi
Ttnw-hauit
5b.
ISO
5e.
4000,
J 30»
2000
ton
tSO
155
160
Figure 5. Fluoride Concentration, Inflow
and Outflow Models for Prospect Tank
189
-------
Modeling The Kinetics Of Chlorination By-Product Formation:
The Effects Of Bromide
Robert M. Clark and Ronald C, Dressman
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Cincinnati, OH 45268
Hossein Pourmoghaddas
University of Medical Sciences
Esfahan, Iran
Larry J. Wymer
DynCorp. Inc.
Cincinnati, OH 45268
Abstract
'» n 'I1 ii| . : • • !'' „':''• ,
Using a data set generated by Pourmoghaddas, et al. a model for characterizing the formation of THM and non-
THM chlorinatlon by-products and their speciation was developed. The model which takes into consideration pH,
tpie and chlorine and bromide concentration shows the effect of bromide concentration on CHCIs and CHBra
formation. The model demonstrates that the concentration of CHB^CI and CIBrCl2 increases to a maximum level
for bromide concentrations of 2.5 mg/L and 0.5 mg/L respectively and then declines with increasing bromide
concentration. Concentrations of CHCIa decline with increasing bromide concentration and CHBr3 increases with
increasing bromide concentration.
190
-------
Introduction
Chloroform and other trihalomethanes have been reported in finished drinking water for over 20 years. !>2=3 Based
on available toxicology data, chloroform was determined to be a carcinogen in mice and rats at high dose levels.
Because of suspected health effects, in November of 1979, the U.S. Environmental Protection Agency (U.S. EPA)
proposed to limit the concentration of total trihalomethanes (TTHMs) in distribution systems serving 10,000 or
more persons (4 years after promulgation) to 0.10 mg/L (100 |J.g/L).4
While research was being conducted on TTHM formation, the formation and control of other disinfection by-
products (DBFs) was also being studied. Many of these compounds (primarily formed during chlorination) are
found at (o.g/L concentrations and lower, and the majority of these compounds have not yet been identified.5.6
A major concern is to control chlorination disinfection by-products, which requires an understanding of the
factors that influence their formation including pH, time, temperature, chlorine concentration and the concentration of
organic and inorganic matter. Another factor affecting disinfection by-product formation is the presence and
concentration of the bromide ion in the raw or finished water. The absence or presence and concentration of some
organic by-products of chlorination depends on the bromide ion concentration in the water. Bromide is oxidized to
bromine and substitutes for chlorine to produce bromine-containing homologs of the more familiar chlorine species.
Once bromide ion is present in source waters, there are no known economical treatment techniques available for
removing it. The current list of by-products targeted for regulation contains brominated and mixed bromine-chlorine
species of TTHMs and haloacetic acids (HAAs). These are known to form in bromide-containing waters when
chlorinated.
In order to test the hypothesis that bromide ion was a major factor influencing the formation of disinfection by-
products, Pourmoghaddas et al. conducted a series of experiments specifically designed to accomplish the following
objectives:?
1. Investigate the effect of bromide ion concentration on the formation and speciation of certain chlorination
by-products other than THMs.
2. Identify the formation of haloacetic acids containing bromine and chlorine under different bromide ion
concentrations and quantify them.
3. Evaluate the relationship of these organic by-products with TOX
4. Determine some of the conditions required to control the formation of non-THM chlorination organic by-
products resulting from the" disinfection process with
This paper presents an extension of the work by Pourmoghaddas et al. and develops a model for characterizing
the formation of THM and non-THM chlorination by-products and their speciation.
Methods And Experimental Design
The experimental data used in this analysis was derived from a study conducted by Pourmoghaddas, et al.'
The study was performed in two sample blocks. In the first block, a humic acid (HA) solution containing 2.90
mg/L HA measured as non-volatile total organic carbon (NVTOC), was treated with a high chlorine dose of 25
mg/L. For the second block containing 2.83 mg/L NVTOC, a lower chlorine dose of 11.5 mg/L was used. The
independent variables were pH, reaction time and bromide ion level. The-three levels of pH used were 5, 7 and 9.4.
The three reaction times were 6 hours, 48 hours, and 168 hours. The three bromide levels studied were 0.5, 1.5,
4.5 mg/L as Br. Also, the tests at each pH and reaction time were run without bromide ion being present. All of
the tests were conducted at 25°C. A two-block, full-factorial design was used to statistically evaluate experimental
results. The use of a full-factorial design allowed for the effects of each variable to be evaluated for accuracy and
precision. The factorial design further allowed for detection of the main and interaction effects of the variables. The
factorial design incorporated one factor (Br) at four levels and two factors (time, pH) at three levels for each block. A
computer program (SAS) was used for statistical analysis.
191
-------
For block I, high-chlorination-dose (25-mg/L) samples, lodometric Method 1 was used to measure the residual
chlorine.8 For block 2, low-chlorination dose (1 1.5-mg/L) samples, a DPD/FAS titrimetric method was used to
measure the residual chlorine.8
TTHM analyses were performed by EPA Method 501.2; TOX analyses by EPA Method 450.1; HAAs by EPA
Method 552.9 Special investigations were conducted to extend the method of analysis to all nine HAAs (Table 1),
verify their stability, account for the contamination of certain of the HAA reference materials with other HAAs, and
to establish the recovery efficiency of the HAAs by the TOX method. The latter was necessary to be able to
properly relate TOX measurements to the sum of individual compound measurements.
Disinfection By-Product Model
Most modeling efforts related to the formation of disinfection by-products have focused on TTHM formation.
Kavanaugh et al. has suggested an empirical kinetic model of TTHM formation in the format of a differential
equation.10 Engerholm and Amy found that formation of chloroform from humic acid under different conditions of
pH, temperature, precursor concentration and chlorine-to-Non Purgeable Organic Carbon (NPOC) ratio could be
accurately modeled by transforming both dependent and independent variables into natural log form.u.iz Other
investigators have focused on the effect of bromide ion on the formation of disinfection by-products.13>14>15>l6,!7
Although not a TTHM formation model, Quails and Johnson developed a two-component reaction equation for the
utilization of free chlorine by fulvic acid.18 Haas and Karra have proposed a two-phase model for combined chlorine
demand. 19 For purposes of this analysis, we will assume the following equation for DBP formation:
bB M products
'
(1)
'
. . .
in which Z represents the concentration of chlorine reacting with precursor B to form disinfection by-products and a
and b are coefficients. Another model might be:
aZ + bT M I
cZ + dl M P
(2)
(3)
In this model chlorine acts on existing precursor to form intermediate products (I), which in turn are acted upon
by chlorine to form disinfection by-products (P) where c and d are coefficients.
For purposes of the analysis discussed here, the following relationship will be assumed:
eR + fCI2 + gBr M DBP
(4)
in which R represents precursor, CI2 and Br represent chlorine and bromide ion respectively, DBP the products of
the reaction, and e, f and g are coefficients. We will also assume that the increase in formation of DBP is governed
by the following equation:
DBP - DBPu - DBPtu (5)
i . • • ' •• v- . ••.. . •
In Equation 5, DBPU is the ultimate formation potential of an individual disinfection by-product, and DBP*U
represents the remaining formation potential at time t. We also make the first order assumption:
d DBP'u / dt = -kDBPtu (6)
where
DBP - DBPU (gj t = 0 (7)
192
-------
Therefore integrating Equation 6 yields:
DBF = DBPU (1 - e-kt) (8)
Equation 8 will be used to model the formation of individual DBFs in this study.
The first step in estimating parameters for equation 8 is to estimate DBPU, which is the maximum formation
potential for individual disinfection by-products. After considering a number of possibilities, the following
relationship was obtained:
DBPU=
where
H = concentration of Br or CI2
P = MCI2/(MCI2 + Mbr)
Mbr = moles of bromide ion
MCI2 = moles of chlorine
In equation 9, H is the concentration of Br in the purely brominated and mixed species and H is the
concentration for CI2 in the purely chlorinated compounds.
The reaction rate k in equation 8 was determined as follows:
k- = T 4- T P
k 1Q+ ijf
P was defined previously
TO, T, = parameters to be determined
The final form of equation 8 is as follows:
DBF =J3QHflipP2
(9)
(10)
(H)
Parameter Estimation
One method for estimating parameters in equation 1 1 is least squares regression; i.e., estimating values of the
parameters so that the sum of the squared differences between the actual values and the predicted values of the
response (called the sums of squares for error, or SSE) is minimized. For models that can be represented using
matrix algebra, this is a relatively simple calculus problem for which numerous computer packages are available.
However, the above model cannot be represented by linear algebra, i.e., is nonlinear. For this reason an iterative
method of estimating the parameters was used.
One of the most common algorithms for finding parameter estimates for nonlinear models is the Gauss-Newton
method. This method starts with initial estimates of the parameters obtained either through previous knowledge or
a preliminary search, and uses the estimates and the partial derivatives of the model with respect to the various
parameters to compute a Taylor series approximation to produce revised estimates of the parameters. This process
is repeated each time using the new estimates as starting values, until a convergence criterion is satisfied. The
criterion can be either a lack of change in values of the parameters or in the SSE.
The method of steepest descent is another method based on the .utilization of the partial derivatives of the
model. The gradient of the SSE considered as a response surface is calculated and used to determine the direction
(in relation to the parameters) for which SSE decreases. The parameter estimates are then updated and the process is
193
-------
repeated until the convergence criterion is met.
A compromise between the two above methods is the Marquardt method.20 A coefficient lambda determines
whether the method acts like Gauss-Newton or steepest descent; if lambda is zero, Marquardt is equivalent to Gauss-
Newton. As lambda approaches infinity, the Marquardt method behaves like the method of steepest descent.
Lambda is usually adjusted at each iteration, the exact amount depending on the circumstances.
After it was decided to use the Marquardt compromise,20 values for DBPU and k were obtained separately for
each treatment combination by using a grid search in conjunction with Marquardt's method. Each of the compounds
listed in Table 2, as well as various combinations of the compounds, including total trihalomethanes and total
halides, were modeled. Table 2 summarizes the models parameter estimate for a two-block factorial design. The
data points that were modeled for the individual compounds are single-value points, except in those few cases where
replicates were analyzed for quality control purposes. The data points for TOX are the average of duplicate results.
All other data points are (See Table 2) a sum of their component values.
A sensitivity analysis was performed on models thus estimated by omitting the most influential observations
from each dataset and reestimating the model. Most influential observations were determined by means of their
leverage. In general, an observation which has extreme values for the set of independent variables and which is
highly under- or over-predicted by the model will-have a high leverage, thus being most influential on the parameter
estimates. Even when such observations were omitted and the models reestimated? parameter estimates were found
to differ by at most five percent from the estimates using the full dataset. The authors believe that these results
indicate mat resulting models are "stable."
Discussion
Previous research has shown that in the presence of Br and a constant average free chlorine residual,
hypochlorous acid (HOCI) oxidizes the Br, and hypobromous acid (HOBr) is formed.21 Since HOBr is a faster
halogenating species than HOCI, more organic material becomes reactive and more TTHM is produced, on a molar
basis. It has been found that the variable controlling the bromine substitution reaction, in TTHMs, is the initial
Br/average CI* molar ratio. Higher ratios produce more bromine incorporation. The bromine substitution reaction
tends to go to completion first. When excessive amounts of HOCI are available and after the bromine substitution
reaction is complete, if some HOCI remains, TTHMs will continue to form. However, in the absence of any excess
HOCI, virtually no TTHM will form.
Figure 1 illustrates the effects described above. Figure la, Ib and Ic show that as Br concentrations increase,
levels of CHClj decrease, while CHBra increases, indicating higher bromine incorporation at higher bromine levels.
Figures 2a, 2b and 2c show that as Br levels increase, the species formed tend to have higher bromine incorporation.
These figures also show that at higher levels of chlorine residual, more Br is oxidized leading to higher bromine
incorporation. In Figure Ic, the formation curve for C12=11.5 mg/L and C12=25 mg/L lie on the same line.
Figures 3 and 4 compare best fit lines to the experimental data. From Figure 3, it can be seen that CHCIj
levels tend to increase with increasing chlorine residual levels. Figure 3 also illustrates the dramatic impact of Br
concentration on CHCIs formation. Higher Br levels clearly reduce both the rate and ultimate formation of CHCIs
For Br levels of 4.5 mg/L, the formation line lies along the horizontal axis. Figure 3 also shows that higher pH
results in higher levels of CHCl^. Figure 4 shows that as chlorine residuals increase, the rate and ultimate formation
levels of the brominated compounds increases. Figure 4 also illustrates that as Br levels increase, the concentration
of the compounds with higher levels of Br incorporation increase. In these figures pH has little impact on the
formation level or rate of the brominated compounds.
Summary And Conclusions
The regulation and control of disinfection by-products is of major interest to the regulatory community and to
utilities in the United States. In order to control the formation of disinfection by-products, understanding the various
factors that effect their formation is imperative.
• 194
-------
The capability to characterize both the magnitude of formation and formation rate for disinfection by-products in
the presence of various combinations of these factors would be of inestimable value. Few models have been
developed to address this need, and none that incorporate the effect of the bromide ion.
Using data from Pourmoghaddas, the formation rate and formation potential for a wide range of DBFs have been
modeled. This model incorporates pH, chlorine concentration and bromide concentration as variables, all of which
affect both the rate of formation and the ultimate concentration of the compound formed.
The data generated in this experiment were based on the use of super Q water containing commercial humic acid
as the model system. Although the model is not representative of a real system it does demonstrate the kinetics of
the various compounds without interferences. On-going research is being conducted to quantify the effect of raw
water TOC and bromide on the rate of formation and the final concentration of various by-product species formed.
This work demonstrates that modeling to take into account multiple synergistic factors is a complicated and
difficult task that may require a considerably larger body of data than was available to this work, but which, in spite
of the lack of data, is promising. Also, because the factors considered in this work do not effect all of the
compounds tested in the same way, one general model for predicting results for all compounds under the influence of
bromide ion is not possible.
Acknowledgments
The authors acknowledge Ms. Diane Routledge, Ms. Jean Lillie and Mr. Steve Waltrip of the Water Supply
and Water Resources Division for their assistance in preparing this manuscript. The authors would like to extend a
special acknowledgment to Dr. Paul Ringhand and Dr. Hiba Shukairy of the Water Supply and Water Resources
Division for their insightful suggestions.
References
1. Rook, R. J., "Formation of Haloforms during Chlorination of Natural Waters", Water Treatment Exam.
23(2):234-243, (1974).
2. Bellar, T. A., J. J. Lichtenberg, and R. C Kroner, "The Occurrence of Organohalides in Chlorinated Drinking
Water," JAWWA 66(12):703-706, (1974).
3. Symons, J. M., Bellar, T. A., Carswell, J. K., DeMarco, J., Kropp, K. L., Robeck, G.G., Seeger, D. R.,
Slocum, C. J., Smith, B. L. and Stevens, A. A., "Natural Reconnaissance Survey for Halogenated Organics,"
JA WWA, 67(11):634, (Nov. 1975).
4. Symons, J. M., Stevens, A. A., Clark, R. M., Geldreich, E. G., Love, 0. T., Jr. and DeMarco, J., Treatment
Techniques for Controlling Trihalomethanes in Drinking Water, EPA/600/12-81/156, Municipal
Environmental Research Laboratory, Cincinnati, Ohio 45268, Sept. 1981.
5. Stevens, A. A., Moore, Leown A., and Miltner, R. J., "Formation and Control of Non-Trihalomethane By-
Products."
6. Stevens, A. A., Moore, Leown A., Slocum, Clois, J., Smith, B. L., Seeger, D. R. and Ireland, J. A., "By-
products of Chlorination at Ten Operating Utilities," presented at the Sixth Conference on Water Chlorination:
environmental Impact and Health Effects, May 3-8, 1987, Pollard Auditorium, Oak Ridge, Associated
Universities, Oak Ridge, Tennessee.
7. Pourmoghaddes, H., Stevens, A.A., Kinmen, R.N., .Dressman, R.C., Moore, L.A. and Ireland, J.C., "Effect of
Bromide Ion on HAAs during Disinfection With Chlorine." JAWWA, Vol. 85, No. 1, Jan. 1993.
195
-------
8. APHA, AWWA, WPCF, (1985) Standard Methods for the Examination of Water and Wastewater, 16th
Edition, Part 4500—C1I (Idometric Electrode Technique), Part 4500-C1I (DPD Ferrous Titrametric Method).
9. EPA Method 501.2, "The Analysis ofTrihalomethanes in Drinking Water by Liquid/Liquid Extraction;" EPA
Method 450.1, "Total Organic Halide;" EPA Method 552, "Determination of Haloacetic Acids in Drinking
Waiter by Liquid/Liquid Extraction, Derivatization, and Gas Chromatography with Electron Capture
Detection." Environmental Monitoring Systems Laboratory, Office of Research and Development, Cincinnati,
OH 45268. " ' ,
vi if; ' '•.'•...,,..
ill , : ' ' i, • ,'.'•• '.' . "
10. Kavanaugh, Michael C., Trussell, A. R., Croiner, J., and R. Rhodes Trussel, "An Empirical Kinetic Model of
Trihalomethane Formation: Application to Meet the Proposed THM Standard," JAWWA, October 1980, pp
578-582.
11. Engerholm, B. and Amy G., "Predictive Model for Chloroform Formation from Humic Acids" JAWWA
7S(8):418,(Aug. 1983).
12. Engerholm, B. and Amy G., "An Empirical Model for Predicting Chloroform Formation for Chloroform
Formation from Humic and Fulvic Acid," Water Chlorination: Environmental Impact and Health Effect, Vol.
4 (R. Tolley Editor) Ann Arbor SciJPubl., Ann Arbor, Mich. (1983).
13, Minear, R.A. and Bird, J.C., "Trihalomethanes: Impact of Bromide Ion Concentration on Yield, Species
Distribution, Rate of Formation, and Influence of Other Variables," in Water Chlorination: Environmental
Impact and Health Effects, Vol. 3, Jolley, R. L., Brungs, W.A.,, and Cummings, R.B., Eds. (Ann Arbor, MI:
Ann Arbor Science Publishers, Inc., 1978), pp. 151-160.
14, Oliver, G.B., "Effect of Temperature, pH, and Bromide Concentration on the Trihalomethane Reaction of
CJilprine with Aquatic Humic Material," in Water Chlorination Environmental Impact and Health Effects,
Vol 3, Jolley, R. L., Brung, W.A. and Cummings, R. B., Eds. (Ann Arbor, MI: Ann Arbor Science
Publishers, Inc., 1978), pp. 141-149.
15, Cooper, W.J., Meyer, L.M., Bofill, C.C. and Cordal, E., "Quantitative Effects of Bromine on the Formation
and Distribution ofTrihalomethanes in Groundwater with a High Organic Content," in Water Chlorination
Environmental Impact and Health Effects, Vol. 3, Jolley, R.L., Brung, W.A. and Cummings, R.B., Eds.
(Ann Arbor, MI: Ann Arbor Science Publishers, Inc:, 1978), pp. 285-296.
i i:"1 , ' i . • '• " , liii" • • ,• r"'.,' '!•,!• ," -. • ::
16. Amy, G.L., Chadik, P.A., Chowdury, Z.K., King, P.H. and Cooper, W.J., "Factors Affecting Incorporation of
Bromide into Brominated Trihalomethanes During Chlorination," in Water Chlorination Chemistry,
Environmental Impact and Health Effects, Vol. 5, Jolley, R.L., Bell, W.P., Davis, W.P., Katz, S., Roberts,
M.H., Jr., Jacobs, V.A., Eds. (121 S. Main St., P.O. Drawer 519, Chelsea, MI 48118: Lewis Publishers, Inc.,
1985).
, •',, • i: ;i , • ::•.•.• • . ' . • ' ';; '.• '" ,
17. Amy, G.L., Chadik, P.A. and Chowdury, Z.K., "Developing Models for Predicting Trihalomethane Formation
Potential and Kinetics." JAWWA, July 1987, pp. 89-97.
18. Quails, R. G. and Johnson, J. D., "kinetics of the Short-Term Consumption of Chlorine by Fulvic Acid."
Environ. Sci. Techn..s Vol. 17, No. 11, 1983, pp 692-695.
19. Haas, C. R and 'Kami, S. B. "Kinetics of Wastewater Chlorine Demand Exertion," JWPCF, 56(2): 120-173,
1984. _ ,' " _ ' .. ^ , _ ."' : ;, i;
20l SAS Users Guide: Statistics, Version 5 Edition, Copyright 1985 by SAS Institute, Inc., Gary, NC, USA, p.
586.
21. Symon, J.M., Krasner, S.W., Simms, L.A. and Sclimenti, M., "Measurement of THM and Precursor
Concentrations Revisited: The Effect of Bromide Ion." JA WWA, Januaiy 1993, pp. 51-62.
196
-------
TABLE I. Proposed Haloacetic Acid (HAA) Target Compound List
HALOACET1C ACIDS fHAAs):
Monochloroacetic Acid (MCAA)
Dichloroacetic Acid (DCAA)
Trichloroacetic Acid (TCAA)
Monobromoacetic Acid (MBAA)
Dibromoacetic Acid (DBAA)
Tribromoacetic Acid (TBAA)
Bromochloroacetic Acid (BCAA)
Dibromochloroacetic Acid (DBCAA)
Dichlorobromoacetic Acid (DCBAA)
CH2CI-CO2H
CHCI2-CO2H
CCI3-CO2H
CH2Br-CO2H
CHBr2-CO2H
CBr3-CO2H
CHBrCI-COjH
CBr2CI-CO2H
CCI2Br-C02H
197
-------
Table 2. Parameter estimates for Equation 11
Species
CHCU, pH=5,7
CHC!,, pH=9.4
CHBr,
CHBr,CI
CHBrCi,
MBAA
DBAA
DCAA, pH=5,7
DCAA, pH=9.4
TCAA, pH=5,7
TCAA, pH=9.4
BCAA, pH=5,7
BCAA, pH=9.4
DBCAA, pH=5,7
DBCAA, pH=9.4
DCBAA, pH=5,7
DCBAA, pH=9.4
TOX,'pH=5,7
TOX, pH=9.4
Parameters
Cl,
CI,
Br
Br
Br
Br
Br
Ci,
Cl,
Cl,
Cl,
Br
Br
Br
Br
Br
Br
^ ^
—
A
Po
102
135
198
556
276
5.03
64.5
81.4
60.3
131
42.7
186
104
330
44.1
582
88.9
945
729
A
P,
0.34
0.34
1
1.76
0.49
0.59
0.73
0.21
0.21
0.39
0.39
1
1
1
1
1
1
__
—
A
P2
__
__
__
6.25
11.27
„
«
...
-.
__
•M
11.65
11.65
13.13
13.13
40.98
40.98
2.44
2.44
A
Pa
-1.25
-1.25
_
-0.59
-0.35
«
„
-0.64
-0.64
-1.26
-1.26
-0.29
-0.29
~*
—.
—
—
0.10
0.10
A
T.
0.099
0.099
0.18
0.17
0.13
0.18
0.12
0.12
0.12
0.11
0.11
1.22
1.22
0.38
0.38
0.089
0.089
0.212
0.212
A
r,
—
—
—
—
—
_.
—
—
..
—
__
-1.17
-1.17
-0.27s
-0.278
«
..
__
—
* Br or CI2 (in mg/L)
B Not significant
198
ilili it . ; • [„». , iJil ., •':' ., ,-,i,,, 'iil'i, .'fj
. ,; jj.i,,!,,; ,fi ,; 'illfci1 • i it'.'.' i:./, -ai..' •».'.:' i)
-------
co
1/Brt
199
1/Brt
-------
, 'I,
6
-I/6rt
ow
ffl
o
'to'
1/6ri
200
c
o
*-•
co
i_
•«-•
C
0)
o
C
o
u
0)
E
O
C
o
'4^
u
C
CO
u.
-------
CD 0)
o
i- CO
Hi
o
- CD
0
CM
m
CM
g
~^ in
CD II
o E X
CO j— O.
O
I
O
O
- CO
o --.
*•" e»
o E X
co :•"• o.
o
"co
o
03
v_
c
10
7 T
ffi m
•4 <
0
CO
o
CO
n/Bri
0
CM
o>
m
o —~
o w ^
»- ^ lx
•— U?
8
8
o
5
O
i- co
o
- CO
o
CM
§
o E
00
CO
CD
3
O)
u_
l/6ri
201
-------
'&.' -*>:
i
< <
oo
I-
o
CO
S
o
CM
e
o
a,
- 0
o
co
o
co
o
tJ-
o
cM
o
o
o
co
o
co
O
^r
o
cM
1/Brt
«« «
| 1 |
"? ^ 2'
£Q CO CQ
O -« ^
1
1
« * '
<
,.! ,
f
«a oa
1
4 -4
•-
OOOOOOOOOO
COffl'-J-CMOOOaJ^CM
202
1/Brl
-------
Chlorine Demand and TTHM Formation Kinetics:
A Second-Order Model
Robert M. Clark
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Cincinnati, Ohio 45268
Abstract
Much effort has been expended in attempting to develop mathematical models for chlorine demand in water
and waste water. Most of these efforts have centered around the use of first-order functions or modifications of first-
order functions. Recently there has been interest in also characterizing the formation of Total Trihalomethanes.
These efforts have taken on new meaning because of the importance of maintaining chlorine residuals for microbial
protection and the concern over minimizing the formation of Trihalomethanes in drinking water distribution
systems. This paper utilizes second-order kinetics to describe both of these relationships using data collected from a
recent collaborative study between the USEPA and the American Water Works Association Research Foundation.
It also demonstrates that TTHM formation can be characterized as a function of chlorine demand.
203
-------
Introduction
Disinfection of public drinking water supplies to prevent the occurrence of waterborne disease from
micfobial pathogens is generally considered to be one of mankind's most successful public health interventions.
Ch|orination of drinking water has been successfully used to inactivate pathogenic organisms since the turn of the
cerfbry and is the most widely used method of disinfection in the United States (US). Currently, more than 200
million people in the United States consume disinfected drinking water and this practice has virtually eliminated
serious waterborne outbreaks.
I; ', Rules and regulations have been established under the Safe Drinking Water Act and its Amendments
(SDWAA) of 1986 to ensure that this protection is maintained (Vasconcelos et. al. 1996). For example, the Surface
Water Treatment Rule (SWTR) requires a minimum level of 0.2 mg/L at the entrance to the distribution system. It
also requires that a detectable chlorine residual level (or Heterotrophic Plate Count < 500/mL) be maintained
throughout the system. However, when chlorinated water is introduced into the distribution system the chlorine
resfSual tends to dissipate which may then result in microbiological regrowth and which may increase the system's
vulnerability to contamination. Three factors which frequently influence chlorine consumption are as follows: (1)
reaction with organic and inorganic chemicals (e.g., ammonia, sulfides, ferrous iron, manganous ion, humic
material) in the bulk aqueous phase; (2) reactions with biofilm at the pipe wall; (3) consumption by the corrosion
process.
t ' ,h ' ' ^'i; ' „ , " '.J.,, - .' ' ' ' '"' ,
i A by-product of chlorination is the formation of total trihalomethanes (TTHMs) hi waters containing
organic precursor compounds, such as humic and fulvic acid substances. Generation of TTHMs have been shown to
be a function of various water quality parameters and chlorination conditions including total organic carbon (TOC),
the type of organic precursor, chlorination level, pH, temperature, bromide level, reaction tune, and UV-254
absorbance. TTHMs are also regulated under the SDWAA. (Amy, et al., 1987; Clark, et al., 1996). Chlorine
decay in distribution systems is generally considered to be composed of two components. One component is wall
demand while the other is associated with decay in the bulk phase of the water (Clark, et al., 1993).
Frequently, chlorine decay in the bulk phase is characterized by a first-order kinetic model as follows:
dc/dt = -kc
(1)
where c = the chlorine concentration (mg/L), k is the first-order decay, constant (min-1) and t time hi minutes.
Integrating equation (1) yields:
C(t) = C0e-kt (2)
where C(t) is the chlorine concentration (mg/L) at time t, CQ is the initial chlorine concentration (mg/L), and t is the
tinfe of reaction (Clark, et al., 1993).
Several other types of kinetic models have been utilized to characterize chlorine decay kinetics. For
example, Haas and Karra (1984) evaluated the reliability and performance of five different chlorine decay models.
They investigated first-order decay, power-order decay (n* order), first-order decay with stable component, power
la\y decay with stable component^111) and parallel first-order decay. They found that the parallel first- order decay
model which assumes parallel decay of two components of chlorine residual, one decaying more rapidly the other,
provided the best results. To this point few models have been developed to predict the formation of THMs in
chlorinated water (Amy, et al., 1987; Clark, et al., 1996).
In 1993 a collaborative study between the US Environmental Protection Agency (USEPA) and the
American Water Works Association Research Foundation (AWWARF) was initiated (Vasconcelos, et al., 1996).
The study was conducted in order to gain an understanding of the kinetic relationships that describe chlorine decay
and Total Trihalomethane (TTHM) formation in water distribution systems. Using the EPANET distribution
nefjsvprk model, several kinetic models were tested and validated based on data collected from the utilities,that
participated in the collaborative study (Rossman, et al., 1994). Controlled tests were conducted at each utility for
chlorine decay and TTHM formation. For the chlorine decay tests, samples of water were stored in head space free,
204
-------
darkened bottles at ambient temperature. Individual bottles were then opened at different times, chlorine residuals
measured and the residual values plotted against time. TTHMs were measured using the liquid-liquid extraction
process (Vasconcelos, et aL, 1996) Various chlorine decay and TTHM formation kinetic models were evaluated to
find the one which best characterized the data. For chlorine decay, as with the Haas and Karra (1984) study, models
evaluated included a first-order decay, an n* order model, a first-order model with a stable component, and a parallel
first-order decay model. In the Vasconcelos study (1996) the nth order model yielded the best fit in about half the
cases and the parallel first-order model yielded the best fit in the other half. For TTHM formation a formation
potential model was developed similar to the one developed by Clark, et al. (1996).
This paper presents alternative models for bulk chlorine decay and TTHM formation based on second-order
reaction kinetics. The model also shows that TTHM formation can be characterized as a function of chlorine
demand. The rate of reaction is assumed to be proportional to the first power of the product of the concentration of
two different species. One component is assumed to be hypochlorous acid and the second represents the constituents
in water which cause the "chlorine demand." Among the products of this reaction are the formation of TTHMs.
The model is validated using data collected from the AWWARF/EPA study and from an earlier study conducted at
the North Marin Water District (NMWD) only.
Data Sources
The AWWARF/EPA joint study involved collecting field data from five sites in the United States and two
in France (Vasconcelos, et al. 1996). The earlier study was conducted approximately one year prior to the
AWWARF/EPA study (Clark et. al 1994). North Marin data from the earlier study is designated as Study A.
North Marin from the AWWARF/EPA study is designated as Study B. These sites are as follows:
United Water Resources (UWR), Harrisburg PA
Oberlin Pump Station (OPS)
• North Perm Water Authority (NPWA), Lansdale PA
Keystone Tie-in (KTI)
- Forest Park Treatment Plant (FPTP)
- 50/50 Blend of Keystone and Forest Park Water (50/50 KFP W)
- Well W17 (W17)'
- Well W12 (W12)
City of Fairfield Water Department (FWD), Fairfield CA
Waterman Treatment Plant (WTP)
• North Marin Water District (NMWD-A), Novato, CA - Study A
Russian River Aqueduct (RRA-A)
Stafford Lake Treatment (SLTP-A)
- 50/50 Blend of Aqueduct and Stafford Lake Water (50/50 ASLW-A)
North Marin Water District (NMWD-6), Novato, CA - Study B
Russian River Aqueduct (RRA-6)
Stafford Lake Treatment (SLTP-B)
- 50/50 Blend of Aqueduct and Stafford Lake Water (50/50 ASLW-B)
City of Bellingham Water Department(BWD), Bellingham, WA
Welcome Water Treatment Plant (WWTP)
• Parisienne des Eaux System (PES), Paris (four different sampling dates) -
- 5/23/94
- 5/31/94
205
-------
* - 6/8/94
- 6/16/94
* City of Orleans Water System (OWS)
Chlorine reaction kinetics were determined by analyzing the results of bottle tests made on treated waters at each
l sife, Chlorine disappearance was measured over time in a series of TOC (mg/L) of samples stored in
nonreacting containers such as amber glass bottles. These data were used to estimate the rate coefficient associated
with kinetic models describing the reactions of chlorine. TTHM measurements were made using the liquid-liquid
extraction gas chromatic method. Samples were collected in septum-sealed screw-cap bottles with care to omit air
bubbles. A reducing agent and acid were added to prevent further TTHM formation after sample collection
(Vasconcelos, etal., 1996).
For purposes of this study only the systems using free chlorine were used for analysis, which excluded the
North Penn Water Authority. No data was available from the OWS and the 5/31/94 study in the PES and TTHM
data was not collected for UWR, and the French water systems and no chloroform data was collected for NMWD-
SLTP and the Bellingham system. In most cases TTHM samples were collected from a clear well slightly removed
in time from the point of chlorination. Therefore, by the time the sample was taken, a small amount of TTHM had
already formed. This anomaly was compensated for in the TTHM models. In several cases the variation in
analytical results were so great, that the data was virtually meaningless. The location and the nature of the waters
from each utility are described in Table 1.
Model Development
Chlorine dissolved in water yields:
C12 + H2O/£ HOC1 + H++ CI~
HOCI generally reacts with the various components mat make up chlorine demand as follows:
HOC1 + CldemandJE products
In equation (4), if we let the balanced reaction equation be represented by:
aA + bBsE pP
where A and B are the reacting substances and P is the product of the reaction then the rate of
reaction is given by:
^-w
dt A
where k^la=kj/b=
or
dC,
dt
or
dCn
IT
(3)
(4)
(5)
(6)
Equation for Chlorine Decay
Since both CA and C# are changing with time, we must write a relation connecting them in order to integrate
the differential equation. If CAO an<^ CBO represent the initial concentrations of A and B, respectively, at t = 0 and x
represents the concentration of A that has reacted, then the concentration of B that has reacted is given by for/a.
Consequently,
C=CA-x C = CB-- (7)
206
j i:
illiiiii: ..
Jiiiliii;
-------
and from equation (7)
dCA = -dx
_ a
By substitution in equation (6) we have
=-kA(C/S)-x) (CB(|-bx/a)
Rearranging equation (9) yields:
dCA
= kAdt
(C^-x) (CB()-bx/a)
Integrating equation (10) yields:
ln
l-bx/aCn a
Let
bC
A,
a »
and
u=rkA
then
1-x/C
A,
l-bx/aCD
If we let
then
-C -C
b BQ
aCD
bCA
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
or
207
-------
CA=-
an
-£c«»
ac%
1"-"' Q
(17)
If we let
K=C -1_
*» b
(18)
and
aC
r?— •
! r " be.
(19)
then equation (18) can be rewritten as follows:
„ _
A~l-Re-«
(20)
Equation (20) will be applied to the systems described in Table 1. In equation (20), CA is the concentration of
free chlorine. Rewriting equation (20) yields:
Cl (t)=-
K
1-Re-
(21)
where Cl(t) = the chlorine concentration in mg/L at time t, R (dimensionless), K (mg/1) and u (min-1) are
parameters to be estimated, and t = the time of reaction in minutes.
TTfJM Formation Equation
I' ,"'..' ' „ ' ' , '• ,lil, 'I ' I
Based on equation (5), a TTHM formation equation was developed. Using equations (6), and eliminating time
as « variable yields:
dC.
= —T
(22)
where
(23)
if we let
'
(24)
then
208
-------
(25)
We can assume
therefore Cp = T(CA -CA)
If we let Cp = TTHM
where TTHM= Total Trihalomethanes
then
j. = the initial chlorine residual in mg/L
and
<> vl-/fe-"'
where T = Dimensionless parameter
(26)
(27)
C A = Chlorine residual in mg/L
(28)
R,u= parameters from the chlorine decay equation
Equation 28 demonstrates that using second-order kinetics TTHM formation can be characterized as a function of
chlorine demand.
As mentioned earlier, in several cases samples were taken from a clear well or some point after initial
chlorination and therefore a small amount of TTHMs had already formed by the time the sample was taken.
Therefore, equation (28) was modified to the form shown below:
TTHM = T(CA - (-
(I-/?)
(29)
l-Re-' "
where M is the estimated value for TTHM at time zero.
Parameter Estimation for Chlorine Decay Equation
It will be demonstrated that the use of equation 21 provides results equivalent to or better than the models
utilized in the Vasconcelos, et. al (1996) study. Reformulating equation (21) yields:
Cl (0 =
Cln(l-R)
In — ut
— Re
assuming Cl(t)=Clo @ t=0
where Clo = the initial value of free chlorine in mg/L.
(30)
209
-------
From equations (12) through (19), it can be seen that if u>0, then R<1 and K > 0. If u<0, then R>1 and K< 0.
The modified Gauss-Newton method was used to compute the least squares estimates of the two nonlinear
models (Hartly, 1961). This method uses the first-order Taylor series expansion for the nonlinear functions of the
independent variable (time) and the parameters. This method and the SAS procedure NLIN are used to estimate the
mCKiel parameters. The procedure regresses the residuals (observed-predicted) against the partial derivatives of the
model with respect to the parameters until the estimates converge.
Parameters were estimated for equation (30) (second-order) and equation (2) (first-order) for the cases contained
in Table 1. The estimates for these parameters are shown in Table 2.
In all of the cases examined the r2 values were higher for the second-order model then those for the first order
mddcl. Figures 1-1 1 compare the first-order fit to the second-order fit for the data in Table 1. As can be seen from
both Table 2 and Figures 1-1 1, the second-order equation provides the "best fit" for most of the data sets.
Vasconcelos, et al. (1996) estimated parameters for the various equation forms mentioned earlier. To compare
the results from Vasconcelos, et al. (1996) and the analysis presented in this paper, Table 3 was created. In some
caS&S the r2 values for the first order decay model obtained hi this study were lower then those reported by
Vasconcelos, et al. (1996). In those cases the r2 values are shown in parenthesis in column 2. The second-order
model developed in this study was better or at least as good as the models analyzed in the Vasconcelos, et al.
(1996) study.
If we assume that Cg, the second reactant in equation (5) is TOC, then from equations (6) through (21) we
would expect a value for R>1 to result in a very rapid decline in chlorine residual. Figures 3,4, 5, 6 and 7 for the
NMWD illustrate this effect. The RRA is a high quality, low TOC source water as compared to SLTP water which
is of poor quality with a high TOC. As can be seen the STLP water experiences a much more rapid loss of chlorine
residual than does RRA water.
As can be seen from Table 3 the second-order equation is comparable to both the nth and parallel order
equations which were found to be "best" in the Vasconcelos, et al. study (1 996), and superior to five out of six of
the first order fits.
Estimation of Parameters for Chlorine Decay and TTHM Formation
In order to illustrate the principle that the models developed can be used to characterize TTHM formation as a
function of chlorine demand, the parameters in equations 21 and 29 will be estimated using data from the
Vasconcelos e. al. (1 996) study and from the Clark, et al. (I 994) study. Consistent and reliable data was available
for FWD- WTP, SLTP, RRA-A, RRA-B, 50/50 ASLW-B, and BWD-WWTP. The parameters for equations (21)
and (29) are contained in Table 4.
j'," " ; •'" • ":"'" ' !•'.•: ,i ", 1 ." :"" ' ' ',r . I,' i ;:; , • , .' ' /i" ':i":X
Figures 12 to 17 display the results from estimating the parameters in equations (21) and (29).
"'' . ' ' ']:.1j •• •' " ;: •'-.. "::""" ' ' . ';;' it1
Discussion
Most municipal water systems in the United States practice some form of disinfection to prevent waterborne
infectious disease. These disinfectants may include chlorine, chloramines, ozone, or chlorine dioxide. Frequently
one disinfectant may be used as a primary disinfectant followed by another. In 1974 chloroform, a product of the
reaction of chlorine and naturally occurring organic matter was identified in disinfected drinking water. Since that
time a number of other disinfection byproducts (DBPs) have been identified including trihalomethanes (THMs), by-
products-other then chloroform (e.g., bromodichloromethane), haloacetic acids (HAAs), haloacetonitriles,
haloketones, and haloaldehydes. All of the disinfectants mentioned are reactive and ozone, chloramines, and
chlorine dioxide result in both inorganic and inorganic by-products. More then 500 DBPs have been identified in
tap water (Clark et al.,1996).
210
-------
Chlorine is the most widely used disinfectant in the United States and research has therefore focused on
exposure issues related to chlorinated drinking water. A key issue is the complex issue of balancing the microbial-
and DBF-risks associated with disinfecting drinking water.
By-product concentrations vary both spatially and temporally within a distribution system. These variations
are due to source water quality variability, variations in water treatment efficiency, and the dynamic nature of
byproduct formation during distribution. Two and four fold differences may occur within a water distribution
system and the pattern of by-products from different sources can vary significantly over the course of a day as well as
during a year.
The effects of by-product exposure are assumed to be linked to both the concentration and the nature of the by-
products in the water consume. Therefore there is a great deal of interest in understanding the factors that effect the
formation of disinfection by-products during the distribution of drinking water.
This paper presents a model that links the formation of trihalomethanes to the consumption of chlorine and the
maintenance of chlorine residuals in drinking water. The model presented should assist in providing a quantitative
framework for assessing the risk-risk tradeoffs between the microbial and DBF risks associated with disinfecting
drinking water.
Data used in developing this model was collected under laboratory conditions in order to eliminate the site
specific effects that individual distribution systems may have on the loss of chlorine residuals. These distribution
system effects are well documented in papers by Rossman, et al. (1994) and Clark, et al. (1995) and the equations
presented in the Rossman, et al. and Clark, et al. paper can be used in conjunction with the model developed in this
paper to simulate the effect of chlorine residual loss in a specific distribution network.
Summary and Conclusions
Rules and regulations established under the Safe Drinking Water Act and its Amendments of 1986 and 1996
have emphasized the importance of maintaining disinfectant residuals in drinking water distribution systems. There
is also concern over the formation of potentially carcinogenic compounds which result from the application of
disinfectants. Chlorine is one of the most commonly used and effective disinfectants in use today. Many
investigators have examined the kinetic relationships associated with chlorine decay in water. Chlorine demand in
distribution systems is generally assumed to be composed of a wall decay and a bulk decay component which is
usually assumed to be first-order in nature. A study conducted by Vasconcelos, et al. (1 996) has shown that the
nth order and parallel order decay equations best described chlorine decay in the study described in this paper. In
this paper a second-order kinetic model has been shown to provide an equivalent fit.
Acknowledgments
\
The author would like to gratefully acknowledge the assistance of Dr. Mano Sivaganesan,
Statistician, Lockheed Environmental Systems Technologies Co., Ms. Jean Lillie, Ms. Toni Frey, Mrs. Sandra
Taylor and Mr. Steven Waltrip of the WSWRD for their assistance in preparing this paper.
} :;
References
Amy, G. L., Chadik, P.A., and Chowdhury, Z. K., "Developing Models for Predicting Trihalomethane Formation
Potential and Kinetics", Jour AWWA, 79 (7), 1987, pp 89-97.
Clark, R.M., Goodrich, J.A., and Wymer, L.J., "Effect of the Distribution System on Drinking Water Quality,"
Journal of Water Supply Research and Technology - AQUA, Vol. 42, No. 1, Feb. 1993, pp 30-38.
Clark, R.M., Smalley, G., Goodrich, J.A., Tull, R., Rossman, L.A., Vasconcelos, J.J. and Boulos, P.F.
211
-------
"Managing Water Quality In Distribution Systems: simulating TTHM and Chlorine Residual Propagation,"
Journal of Water Supply Research and Technology - AQUA, Vol. 43, No. 4, pp 182-191, 1994.
Clark, R.M., Rossman, L., and Wymer L., "Modeling Distribution System Water Quality: Regulatory
Implications", Journal 'of Water Resources Planning and Management, Vol. 121, Nov/Dec 1995, No. 6, pp 423-
428. . i . i , ,i, ,,_,
Clarjc, R.M., Pourmoghaddas, H., Wymer, L.G., Dressman, R.C., "Modeling the Kinetics of Chlorination By-
product Formation: The Effects of Bromide," Journal of Water Supply Research and Technology - Aqua, Vol. 45,
No, I, pp 1-8, 1996.
Haas, Charles N., and Karra, S. B., "Kinetics of Wastewater Chlorine Demand Exertion," Journal ofWPCF,
Volume 56, Number 2, Feb 1984 pp 170-173.
llartly, H.O., "The Modified Gauss-Newton Method for the Fitting of Nonlinear Regression Functions by Least
Squares," fechnometrics, 3, 196X pp 269-280.
Rossman, L. A,, Clark, R. M. and Grayman, W. M. "Modeling Chlorine Residuals in Drinking-Water
Distribution Systems" Journal of Environmental Engineering, Vol. 120, No. 4, July/August, 1994, pp 803-820.
Rossman, L.A., Clark, k.M., and Grayman, W.M., "Modeling Chlorine Residuals in Drinking Water Distribution
Systems", lottrnal of Environmental Engineering, Vol. 120, July/August 1994, pp 803 -820.
Vasconcelos, J.J., Boulos, P.'F., Grayman, W.M., Kiene, L., Wable, O., Biswas, P., Bhari, A., Rossman, L.A.,
Clark, R.M., and Goodrich, J.A., Characterization and Modeling of Chlorine Decay in Distribution Systems,
AWWA, Research Foundation, 6666 West Quincy Avenue, Denver, CO 80235, 1996
. , , ii- .-S i. liliii 1 >,.:<'<>: :"'!!'i'', l:',:> !' ii •,.'.' T! v IS.!! "V. '.i :'. •
,i iii-iiiuiiT-ii !. i:1 i:.iiii,il'iiit •,;• i;.:1."!'1!.'.!1!!; ..^1
-------
TABLE i. Summary of Finished Source Waters Subjected to
Chlorine Decay Teate
. Finished
water
source
CD
UWR-OPS
FWD-WTP
RRA-A
SLTP-A
RRA-B
SLTP-B
50/50 ASLW-B
BWD-WWTP
PES-3/23/94
PES -6/8/94
PES-6/l6m
PH
(2)
7.52*
8.15-
7.4*
8J5*
7.42*
8.85"
7.92-
8.05*
7.8
8.00
7.80
Temperature
CO
(3)
16.4
17.y
25.0*
25.0*
22.2-
21.9s
22.1'
17.4'
14
16
16
TOG
(mo/L)
(4)
1.73*
1.87*
0.55*
3.55*
0.56*
3^5*
2.05*
0.84*
—
_
—
Residual
free chlorine
(mg/L)
(5)
0.98
1.73
3.9
10.0
0.31
0.49
0.40
0.72
0.6
0.62
0.67
'Measured in field during tunpling study.
*EPA laboratory study.
215
-------
jgIT '"';;,! :•«,»*
TABLE 2. PARAMETER ESTIMATES FOR FIRST
AND SECOND ORDER EQUATIONS
Site
(D
UWR-OPS
FWD-WTP
R!|A-A (
STLP-A
RRA-B
STLP-B
50/50 ASLW-B
BWD-WWTP
PES-3/23/94
PES-6/8/94
PES-6/16/94
First-Order
K
(2)
0.0097
0.048
0.0022
0.01455
0.044
0.73
0.45
0.035
0.001
0.07
0.06
r2
(3)
0.944
0.834
0.158
0.957
0.707
0.982
0.948
0.788
0.931
0.729
0.778
Second-Order
K
(4)
0.065
1.0002
3.594
9.854
0.099
-0.33
-0.02
0.266
-0.01
0.22
0.23
R
(5)
0.933
0.4207
0.078
0.9027
0.677
1.67
1.05
0.631
1.02
0.65
0.66
U
(6)
9x10^
0.12
0.1557
0.00257
0.038
-0.42
-0.04
0.051
-0.0025
0.08
0.06
R2
(7)
0.960
0.961
0.879
0.989
0.812
0.989
0.990
0.987
0.958
0.907
0.948
216
-------
TABLE 3. COMPARISON OF VARIOUS
CURVE FITTING APPROACHES
Site
0)
UWR-OPS
FWD-WTP
RRA-B
STLP-B
50/50 ASLW-B
BWD-WWTP
First Order
(2)
0.994
0.834
0.707
(0.839)
0.982
(0.990)
0.948
(0.946)
0.788
(0.931)
Nth Order
<3)
0.995
0.966
0.827
0.995
0.989
0.991
Limited First
Order
(4)
0.960
0.948
0.824
0.993
0.974
0.973
Parallel
First Order
(5)
-
0.978
0.808
0.986
0.986
0.993
Second
Order
(6)
0.960
0.961
0.812
0.9.89
0.990
0.987
() r2 from the Vasconcelos study
217
-------
1
TABLE 4. PARAMETERS FOR CHLORINE DECAY
AND TTHM FORMATION MODELS
, »»' ' , '• • ' , , .,„ I ',' " X ;„' 'J" ' ', ,' ,1, 'i
Site
(1)
SLTP-A
RRA-A
FWD-WTP
RRA-B
SO/50 ASLW-B
BWD-WWTP
CHLORINE DECAY
K
(mg/L)
(2)
7.03
3.58
1.298
0.160
0.028
0.160
R
(mg/L)
(3)
0.2974
0.0828
0.2500
0.4829
0.9299
0.7773
u
(mirr1)
(4)
0.252
0.072
0.440
0.293
0.193
0.017
TTHM FORMATION
M
(5)
-
-
10.0
-
-
17.0
T
(6)
56.43
72.52
41.29
126.73
188.60
36.27
218
Sail!;.
-------
1.0
0.9 -
0.8 -
c.
15
o>
h.
o>
*£Z
O
0.6 -
0.5 -
0.4
First order model
Second order model
10 20 30 40 50 60 70 80
Figure 1. Chlorine Demand In mg/L Versus Time In Hours
at United Water Resources - Oberlin Pump
Station (UWR - OPS)
219
-------
1.8
1.6 -
D>
•£
£ 1.4
co
£
8 1.3
*C
J2
jc
O
0.9
—— First order model
--• Second order model
1.1 -
0
Figure 2. Chlorine Demand In mg/L Versus Time In Hours
for City of Fairfield Water Department - Waterman
Treatment Plant (FWD - WTP)
* f J f '! 1." „' «",«,
220
-------
4.0
3.8
CD
E
•E 3.6
CO
CO
CD
CO
c 3.4 -
o
JC
a
3.2 -
3.0
First order model
Second order model
0
10
20
30
Hours
40
50
Figure 3. Chlorine Demand in mg/L Versus Time in Hours
for North Marin Water District - Russian River
Aqueduct (RRA - A)
60
221
-------
10
9
8 -
O)
E
CO
3
G>
t_
G>
I5
JC
O
4 -
3 -
First order model
Second order model
0
10 20 30 40 50 60 70 80
Hours
Figure 4. Chlorine Demand In mg/L Versus Time in Hours
for North Marin Water District • Stafford Lake
Treatment Plant (SLTP-A)
90 100
222
-------
0.35
0.30
0.25 -
ca
=i
0.20 -
o>
.E
1 0.15
O
0.10 -
0.05
First order model
-•• Second order model
10
20
30
Hours
Figure 5. Chlorine Demand in mg/L Versus Time in Hours
for North Marin Water District - Russian River
Aqueduct (RRA - B)
223
-------
0.5
0.4
e
z °-3
cd
3
15
"w
CD
-------
0.4
0.3
o>
E
CO
Q)
o
0.1 -
0.0
First order model
— Second order model
10
20
30
Hours
Figure 7. Chlorine Demand in mg/L Versus Time in Hours
for North Marin Water District - 50/50 Blend of
Aqueduct and Stafford Lake Water (50/50 ASLW - B)
•225
-------
0.8
0.7
0.6 ~
E
c
•5 0-5
to
£
(D
D
0.4 -
0.3 ~
0.2 -
0.1
First order model
Second order model
0
Figure
10
20
30
40
50
Hours
8. Chlorine Demand in mg/L Versus Time in Hours
for City of Belllngham Water Department - Welcome
Water Treatment Plant (BWD-WWTP)
226
-------
0.6 t
0.5 -
0.4 -
CO
lo-3
©
-------
0.7
0.6
0.5 -
£
c
3
1°
"35
2
0
'O
.c
O
0.4 -
0,3 -
0,2 -
0.1 -
0.0
First order mode!
Second order model
0
10 15
Hours
20
25
Figure 10. Chlorine Decay in mg/L Versus Time in Hours
for Parisienne des Eaux System on 6/8/94
(PES-6/8/94)
228
„ •>.!, ,li,ai.
.
i ,,i[;,:l-
-------
0.7
0.6
0.5 -
D>
E
75 0.4
CO
-------
20
15 -
o>
E
c
13
1 10
©
V.
o
.S
5
O
5 -
0 O
0
175
10
Hours
15
- 150
- 125
- 100
3.
c
- 75
- 50
- 25
0
20
Figure 12. TTHM Formation (M0/L.) and Chlorine Demand (mg/L)
Versus Time in Hours for North Marin Water District
. ' "• „ :,, f !"''!i .iiiiLS;"':!1,!!!! '':• ,; n ' ' • film V" '" I ', I' !' „ if! "I f, i '" i »!!':", » H' J , I! it 'i,1!,!, ... ii.li,'!:1'1"11!. 'i,1 • , • ' • , . '"i* .1 • ,h
Stafford Lake Treatment Plant (SLIP - A)
••: "/• ;• - ::ffv '.;•' ; ,, ":':,::,;« •-• ••* ,! iiVrrt ,- v '.••
230
-------
_,
f^l
E
c
"55
3
TJ
"co
0)
0)
c
jo
.c
O
1U
9
8
7
6
5
4 <
3
2
1
0 <
...-O-
^--"
+
~ *
_ /
t
t
t
/O
* • Chlorine decay model
t
/ 0 TTHM model
— , *
i
i
o ;
~ i
t
— i -
i
—t
i
i
D \ \ I I I
^D
20
15 i
CD
C
X
10 t
5
0
0 10 20 30 40 50 60
Hours
Figure 13. TTHM Formation (pg/L) and Chlorine Demand (mg/L)
Versus Time In Hours for North Marin Water District
Russian River Aqueduct (RRA - A)
231
-------
35
-cr
30
O)
E
(0
=5
I 2
£
o>
.E
J!
JC
O
1
I I
Chlorine decay model
o TTHM model
25
20
15
10
0
0
10
20
30
Hours
40
0
50
Figure 14. TTHM Formation (pg/L) and Chlorine Demand (mg/L)
Versus Time In Hours for City of Bellingham Water District
Welcome Water Treatment Plant (BWD - WWTP)
232
-------
Chlorine decay model
o TTHM model
Figure 15. TTHM Formation (ug/L) and Chlorine Demand (mg/L)
Versus Time In Hours for North Marln Water District
50/50 Blend of Aqueduct and Stafford Lake Water (50/50 ASLW - B)
233
-------
1.0
0.9
0.8
?i 0.7
£
•S 0.6
ca
1 0.5
£
.1 0.4
in
0 0.3 '
. • '.
CX2
•'*' O.l"
0"n
.U
c
o
_ o
- .*"""•"
t
*
9oo o
i
O * • Chlorine decay model
,' o TTHM model
i
tf.
Xw *• • .
* ^^*^— *
i
i
i~ •
i i i i i
) 5 10 15 20 25 3
Hours
- 20
- 15
- 10
5
0
Figure 16. TTHM Formation (ug/L) and Chlorine Demand (mg/L)
Versus Time in Hours for North Marin Water District
Russian River Aqueduct (RRA - B)
234
-------
co
TO
2
15 ^
10
0 12 34 5 6 7 8 9 10 11 12
Hours
Figure 17. TTHM Formation (MB/L) and Chlorine Demand (mg/L)
Versus Time In Hours for City of Falrfield Water Department
Waterman Treatment Plant (FWD - WTP)
235
-------
"ill
!, i
'£); ' • ' ! APPENDIX A ' '
"RELATED PROFESSIONAL JOURNAL ARTICLES
1. Fox, K.R. and Lytle, D.A. "Milwaukee's Crypto Outbreak:
Investigation and Recommendations," JAWWA, Vol. 88, No. 9,
pp 87-94, September 1996.
.»!,
2. Clark, R.M., Geldreich, E.E., et al. "Tracking a Salmonella
serovar typhimurium Outbreak in Gideon, Missouri: Role of
Contamination Propagation Modeling." J. Water SRT-Aqua,
Vol. 45, No. 4, pp 171-183, 1996.
3. Rice, E.W., Fox, K.R., Miltner, R.J., Lytle, D.A., and
Johnson, C.H. "Evaluating Water Treatment Performance Using
a Microbial Surrogate System," Jour. AWWA, 88:9:122-130
(1996).
4. Sethi, V., Patnaik, P., Biswas, P., Clark, R.M., and Rice,
E.W. "Evaluation of Optimal Detection Methods for
Characterizing Suspensions in Drinking Water," Journal of
American Water Works Association, Vol. 89, No. 2, February
1997, pp 98-112.
5. Li, S.Y., Goodrich, J.A. , Owens, J.H., Wallace, G.E.,
Schaefer, F.W. Ill and Clark, R.M. "Reliability of
Surrogates for Determining Cryptosporidium Removal,"
Jour.AWWA. Vol. 89, No. 5, May 1997, pp 90-99.
6. Lytle, D.A., Schock, M.R., Clement, J.A. & Spencer, C.M.
"Using Aeration for Corrosion Control". Journal of the
American Water Works Association, 90:3:74 (1998).
Lytle, D!A., Schock, M.R. , Clement, J.A. & Spencer, C.M.
"Using Aeration for Corrosion Control-Eratum". Journal of
the American Water Works Association, 90:5:4 (1998).
7. Lytle, b'.A. & Schock, M.R. "An Investigation of the Impact
of Alloy Composition and pH on the Corrosion of Brass in
Drinking Water". Advances in Environmental Research, 1:2:1
(1997). Available from Internet @
http://WWW.sfo.com/~aer/518-97.pdf
236
-------
9.
Clark, R.M., Abdesaken, P., Boulos, P.P. and Mau, R.
"Mixing in Distribution System Storage Tanks: Its Effect on
Water Quality," Journal of Environmental Engineering.
Vol. 122, No. 9, September 1996, pp 814-821.
Clark, R.M., Pourmoghaddas, H., Wymer, L.G. and Dressman,
R.C. "Modeling the Kinetics of Chlorination By-Product
Formation: The Effects of Bromide," Journal of Water Supply
Research and Technology - Aqua, Vo-1. 45, No. 1, pp 1 -8,
1996.
10. Clark, R.M., "Chlorine Demand and TTHM Formation Kinetics:
A Second-order Model," Journal of Environmental Engineering.
Vol. 124, No. 1, Jan. 1998, pp 16-24. Errata, Vol. 124,
No. 5, May 1998, p. 485
237
-------
"I F
.;»!' iis
APPENDIX B
EXPERTISE/POiNT-OF-CONTACT LIST
ORGANIZATION - U.S. ENVIRONMENTAL PROTECTION AGENCY
NATIONAL RISK MANAGEMENT RESEARCH LABORATORY
WATER SUPPLY AND WATER RESOURCES DIVISION
26 W. MARTIN LUTHER KING DRIVE
CINCINNATI, OH 45268
2/11/98
FAX 513-569-7658
RESEARCH AREA
' ; ', r"1: •
in
CONTACT
Robert Clark, Director
Walter Feige, Tech. Asst.
Susan Campbell, Program Analyst
Phyllis Jones, Mgt. Analyst
PHONE NO.
(513-569-1
7201
7496
7426
7205
DRINKING WATER TREATMENT
Organics Control Technology
Disinfection Byproducts
Engineering Research
Analytical Chemistry
Research
I!
Membrane Processes
Synthetic Organic Chemicals
Granular Activated Carbon
Adsorption
Biofiltration
1 !'' '!
Ozone
Chlorine Dioxide
Ozone/UV
Other Disinfectants
Air Stripping
Richard Miltner
Ben Lykins
Matthew Magnuson
Edward Urbansky
Tom Speth
Jeffrey Adams
Carol Ann Frank
Tom Speth
Tom Speth
Ben Lykins
Riqhard Miltner
Eugene Rice
Richard Miltner
Michael Elovitz
Ben Lykins
Richard Miltner
Richard Miltner
Richard Miltner
Ben Lykins
Jeffrey Adams
Tom Speth
7403
7460
7321
7655
7208
7835
7592
7208
7208
7460
7403
7204
7403
7642
7460
7403
7403
7403
7460
7835
7208
238
-------
RESEARCH AREA
CONTACT
PHONE NO.
(513-569-)
Inorganics Control Technology
Nitrate
Fluoride
Arsenic
Radionuclide Control
Technology
Corrosion/Lead/Copper
Corrosion/Secondary Impacts
Coagulation and Filtration
Small Systems
Organics
Inorganics
Performance Verification
Waterborne Disease Outbreaks
Costs
Distribution Systems and
Modeling
Point-of-Use/Point-of-Entry
Treatment
Organics
Inorganics
Computer Activities
Mutagenicity
Thomas Sorg
Thomas Sorg
Thomas Sorg
Kim Fox
Tom Sorg
Mike Schock
Darren Lytle
Thomas Sorg
Darren Lytle
Mike Schock
Kim Fox
Richard Miltner
Darren Lytle
Nicholas Dugan
Ben Lykins
James Goodrich
Thomas Sorg
Kim Fox
Jeffrey Adams
Kim Fox
Don Reasosner
Eugene Rice
Jeffrey Adams
Lewis Rossman
James Goodrich
Roy Haught
Mark Meckes
Ben Lykins
James Goodrich
Tom Sorg
Kim Fox
Steve Waltrip
Maura Lilly
Kathleen Schenck
7370
7370
7370
7820
7370
7412
7432
7370
7432
7412
7820
7403
7432
7239
7460
7605
7370
7820
7835
7820
7234
7204
7835
7603
7605
7067
7348
7460
7605
7370
7820
7386
7245
7947
239
-------
1 '.:..''!" T 'i 'I'Ull ' :
RESEARCH AREA
Publications
GIS
Constructed Wetlands
Hydrology
Microbiology Treatment
Coliform Methodology
' : , f!
Criteria and Standards
.I, ,-ii
Disinfection Treatment
Species Identification
Pigmented Organisms
Cone. X Time Concept
Giardia/Cryptosporidium
Research
Home Treatment Devices
Rapid Bacteriological Methods
Raw & Potable Water Quality
Sample Transit Time
Viruses in Water
Virus Methods
Distribution Water Quality
Microbial Growth
Assimilable and Biodegradable
Organic Carbon
CONTACT
Walter Feige
James Goodrich
Jill Neal
Don Brown
Bill Sidle
Donald Reasoner
Eugene Rice
Donald Reasoner
Eugene Rice
Christon Hurst
Donald Reasoner
Donald Reasoner
Christon Hurst
Eugene Rice
Eugene Rice
Jim Owens
Kim Fox
James Goodrich
•' fl, :' i": ' '• ' •
Donald Reasoner
Donald Reasoner
Donald Reasoner
Lucille Garner
Donald Reasoner
Eugene Rice
Christon Hurst
Christon Hurst
Donald Reasoner
Mark Meckes
Eugene Rice
Donald Reasoner
Eugene Rice
Donald Reasoner
Lucille Garner
PHONE NO.
(513-569-'*
7496
7605
7277
7630
7212
7234
7204
7234
7204
1
7461
7234
7234
7461
7204
7204
7235
7820
7605
7234
# ;
7234
7234
7417
7234
7204
•7461
7461
7234
7348
7204
7234
7204
7234
7417
240
-------
2/11/98
EXPERTISE/POINT-OF-CONTACT LIST
ORGANIZATION - U.S. ENVIRONMENTAL PROTECTION AGENCY FAX 732-321-6640
NATIONAL RISK MANAGEMENT RESEARCH LABORATORY
WATER SUPPLY AND WATER RESOURCES DIVISION
URBAN WATERSHED MANAGEMENT BRANCH
EDISON, NJ 08837
RESEARCH AREA
CONTACT
PHONE NO.
(732-^
Urban Watershed Management
Wet Weather Flow fWWF)
WWF; all areas
Watershed Management
Strategies, including Urban Hydrology
WWF Characterization
WWF Drainage Systems Design
WWF Treatment Technologies
WWF Disinfection
WWF Best Management Practices
WWF Modeling
WWF Databases
Eutrophication
Infrastructure Rehabilitation & Replacement
Prevention, Detection & Location
of Releases in Storage & Conveyance
Systems
Daniel Sullivan
Richard Field
Michael Borst
Thomas O'Connor
Joyce Perdek
Michael Royer
Evan Fan
Mary Stinson
Thomas O'Connor
Carolyn Esposito
Joyce Perdek
Mary Stinson
Evan Fan
Michael Borst
Richard Koustas
Ray Frederick
Richard Koustas
Thomas O'Connor
Marie O'Shea
Anthony Tafuri
James Yezzi
321-6677
321-6674
321-6631
321-6723
321-4380
321-6683
906-6924
321-6683
321-6723
906-6895
321-4380
321-6683
906-6924
321-6631
321-6639
321-6627
906-6898
321-6723
321-4468
321-6604
321-6703
241
-------
RESEARCH AREA
CONTACT
PHONE NO.
(732-)
Infiltration/inflow & Exfiitration
in Wastewater Collection Systems
Advanced Oxidation Treatment
of Water & Wastewater
Urban Highway and Roadway
Runoff Controls
Anthony Tafuri
Chien Chen
Michael Gruenfeld
321-6604
906-6985
321-6625
242
-------
Water Supply and Water Resources Division
Office of the Director
Director - Mr. E. Timothy Oppelt
Deputy Director - Mr. Cal Lawrence
February 18, 1998
Water Supply & Water
Resources Div.
Director - Dr. Robert M. Clark
Technical & Administrative
Support Staff
Mr. Walter A. Fejge
Treatment Technology
Evaluation Branch
Chief-Robert C.Thurnau
Microbial Contamina
Control Branch
Chief-Donald J. Reas
Urban Watershed
Management Branch
Chief-Daniel Sullivan
Water Quality
Management Branch
Chief-Ben W. Lykins.Jr.
Organics Cntrl.Tech.
-Bench/pilot evaluations
-Analytical chemistry research
-Activated carbon adsorption
-Biofiltration
-Membrane processes
-Ozone
-Ozone/GAC
-Ozone UV
-Disinfectants
-Disinfection byproducts
-Oxygenates
Inorganics Cntrl.Tech.
-Arsenic
-Perchlorate
Corrosion Control
-Lead
-Copper
Filtration
Residuals management
Microbiology Trtmt.
-Coliform methodology
-Disinfection trtmt.
-Concentration X time
concept
-Giardia disinfection
-Cryptosporidium
disinfection
Bacteriological methods
& Sampling
Viruses in water
-Virus methods
Distribution Wtr.Qual.
-Microbial growth
-Species identification
-Pigmented organisms
-Assimilable organic
carbon
-Biodegradable organic
carbon
Wet-Weather Flow(WWF)Resh.
Watershed management
strategies, including
urban hydrology
-WWF characterization
• -Drainage system design
-Treatment technologies
-Disinfection
-WWF best management
practices
-WWF modeling
-WWF databases
-Eutrophication
Infrastructure Rehab.Research
-Releases in water supply
systems
-Infiltration/inflow in
wastewater systems
System Evaluation Costs
-Organic field activities
-Unit process modeling
-Distribution systems
-Point-of entry/Point-of
use
-Disinfection byproducts
-Mutagenacity
-Small systems
-Computer activities
-Telemetry
-Geographic Information
Systems
-Constructed Wetlands
-Watershed Hydrology
-Environmental Technology
Initiative (ETI)
243
•&U.S. GOVERNMENT PRINTING OFFICE: 1998 -650-070/60033
-------
-------
|