ANALYSIS OF THE 1998 WATER-DISTRIBUTION SYSTEM
SERVING THE DOVER TOWNSHIP AREA, NEW JERSEY:

Field Data Collection Activities and
Water Distribution System Modeling

AUTHORS

Morris L. Maslia, MSCE, P.E.
Project Officer, Exposure-Dose Reconstruction Project
Division of Health Assessment and Consultation
Agency for Toxic Substances and Disease Registry

Jason B. Sautner, MSCE
Post-Graduate Research Fellow
Oak Ridge Institute for Science and Education
Division of Health Assessment and Consultation
Agency for Toxic Substances and Disease Registry

Mustafa M. Aral, Ph.D., P.E.
Director, Multi-Environmental Simulations Laboratory
School of Civil and Environmental Engineering
Georgia Institute of Technology

For additional information write to:

Project Officer
Exposure-Dose Reconstruction Project
Division of Health Assessment and Consultation
Agency for Toxic Substances and Disease Registry
1600 Clifton Road, Mail Stop E-32
Atlanta, Georgia 30333

FOREWORD

The Agency for Toxic Substances and Disease Registry (ATSDR) and the New Jersey Department of Health and Senior Services (NJDHSS) are investigating the elevated incidence of childhood cancers in Dover Township, Ocean County, New Jersey. In 1996, ATSDR and NJDHSS developed a Public Health Response Plan in cooperation with the Ocean County Health Department and the Citizen's Action Committee on Childhood Cancer Cluster. The plan outlined a series of public health activities including an updating and detailed re-evaluation of childhood cancer incidence statistics, and assessments of potential environmental exposures in the community. In 1997, ATSDR and NJDHSS determined that an epidemiologic study was warranted, and that the study would include assessments of the potential for exposure to specific drinking water sources.

ATSDR developed a workplan in February 1997 to reconstruct historical characteristics of the water-distribution system serving the Dover Township area by using water-distribution system modeling techniques. The model chosen by ATSDR for this effort, EPANET, is available in the public domain and is described in the published scientific literature. To test the reliability of model simulations, investigators need historical or present-day data with which to compare model results. Lacking such data, investigators initiated a field-data collection effort to obtain pressure measurements, storage-tank levels, and system operation schedules (the on/off cycling of pumps and wells) during winter-demand (March 1998) and peak-demand (August 1998) operating conditions. Using these data, ATSDR investigators calibrated the water-distribution system model to present-day (1998) conditions.

This report, therefore, presents and describes the following: (1) data gathered during field tests conducted in March and August 1998, (2) the development, calibration, and testing of the water-distribution system model for 1998 conditions, (3) a water-quality simulation of a naturally occurring conservative element, barium, to further test the reliability of the model calibration, and (4) the simulation of the proportionate contribution of water from points of entry to various locations throughout the distribution system for 1998 conditions.

ABSTRACT

The New Jersey Department of Health and Senior Services (NJDHSS) and the Agency for Toxic Substances and Disease Registry (ATSDR) are conducting an epidemiologic study of childhood leukemia and nervous system cancers that occurred in the period 1979 through 1996 in Dover Township, Ocean County, New Jersey. The epidemiologic study is exploring a wide variety of possible risk factors, including environmental exposures. ATSDR and NJDHSS have determined that completed human exposure pathways to groundwater contaminants have occurred in the past (through private and community water supplies) in some parts of the community. Because of this, ATSDR is developing a water-distribution model using the EPANET software to assist with environmental exposure assessment in the epidemiologic study. Results obtained from the model will be considered as one of the risk factors in the epidemiologic investigation.

As an important first step, the model was calibrated to the present-day (1998) water-distribution system characteristics. Pressure data were gathered simultaneously at 25 system hydrants using continuous pressure recording data loggers during field tests in March and August 1998. Data for storage-tank water levels, system demand, and pump and well status (on/off cycling) were also obtained. Data collected during the March 1998 test represent low, winter-time demand of 7.6 million gallons per data (MGD) and data collected during the August 1998 test represent peak-demand of 16.1 MGD. Measured pressure data and model simulations indicate that for the water-distribution system, pressures range from 40 pounds per square inch (psi) to slightly more than 100 psi for both tests.

The model network consists of 16,071 pipe segments ranging in diameter from 2 inches to 16 inches, and 14,987 junctions (nodes). The model was calibrated using the present-day (1998) data collected during winter-demand conditions of March 1998. The calibration was then tested against data collected under peak-demand conditions during August 1998. The absolute difference between measured and simulated hourly average pressures (pressure difference) for all measurement locations for March 1998 ranges 1.4-5.3 psi, and for August 1998, ranges 2.9-6.6 psi. For the March 1998 test and simulation, analysis indicates that 90% of hourly data at all test locations have an absolute pressure difference of approximately 5 psi or less. For the August 1998 test and simulation, analysis indicates that 90% of hourly data at all test locations have an absolute pressure difference of approximately 7.5 psi or less. These small pressure differences support the assertion that the model is calibrated and an acceptable and reliable representation of water-distribution system conditions during 1998.

As further evidence of the reliability of the model calibration, a simulation of the transport of a naturally occurring conservative element, barium, was conducted and compared with data gathered at 21 schools and 6 points of entry to the water-distribution system for March and April 1996. Measured concentrations of barium ranged 13-51 micrograms per liter (mg/L). Comparison of measured and simulated barium concentrations at the 21 school locations indicates a difference ranging 0.2-12.4 mg/L, which results in a mean relative difference of 13.6% with a range of 0.6-25.6%. Additional analyses comparing measured and simulated concentrations of barium show a geometric bias of 0.93, indicating a slight under prediction by the model (1.00 indicates perfect agreement), and a correlation coefficient of 0.81, indicating a high correlation between measured concentrations and simulated values. Therefore, this water-quality simulation is further evidence that the model is reasonably calibrated and an acceptable representation of the present-day water-distribution system characteristics.

The calibrated model of the water-distribution system makes it possible to conduct trace analyses for each point of entry (well fields) to the distribution system. These analyses provide an estimate of the percentage of water that any location of interest receives from the 8 points of entry to the distribution system. The results are presented in a series of 10 maps, a graph, and a table showing the percentage of water contributed by specific wells and storage tanks to locations in the Dover Township area for 1998 conditions. Based on residence histories, the trace-simulation results will be used in an epidemiologic investigation to estimate exposure of participants to specific water sources by determining the percentage of water they may have received from each of the points of entry to the distribution system.

INTRODUCTION

The Agency for Toxic Substances and Disease Registry (ATSDR), an agency of the United States Department of Health and Human Services, is required, among several other congressional mandates, to evaluate the public health threat of hazardous waste sites using environmental characterization data, community health concerns, and health outcome data. In the spring of 1996, ATSDR and the New Jersey Department of Health and Senior Services (NJDHSS) began to investigate health concerns of the Dover Township, Ocean County, New Jersey, community. In particular, community members feared that exposure to environmental contaminants from the area's hazardous waste sites, including two National Priorities List (Superfund) sites (Plate 1) was related to the elevated incidence of childhood leukemia and brain and central nervous system (CNS) cancers.

In 1997, ATSDR and NJDHSS began designing a case-control epidemiologic study of childhood cancers that occurred in the period 1979 through 1996 (Berry and Haltmeier 1997) in Dover Township. In a case-control study, a population is delineated and cases of diseases arising in that population over a specified time period are identified. The exposure experiences of the case group are compared to the exposure experience of a sample of the non-diseased persons in the population from which the cases arose. Exposures that are more common among the cases may be considered as possible risk factors for the disease (Rothman and Greenland 1998).

The study, which began data collection in 1998, is exploring multiple possible risk factors, including environmental exposures. One of the environmental factors of community concern that is being investigated in the study is the potential for exposure to certain drinking water sources. ATSDR and NJDHSS have determined that completed human exposure pathways to groundwater contaminants have occurred in the past (through private and community water supplies) in some parts of the community (NJDHSS 1999a, b, c).

To assist with the exposure assessment component of the epidemiologic study, ATSDR is developing a water-distribution model using the EPANET software (Rossman 1994) to reconstruct historical patterns of water distribution. Given the paucity of historical contaminant-specific concentration data during the time frame relevant to the epidemiologic study, ATSDR and NJDHSS have determined that modeling would estimate the percentage of water that a study subject might have received from each of the points of entry to the water-distribution system (Plate 2). This would allow epidemiologists to assess the association between the occurrence of childhood cancers and exposure to each of the sources of potable water entering the distribution system, including ones known to have been historically contaminated.

A detailed literature review of epidemiologic investigations relating water-supply contamination with health effects is beyond the scope of this report. However, a brief review is provided below. Lagakos et al. (1986) describe an association between exposure to trichloroethylene-contaminated drinking water and increased prevalence of stillbirths and CNS defects, oral defects, and chromosomal defects. To investigate the potential reproductive health effects of long-term, low-dose exposure to waterborne chloroform, Kramer et al. (1992) conducted population-based case control analyses to study the association of trihalomethanes with low birth-weight, prematurity, and intrautarine growth retardation using state of Iowa birth certificate data. Bove et al. (1995) used environmental and birth-outcome databases for a four-county area in northern New Jersey to study the effects of public drinking water contamination on birth outcomes.

This report will focus on the four aspects of the overall exposure assessment effort, being conducted jointly by ATSDR and NJDHSS, that will eventually use a calibrated model for historical reconstruction of the hydraulic characteristics of the water-distribution system. These aspects are: (1) data gathered during field tests conducted in March and August 1998, (2) the development, calibration, and testing of the water-distribution system model for 1998 conditions, (3) a water-quality simulation of a naturally occurring conservative element, barium, to further test the reliability of the model calibration, and (4) the simulation of the proportionate contribution of water from points of entry to various locations throughout the distribution system for 1998 conditions.

BACKGROUND

To reconstruct the historical flow of water from different sources into and through a system of interconnected pipelines, we have chosen to use a water-distribution system model. In a distribution system such as the one serving residents of the Dover Township area, not all public-supply wells were contaminated. Furthermore, some supply wells affect certain areas more than other wells do. Thus, at any given point in the distribution system, water may be derived from one or more sources in differing proportions, i.e., the concept of "proportionate contribution." Therefore, a water-distribution model is a useful tool to estimate the "proportionate contribution" of water sources through time.

Because the focus of the epidemiologic investigation is on children, exposure at residential locations is deemed as the most important exposure source to investigate, although other exposure sources may be present. Based on residence histories, reconstruction of historical water-distribution system characteristics can be used to estimate exposure to specific water sources by determining the percentage of water study subjects may have received from each of the points of entry (i.e., well fields) to the water-distribution system.

Table 1. Description water-distribution system tanks, wells, and pumps, Dover Township area, New Jersey

¹Data provided by United Water Toms River, Inc.
²Aquifer ID: KC, Kirkwood-Cohansey; PP, Piney Point; PRM, Potomac/Raritan/Magothy
³Not applicable.

The use of water-distribution system modeling for estimating exposure has been described in the literature by several investigators. Murphy (1986) calculated exposure to TCE from Wells G and H in Woburn, Massachusetts, by using a water-distribution model to assess various pumping and water use configuration patterns during each month that the wells were in operation. Clark et al. (1991) and Geldreich et al. (1992) used extended period simulation hydraulic and dynamic water-quality models to investigate the distribution of occurrences of illness due to waterborne contaminants (escherichia coli sterotype 0157:H7) found in the Cabool, Missouri, distribution system. Clark et al. (1996a, b) used the EPANET water-distribution system model (Rossman 1994) to develop several scenarios to explain possible pathogen transport of waterborne Salmonella typhimurium outbreak in the Gideon, Missouri, municipal water system. Aral et al. (1996) and Aral and Maslia (1997) used the EPANET water-distribution system model in conjunction with a geographic information system (GIS) to simulate four exposure scenarios for the Southington, Connecticut, water-supply system that used groundwater contaminated with VOCs during the 1970s. For the current investigation, the EPANET water-distribution model, integrated with spatial analysis technologies, is being used to model the water-distribution system serving the residents of the Dover Township area. A description of the model is presented in a subsequent section of this report.

Water-distribution system modeling can be used in a predictive sense such as scenario testing described in Aral et al. (1996) and Aral and Maslia (1997), or as a diagnostic tool such as finding the cause for disease outbreak described by Clark et al. (1996a, b) and Geldreich et al. (1992). For the current study, we will eventually be using the model in a diagnostic mode-reconstructing historical water-distribution system characteristics. To accomplish this, it is critical that investigators understand the reliability of model generated (or simulated) results. To assess the reliability of a model, a calibration process is undertaken. That is, investigators must be able to quantify the difference between measured parameter values (e.g., pressures, storage tank water levels) and simulated parameter values. (Details of the calibration process are described in the section on "Hydraulic Model Calibration.") Thus, the first step in our investigation is to establish the reliability of the water-distribution system model for the Dover Township area by undertaking a calibration process. Once the reliability of the model has been established, then the model can be used in a diagnostic mode to examine (reconstruct) historical characteristics of the water-distribution system serving the Dover Township area.

Description of the Water-Distribution System

The water-distribution system being analyzed has been operating since 1897 and is currently operated by United Water Toms River, Inc. (UWTR). It serves the residents of Dover Township, New Jersey, and communities outside of Dover Township including the borough of South Toms River and a portion of Berkeley Township (Plate 2). At the end of 1997, the water-distribution system served a population of 92,160 that consisted of 44,510 customers. The distribution system consists of 488.2 miles (mi) of mains, ranging in diameter from 2 inches (in.) to 16 in., 3 elevated and 6 ground-level storage tanks with a total rated storage volume of 7.35 million gallons (Mgal), 23 municipal groundwater wells in 8 well fields (or points of entry) with a total rated capacity of 27 million gallons per day (MGD), and 12 high service or booster pumps (Board of Public Utilities 1997). A list and description of the water-distribution system tanks, wells, and pumps serving the Dover Township area is provided in Table 1. In the distribution system as presently configured (1998), 7 of these wells pump directly into the distribution system (e.g., wells 20, 31), whereas the remaining 16 wells are used to fill storage tanks (e.g., Parkway well field ground-level storage tank) and then high service or booster pumps are used to supply the distribution system with water from the storage tanks (Plate 2, Table 1).

Demand for water in the Dover Township area is characterized by two typical demand patterns. A winter-time demand pattern, typical of data collected in March 1998 (Figure 1A), generally occurs from October through mid-May. Data collected in March 1998, show that demand was equal to 7.6 MGD (Sautner and Maslia 1998). These data were obtained from the water utility's supervisory control and data acquisition (SCADA) system and recorded by ATSDR staff stationed in the water utility's control room during the test. Peak-demand conditions, typical of field data collected in August 1998 (Figure 1B), and equal to 16.1 MGD, generally occur during the summer season from the end of May (Memorial Day) through September (Maslia and Sautner 1998b). Thus, for the water-distribution system serving the Dover Township area, average annual demand, based on data obtained during the tests from the water utility, is approximately 12 MGD.

Initial Model Simulation

In February 1997, ATSDR was provided with a database by the water utility (UWTR) that was used to described an equivalent pipe (or hydraulic) network of a 1993 water-distribution system network. (The 1993 distribution system was configured similarly to the present-day [1998] system, and therefore, the present-day system shown on Plate 2 can be used as a reference.) An equivalent pipe network is one in which smaller diameter pipes, bends, and valves are eliminated from the network and replaced by a simple pipe of uniform diameter. In this equivalent pipe or hydraulic network, the head loss and discharge are the same as the head loss and discharge in the multiple pipe network (Bhave 1991). Thus, equivalent pipe networks can be used to model the generalized characteristics of a water-distribution system. The equivalent hydraulic network constructed by the water utility for the Dover Township area consisted of approximately 950 pipeline segments ranging in diameter from 6 in. to 16 in. This equivalent hydraulic network represented, according to the water utility (Flegal 1997), the salient characteristics of the 1993 water-distribution system network, and is shown in comparison with the existing 1998 water-distribution system network on Plate 3.

Figure 1: Time pattern for water usage: (A) winter-time demand, March 24-25,1998, and (B) peak-demand, August 14-15, 1998

Using the equivalent hydraulic network database provided by the water utility in conjunction with information on storage tanks, wells, and pumps (Table 1), initial simulations were conducted using the EPANET water-distribution system model (Rossman 1994) to simulate typical hydraulic conditions for 1993^(1),(2). Evaluation of the results caused three concerns:

1. Simulated pressure using the equivalent hydraulic network in the southern Dover Township, the Berkeley Township, and the borough of South Toms River areas appeared to be unusually high; many values were near 110 pounds per square inch (psi) and some exceeded 125 psi;
   
   
2. The goal of the ATSDR and NJDHSS investigation is to relate study subjects to pipeline segments and sources that may have historically serviced their water-supply needs; therefore, using a skeleton of the distribution system represented by the equivalent hydraulic network could result in misclassification; and
   
   
3. The distribution system has expanded substantially since the early 1960s; using the equivalent hydraulic network to reconstruct characteristics of historical water-distribution system networks could result in residence areas of study subjects being serviced by an unrealistically limited number of pipelines, again giving rise to the possibility of misclassification.

Because of these concerns, ATSDR investigators decided to refocus their analysis of the water-distribution system on the actual or "street-level" distribution system, which is characterized by pipelines ranging in diameter from 2 in. to 16 in. Therefore, a model would need to be developed using all "street-level" pipelines as opposed to the limited number of pipelines of the equivalent hydraulic network (see Plate 3 for a comparison of "street-level" pipelines as of 1998 with the equivalent hydraulic network pipelines). In addition, ATSDR determined that it would be necessary to obtain field data for 1998 conditions to calibrate the water-distribution system model and confirm or negate the unusually high pressures simulated by the equivalent hydraulic network model in the southern Dover Township, Berkeley Township, the borough of South Toms River areas. (A detailed discussion of model calibration is provided in the section on "Hydraulic Model Calibration.")

Technical Work Group and Expert Panel Review

Throughout this investigation, ATSDR has sought outside technical input and expert peer review for this effort. In September 1997, ATSDR convened a technical workgroup to review the initial model simulations described above. On the basis of their discussions and review, the following recommendations were made (ATSDR 1999, p. 10):

1. Model simulated pressures using the water utility's "skeletonized" or equivalent hydraulic network in southern Dover Township and in the South Toms River areas appear to be exceedingly high . . . thus a "reality check" is needed to either confirm or negate model simulated pressures;
   
   
2. To use the model for simulating present-day conditions and reconstructing historical conditions, a set of spatially distributed pressure measurements, occurring under varying operating conditions, should be obtained;
   
   
3. To use the model to simulate water-quality characteristics, a set of water-quality calibration data would be needed;
   
   
4. The equivalent hydraulic network provided to ATSDR by the water utility should be refined down to a "street-level" network; and
   
   
5. As much of the network as possible should be geo-referenced.

After implementing items (1), (2), (4), and (5) of the recommendations above, ATSDR held an expert panel meeting in December 1998 (ATSDR 1999). The meeting convened nine scientific and technical experts from academia, government, private consulting, and industry, who discussed the status of ATSDR's water-distribution system model and its intended use. The salient recommendations resulting from the meeting can be found in an ATSDR report (1999, p. 34) and are summarized below:

1. ATSDR should re-calibrate the model using additional available data about pumping schedules during the March and August 1998 pressure tests and make appropriate modifications if needed;
   
   
2. Because the UWTR system contains a large amount of polyvinyl chloride pipe, roughness coefficients are not believed to be an especially significant factor for modeling; to test this assumption, a sensitivity analysis should be performed on the effect on water flow of roughness coefficients in the UWTR system;
   
   
3. ATSDR must fully define system operating rules, including those for filling tanks in the early morning hours; once defined, these rules must be incorporated into the model;
   
   
4. The distribution system model should be operated for sufficient time to reach dynamic equilibrium for the various demands that characterize the system; and
   
   
5. A tracer should be used to generate data on transport of water in the distribution system from points of entry; ideally the selected tracer would be introduced at entry points of the wells through which contaminants are suspected of entering the distribution system.

ATSDR's implementation of recommendations from the technical work group and expert panel review follows in the remaining sections of this report. Details of field collection activities, model development, and procedures used to calibrate and further test the water-distribution system model for the Dover Township area are presented below. The calibrated model is believed to be a reasonable representation of the present-day (1998) characteristics of the water-distribution system operating under winter-demand (March 1998) and peak-demand (August 1998) conditions.

Next Section     Table of Contents